Abstract: A new method is used to model the class probability from data that is based on a novel multiplicative adjustment of the class probability by a plurality of items of evidence induced from training data. The optimal adjustment factors from each item of evidence can be determined by several techniques, a preferred embodiment thereof being the method of maximum likelihood. The evidence induced from the data can be any function of the feature variables, the simplest of which are the individual feature variables themselves. The adjustment factor of an item of evidence Ej is given by the ratio of the conditional probability P(C|Ej) of the class C given Ej to the prior class probability P(C), exponentiated by a parameter aj. The method provides a new and useful way to aggregate probabilistic evidence so that the final model output exhibits a low error rate for classification, and also gives a superior lift curve when distinguishing between any one class and the remaining classes.

Abstract: A new method is used to model the class probability from data that is based on a novel multiplicative adjustment of the class probability by a plurality of items of evidence induced from training data. The optimal adjustment factors from each item of evidence can be determined by several techniques, a preferred embodiment thereof being the method of maximum likelihood. The evidence induced from the data can be any function of the feature variables, the simplest of which are the individual feature variables themselves. The adjustment factor of an item of evidence E1 is given by the ratio of the conditional probability P(C|E1) of the class C given E1 to the prior class probability P(C), exponentiated by a parameter a1. The method provides a new and useful way to aggregate probabilistic evidence so that the final model Output exhibits a low error rate for classification, and also gives a superior lift curve when distinguishing between any one class and the remaining classes.

Abstract: A method of boosting of predictive models, called cascade boosting, for resolving the interpretability problem of previous boosting methods while mitigating the fragmentation problem when applied to decision trees. This method of cascade boosting always applies a single weak model to any given data point. An improvement to the common method of boosting lies in how weak models are organized in a decision list rather than a weighted average. Cascade boosting resolves the interpretability problem of previous boosting methods while mitigating the fragmentation problem when applied to decision trees. Cascade boosting is simplest when applied to segmented predictive models but may also be applied to predictive models that do not explicitly segment the space of possible data points. The predictive model resulting from cascade boosting has fewer rules, or tree leaves, thereby enabling a modeler to better understand the correlations among the data.

Abstract: A Web site serves as a portal to only certified Web sites where any products or services listed through these Web sites are certified as genuine. The Web site (named, for example, certified.com) assures the claims by verifying and obtaining strict guarantees from the client companies whose Web sites are listed in the certified.com Web site. The certification can be restricted to apply only to a portion of a client Web site. The invention solves the problem of lack of knowledge about the trustworthiness of Web sites by providing “checking” services that verify the claims made for a Web business. The checking is done via normal audit procedures, as is currently customary, and the keeping of records. The Web site is an Internet version of such assurances as UL listings and Good Housekeeping Seals of Approval so that customers can trust what they get from Web sites; however, the Web site goes beyond just putting a seal on a product or service.

Abstract: When a customer is in the process of filling a market basket for purchase on an Internet commerce site, a method makes prioritized recommendation of items so as to maximize the likelihood that the customer will add to the basket those items that are in the list with higher priorities. The method separately considers in turn preferences due to a current set of items in the market basket and also preferences due to a new choice independent of what is in the market basket. In this way, the method recognizes that not all items in the market basket are selected because of their affinity with some other item already in the basket. The two preferences are estimated separately from training data and combined in proper proportions to obtain an overall preference for item not yet in the market basket.

Abstract: A fast way for determining the best subset test for a nominal attribute in a decision tree. When a nominal attribute has n distinct values, the prior art requires computing the impurity functions on each of the 2.sup.n-1 -1 possible subset partitioning of the n values and finding the minimum case among them. This invention guarantees the minimum impurity test on the attribute by computing only (n-1) impurity function computations. This reduction of computational complexity makes it practically possible to find the true best tests for many real data mining application, where a binary decision tree is used as the classification model.

Abstract: The invention discloses steps to compute a new contextual merit measure for feature variables based on a two-fold discovery. First, a feature variable does not exist by itself in distinguishing classes, and second, what is desired is to obtain its correlation to a class in the context of other feature variables. The invention specifies a merit measure that captures the contextual correlation implicitly.

Abstract: This specification describes the testing of interconnections between modules mounted on a card and between the modules and the input and output terminals of the card. Each of the modules has an Exclusive-OR circuit which receives an input from each of the input pins of the module and has a single output which is taken off an output pin of the module. Also, each of the modules has a test input circuit for accessing all of the output pins of the module in parallel from a single input terminal. The test input circuits are used to apply a binary 0 followed by a binary 1 to all the outputs of all the modules. The Exclusive-OR circuits are used to monitor the response to those signals. By testing in this manner, all the connections between the modules and also between the modules and the card terminals can be checked for stuck ones and zeros.