Abstract: Systems and methods for identifying recommended topics are disclosed. An example method may be performed by one or more processors of a system and include identifying one or more attributes of a system user, identifying a subset of topics relevant to the system user based on analyzing the one or more attributes of the system user using an analysis model trained with a machine learning process to identify relevant topics for system users based on historical user attributes, generating, for each respective topic of the subset of topics, using the trained analysis model, a relevance score for the respective topic based at least in part on a most recent system page previously accessed by the system user and a current system page accessed by the system user, and generating one or more recommended topics for the system user based on the relevance scores.

Abstract: In general, in one aspect, one or more embodiments relate to a method including receiving, in a business rules engine, input data from disparate data sources. The input data describes a merchant and an application by the merchant to use an electronic payments system for processing transactions between the merchant and customers. Featurization is performed on the input data to form a machine readable vector. By applying the machine readable vector as input to a machine learning model in a machine learning layer, a risk score is predicted. The machine learning model is trained using training data describing use of the electronic payments system by other merchants. The risk score is an estimated probability of the merchant being unable to satisfy an obligation of using the electronic payments system. A business rules engine, based on the risk score, limits use of the electronic payments system by the merchant.

Abstract: Certain aspects of the present disclosure provide techniques for access control. Embodiments include receiving, by a satellite component of an access control system, a request from a computing device to verify an identity of the computing device, wherein the request comprises one or more characteristics of the computing device. Embodiments include verifying, by the satellite component, that the one or more characteristics of the computing device are valid, the verifying comprising one or more interactions with a management entity related to the computing device. Embodiments include generating, by the satellite component, a signed document that is trusted by a control component of the access control system. Embodiments include providing, by the satellite component, the signed document to the computing device for use in requesting credentials from the control component to access a secure resource.

Abstract: A method including receiving natural language text. A negative review is extracted from the natural language text using a first machine learning model (MLM). A first input to the first MLM is the natural language text and a first output of the first MLM is a first probability that the negative review has a negative sentiment. The negative review includes an instance of the natural language text having a corresponding negative sentiment probability above a threshold value. The negative review is categorized by executing a second MLM. A second input to the second MLM is the negative review. A second output of the second MLM is a second probability that the negative review is assigned to a category. A name of a target of the negative review is identified using the name recognition controller and the negative review. The name of the target and the category are provided.

Abstract: Systems described herein apply an ordered combination machine-learning models to identify users who are likely to abandon use of an application, predict the reasons why those users are likely to abandon, and identify intervening actions that the application can perform to reduce the probability that the users will abandon the application. A first machine-learning model determines a retention-prediction value indicating a probability that the user will complete a target action in the application before a session terminates. If the retention-prediction value satisfies a threshold condition, a second machine-learning model determines a reason why the session is likely to terminate before the user completes the target action. A third machine-learning model determines an intervention action for the application to perform to increase the probability that the user will complete the target action before the session terminates.

Abstract: A method including generating a first overlap matrix from a first attribute having first measurements of data items, and generating a second overlap matrix from a second attribute having second measurements of the data items. Samples of weights are generated, each of the samples of weights including a corresponding first weight for the first overlap matrix and a corresponding second weight for the second overlap matrix. For each of the samples of weights, the first overlap matrix is combined with the corresponding first weight and the second overlap matrix is combined with the corresponding second weight. Similarity matrices are generated by combining, for each of the samples of weights, a weighted first overlap matrix with a weighted second overlap matrix. A cluster analysis is performed on the similarity matrices to generate groupings the similarity matrices. The groupings represent groups of the data items. A selected grouping is chosen.

Abstract: Aspects of the present disclosure provide techniques for classifying a trip. Embodiments include receiving, from a plurality of users, a plurality of historical trip records. Each of the plurality of historical trip records may comprise one or more historical trip attributes and historical classification information. Embodiments include training a predictive model, using the plurality of historical trip records, to classify trips based on trip records. Training the predictive model may comprise determining a plurality of hot spots based on the historical trip records, each of the plurality of hot spots comprising a region encompassing one or more locations, and associating, in the predictive model, the plurality of hot spots with historical classification information. Embodiments include receiving, from a user, a new trip record comprising a plurality of trip attributes related to a trip and using the predictive model to predict a classification for the trip based on the trip record.

Abstract: A method includes training, using first real data objects, a generative adversarial network having a generator model and a discriminator model to create a trained generator model that generates realistic data, and training, using adversarial data objects and second real data objects, the discriminator model to output an authenticity binary class for the adversarial data objects and the second real data objects. The method further includes deploying the discriminator model to a production system. In the production system, the discriminator model outputs the authenticity binary class to a system classifier model.

Abstract: Certain aspects of the present disclosure provide techniques for orchestrating communications between different application services through a unified connector platform. Embodiments include receiving, via a connector between a first system and a connector platform, a request to pull a specified data set from one or more second systems. The specified data set is obtained from each of the one or more second systems via a connector between each of the one or more second systems and the connector platform. For each of the obtained data sets, intermediary data sets are generated by converting the obtained data set to a common data format, and a result data set is generated by converting the intermediary data sets are converted to a format associated with the first system. The result data set is transmitted to the first system via the connector between the first system and the connector platform.

