Abstract: In some examples, personality trait-based customer behavior prediction may include extracting personality features from images associated with a user. Social style features may be extracted from social data associated with the user. Consumer demographics features may be extracted from consumer demographics data associated with the user. Based on a probability analysis of the extracted features, relevant features may be selected from the images, the social data, and the consumer demographics data. Historical purchase features may be extracted from historical purchase data associated with the user. At least one machine learning model may be trained based on the extracted features, and used to generate a next best offer for the user for purchase of a product or a service. A purchase of the product or the service may be performed based on the generated offer.

Abstract: In some examples, sequence, frequency, and time interval based journey recommendation may include generating a plurality of clusters of entities, and generating a network that identifies a time interval to a next interaction that leads to success. An estimated time interval to a specified number of conversions may be determined. A success criterion that represents a positive outcome in the estimated time interval may be determined. A historical sequence of events may be partitioned into a plurality of sequences of events leading to success or failure. The plurality of sequence of events may be mapped based on analyzed probabilities, a determined waiting interval, and determined frequency contributions, and evaluated as to whether a mapped sequence of events duration is less than a planned duration. If so, a journey may be generated and include a determined sequence of events, a corresponding frequency, and a corresponding waiting interval.

Abstract: In some examples, personality trait-based customer behavior prediction may include extracting personality features from images associated with a user. Social style features may be extracted from social data associated with the user. Consumer demographics features may be extracted from consumer demographics data associated with the user. Based on a probability analysis of the extracted features, relevant features may be selected from the images, the social data, and the consumer demographics data. Historical purchase features may be extracted from historical purchase data associated with the user. At least one machine learning model may be trained based on the extracted features, and used to generate a next best offer for the user for purchase of a product or a service. A purchase of the product or the service may be performed based on the generated offer.

Abstract: A device receives an optimization problem and a target value and input data for the optimization problem, and specifies constraints for the optimization problem based on the input data. The device identifies an optimization space for the optimization problem based on the constraints and the input data, and divides the optimization space into sub-regions based on the constraints and the input data. The device performs optimizations of a set of the sub-regions, and determines a respective distance of each sub-region, of the set of the sub-regions, from the target value. The device selects a particular sub-region that is a shortest distance from the target value, and selects a vector from the particular sub-region. The device executes the optimization problem using the vector as an initial parameter and to generate results, and utilizes the results to recommend one or more decisions or modify a process.

Abstract: In some examples, sequence, frequency, and time interval based journey recommendation may include generating a plurality of clusters of entities, and generating a network that identifies a time interval to a next interaction that leads to success. An estimated time interval to a specified number of conversions may be determined. A success criterion that represents a positive outcome in the estimated time interval may be determined. A historical sequence of events may be partitioned into a plurality of sequences of events leading to success or failure. The plurality of sequence of events may be mapped based on analyzed probabilities, a determined waiting interval, and determined frequency contributions, and evaluated as to whether a mapped sequence of events duration is less than a planned duration. If so, a journey may be generated and include a determined sequence of events, a corresponding frequency, and a corresponding waiting interval.

Abstract: A device receives an optimization problem and a target value and input data for the optimization problem, and specifies constraints for the optimization problem based on the input data. The device identifies an optimization space for the optimization problem based on the constraints and the input data, and divides the optimization space into sub-regions based on the constraints and the input data. The device performs optimizations of a set of the sub-regions, and determines a respective distance of each sub-region, of the set of the sub-regions, from the target value. The device selects a particular sub-region that is a shortest distance from the target value, and selects a vector from the particular sub-region. The device executes the optimization problem using the vector as an initial parameter and to generate results, and utilizes the results to recommend one or more decisions or modify a process.

Abstract: An unsupervised machine learning model can make prediction on time series data. Variance of time-varying parameters for independent variables of the model may be restricted for continuous consecutive time intervals to minimize overfitting. The model may be used in a control system to control other devices or systems. If predictions for the control system are for a higher granularity time interval than the current mode, the time-varying parameters of the model are modified for the higher granularity time interval.

Abstract: An unsupervised machine learning model can make prediction on time series data. Variance of time-varying parameters for independent variables of the model may be restricted for continuous consecutive time intervals to minimize overfitting. The model may be used in a control system to control other devices or systems. If predictions for the control system are for a higher granularity time interval than the current mode, the time-varying parameters of the model are modified for the higher granularity time interval.