ECONOMIC OPTIMIZATION FOR PRODUCT SEARCH RELEVANCY
In one embodiment, a method is illustrated as including defining a set of perspective objects capable of being placed onto a modified web page, monitoring parameters of a web page, the parameters including a number of times a current object is executed on the web page, using an Artificial Intelligence (AI) algorithm to determine a perspective object with a preferred Return On Investment (ROI), and selecting the perspective object to be placed onto the modified web page.
Latest Patents:
This application is a divisional of U.S. patent application Ser. No. 11/821,928, filed on Jun. 26, 2007, which is a non-provisional patent application that is related to the United States Patent Application titled: “DETERMINING RELEVANCY AND DESIRABILITY OF TERMS” (application Ser. No. 11/679,973), both of which are incorporated by reference in their entirety.
TECHNICAL FIELDExample embodiments relate generally to the technical field of commercial uses of business rule algorithms implemented on a computer and, in one specific example, the use of such algorithms to display web pages.
BACKGROUNDWeb pages, and the business rules that govern them, are typically static in nature. That is, the logic dictating when certain web pages are to be displayed in response to a user's query must be re-written every time a designer of a web page would like to have the user view new web pages. The same holds true for the objects or widgets displayed on a web page. Namely every time a new object is introduced on a web page, the logic governing the page upon which the new objects are introduced must be changed.
Some embodiments are illustrated by way of example and not limitation in the figures of the accompanying drawings in which:
In the following description, for purposes of explanation, numerous specific details are set forth to provide a thorough understanding of example embodiments. It may be evident, however, to one skilled in the art, that the present invention may be practiced without these specific details.
In one embodiment, the method and system illustrated herein may be thought of as a series of modules executed through a feedback loop to optimize a web page, and the selection of screen objects and widgets (collectively objects) to be displayed thereon. The data provided to this feedback loop, in the form of a content request, serves as a training data set for training particular types of AI algorithms. This training data set may then be used by a plurality of AI algorithms for the purpose of making predictions as to which objects or screen widgets would be best suited for the purposes of optimizing a web page. The concept of “optimizing” may be based upon such concepts as Return On Investment (ROI) values, or other type of values.
In some embodiments, web pages, and the object contained thereon, are optimized (e.g., optimizing) using certain algorithms (e.g., AI algorithms) so as to maximize the ROI for each web page. These algorithms may use histograms containing aggregations of data relating to, for example, keyword usage, traffic optimizer module usage, and/or object/widget usage for a particular webpage. Put another way, this histogram may be used as the basis upon which to determine the allocation of space (e,g., floor plan space) within an optimized web page. In some embodiments, a web page is akin to a piece of real estate, and, in one embodiment, the system and method illustrated herein seeks to maximize the ROI for the displayed area within a web page. The histogram may be used to chart a particular user profile by aggregating data relating to the particular user. These histograms may then be combined together so as to create a set of user data that can then be analyzed for certain trends so as to determine an ROI for a particular allocation of space on a web page and respective user or group of users.
Some embodiments may include ROI understood as, for example, the difference between the cost of generating a web page (e.g., labor costs associated with coding the web page, modifying the page to add or remove content from the web page, presenting search results, cost of access to system resources in terms of time, memory, Central Processing Unit (CPU) cycles etc., making a web page scalable etc.), and the revenue generated by the web page. Example embodiments may include the business rules governing such web pages that are optimized in an automated or an organic manner such that rather than being static, the web pages and content contained therein are dynamic, reflecting, for example, the information relating to the optimal way to present a web page in order to market a particular good (e.g., diamonds) or service. For example, rather than having a web page always contain pictures of diamonds, a web page may only contain such picture, if it is shown that web pages containing such pictures are more successful in selling diamonds than web pages without such pictures. And again, rather than, for example, having a web page only contain pictures of diamonds, certain objects may also be presented on the page, where the combination of the pictures of diamonds and these screen objects yields a higher ROI than a picture with pictures of diamond alone. The determination of what constitutes success, however, may, in some cases, be based upon certain algorithms seeking to maximize ROI for a web page, rather than strictly based upon the decisions of a marketing professional. Further, success may be a preferred ROI such that a high, lower, or some other suitable ROI is sought based upon the desires of a user, or other implementing the system and method illustrated herein.
Web pages are typically governed by static business rules (e.g., logics) such that anytime a new web page is introduced, a web page containing additional objects, or most any other modification is made, these business rules must also be changed. This results in many hours being spent by designers of web pages and websites changing a web page or web site to meeting the marketing needs of a particular company. In the commercial setting (e.g., the e-commerce setting), the time spent by designers of web pages to add or remove a web page, object, or widget contained on a web page can be very costly financially to the party paying for the services of the designer. More to the point, not only may the designer have to write the new web page, but they may also have to write new business rules to govern this new web page. For example, if a new object in the form of a new link is added to an existing web page, then the scripting language governing this page (e.g., JavaScript, VBScript) may also have to be rewritten.
Further, from an e-commerce marketing perspective, the approach of having a designer rewrite the business rules is always done after a new marketing scheme or approach has become “predominant.” That is, in many cases, a designer of a web page is only instructed to modify a web page after a particular use of a web page has been observed to be advantageous from a marketing perspective. This observation, however, may not occur for some period of time and only after several types of analysis and several layers or reviews have occurred. For example, if it is observed that sales of diamonds are more successful on a web page that provides objects or widgets that allow a user to better understand cut, clarity, and color, then the designer may be instructed to generate web pages containing such objects or widgets. However, this observation may only occur after the analysis, by a marketing or financial analyst, of sates figures for pages containing such objects or widgets. Only when such an analysis is conducted can a designer be asked to make changes to the web pages such that they contain these objects or widgets. Such an approach costs an e-commerce site valuable time during which they could have been implementing the objects or widgets relating to cut, clarity, and color.
