Comprehensive Analysis of Deep Learning Architectures for Temporal Prediction in Financial Markets

Abstract

A set of data arranged by the time of the year is referred to as a time series, and the standard method of the presentation of the changes in the thing during time. To help in decision-making, time series forecasting tries to predict the future values of time series by attempting to predict how these values alter with time. Four large companies on the Nasdaq index are Amazon, Google, Microsoft and Apple. In this work, perform an analysis of architectural deep learning models in the prediction of stock prices on the financial market, based on its stock projections. This included the normalization of data, trimming of outliers and sentimenting of data since getting Ticker Stock via Yahoo Finance until April 2025. Four models were analyzed using traditional measures, including LSTM, GRU, RNN and ARIMA. RMSE, MSE and MAE were performance indicators. It has been shown that LSTM models can perform better than rival models, when required to model the long-term dependencies of time series data. Sentiment analysis and consideration of technology substantially boost the ability of the model to predict. It is also seen that deep learning models have the potential to predict financial time series, and it is possible to propose methods to enhance the present system of predicting market volatility.