A Hybrid Approach for Stock Price Prediction Using Support Vector Regression and Multi-View Learning

Abstract

The stock market is an industry that is constantly changing and very complex. Such a situation is characterized in a new intra-day high and low by high volatility and low predictability. To make accurate forecasts regarding the stock prices is indeed vital for many shareholders, financial analysts, as well as financial organizations. The present study presents a combination of two machine learning algorithms, Multi-View Learning (MVL) and Support Vector Regression (SVR), to improve stock price forecasting. The model is applied to five major stocks like Apple (AAPL), Microsoft (MSFT), Google (GOOGL), TCS (TCS.NS), and Netflix (NFLX) uses technical indicators (14-day relative resistance index and 50-day simple moving average) as well as price-related data (Open, High, Low, Close, and Volume). The grid search method (GridSearchCV) is then employed to select the optimal combination of hyperparameters for SVR which is applied via structuring these features into two separate views, the algorithm allows the GridSearchCV model to tune hyperparameters and, to improve the SVR performance. The result of evaluation showcase model's capability, correctness and reliability by means of R-squared, MSE, and RMSE as performance metrics.