Machine Learning Applications in Short-term Financial Forecasting: A Comparative Analysis with Hybrid Model

Financial analysis and trading are very crucial in any economy. It holds a significant importance to generate income and reducing income equality, especially in developing countries where inequality and employment are major concerns. Forex analysis deals with the currency exchange and plays a major role in world economy. Accurate prediction of currency market is almost impossible as there are many unpredictable factors may instantly come to play and greatly affect the market scenario. Hence, long term predictions may not be very successful however, short term predictions or very short-term prediction may be accurate up to some extent. This work tries to predict the next day currency exchange rate (US Dollar to Indian Rupees) using machine learning models. Three machine learning models such as ANN-PSO, RNN and LSTM have been implemented to predict the next day currency exchange rate. The data set used in this work is the daily trading data downloaded from the www.investing.com for the period 1st January 2020 to 20th May 2024. Using the close price of these values the models were trained and the value for the day 21st May 2024 has been predicted. The predicted value for 21st May is $83.39,83.27$ and 83.47 by RNN, LSTM and ANN-PSO models respectively. Comparing this with the actual value of 21st May 2024, the accuracy thus calculated for the models are $\mathbf{9 9. 8 6 \%}, \mathbf{9 9. 9 8 \%}$ and $\mathbf{9 9. 7 8 \%}$ by RNN, LSTM and ANN-PSO models respectively.