Combining cloud computing, machine learning and heuristic optimization for investment opportunities forecasting

Prediction of stock market is a challenging task that has attracted researchers in various fields including the computational intelligence and finance. Since stock market data sets are intrinsically large, nonlinear and time-varying, it is extremely difficult to design models for forecasting the future directions with an acceptable accuracy. In this paper, an integrative and intelligent machine learning framework is proposed through combining cloud computing, machine learning and heuristic optimization. Essentially, the Support Vector Machine (SVM) method is extended with the Grid Search (GS) or Chemical Reaction Optimization (CRO) as a heuristic optimization method together with Principal Component Analysis (PCA) and Feature Noise Filter (FNF) to construct quantitative investment forecasting models for efficient executions on cloud computing platforms. To demonstrate the effectiveness of the proposed framework, the Hang Seng Index and some major stocks listed on the Hong Kong Exchange are predicted using the constructed models on a daily basis. The empirical results clearly indicate that the proposed integrative approach is promising and gives impressive performance in terms of the prediction accuracy.