Decision optimization utilizing tabular data

A computer program product for automatically generating optimized decision policies, the computer program product comprising one or more computer readable storage media and program instructions stored on the one or more computer readable storage media, the program instructions including instructions to:

• receive a tabular dataset from a database, wherein the tabular dataset comprises state variables, action variables, and reward variable;
• generate Q-function candidates, environments, and transitions by applying a plurality of the state variables, the action variables, and the reward variables to a trained regression model;
• apply the generated Q-function candidates, environments, and transitions to a reinforcement learning algorithm to learn a decision policy for each of the Q-function candidates;
• determine if one or more of the learned decision policies are optimal decision policies based on an outputted expected reward from each of the one or more learned decision policies;
• analyze available policies to determine when there is an improvement of any prior determined expected reward to ensure a selected policy is a most favorable decision policy amongst available policies, wherein the most favorable policies are determined by iteratively selecting a different plurality of state variables, action variables, and reward variables from the tabular dataset;
• upon determining that one or more of the learned decision policies are optimal decision policies, generate different Q-function candidates, environments, and transitions by applying the different plurality of the state variables, the action variables, and the reward variables to a trained regression model and selecting a final decision policy wherein said final decision policy provides prediction of a next best action and/or sequence of actions that improve processing performance, storage capability and time of execution;
• and
• execute said final decision policy using an auto machine learning system.

A computer system for automatically generating optimized decision policies comprising:

• one or more computer processors;
• one or more computer readable storage media;
• and
• computer program instructions, the computer program instructions being stored on the one or more computer readable storage media for execution by the one or more computer processors, the computer program instructions including instructions to:
• receive a tabular dataset from a database, wherein the tabular dataset comprises state variables, action variables, and reward variables;
• generate Q-function candidates, environments, and transitions by applying a plurality of the state variables, the action variables, and the reward variables to a trained regression model;
• apply the generated Q-function candidates, environments, and transitions to a reinforcement learning algorithm to learn a decision policy for each of the Q-function candidates
• determine if one or more of the learned decision policies are optimal decision policies based on an outputted expected reward from each of the one or more learned decision policies;
• analyze available policies to determine when there is an improvement of any prior determined expected reward to ensure a selected policy is a most favorable decision policy amongst available policies, wherein the most favorable policies are determined by iteratively selecting a different plurality of state variables, action variables, and reward variables from the tabular dataset;
• upon determining that one or more of the learned decision policies are optimal decision policies, generate different Q-function candidates, environments, and transitions by applying the different plurality of the state variables, the action variables, and the reward variables to a trained regression model and selecting a final decision policy wherein said final decision policy provides prediction of a next best action and/or sequence of actions that improve processing performance, storage capability and time of execution;
• and
• execute said final decision policy using an auto machine learning system.