REAL-TIME STOCK PRICE PREDICTION WITH A HYBRID DEEP LEARNING MODEL USING YFINANCE API

Abstract

The study is a deep learning-driven hybrid model is developed for stock price prediction by integrating historical
market data with real-time news sentiment analysis. The system leverages a combination of classical time series
models (ARIMA, Linear Regression) and advanced deep learning architectures (LSTM-GRU hybrid networks)
to accurately forecast future stock prices. Real-time financial news headlines are retrieved from Yahoo Finance
and analyzed using Natural Language Processing (TextBlob) to classify sentiment as Positive, Negative, or
Neutral, enriching the prediction model with external market factors. The resulting forecasts, along with
actionable investment recommendations (Buy, Sell, or Hold), are delivered through an interactive, web-based
application built using the Flask framework. Model performance is rigorously evaluated using metrics such as
Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE), demonstrating that the hybrid LSTM-GRU
model achieves superior accuracy, especially in capturing non-linear and volatile market behaviors. This project
highlights the effectiveness of combining historical data, deep learning techniques, and real-time sentiment
analysis to build a robust, adaptive, and user-friendly stock forecasting system.