Simulation-based controls optimization using time series data forecast

Abstract

A trained neural network that is configured to generate predictions for periods of time in the future based on input data can be received, where the neural network is trained using training data that includes time series data segmented into windows. Observed time series data can be processed to generate the input data. Using the trained neural network and the generated input data, data predictions can be generated. The predictions can be provided to a reinforcement learning model configured to generate predicted outcomes, where the reinforcement learning model varies parameters to simulate conditions for a first and second entity, and an artificial intelligence agent simulates actions performed by one of the first and second entities, the data predictions being a parameter for the simulation. Parameters for the first and second entities can be selected, where the selected parameters correspond to a predicted outcome that meets a criteria.