Prediction of Stock Market Price using Bi-directional Recurrent Neural Network

Recently, an accurate prediction of stock market returns is a very challenging task due to volatile and non-linear nature of the financial stock markets. The advent of artificial intelligence and enhanced processing power has led to the realisation that preprogrammed techniques are more effective at forecasting stock values. The stock market is currently one of the most researched and fastest developing subj ects, and predicting its behaviour is crucial. The primary challenge in this area of research remains improving forecast precision. Strong shareholder returns are made possible by stock price forecasting. Despite this, researchers use these algorithms to predict continuing market trends based on Stock Technical Indicators (STIs), due to recent advancements in machine learning techniques. This study analyses Yahoo Finance data from STIs over a decade to predict stock prices using a Bi- directional Recurrent Neural Network (BI-RNN). Initially, the analysed STIs were used as input for an autoencoder during the dimensionality reduction procedure, which reduced the correlation between the STIs. These STIs and the Yahoo Finance data were then provided to BI-RNN. Next, in order to forecast stock prices, the BI-RNN's outcome attributes were fed into the soft max layer. From the results, it clearly shows that proposed Bi-RNN shown better results in terms of a minimal MAPE value of 0.41, MAE of 4.42, RMSE of 0.10, and MSE of 212.25 in the experiments, the proposed strategy surpassed traditional methods.