A System for Economic Optimization for Product Search RelevancyIn one embodiment, the example system 400 may be thought of as a series of modules executed through a feedback loop, such as a feedback loop 418. The data provided to this feedback loop 418 in the form of a content request serves as a training data set for training particular types of AI algorithms. Specifically, this data set may then be used by a plurality of AI algorithms for the purpose of making predictions as to which objects or screen widgets would be best suited for the purposes of optimizing a web page. The concept of “best suited” may be based upon such things as ROI values or other type of values. This feedback loop 418 may be more fully illustrated below. Once the batch module 405 is executed, a histogram is generated that contains aggregated data that outlines which traffic optimization modules are to be used in response to the content request 102. Further, contained within this histogram is a description of the various screen objects or widgets that may be used during the course of the generation of an optimized web page 112. In some embodiments, an individual or even a series of histograms is then stored in a demand service backend 406. This demand service backend 406 may be any one of a number of commercially available database applications and may even include the previously illustrated On Line Analytical Processing (OLAP) database and/or a relational or object relational database.
Some embodiments may include the histogram, or series of histograms, being stored into the demand services backend 406, for use by a traffic optimizer engine 409. This traffic optimizer engine 409, in some cases, makes a data request of the demand service backend 406 seeking to request the histograms that are stored within the demand services backend 406. This request may be in the form of an Structured Query Language (SQL) query, an Multidimensional Expression (MDX) query, or some other suitable type of query written in a query request language. Additionally, the traffic optimizer engine 409 may make a request of a search backend 408 wherein the search backend 408 contains a variety of data relating to items that may be, for example, offered for sale by the website supported by this system 400.
In some embodiments, the data stored into the search backend 408 may also be optimized using a classifier 407, where a classifier 407 uses any one of a number AI algorithms to optimize the various items for sale and categories relating to items for sale. Further, the traffic optimizer engine 409 may be governed by a configure rule set 416 generated by, for example, a system administrator 415 where this configure rule set 416 is provided to rules engine 417. In one embodiment, the rules engine 417 dictates to this traffic optimizer engine 409 certain rules that cannot be violated through utilization of various AI algorithms. Put another way, the rules engine 417 dictates to the traffic optimizer engine 409 certain ways in which an optimized web page may not be optimized. These rules provided by the rules engine 417 to the traffic optimizer engine 409 may relate to, for example, certain advertisements that must be paired with a particular web page or may seek to enforce other types of contractual arrangements that a particular business implementing, or otherwise using this system, must adhere to in the providing of an optimized web page 112.
Once a traffic optimizer engine 409 retrieves a histogram containing aggregated data relating to screen objects or widgets, it provide this data set (e.g., the item data set and the screen objects data set) to one or more traffic optimizer modules. These traffic optimizer modules may seek to optimize certain portions or aspects of a web page 112. For example, a page type module 410 may seek to optimize the type of page that is presented to the user for viewing. Further, a merchandising module 411 may seek to optimize (e.g., to optimize the contents of merchandising placements) a particular web page, such as 112, based upon the type of screen objects or widgets that are actually presented on that page. For example, the screen objects or widgets 305 through 307 may result from an application of a merchandising module 411. Further, an ad module 412 seeks to optimize for the purposes of advertisements. For example, the advertising object 304 may reflect this application of this ad module 412, where advertisements related to a search query are placed on an optimized web page such as 302. Further, a navigation module 413 may seek to insert certain links into an optimized web page for the purposes of allowing navigation to related web pages. As with the decisional engine 412, each of these various traffic optimizer modules (e.g., 410-414) may use various AI algorithms in combination with, for example, certain decisional logic as embodied in a decision tree. In some embodiments, each of the traffic optimizer modules (e.g., 410-414) may optimize specific screen objects or widgets related to their particular illustrated area or subject matter. For example, the page type optimizer 4110 may only optimize certain screen objects or widgets in the form of, for example, various available CSS that may be used to optimize a page, whereas the ad module 412 may seek to only optimize certain screen objects or widgets in the form of advertisements.
Once the various traffic optimizer modules have been selected by the traffic optimizer engine 409 and have, in turn, optimized a web page (e.g., 112, or 302) based upon their particular area of application (e.g., page type optimization, merchandising optimization, ad optimization, navigation optimization), a page optimization instruction set is passed to the presentation module 401 wherein the presentation module 401 assembles the various optimized screen objects or widgets and presents them in the form of an optimized web page 112. The various functionalities of the above-illustrated modules (e.g., 401-418) are more fully outlined below.
In some embodiments, a method may optimize a web page using one or more Artificial Intelligence (AI) algorithms, wherein the one or more AI algorithms are optimized based upon a predefined criteria including ROI, and presenting the web page to a user to be viewed. As may be more fully illustrated below, these AI algorithms may be deterministic or non-deterministic in nature, and may be used by decisional engine 404, batch module 405, traffic optimizer engine 409, and/or any one of the traffic optimizer modules (e.g., 410-414). The predefined criteria for optimization may be ROI, but may also include user-segment base success or some type of dynamic measure of success. As previously illustrated, screen objects or widgets may be optimized, as may the layout of a web page and the items (e.g., items for sale) displayed on the web page.
An Architectural DescriptionIn some embodiments, a client-based browser application using a GUI, such as GUI 303, is implemented, whereas in some other embodiments, a command line interface is implemented. Some well-known client-based browser applications include NETSCAPE™, INTERNET EXPLORER™, MOZILLA FIREFOX™, OPERA™, or some other suitable browser application. Common to these browser applications is the ability to use HTTP or Secured Hyper-Text Transfer Protocol (HTTPS) to get, upload (e.g., PUT), or delete web pages and interpret these web pages, which are written in a Hyper-Text Markup Language (HTML) and/or an XML. HTTP and HTTPS are well known in the art, as are HTML and XML. HTTP and HTTPS are used in conjunction with a TCP/IP protocol as illustrated in the OSI model, or the TCP protocol stack model, both of which are well known in the art. The practical purpose of the client-based browser application is to enable a user to interact with the application through the display of plain text, and/or interactive, dynamic functionality in the form of page links, buttons, text boxes, scroll down bars, widgets, or other objects (collectively objects or widgets) contained on one or more web pages (e.g., optimized web pages 302 and 112) constructed using the aforementioned HTML and/or XML. Further, these object and widgets may be used to execute certain code components or modules (e.g., modules written as Java Enterprise Beans (EJB), Component Object Model (COM), or Distributed Component Object Model (DCOM) or some other suitable module type used to, for example, make search queries or to execute other functionality.
