Comparative Study between FPA, BA, MCS, ABC, and PSO Algorithms in Training and Optimizing of LS-SVM for Stock Market Prediction

In this Paper, five recent natural inspired algorithms are proposed to optimize and train Least Square- Support Vector Machine (LS-SVM). These algorithms are namely, Flower Pollination Algorithm (FPA), Bat algorithm (BA), Modified Cuckoo Search (MCS), Artificial Bee Colony (ABC), and Particle Swarm Optimization (PSO). These algorithms are proposed to automatically select best free parameters combination for LS-SVM. Six financial technical indicators derived from stock historical data are used as inputs to proposed models. Standard LS-SVM and ANN are used as benchmarks for comparison with proposed models. Proposed models tested with six datasets representing different sectors in S&P 500 stock market. Proposed models were used to predict daily, weekly, and monthly stock prices. Results presented in this paper showed that the proposed models have quick convergence rate at early stages of the iterations. They achieved better accuracy than compared methods in price and trend prediction. They also overcame over fitting and local minima problems found in ANN and standard LS-SVM.