Forecasting volatility trend of INR USD currency pair with deep learning LSTM techniques

Volatility is an important and most discussed topic in finance. Many of financial trades and applications are based on the volatility. In the recent times currency pair conversion trades are new found interest among financial traders due to high instability in the financial market. The advancement in technologies, increased computing speed and capability to handle large data has given rise to deep learning techniques. In this paper, deep learning LSTM techniques have been used to solve volatility forecasting problem of INR USD currency pair. The research uses an innovative approach in arrangement of data to make use of recent 25 values for forecasting volatility trend. The algorithm forecasts uptrend or downtrend movement of volatility a day ahead. The experiments were conducted to forecast volatility using machine learning and deep learning techniques. The LSTM technique is experimented with several epochs and configurations to yield better accuracy. The results show that LSTM techniques produced better accuracy compared with neural networks, SVM, random forest, regression, decision trees and boosting techniques. One of the main application of this research paper is forecasting the rise and fall of INR versus USD. The approach can also be applied to forecasting problems in algorithmic trading, churn predictions, Lead optimization and Fraud detections.