Some embodiments may include a user selection of screen objects using an input device (e.g., a mouse) to select and send an HTTP-based query (e.g., a content request) to a website. This content request may have the form of “http://antiques.acme.com/” or, for example, “http://books.actne.com/”. More sophisticated content requests may also be made wherein not only are web pages requested, but specific objects on web pages are requested, and the functionality associated therewith. The following are examples of these more sophisticated queries:
- http://books.listings.acme.com/Accessories_Address-Bookfromdatabase1
- http://books.listings.ebay.com/Accessories_Address-Book2fromdatabase2
With respect to these more sophisticated content requests, in some embodiments, objects or widgets that are a part of these requests (e.g., Accessories_Address-Book and Accessories_Address-Book2) are parsed and tracked based upon the number of times they are selected and the logic used (e.g., AI algorithms) to generate the web page containing the objects or widgets.
Some embodiments may include a method that further comprises listening to a traffic exchange between the user and a website, wherein the traffic exchange includes content request data relating to a specific transaction between the user and the website and capturing the content request data and storing into a data structure. The model 801 may conduct this listening, while the operation 802 may actually capture what is listened to by the operation 801. This listening may be over a network connection including an Ethernet connection or some other suitable network connection. Further, the capturing of the content request may be by way of a recording application that simply records content requests by a user such as user 101 and records such content request data.
In some embodiments, an additional field of data (not pictured) for each tuple entry is maintained in the table 901 that contains data providing a context within which the content request appearing in content request field 904 has been accessed. This context field may reference other pages (e.g., containing a link to them) that link to the web page upon which the content request 102 may appear. Moreover, this context field may include additional object, widgets, layout and other information present on web page upon which the content request 102 appeared.
Some embodiments may include a method wherein the table 901 or other suitable data structure contains data including a content request data relating to a traffic optimizer module used to generate the content request data, a name of an algorithm used to optimize the content request data, the content request data itself, and a value reflecting number of times the content request data has been selected by the user. In some embodiments, some other suitable data structure is used, including a tree, hash table, heap, or array.
Optimizing Presentation DecisionsIn some embodiments, an AI library 1005 is implemented so as to allow for a variety of optimization techniques and algorithms implementing such techniques. This AI library 1005 allows for a dynamic approach to the business rules implementing a website such that web pages, and the objects or widgets displayed thereon, may be presented based upon the criteria including ROI, the orientation of the particular site (e.g., a Japanese, American, or German site), the query (e.g., IPOD™, HARRY POTTER™), user segment (e.g., buyer, seller, etc.), category (e.g., consumer electronics, etc.), previous page viewed by user, referrer (e.g., from GOOGLE™, YAHOO™), a user's session behavior, results context (e.g., number of results, primary category, etc.), time of day, day of week, season of year, or some other suitable criteria.
The AI algorithms contained in this AI library 1005 may, in an example embodiment, fall into one of two categories of AI algorithms: deterministic or stochastic. In order to understand these two categories as they relate to the issue of users purchasing good or services off a website, it should be understood that a wholly deterministic world may be capable of analysis through one or more deterministic algorithms. However, where there exist actions of agents (e.g., users of the website) that do not lend themselves to prediction, then stochastic algorithms may need to be implemented by themselves, or, in combination with other deterministic, or stochastic algorithms. One distinguishing feature of stochastic algorithms, as compared to deterministic algorithms, is that the former may use some sort of nondeterministic feature such as random number generation, or random node selection (sec, e.g., genetic algorithms) in its optimization analysis, while the later avoids such random, non-deterministic features. This random number generation may allow for the avoidance of pitfalls such as slavish adherence to predictability criteria during the optimization process. Some embodiments may include the use of deterministic or stochastic algorithms in combination (see e.g., genetic fuzzy algorithms).
Some example embodiments may include any number of deterministic algorithms implemented in the AI library 1005, including case-based reasoning, bayesian networks (including hidden markov models), neural networks, or fuzzy systems. The bayesian networks may include: machine learning algorithms including—supervised learning, unsupervised learning, semi-supervised learning, reinforcement learning, transduction, learning to learn algorithms, or some other suitable bayesian network. The neural networks may include: kohonen self-organizing network, recurrent networks, simple recurrent networks, Hopfield networks, stochastic neural networks, boltzmann machines, modular neural networks, committee of machines, Associative Neural Network (ASNN), holographic associative memory, instantaneously, trained networks, spiking neural networks, dynamic neural networks, cascading neural networks, neuro-fuzzy networks, or some other suitable neural network.
In some embodiments, any number of stochastic algorithms may be implemented including: genetic algorithms, ant algorithms, tabu search algorithms, or monte carlo algorithms (e.g., simulated annealing). Common to these algorithms is the use of randomness (e.g., randomly generated numbers) to avoid the problem of bring unduly wedded to a local minima or maxima.
The decision to use one AI algorithm versus another may be based upon the relative ROI of optimizing web pages using one set of modules as compared to another set of modules. For example, while a projected ROI using a bayesian network to optimize a set of web pages my be $75,000, the projected ROI, using a fuzzy genetic algorithm may be $80,000, assuming equal costs. In some embodiments, the various AI algorithms are used to optimize the individual nodes of a decision tree.
Some embodiments may include a decision tree generated from or optimized by one or more of the above illustrated AI algorithms applied separately or in combination. A decision tree, for example, allows for the modeling of decisions and their possible consequences. For example, if a first combination of modules (e.g., the merchandising module 411 and the ad module 412) allows for a larger ROI relative to cost than a second combination of modules (e.g., navigation module 413 and page type optimizer module 410), then the first combination may be the preferred combination to implement.
In some embodiments, the method 404 further generates a decision tree at operation 1002 to determine which traffic optimizer modules are best suited to optimize the web page, and operations 1004-1006 to optimize the decision tree using the one or more AI algorithms to determine an selected traffic optimizer module set, an operation 1007 to select the traffic optimizer module set, and an operation 1008 to transmit the selected traffic optimizer module set. As previously illustrated, these AI algorithms may be deterministic or non-deterministic algorithms.
