System and method of machine learning based deviation prediction and interconnected-metrics derivation for action recommendations

A system of machine learning based deviation prediction and interconnected-metrics derivation for action recommendations, the system comprising:

• a client device (116) that is used to specify a meta data model of a use-case, select at least one metric of interest and capture a use-case action;
• a deviation management environment (108) that provides a deviation management service of the use-case to the client device (116), wherein the deviation management environment (108) further comprises:
• a deviation computation module (302) that obtains a specification information of a deviation management service from a client device (116) and computing a plurality of deviations of at least one metric of interest;
• a deviation factor discovery module (304) that identifies at least one trend and a set of factors which relate to the plurality of deviation of the at least one metric of interest;
• a deviation interconnections derivation module (306) that derives a plurality of interconnections among the plurality of deviations of the at least one metric of interest using interconnected machine learning models;
• a deviation prediction module (308) comprising a plurality of interconnected prediction models, wherein the deviation prediction module (308) determines at least one predicted deviation based on the plurality of deviations and a probability distribution associated with the at least one predicted deviation;
• a predicted deviation classifier and scoring module (310) that receives the at least one predicted deviation determined by the deviation prediction module (308), determines if the at least one predicted deviation qualifies above a predefined threshold, classifies the at least one predicted deviation and assigns a score to the at least one predicted deviation on the basis of a criticality;
• a predicted deviation attribution module (312) that utilizes data available in an external information database (112) and a repository (210) to determine a set of factors which attribute to the at least one predicted deviation;
• a predicted deviation impact computation module (314) that determines a quantified impact of the at least one predicted deviation using interconnected machine learning models;
• a recommendation module (316) that generates action recommendation for the use-case based on the probability distribution of the at least one predicted deviation, a probability distribution of an outcome of at least one action alternatives and the quantified impact, wherein the action recommendation is of either mitigation or leverage for the use-case;
• an action tracking module (318) that tracks an action taken in the use-case from the client device (116) to enable reinforcement learning for providing accurate predictions on the plurality of deviations of the at least one metric of interest.

A one or more non-transitory computer-readable storage mediums storing the one or more sequences of instructions, which when executed by the one or more processors, causes to perform method of machine learning based deviation prediction and interconnected-metrics derivation for action recommendations, the method comprising:

• specifying, using a client device (116), a meta data model of a use-case and at least one metric of interest;
• obtaining a specification information of a deviation management service from a client device (116) and computing a plurality of deviations of at least one metric of interest;
• identifying at least one trend and a set of factors which relate to the plurality of deviations of the at least one metric of interest;
• deriving a plurality of interconnections among the plurality of deviations of the at least one metric of interest using interconnected machine learning models;
• determining at least one predicted deviation based on the plurality of deviations and a probability distribution associated with the at least one predicted deviation based on a plurality of interconnected prediction models;
• classifying the at least one predicted deviation if the at least one predicted deviation qualifies above a predefined threshold and assigning a score to the at least one predicted deviation on the basis of a criticality;
• determining a set of factors which attribute to the at least one predicted deviation based on data available in an external information database (112) and a repository (210);
• determining a quantified impact of the at least one predicted deviation using interconnected machine learning models;
• generating an action recommendation for the use-case based on the probability distribution of the at least one predicted deviation, the probability distribution of an outcome of at least one action alternative and the quantified impact, wherein the action recommendation is at least one of mitigation or leverage for the use-case;
• tracking an action taken in the use-case from the client device (116) to enable reinforcement learning;
• and
• providing accurate predictions on the plurality of deviations of the at least one metric of interest based on the reinforcement learning.