Abstract: Techniques for model interpretability for anomaly detection are provided. In one technique, a machine-learned (ML) model is trained based on training data that comprises first input data and first target data. Using the ML model, (1) first output data is generated based on the first input data and (2) second output data is generated based on second input data. For each data item in the second output data, a difference is generated between the data item and a corresponding data item in second target data that corresponds to the second output data. If the difference is greater than a threshold, then the data item, the corresponding data item, and a corresponding data item in the second input data are identified as an anomalous set. Second training data is generated based on the first training data and the identified anomalous sets and is used to train a second ML model.

Abstract: Techniques for enabling the building of general input data ML flows using a serverless data-representation-as-a-service (DRaaS) are provided. In one technique, in response to receiving a first data representation (DR) generation request from a first calling entity, first input data is retrieved based on the first DR generation request, a first set of DRs is generated (by a DR generator) based on the first input data, and the first set of DRs are made available to the first calling entity. In response to receiving a second DR generation request from a second calling entity that is different than the first calling entity, second input data is retrieved based on the second DR generation request, a second set of DRs is generated based on the second input data, and the second set of DRs are made available to the second calling entity.

Abstract: A computer system computes a score for a received data exchange and, in accordance with a neural network and input variables determined by received current exchange and history data, the computed score indicates a condition suitable for a denial. A set of attribution scores are computed using an Alternating Decision Tree model in response to a computed score that is greater than a predetermined score threshold value for the denial. The computed score is provided to an assessment unit and, if the computed score indicates a condition suitable for the denial and if attribution scores are computed, then a predetermined number of input variable categories from a rank-ordered list of input variable categories is also provided to the assessment unit of the computer system.

Abstract: A computer system computes a score for a received data exchange and, in accordance with a neural network and input variables determined by received current exchange and history data, the computed score indicates a condition suitable for a denial. A set of attribution scores are computed using an Alternating Decision Tree model in response to a computed score that is greater than a predetermined score threshold value for the denial. The computed score is provided to an assessment unit and, if the computed score indicates a condition suitable for the denial and if attribution scores are computed, then a predetermined number of input variable categories from a rank-ordered list of input variable categories is also provided to the assessment unit of the computer system.

Abstract: A computerized system that processes a fraud score for a financial transaction in connection with an account is computed from retrieved data to indicate a probability of the account being in a compromised condition. A self-similarity score is computed if the computed fraud score is above a predetermined threshold to indicate similarity of the received transaction to other transactions of the account in the set of prior transactions. A suggested action to authorize or decline the transaction is determined based on the computed fraud score and the computed self-similarity score.

Abstract: Provided is a system for estimating price sensitivities and determining aggregate price adjustments for a population of items, the population comprising a plurality of sub-populations.

Abstract: A method of modeling includes quantifying a co-operative strength value for a plurality of pairs of variables, and identifying a clique of at least three variables based on a graph of the co-operative strength values of a plurality of pairs of variables. The method also includes selecting a first pair of variables of the plurality of pairs of variables having a high co-operative strength value. A second clique may also be identified. A model of the first clique and a model of the second clique are made. The outputs of these models are combined to form a combined model which is used to make various decisions with respect to real time data.

Abstract: A system for classifying a transaction as fraudulent includes a training component and a scoring component. The training component acts on historical data and also includes a multi-dimensional risk table component comprising one or more multidimensional risk tables each of which approximates an initial risk value for a substantially empty cell in a risk table based upon risk values in cells related to the substantially empty cell. The scoring component produces a score, based in part, on the risk tables associated with groupings of variables having values determined by the training component. The scoring component includes a statistical model that produces an output and wherein the transaction is classified as fraudulent when the output is above a selected threshold value.

Abstract: A method of modeling includes quantifying a co-operative strength value for a plurality of pairs of variables, and identifying a clique of at least three variables based on a graph of the co-operative strength values of a plurality of pairs of variables. The method also includes selecting a first pair of variables of the plurality of pairs of variables having a high co-operative strength value. A second clique may also be identified. A model of the first clique and a model of the second clique are made. The outputs of these models are combined to form a combined model which is used to make various decisions with respect to real time data.

Abstract: A system for classifying a transaction as fraudulent includes a training component and a scoring component. The training component acts on historical data and also includes a multi-dimensional risk table component comprising one or more multidimensional risk tables each of which approximates an initial risk value for a substantially empty cell in a risk table based upon risk values in cells related to the substantially empty cell. The scoring component produces a score, based in part, on the risk tables associated with groupings of variables having values determined by the training component. The scoring component includes a statistical model that produces an output and wherein the transaction is classified as fraudulent when the output is above a selected threshold value.