Cloud-Based Time Series Forecasting Using LSTM for Real-Time Predictive Analytics

This research presents a comprehensive end-to-end MLOps pipeline designed for financial time-series forecasting, specifically applied to the volatile nature of stock market data. The project addresses the limitations of traditional statistical models like ARIMA and Exponential Smoothing, which often fail to capture the non-linear and non-stationary patterns of real-world financial dataset Technical Workflow Cloud Infrastructure: Built on AWS, using S3 for data storage and Amazon SageMaker for scalable model training and real-time deployment. Data Processing: Includes feature selection (closing prices), MinMax scaling, and 60-day sliding window sequence generation. Deployment: The model is hosted as a real-time HTTPS endpoint on an ml. t2. medium instance. Key Results Accuracy: The model achieved a directional prediction accuracy of 95% on Tesla (TSLA) stock data. Error Metric: Recorded a Root Mean Squared Error (RMSE) of 90. 49