- “Z” 1101 is an initial cost of investment in the building of a web page, say $20,000 covering, for example, cost associated with serving data, serving a UI, and facilitating the scaling of usage, and the like;
- “A” 1102 is a first possible ROI over some period of time (e.g., two weeks) in the amount of $100,000 using optimized web pages generated by way of the merchandising module 411 used in combination with the ad module 412;
- “B” 1105 is a second possible ROI over some period of time (e.g., two weeks) in the amount of $75,000 using optimized web pages generated by way of the navigation module 413 used in combination with the page type module 410;
- “Y” 1103 is a probability value, say 60%, that the ROI for node “A” 1102 occur;
- “C” 1104 is a probability value, say 40%, that the ROI for node “A” 1102 occur;
- “E” 1106 is a probability value, say 30%, that the ROI for node “B” 1105 occur;
Given the above nodes and associated values, it is clear that by selecting node “A” 1102 that one is provided a high probability of a greater ROI than by selecting node “B” 1105. Assume, however, that through applying an AI algorithm to the decision tree, a new node, call it “G” 1107 is generated, wherein “G” 1107 represents a probability value, say 70%, that the ROI for node “B” 1105 may occur. Given the new node “G” 1107, the combination of optimizing the navigation module 413 with the page type module 410 may be an optimal decision, relative to choosing node “A” 1102. In one example embodiment, a bayesian network is used to predict the probability for a ROI given some combination of modules (e.g., modules 410-414).
Some embodiments may include generating a decision tree who nodes reflect the comparative ROI of using individual modules or even a plurality of modules (e.g., modules 410-414). For example, rather than building a decision tree based upon ROI as reflected in the combining of two modules, and comparing this ROI against the ROI for two other modules that have been combined, a decision tree may be built that reflects the ROI based upon one module, a module that is then compared to another module and its associated ROI. Other comparisons and combinations are also possible including, for example, basing ROI on the combination of more than two modules and comparing this ROI against the ROI for one or more than two modules.
Example embodiments may also include building a decision tree that compares the ROI of using one or more of the modules (e.g., modules 410-414) to the ROI of letting a third party provide the services otherwise provided by one or more of the modules. For example, rather than using the ad module 412 as a basis upon which to determine the ROI using this module by itself or in combination with other modules, an ROI for advertising may be used upon using data provided by a third party advertiser who, for example, provides the content to be placed within the available space on a web page. In some embodiments, however, these services may be provided through the module (e.g., module 412).
- “P” 1201 is a probability value of 10% that optimized web pages generated using a navigation module 413 in combination with the page type module 410 may result in a ROI of greater than or equal to $75,000;
- “M” 1202 is a probability value of 10% that optimized web pages generated using a navigation module 413 in combination with a merchandising module 411 may result in a ROI of greater than or equal to $75,000;
- “I” 1203 is a probability value of 10% that optimized web pages generated using a navigation module 413 in combination with an ad module may result in a ROI of greater than or equal to $75,000;
- “G” 1107 is a probability value of 10% that optimized web pages generated using the navigation module 413 by itself may result in a ROI of greater than or equal to $75,000;
Given the above nodes and associated values, one result of selecting nodes “P” 1201, “M” 1202, “I” 1203, and “G” 1107 (e.g., combining probability values arising from the use of each of the referenced modules to generate a web page) is that there may be a 40% chance that an ROI of $75,000 may occur. Assume, however, that new data represented by a node “K” 1205 is added into the network, wherein “K” 1205 is a probability value of 10% that, by optimizing a web page using the navigation module 413 used in combination with the page type module 410, an ROI of $75,000 may occur. By inserting “K” 1205 into the path between “P” 1201 and “G”, there may exist a 60% probability value that the ROI may be $75,000.
In addition to bayesian networks, other types of AI algorithms may be available as a part of the AI library 1005 to optimize nodes of a decision tree built through operation 1004. For example, a fuzzy genetic algorithm may be applied that can be understood as incorporating the features of both a deterministic and stochastic algorithm. Applied in the present context, a genetic algorithm could be used to optimize the set of modules (e.g., 410-414) through selecting and combining successful modules (e.g., those with a high ROI value, relative to cost) using principles of crossover and mutation that incorporate certain concepts of randomness in the selection of these modules. This process of crossover and mutation may be repeated until diversity is established, based upon some predefined criteria. Once diversity is established, then a fuzzy associative matrix may be generated illustrating the success percentages for each of the combinations of modules based upon a particular data set (e.g., the ROI value over a period of time). An example fuzzy associative matrix is provided below:
This table may then be used to generate the decisional logic of each node (e.g., “B” 1105) for in a decision tree such that, for example:
- If the Merchandising Module/Page type Module is to be optimized and the optimization may more than likely result in the ROI 20%-80% of the time, then implement;
- If the Merchandising Module/Page type Module is to be optimized and the optimization may result 10%-20% of the time, but may not result 10%-20% of the time in ROI value, then implement only if most likely not a higher percentage for attaining an ROI value.
- In some embodiments, the values in the above matrix are inserted into a data structure such as an array, tree (e.g., a binary search tree, red-black tree), list, stack or query, wherein the type of data structure used may be dictated by the efficiency of searching the data structure for a given data set size. Further, a table may be generated for every combination of modules created using the genetic algorithm.
In some embodiments, the batching of histograms may result in a statistical view (e.g., a data distribution for past performance) of some set of persons (e.g., buyers) and their habits. For example, the batch of histograms may reflect the percentage of persons who purchased a particular item, given a certain number of advertisements appearing on a web page selling the item such that if two ads appear then there is a 33% chance of making a sale, while if four ads appear then there is a 66% chance of making a sale. And again, this statistical view may reflect, for example, a statistical relationship between searches performed in a certain category of items of the web page, and the probability for making a sale.
A further example of how batches of histograms may be used is in the case where histograms are combined and conclusions drawn based upon this combination. A first histogram containing data in the form of the number of times a set of key words have been used over a period of time could be compared to a second histogram containing data relating to sales of a particular item over the same time period. These two histograms may be combine such that certain conclusions could be drawn regarding the relationship between the data contained in the two histograms. Through knowing the probability of making a sale, or even more broadly an ROI value, the histogram or batch of histograms can be used by, for example, the traffic optimizer engine 409 and the various traffic optimizer modules 410-414 to determine how the space (e.g., the real estate on a web page may be divided up or organized to facilitate the highest ROI relative to other ways of organizing the web page.
