Forecasting Stock Prices: Application of Econometrics and Machine Learning with an Impact from Climate

Roy, Prince (2021) Forecasting Stock Prices: Application of Econometrics and Machine Learning with an Impact from Climate. Masters thesis, Indian Institute of Science Education and Research Kolkata.

Text (MS dissertation of Prince Roy (14MS107)) 14MS107_Thesis_file.pdf - Submitted Version Restricted to Repository staff only Download (3MB)

Abstract

For the purposes of this study, we have compared the accuracy of time series models in predicting the stock prices of five different firms from five distinct industries. Forecasting and analysis of stock prices are essential skills in the disciplines of finance and economics. In the case of a collection of variables that change over time, time series forecasting may be quite precise. Several research investigations have been conducted in this area of expertise. Using 15 years of daily stock price data from TCS, BAJAJ FINANCE, SBI, TATA MOTORS, and ITC, this research conducted a comparative univariate analysis of time series forecasting using one econometrics model ARIMA (autoregressive integrated moving average) and three machine learning models PROPHET, KNN(K-Nearest Neighbour), and Feed Forward Neural Network. Historically recorded stock price data was obtained from the National Stock Exchange (NSE) and used to create these models, which were then used for comparative purposes. Based on the historical data samples that were used to construct the models, the obtained findings indicate that all four models have substantial potential for prediction and forecasting. All the models performed better on larger data sets, with the Feed Forward Neural Network model proving to be the most accurate in terms of projecting future stock price movement. The second part of the study involved multivariate time series analysis. The impact of the weather on stock price returns and volatility has been observed here. 12 years Mumbai Temperature data has been collected and fed into a GARCH model along with above mentioned five stocks’ 12 years daily stock price return. The results show that stock temperature of Mumbai has no correlation with stock price return but it affects the stock price volatility.

Actions (login required)

View Item