Bitcoin Prediction Tracker

The premise

Predicting Bitcoin’s price with a single model is a fool’s errand. Markets are noisy, models overfit, and any one approach captures only a slice of the signal. The Bitcoin Prediction Tracker takes a different approach: run multiple prediction models in parallel, track their performance over time, and let the ensemble tell you which signals are worth paying attention to.

How it works

The system is built around an ensemble service that orchestrates multiple prediction models. Each model produces its own forecast, and the ensemble layer aggregates, weights, and tracks accuracy across all of them. A dashboard surfaces the predictions, model performance metrics, and historical accuracy so you can see which models are earning their keep and which are noise.

The entire pipeline is automated with n8n workflows that handle data collection, model execution, result aggregation, and alerting. Data archives and backups ensure nothing is lost as the prediction history grows.

The stack

The backend is a full-stack system with an API layer, the ensemble prediction service, and a frontend dashboard. n8n handles the workflow automation that ties everything together, from scheduled data pulls to model runs to notification triggers.

What it demonstrates

This project shows the data and ML side of my range. It is a completed, feature-complete system that combines data science (multiple prediction models), backend engineering (API + ensemble service), frontend (dashboard), and workflow automation (n8n). It is in maintenance mode now, but the architecture is designed to absorb new models without structural changes.

The biggest lesson: the value of ensemble approaches is not in any single prediction being right. It is in the meta-layer that tracks which approaches work under which conditions. The tracker became more useful as a model evaluation tool than as a price prediction tool.