Some embodiments may include a method including an operation 1701 for retrieving a batch of histograms and items to be optimized, an operation 1702 for categorizing each histogram of the batch of histograms based upon item type, an operation 1703 for filtering each histogram based upon a configuration rules set, an operation 1704 for sorting the histograms, and an operation 1705 for optimizing each category of histogram using one or more AI algorithm, wherein the optimizing is based upon some predefined criteria including ROI. The sorting of the histograms may be achieved using any one of a number of sorting algorithms.
In some embodiments, the traffic optimizer engine 409 may query the demand services back end 406 using a query language such as SQL, MDX, Data Mining Extensions (DMX), or some other suitable query language. These queries may seek results that optimize combinations of web page and modules, and/or optimal page and module, and object and/or widget combination. This optimization, in some case, may be based upon certain AI algorithm or combination of AI algorithms. An example query for an optimal page and module combination may have the form of:
This request can be explained in that using performance data for the US site (site 0) and the query “ipod” for each page variant (page type) returns the associated number of impressions (imp), the number of item pageviews (vi), and the number of bids (bid).
- In response to the above query, the demand service backend 406 may provide the following result:
In addition to making queries regarding a particular optimal page, module, and object and/or widget combination, other types of queries may be made. For example, a query may have the following form:
In response to such a query, the demand services backend 406 may return the following type result:
Some embodiments may include an operation 1801 for dividing the batch of histograms by optimization classification, wherein optimization classification includes classification by traffic optimizer module type, an operation 1802 for selecting a histogram from the batch of histograms, and operations 1803-1805 for applying an AI algorithm to the histogram to optimize the histogram and transmitting the histogram as an optimized histogram. In some embodiments, these operations 1803-1805 apply a genetic algorithm to optimize the histogram. In some cases, some other suitable algorithm is applied.
Example embodiments may include the use of operation 1705 to develop new functions utilizing a certain combination of histograms such that new relationships between the data contained within the histograms may be revealed. Further, new combinations of histograms may be generated such that such that these combinations may be provided to existing functions such that new relationships between the data contained in the histograms may be revealed. For example, a mapping of keywords to a product may be created based upon some probability value generated using an existing function. Or percentage values supporting a mapping of keywords to category of items for sale might be generated based upon a newly created function that is created specifically for the purpose of this mapping.
Some embodiments may include a method wherein the traffic optimizer module type (e.g., modules 410-414) is selected from the group consisting of page type, merchandising, ad, and navigation modules. Further, in some embodiments, an operation 1901 is implemented for receiving an optimized histogram, an operation 1902 fur optimizing object and widget data corresponding to a web page illustrated in the optimized histogram using the one or more AI algorithms, an operation 1903 for retrieving optimized object and widget data, and transmitting the optimized object and widget data as part of a page optimization instruction set. In some embodiments, the transmission is over a network (e.g., an internet) to a presentation module 401.
- “S” 2001 is a probability value of 10% that a PEG of a princess cut diamond on an optimized web page generated using an ad module 412 may result in a ROI of greater than or equal to $75,000;
- “L” 2002 is a probability value of 10% that a REG of a round cut diamond on an optimized web page generated using an ad module 412 may result in a ROI of greater than or equal to $75,000;
- “W” 2003 is a probability value of 10% that a PEG of a triangle cut diamond on an optimized web page generated using an ad module 412 may result in a ROI of greater than or equal to $75,000;
- “K” 2004 is a probability value of 10% that a Joint Photographic Experts Group (JPEG) photo of an oval cut diamond on an optimized web page generated using an ad module 412 may result in a ROI of greater than or equal to $75,000;
Given the above nodes and associated values, one result of selecting nodes “S” 2001, “L” 2002, “W” 2003, and “K” 2004 (e.g., combining the referenced modules) is that the most likely result may be a 40% chance that an ROI of $75,000 may occur. Assume, however, that new data represented by a node “Z” 2005 is added into the system, wherein “Z” 2005 is a probability value of 20% that by optimizing a web page using the ad module 412, used in combination with the page type module 410, that an ROI of $75,000 may occur. By inserting “Z” 2005 into the path between “S” 2001 and “K” 2004, there then exists a 60% pessimistic probability value that the ROI may be $75,000.
Some embodiments may include the various databases (e.g., 105, 106, 804, 1003, and 2203-2206) being relational databases, or in some cases On-Line Analytical Processing (OLAP) based databases. In the case of relational databases, various tables of data are created and data is inserted into, and/or selected from, these tables using a Structured Query Language (SQL) or some other database-query language known in the art. In the case of OLAP databases, one or more multi-dimensional cubes or hypercubes containing multidimensional data from which data is selected from or inserted into using MDX may be implemented. In the case of a database using tables and SQL, a database application such as, for example, MYSQL™, SQLSERVER™, Oracle 81™, 10G™, or some other suitable database application may be used to manage the data. In this, the case of a database using cubes and MDX, a database using Multidimensional On Line Analytic Processing (MOLAP), Relational On Line Analytic Processing (ROLAP), Hybrid Online Analytic Processing (HOLAP), or some other suitable database application may be used to manage the data. These tables or cubes made up of tables, in the case of, for example, ROLAP, are organized into a RDS or Object Relational Data Schema (ORDS), as is known in the art. These schemas may be normalized using certain normalization algorithms so as to avoid abnormalities such as non-additive joins and other problems. Additionally, these normalization algorithms may include Boyce-Codd Normal Form or some other normalization, optimization algorithm known in the art.
