A Comparative Analysis of the performance of the LSTM Network and ARIMA Model for the forecasting of a non-stationary financial timeseries.

Abstract- Forecasting the stock-market is an age-old requirement in an investor's tool-kit for successful investments. The stock market is affected by various factors that include (but are not limited to) sentiments, demand, supply, policy, political climate, among others. This makes predicting the dynamics of the stock market an incredibly challenging task. It is a combination of both random and deterministic phenomena. For this task, a variety of deep learning and stochastic models have been developed. This paper attempts to compare two of the most used predictive models: LSTM-RNN and ARIMA over data collected from the State Bank of India's historical stock pric