Real-Time Stock Market Forecasting using Ensemble Deep Learning and Rainbow DQN

After years of study by researchers and finance experts on stock market prediction, there is no definite method that seems to predict stock price both accurately and is long- lasting at the same time. This is due to the uncertain behavior of stock movement and numerous parameters that take part in deciding market performance. Lots of techniques have been examined for stock market price prediction. Also, various comparative studies have been done to find the best techniques which can help traders make decisions. This paper proposes the use of ensemble techniques like Rainbow DQN, LSTM, and GRU for real time stock market forecasting and indicating buy/sell signals. The results of these techniques used were closely examined keeping a check on overfitting. The performance of the techniques was evaluated using accuracy, return of investment. Few well-known indicators like MACD and RSI were also used for verifying results. Tests were run on NIFTY50, Microsoft and Google with real-time data structured in 1-minute intervals