JSLL Stock Prediction Platform

ML forecasting & monitoring platform (live) with a per-horizon model tournament, confidence bands and automatic drift-triggered retraining.

The JSLL Stock Prediction Platform is a Django application for forecasting and monitoring Jeena Sikho Lifecare Limited (JSLL.NS). It combines market-price APIs, a dashboard and a model-selection pipeline. The engineering problem extends beyond producing a forecast: the application must align observations with NSE trading sessions, compare models across forecast horizons and show when available history is insufficient to judge a prediction. A tournament pipeline selects champions by horizon, generates predictions with confidence and uncertainty bands, and tracks predictions against matched actual prices. Data-quality checks, repair jobs and drift checks support the pipeline, with automatic retraining configurable when drift is detected. The documented feature inputs include Nifty, India VIX and USD/INR; event-calendar features are configurable separately. Runtime predictions, scoreboards and OHLCV history use SQLite-backed storage. The dashboard and JSON APIs expose tournament summaries, scoreboards, latest predictions and run status. Operator-facing reports distinguish capability status, baseline accuracy, matched-history sufficiency and promotion decisions. This makes an important limitation visible: longer-horizon predictions may exist before enough corresponding actual observations have accumulated. A pending or insufficient-sample state should not be read as proven forecast accuracy. The documented VPS deployment uses Gunicorn behind Nginx under a /jsll/ path prefix. Scheduled pipeline work is separate from public dashboard polling, and mutation endpoints for tournament runs and prediction refreshes are admin-token gated outside debug mode. The project demonstrates forecasting infrastructure and model monitoring; its confidence bands and model rankings are not guarantees of market performance.

Python, Django, Machine Learning, pandas, yfinance, Nginx

Data Science · ML