Feature recognition on stock price and indices movements using wavelets

Abstract

Feature recognition has wide range of applications but no single approach focusses on stock market analysis This work analysed the new feature trading interval for prediction of stock prices which is complex challenging and need tremendous efforts In current scenario machine learning has drawn increasing interest in stock market analysis To analyse the stock market data for identifying significant feature data mining and machine learning techniques can be applied Identifying domain specific feature is a continuous iterative and logical process Wavelets have its own advantage in tremendous applications however remains less explored in the field of economics and finance This work used wavelets for identifying domain specific feature in stock market data General forecasting models were used to forecast the denoised signals Further the forecast model is selected based upon the performance measure coefficient of determination with high values The selected model is used to forecast the share prices Then performance measures such as RMSE MAE MAPE and Theil U were calculated for which trading length consider for analysis The optimal trading length was found out based on the lowest values of performance measures Finally purchase decision Making rules were applied to evaluate the accuracy of the selected model and based on that the recommendation of buy or sell was given This work has three proposed approaches; the first approach identifies and removes noise from the stock data efficiently The second approach involves feature recognition using wavelets on stock market data and the third approach concentrates on analyzing stock data which identifies new feature for economic and financial applications Finally to assist investors in making stock market decision a decision support system with trading interval is presented newline