Vol: 14, Issue: 1 2024

Page: 078-087

Developing a Time Series Financial Market Forecasting Model Based on Machine Learning Tools and Techniques

Jaideep Singh Bhullar

http://doi.org/10.37648/ijtbm.v14i01.010

One critical research area in this regard is financial market forecasting, where the outcome bears critical implications for both investors and policy makers and various financial institutions. These markets, represented in the stock markets and other types of financial products, show complexities in nonlinearities and dynamics brought about by multitudes of interacting factors such as macroeconomic variables, investor emotions, and general global events. The traditional ARIMA and GARCH models have found extensive application in financial forecasting. However, these models are not able to capture the intricate dependencies and nonstationary nature of financial time series. Recent improvements in ML and DL have led to very strong analytic and predictive powers of analyzing the trends in the financial market in much greater precision. SVM, RF, k-NN-based algorithms have performed significantly well to describe the complicated interaction patterns within the financial data. Furthermore, deep learning techniques, such as RNNs, LSTM networks, and CNNs, have been proven to have better performance in time series forecasting by capturing long-term dependencies and hierarchical patterns in the data. This paper carries out an all-rounded review of the applied ML techniques for financial market time series forecasting from 2013 to 2022. To that end, we give room to the strengths and weaknesses of different ML models by offering a comparative analysis in terms of performance metrics while focusing on the integration of alternative data sources such as news sentiment, social media analytics, and economic indicators for improving the accuracy of predictions. We discuss the issues associated with financial forecasting, such as overfitting, data quality issues, and model interpretability. This also includes empirical comparisons and tabular analyses of the performance of various models using different datasets. We provide a roadmap for future research in this domain through a comprehensive investigation of recent IEEE publications and other sources. The present contributions have the potential to aid further advancements in financial analytics and decision-making processes, which may improve investment strategies and risk management frameworks