An Application of Decision Tree Algorithms in Financial Markets

In this paper, we investigate the predictive and anticipatory capabilities of stacked Long Short-Term Memory (LSTM) models for stock market patterns. Pre-processing of historical stock market data from several industries included data normalization, cleaning and partitioning into training and testing sets. With the help of the TensorFlow backend and the Keras package, we constructed a stacked LSTM model that has dense layers for output prediction after several LSTM layers. Cross-validation was used to fine-tune hyperparameters such batch sizes and dropout rates, which enhanced the performance of the model. The model's assessment on the test dataset demonstrated its high prediction accuracy and low error rates. The model's ability to detect underlying patterns and volatility was tested by comparing real stock prices to forecasts. Forecasting over the following fifty days confirmed the applicability of the method. Our findings highlight the use of stacked LSTM models in financial forecasting, providing insights into market operations and directing investment strategies in ever-changing financial markets.