Predicting Stock Market Trends Using Long Short-Term Memory (LSTM) Networks: A Deep Learning Approach for Financial Time-Series Forecasting

Investments in a diversified basket of assets have always been a challenge because financial markets are intrinsically unpredictable. Sometimes, simplicity cannot make good predictions. It is only because of advances in machine learning, where computers can be made to perform tasks typically requiring human intelligence, that it has advanced to become one of the leading approaches in financial prediction research. This paper validates various machine learning techniques with an emphasis on the RNN model coupled with a specific LSTM model set up for making predictions of stock market prices using history stock prices and volume trading. Different metrics of the model have been tested based on requirements to check the performance, such as mean absolute error, the root mean squared error, and comparison of accuracy with actuals based on calibration of epochs during training. Our results further stress the fact that LSTM models constitute several advantages as predictors because they can seize sequential dependencies of financial time series data and therefore may hold promise for machine learning applications towards forecasting asset values.