Intraday Market Analysis and Forecasting with LSTM Networks

Intraday price estimation is a significant analytical operation to a trader by depending on the short run movements to optimize the entry and exit positions. This paper is an exploration of how Long Short-Term Memory (LSTM) networks can be used to forecast the daily minimum and maximum price of Infosys Ltd. with the help of past information available in Yahoo Finance between the years 2015 and 2025. The data was scaled with Min-Max scaling and each time series of $\mathbf{1, 5 0 0}$ timesteps was generated to form inputs of two dedicated LSTM models one high-price prediction and one low-price prediction. All the models were trained with the Adam optimizer and the mean squared error loss. High-price model had an R2 of 0.954, MAE of 34.01 and RMSE of 44.26 whereas the low-price model had an R2 of 0.947, MAE of 35.26 and RMSE of 46.41. The above findings suggest that LSTM models are capable of learning long-term temporal structure and give high quality short-term price predictions when the conditions are good.