Component Design and Distributed Computing ModulesIn some embodiments, a method is illustrated as implemented in a distributed or non-distributed software application designed under a three-tier architecture paradigm, whereby the various components of computer code that implement this method may be categorized as belonging to one or more of these three tiers. Some embodiments may include a first tier as an interfere (e.g., an interface tier) that is relatively free of application processing. Further, a second tier may be a logic tier that performs application processing in the form of logical/mathematical manipulations of data inputted through the interface level, and communicates the results of these logical/mathematical manipulations to the interface tier, and/or to a backend, or storage tier. These logical/mathematical manipulations may relate to certain business rules, or processes that govern the software application as a whole. A third, storage tier, may be a persistent storage medium or, non-persistent storage medium. In some cases, one or more of these tiers may be collapsed into another, resulting in a two-tier architecture, or even a one-tier architecture. For example, the interface and logic tiers may be consolidated, or the logic and storage tiers may be consolidated, as in the case of a software application with an embedded database. This three-tier architecture may be implemented using one technology, or, as may be discussed below, a variety of technologies. This three-tier architecture, and the technologies through which it is implemented, may be executed on two or more computer systems organized in a server-client, peer to peer, or so some other suitable configuration. Further, these three tiers may be distributed between more than one computer system as various software components.
Component DesignSome example embodiments may include the above illustrated tiers, and processes or operations that make them up, as being written as one or more software components. Common to many of these components is the ability to generate, use, and manipulate data. These components, and the functionality associated with each, may be used by client, server, or peer computer systems. These various components may be implemented by a computer system on an as-needed basis. These components may be written in an object-oriented computer language such that a component oriented, or object-oriented programming technique can be implemented using a Visual Component Library (VCL), Component Library for Cross Platform (CLX), Java Beans (TB), Java Enterprise Beans (EJB), Component Object Model (COM), Distributed Component Object Model (DCOM), or other suitable technique. These components may be linked to other components via various Application Programming interfaces (APIs), and then compiled into one complete server, client, and/or peer software application. Further, these APIs may be able to communicate through various distributed programming protocols as distributed computing components.
Distributed Computing Components and ProtocolsSome example embodiments may include remote procedure calls being used to implement one or more of the above illustrated components across a distributed programming environment as distributed computing components. For example, an interface component (e.g., an interface tier) may reside on a first computer system that is remotely located from a second computer system containing a logic component (e.g., a logic tier). These first and second computer systems may be configured in a server-client, peer-to-peer, or some other suitable configuration. These various components may be written using the above illustrated object-oriented programming techniques, and can be written in the same programming language, or a different programming language. Various protocols may be implemented to enable these various components to communicate regardless of the programming language used to write these components. For example, a component written in C++ may be able to communicate with another component written in the Java programming language through utilizing a distributed computing protocol such as a Common Object Request Broker Architecture (CORBA), a Simple Object Access Protocol (SOAP), or some other suitable protocol. Some embodiments may include the use of one or more of these protocols with the various protocols outlined in the Open Systems interconnection (OSI) model, or Transmission Control Protocol/Internet Protocol (TCP/IP) protocol stack model for defining the protocols used by a network to transmit data.
A System of Transmission Between a Server and ClientSome embodiments may utilize the OSI model or TCP/IP protocol stack model for defining the protocols used by a network to transmit data. In applying these models, a system of data transmission between a server and client, or between peer computer systems is illustrated as a series of roughly five layers comprising: an application layer, a transport layer, a network layer, a data link layer, and a physical layer. In the case of software having a three tier architecture, the various tiers (e.g., the interface, logic, and storage tiers) reside on the application layer of the TCP/IP protocol stack. In an example implementation using the TCP/IP protocol stack model, data from an application residing at the application layer is loaded into the data load field of a TCP segment residing at the transport layer. This TCP segment also contains port information for a recipient software application residing remotely. This TCP segment is loaded into the data load field of an IP datagram residing at the network layer. Next, this IP datagram is loaded into a frame residing at the data link layer. This frame is then encoded at the physical layer, and the data transmitted over a network such as an interact, Local Area Network (LAN), Wide Area Network (WAN), or some other suitable network. In some cases, internet refers to a network of networks. These networks may use a variety of protocols for the exchange of data, including the aforementioned TCP/IP, and additionally ATM, SNA, SDI, or some other suitable protocol. These networks may be organized within a variety of topologies (e.g., a star topology), or structures.
Market Place ApplicationsIn some embodiments, web pages are optimized such that the business rules governing these pages are dynamic and allow die content on the website to adapt organically (e.g., using AI algorithms such as genetic algorithms) based upon the history of user preferences in the website. For example, rather than having static business rules governing the presentation of content on a site such that the same content (e.g., screen objects or widgets, and the presentation of these objects or widgets via, for example, a CSS) is most always displayed, screen objects and widget may be selected and displayed based upon their ability to satisfy a predetermined ROI value or some other metric for success. These screen objects or widgets may be combined with certain search methods such that an optimized web page may be provided to the user. This web page may change to reflect the ever-changing needs of users, such that rather than having to manually change the business rules implementing a web page, the rules may change automatically or dynamically to reflect current user's preferences. This ability to reflect these preferences may be facilitated through the use of certain AI algorithms used to anticipate user behavior.
A Computer SystemThe present invention is implemented on a digital processing system or computer system that includes a processor, which may comprise one or more processors and may include one or more conventional types of such processors (e.g., x86, x86-64, ARMx), such as an AMD processor, Intel Pentium or XScale processor, or other suitable processor. A memory is coupled to the processor by a bus. The memory may be a dynamic random access memory (DRAM) and/or may include static RAM (SRAM). The processor may also be coupled to other types of storage areas/memories (e.g., cache, flash memory, disk, etc.), which could be considered as part of the memory or separate from the memory.
In some embodiments, a bus further couples the processor to a display controller, a mass memory or some type of computer-readable medium device, a modem or network interface card or adaptor, and an input/output (I/O) controller. The display controller controls, in a conventional manner, a display, which may represent a Cathode Ray Tube (CRT) display, a Liquid Crystal Display (LCD), a plasma display, or other type of suitable display device. Computer-readable medium may include a mass memory magnetic, optical, magneto-optical, tape, and/or other type of machine-readable medium/device for storing information. For example, the computer-readable medium may represent a hard disk, a read-only or writeable optical CD, etc. In some embodiments, a network adaptor card such as a modem or network interface card is used to exchange data across a network such as an internet. The I/O controller controls I/O: device(s), which may include one or more keyboards, mouse/trackball or other pointing devices, magnetic and/or optical disk drives, printers, scanners, digital cameras, microphones, etc.
