Application of Machine Learning in Economic Optimization

Abstract

In this thesis, we introduced two new applications of Machine Learning in the field of Economic Optimization. The first application addresses the problem of searching for global Self-Optimizing variables. We applied Genetic Programming (GP) to solve this problem and demonstrated how powerful is the new GP-based search method. In the second application, we used Convolutional Neural Networks (CNN) to develop a vision-based steady-state detector (SSD) for steady-state Real-Time Optimizers. It was our purpose to investigate if this vision-based SSD has higher accuracy than established statistical SSD. We found that they have comparable performances, but the CNN-based detector possesses certain advantages that the others do not have.