Regression ensemble techniques with technical indicators for prediction of financial time series data

This chapter explores regression techniques with two regression ensemble technique LSBoosting and Bagging for various time series data: Indian stock market prediction (BSE30), Foreign Exchange data (INR/USD), and Crude Oil data. The data is prepared with four technical indicators EMA, MACD, OBV, and FI. The model was developed with static as well as dynamic partition using k-fold cross validation. It is observed that the model performs better with dynamic partitioning using k-fold cross validation for all data sets. The comparative result shows that the model with the regression ensemble technique with Bagging is performing better, which is measured with Mean Absolute Percentage Error (MAPE) = 0.98 for next day ahead prediction of BSE30 and MAPE = 0.1564 for Foreign Exchange Data and MAPE = 1.231 for crude oil data.