Abstract: Capturing dependencies between variables using a variable agnostic object is disclosed. A system is configured to obtain an indication of a first dependency of a first variable to a second variable via a programming interface and depict the first dependency, the first variable, and the second variable in a first instance of a variable agnostic object in a source code. The system is also configured to obtain an indication of a second dependency of a third variable to a fourth variable via the programming interface and depict the second dependency, the third variable, and the fourth variable in a second instance of the variable agnostic object in the source code. The system is also configured to compile the source code to generate a computer-executable program capturing the first dependency and the second dependency based on the first instance and the second instance of the variable agnostic object.

Abstract: An ensemble of machine learning models used for real-time prediction of text for an electronic chat with an expert user. A global machine learning model, e.g., a transformer model, trained with domain specific knowledge makes a domain specific generalized prediction. Another machine learning model, e.g., an n-gram model, learns the specific style of the expert user as the expert user types to generate more natural, more expert user specific text. If specific words cannot be predicted with a desired probability level, another word level machine learning model, e.g., a word completion model, completes the words as the characters are being typed. The ensemble therefore produces real-time, natural, and accurate text that is provided to the expert user. Continuous feedback of the acceptance/rejection of predictions by the expert is used to fine tune one or more machine learning models of the ensemble in real time.

Abstract: A method for detecting fraud rings involves clustering unknown users into unknown user clusters based on a grouping attribute. The method further involves, for each of the unknown user clusters, determining aggregated features including at least one quantification of at least one homogeneity attribute across the unknown users in the unknown user cluster. The method also involves, for each of the unknown user clusters, determining a predictive suspiciousness score based on the aggregated features, determining that at least one of the unknown user clusters is suspicious based on the determined predictive suspiciousness scores, and taking a protective action for the at least one suspicious unknown user cluster.

Abstract: Systems and methods for monitoring items in publicly available inventories, such as websites, are disclosed. Items to be monitored in the publicly available inventories are identified based on embeddings obtained for the inventory items. For example, matching inventory items may be identified based on proximity of the embeddings, such as the cosine distance between embeddings, or a classification machine learning model may be trained to infer matches, e.g., based the embeddings as well as information related to the user and source of the publicly available inventories. The inventory items in the publicly available inventories may be monitored to detect changes which may be reported. Feedback related to the match between inventory items may be used to adjust one or more parameters used for matching.

Abstract: Systems and methods that may be used to provide multitenant key derivation and management using a unique protocol in which key derivation may be executed between the server that holds the root key and a client that holds the derivation data and obtains an encryption key. In one or more embodiments, the derivation data may be hashed. The disclosed protocol ensures that the server does not get access to or learn anything about the client's derived key, while the client does not get access to or learn anything about the server's root key.

Abstract: A method may include obtaining interactions between users and items, and calculating, for each edge in a bipartite graph, an edge weight using an inverse of the degree of a user node connected to the edge and an inverse of the degree of an item node connected to the edge. The bipartite graph includes user nodes corresponding to the users and item nodes corresponding to the items. The method may further include identifying paths each including an edge connecting the target user node and a common item node, an edge connecting a neighboring user node and the common item node, and an edge connecting the neighboring user node and a neighboring item node. The method may further include calculating, using the edge weights calculated for the edges, scores for the paths, and recommending, to the target user and using the scores for the paths, a recommended item.

Abstract: A system receives a request for payment of a transaction between a vendor and a consumer, and sends a first request to a database associated with the online service for historical transactions and personal attributes of the vendor concurrently with sending a second request to a number of third-party services for credit information and personal attributes of the consumer. The system receives information responsive to the first and second requests from the database and the third-party services, respectively, and obtains a risk score for the transaction based on an application of one or more risk assessment rules to the received information by a machine learning model trained with at least the historical transactions and the personal attributes of the vendor. In some aspects, the system determines whether to advance funds to the vendor, prior to requesting payment from a consumer account, based at least in part on the risk score.

Abstract: A method for metrics-based anomaly detection involves receiving an anomaly analysis request for an asset and obtaining metrics associated with the asset. Each of the metrics includes time series data. The method further involves detecting that one of the metrics is a counter. The detection involves seasonally differencing the metric, obtaining a regression line by performing a linear regression on the metric, and determining that an angle of the regression line exceeds a predetermined threshold angle. The method also involves training models for the metrics, the training including training a counter-specific model for the metric that is a counter. The method further involves determining, using the models after the training, at least one metric that is anomalous.

Abstract: A computer-implemented system and method for generating heterogeneous graph feature embeddings for feature learning and prediction. An application server may receive and process a plurality of feature datasets to generate a graph data structure comprising a plurality of interconnected transaction pairs. The application server processes the graph data structure to determine a first-order transaction pair corresponding to a maximum transaction frequency based on a user identifier; executes a jumping probability algorithm to process the graph data structure to determine a second-order transaction pair jumping from a first-order transaction pair; and generates a transaction sequence associated with the user identifier.

Abstract: A method that includes extracting image features of a document image, executing an optical character recognition (OCR) engine on the document image to obtain OCR output, and extracting OCR features from the OCR output. The method further includes executing an anomaly detection model using features including the OCR features and the image features to generate anomaly score, and presenting anomaly score.

Abstract: Systems and methods for identifying suspected anomalies in time series data are disclosed. An example method may receiving time series data for at least one quantity, the time series data including values of the at least one quantity at each of a plurality of times, determining a list of gradients for the time series data, each gradient in the list of gradients based on two or more values of the time series data separated by a specified number of values of the time series data, deriving a plurality of statistics based on the determined list of gradients, and performing a supervised machine learning process based on the derived plurality of statistics to generate a trained machine learning model for identifying one or more suspected anomalies in the time series data.