The present invention may be implemented entirely in executable computer program instructions that are stored on a computer-readable medium or may be implemented in a combination of software and hardware, or in certain embodiments, entirely in hardware.
Embodiments within the scope of the present invention include computer-readable medium for carrying or having computer-executable instructions or data structures stored thereon. Such computer-readable medium may be any available medium which is accessible by a general-purpose or special-purpose computer system. By way of example, and not limitation, such computer-readable medium may comprise physical storage media such as RAM, ROM, EPROM, CD-ROM or other optical-disk storage, magnetic-disk storage or other magnetic-storage devices, or any other media that can be used to carry or store desired program code means in the form of computer-executable instructions, computer-readable instructions, or data structures and that may be accessed by a general-purpose or special-purpose computer system. This physical storage medium may be fixed to the computer system, as in the case of a magnetic drive, or removable, as in the case of an EEPROM device (e.g., flash memory device).
In some embodiments, when information is transferred or provided over a network or another communications connection (e.g., either hardwired, wireless, or a combination of hardwired or wireless) to a computer system, the connection is properly viewed as a computer-readable medium. Thus, any such connection is properly termed a computer-readable medium. Combinations of the above should also be included within the scope of computer-readable media. Computer-executable or computer-readable instructions comprise, for example, instructions and data that cause a general-purpose computer system or special-purpose computer system to perform a certain function or group of functions. The computer-executable or computer-readable instructions may be, for example, binaries, or intermediate format instructions such as assembly language, or source code.
In this description and in the following claims, a computer system is defined as one or more software modules, one or more hardware modules, or combinations thereof that work together to perform operations on electronic data. For example, the definition of computer system includes the hardware modules of a personal computer, as well as software modules, such as the operating system of the personal computer. The physical layout of the modules is not important. A computer system may include one or more computers coupled via a network. Likewise, a computer system may include a single physical device (e.g., a mobile phone or personal digital assistant (PDA)) where internal modules (e.g., a processor and memory) work together to perform operations on electronic data.
Some embodiments may be practiced in network computing environments with many types of computer system configurations, including hubs, routers, wireless access points (APs), wireless stations, personal computers, laptop computers, hand-held devices, multi-processor systems, microprocessor-based or programmable consumer electronics, network PCs, minicomputers, mainframe computers, mobile telephones, RDAs, pagers, or other suitable environment. Some embodiments may be practiced in distributed system environments where local and remote computer systems, which are linked (e.g., either by hardwired, wireless, or a combination of hardwired and wireless connections) through a network, both perform tasks. In a distributed system environment, program modules may be located in both local and remote memory-storage devices (see above).
In some embodiments, the computer system 2600 may additionally include a first retrieval engine 2611 to retrieve a batch of histograms and item types, and a category engine 2612 to categorize each histogram of the batch of histograms based upon the item type, the item type including items to be sold. Next, a filter 2613 is illustrated to filter the batch of histograms using a configuration rules set, the filtering to remove histograms from the batch of histograms that violate a rule the configuration rules set. Further, a sorting engine 2614 is shown that sorts the batch of histograms. Next, a classification engine 2615 is illustrated that divides the batch of histograms by classification, the classification relating each member of the batch of histograms to one or more members of a traffic optimizer module set. Additionally, a histogram selector 2616 is illustrated that selects a histogram from the batch of histograms, the selecting based upon the histograms relationship to a member of the traffic optimizer module set. In some cases, the one or more members of the traffic optimizer module set includes at least one of a page-type, merchandising, ad, or navigation module. In some cases, a second retrieval engine 2617 is shown to retrieve a histogram from the batch of histograms, and a third retrieval engine 2618 is illustrated that retrieves object data associated with the histogram, the retrieving based upon a preferred ROI value as determined by one or more AI algorithms. Moreover, a transmitter 2619 is illustrated that transmits a object associated with the object data as part of a web page instruction set.
Some embodiments may include one or more computer systems containing the modules outlined in
The example computer system 2700 includes a processor 2702 (e.g., a Central Processing Unit (CPU), a Graphics Processing Unit (GPU) or both), a main memory 2701 and a static memory 2706, which communicate with each other via a bus 2708. The computer system 2700 may further include a video display unit 2710 (e.g., a LCD or a CRT). The computer system 2700 also includes an alphanumeric input device 2717 (e.g., a keyboard), a User Interface (UI) cursor controller 2711 (e.g., a mouse), a disk drive unit 2716, a signal generation device 2725 (e.g., a speaker) and a network interface device (e.g., a transmitter) 2720.
The disk drive unit 2716 includes a machine-readable medium 2722 on which is stored one or more sets of instructions and data structures (e.g., software) embodying or used by any one or more of the methodologies or functions illustrated herein. The software may also reside, completely or at least partially, within the main memory 2701 and/or within the processor 2702 during execution thereof by the computer system 2700, the main memory 2701, and the processor 2702 also constituting machine-readable media.
The instructions 2721 may further be transmitted or received over a network 2727 via the network interface device 2720 using any one of a number of well-known transfer protocols (e.g., HTTP or Session Initiation Protocol (SIP)).
The term “machine-readable medium” should be taken to include a single medium or multiple media (e.g., a centralized or distributed database, and/or associated caches and servers) that store the one or more sets of instructions. The term “machine-readable medium” shall also be taken to include any medium that is capable of storing, encoding, or carrying a set of instructions for execution by the machine and that cause the machine to perform any of the one or more methodologies illustrated herein. The term “machine-readable medium” shall accordingly be taken to include, but not be limited to, solid-state memories, optical and magnetic media, and carrier wave signals.
It is to be understood that the above description is intended to be illustrative and not restrictive. Although numerous characteristics and advantages of various embodiments as illustrated herein have been set forth in the foregoing description, together with details of the structure and function of various embodiments, many other embodiments and changes to details may be apparent to those of skill in the art upon reviewing the above description. The scope of the invention should be, therefore, determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled. In the appended claims, the terms “including” and “in which” are used as the plain-English equivalents of the respective terms “comprising” and “wherein,” respectively. Moreover, the terms “first,” “second,” and “third,” etc., are used merely as labels, and are not intended to impose numerical requirements on their objects.
