Approximate Solution using Machine Learning and Evolutionary Algorithm for Large-Scale Data-Driven Chance Constrained Problems

By using a large-scale data set, Chance-Constraint Problem (CCP) is formulated as a deterministic optimization problem called Deterministic Data-driven CCP (DDCP). DDCP not only determines the feasibility of a solution by whether or not it satisfies the constraints at a specified sufficiency level, but also classifies all given data into two classes. Then an approximate solution method for DDCP is proposed by combining a Support Vector Machine (SVM) and an Evolutionary Algorithm (EA). In order to select a training set for the SVM, a stratified sampling method using principal component analysis is employed. Finally, the performance of the proposed approximate solution method is evaluated through a real-world application, namely, the base-station location problem which is formulated as DDCP.