Analysis of Data Driven Statistical Models for Predicting Trends in Stock Prices

Forecasting results for stock markets exhibit variabilities in forecasting accuracy as the tenure of prediction is varied. Typically, stocks are predicted for long, short as well as mid-term tenures based on the tenure of prediction. While longer tenures have relatively much larger data to be trained as training data, divergences are also potentially large as the forecasting period may render higher randomness due to unprecedented events. The short-term forecasting is relatively less prone to unprecedented events due to the tenure of forecasting. However, the lesser amount of training data may result in less accurate pattern recognition. A common ground is typically found in terms of mid term forecasting. Several financial institutions and individuals leverage stock forecasting results for the purpose of investments in pertinent stocks. Moreover, inaccuracies in forecasting may lead to huge financial losses. Hence, accurate forecasting of stock trends in extremely crucial from a financial standpoint. The purpose of this paper is to present a comprehensive review of existing data pre-processing and contemporary statistical models for stock forecasting. Additionally, to elucidate the pre-processing and pattern recognition steps, a DWT-DNN model has been trained with Tesla stocks. The back propagation based gradient descent algorithm is used to train the deep neural network. The forecasting results have been evaluated for long, mid- and short-term intervals. The Mean Absolute Percentage Error (MAPE) has been computed and tabulated for all the three intervals. The MAPE is found to be 2.64, 5.14 and 4.46 for the long-, mid- and short-term forecasts. A comparative analysis with existing work in the domain indicates that the proposed algorithm attains lower MAPE indicating superior forecasting accuracy. The improvement can be attributed to the transform domain noise removal capability of the DWT which aids the pattern recognition capability of the deep neural network model.