Technical Analysis Based Automatic Trading Prediction System for Stock Exchange using Support Vector Machine

Abstract

Stock exchange trading has been utilized to gain profit by constantly buying and selling best-performing stocks in a short term. Deep knowledge, time dedication, and experience are essential for optimizing profit if stock price fluctuations are analyzed manually. This research proposes a new trading prediction system that has the ability to automatically predict the accurate time for buying and selling stock using a combination of technical analysis and support vector machine (SVM). Technical analysis is used to analyze stock price fluctuation based on historical data by utilizing technical indicators such as moving average, Bollinger bands, relative strength index, stochastic oscillator, and Aroon oscillator. SVM maps inputs into higher dimensional spaces using non-linear kernel functions, making it suitable for various technical indicators implementation as inputs in stock trading prediction. Experimentation on five Indonesian stocks reveals that the combination of technical analysis and support vector machine is best suited for continuously fluctuated stocks, with the highest accuracy of 77.8%.