AI-Driven Business Intelligence: A Case Study on Predicting Market Trends Using LSTM

Abstract

The ability to predict market developments accurately helps in decision-making and planning more effectively. In the financial markets, however, classical methods such as ARIMA and exponential smoothing models are ineffective when the data has nonlinearities and seasonality’s and is highly dependent in the long-term. The proposed paper provides an intelligent solution with the application of artificial intelligence techniques for forecasting market trends in the form of LSTM neural networks. Incorporating past price data, trading volume, and technical indicators such as SMA, EMA, RSI, MACD, the LSTM prediction algorithm uses these inputs to forecast market trends. The daily data of 14 years (2010–2023) have been used for training LSTM network. To generate the input windows, the missing values were imputed, feature variables were normalized, and outlier exclusion was performed during the pre-processing stage; 60 days' windows were formed. Various criteria were used to evaluate the performance of the proposed approach, and the obtained results were: RMSE – 0.97; MAE – 0.72; MAPE – 1.85% and R² – 0.91. The following conclusions can be drawn: The suggested algorithm is more accurate and has better features in terms of trend detection than the other prediction models. Apart from examining the predictive power of the chosen approach, other potential problems examined included data quality, complexity and interpretability. The fields where future research can be undertaken include combining various techniques of deep learning, making use of real-time data present in the market, and applying techniques of explainable AI. In conclusion, it is evident that LSTM is a valuable approach to utilize AI in BI applications with respect to market prediction.