The Abstract of the Disclosure is provided to comply with 37 C.F.R. §1.72(b), requiring an abstract that may allow the reader to quickly ascertain the nature of the technical disclosure. It is submitted with the understanding that it may not be used to interpret or limit the scope or meaning of the claims. In addition, in the foregoing Detailed Description, it can be seen that various features are grouped together in a single embodiment for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the claimed embodiments require more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive subject matter lies in less than all features of a single disclosed embodiment. Thus the following claims are hereby incorporated into the Detailed Description, with each claim standing on its own as a separate embodiment.
Claims
1. A computer system comprising:
- a definition engine to define a set of objects capable of being placed onto a web page;
- a monitor to capture content request data including a number of times an object, of the set of objects, is executed; and
- a selection engine to select the object for placement onto the web page, the selecting including a use of an algorithm to determine a Return On Investment (ROI) associated with the object.
2. The computer system of claim 1, further comprising:
- an update module to update a decision tree to include the captured content request data, the decision tree including members of a traffic optimizer module set as nodes; and
- a traversal module to traverse the decision tree and selecting a member of the traffic optimizer module set, the selecting based upon the ROI by the member of the traffic optimizer module set.
3. The computer system of claim 2, wherein the member of the traffic optimizer module set includes at least one of a page-type, merchandising, ad, or navigation module.
4. The computer system of claim 2, further comprising a calculator to calculate the ROI based upon a difference between a cost associated with generating the web page, and a revenue amount generated by the web page.
5. The computer system of claim 2, further comprising a calculator to calculate the ROI relative to another ROI.
6. The computer system of claim 2, further comprising a histogram engine to create a histogram relating to keyword usage, the histogram including a portion of the content request data that identifies the member of the traffic optimizer module set.
7. The computer system of claim 1, further comprising:
- a first retrieval engine to retrieve a set of histograms and item types, the set of histograms including the content request data;
- a category engine to categorize each histogram of the set of histograms based upon the item type, the item type including items to he sold;
- a filter to filter the set of histograms using a configuration rules set, the filtering to remove histograms from the set of histograms that violate a rule in the configuration rules set; and
- a sorting engine to sort the set of histograms.
8. The computer system of claim 7, further comprising:
- a classification engine to divide the set of histograms by classification, the classification relating at least one member of the set of histograms to at least one member of a traffic optimizer module set; and
- a histogram selector to select at least one histogram from the set of histograms, the selection based upon the relationship between the member of the set of histograms and he least one member of the traffic optimizer module set.
9. The computer system of claim 7, further comprising:
- a second retrieval engine to retrieve a histogram from the set of histograms;
- a third retrieval engine to retrieve object data associated with the histogram, the retrieving based upon the ROI as determined by at least one AI algorithm; and
- a transmitter to transmit an object associated with the object data as part of a web page instruction set.
10. The computer system of claim 7, wherein the algorithm is an AI algorithm that includes at least one of a deterministic, or non-deterministic AI algorithm.
11. A non-transitory machine-readable medium storing instructions, which, when executed by one or more processors of one or more machines, cause the one or more machines to perform a method comprising:
- define a set of objects capable of being placed onto a web page;
- capture content request data including a number of times an object, of the set of objects, is executed; and
- select the object for placement onto the web page, the selecting including a use of an algorithm to determine a Return On Investment (ROI) associated with the object.
12. The non-transitory machine-readable medium of claim 11, storing additional instructions, which, when executed, cause the machine to additionally:
- update a decision tree to include the captured content request data, the decision tree including members of a traffic optimizer module set as nodes; and
- traverse the decision tree and selecting a member of the traffic optimizer module set, the selecting based upon the ROI by the member of the traffic optimizer module set.
13. The non-transitory machine-readable medium of claim 12, wherein the member of the traffic optimizer module set includes at least one of a page-type, merchandising, ad, or navigation module.
14. The non-transitory machine-readable medium of claim 12, storing additional instructions, which, when executed, cause the machine to additionally:
- calculate the ROI based upon a difference between a cost associated with generating the web page, and a revenue amount generated by the web page.
15. The non-transitory machine-readable medium of claim 12, storing additional instructions, which, when executed, cause the machine to additionally:
- calculate the ROI relative to another ROI.
16. The non-transitory machine-readable medium of claim 12, storing additional instructions, which, when executed, cause the machine to additionally:
- create a histogram relating to keyword usage, the histogram including a portion of the content request data that identifies the member of the traffic optimizer module set.
17. The non-transitory machine-readable medium of claim 11, storing additional instructions, which, when executed, cause the machine to additionally:
- retrieve a set of histograms and item types, the set of histograms including the content request data;
- categorize each histogram of the set of histograms based upon the item type, the item type including items to be sold;
- filter the set of histograms using a configuration rules set, the filtering to remove histograms from the set of histograms that violate a rule in the configuration rules set; and
- sort the set of histograms.
18. The non-transitory machine-readable medium of claim 17, storing additional instructions, which, when executed, cause the machine to additionally:
- divide the set of histograms by classification, the classification relating at least one member of the set of histograms to at least one member of a traffic optimizer module set; and
- select at least one histogram from the set of histograms, the selection based upon the relationship between the member of the set of histograms and the least one member of the traffic optimizer module set.
19. The non-transitory machine-readable medium of claim 17, storing additional instructions, which, when executed, cause the machine to additionally:
- retrieve a histogram from the set of histograms;
- retrieve object data associated with the histogram, the retrieving based upon the ROI as determined by at least one AI algorithm; and
- transmit an object associated with the object data as part of a web page instruction set.
20. The non-transitory machine-readable medium of claim 17, wherein the algorithm is an AI algorithm that includes at least one of a deterministic, or non-deterministic AI algorithm.
Type: Application
Filed: Jan 12, 2011
Publication Date: May 5, 2011
Applicant:
Inventors: Eric Noel Billingsley (Campbell, CA), Raghav Gupta (Sunnyvale, CA), Randall Scott Shoup (San Francisco, CA), Neelakantan Sundaresan (Mountain View, CA)
Application Number: 13/005,408
International Classification: G06Q 40/00 (20060101);