Jann/investment-sandbox
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Closes #019 - Live Python Machine Learning Pipeline Integration

Browse Source
This commit is contained in:
Antigravity Agent
2026-06-14 13:07:55 +02:00
parent 02cedd6ee6
commit 9c5cd78801
6 changed files with 920 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
DEV_LOG.md
View File

@@ -214,4 +214,19 @@ This document tracks all modifications, npm packages, active compilation states,
* **Active Bugs**: None.
* **Type Checker Status**: Verified 100% clean type verification (`npx tsc --noEmit` returns exit code 0).
---
## [2026-06-14] - Live Python Machine Learning Pipeline Integration (#ISSUE-019)
### Added
* **Real Data Ingestion**: Added a web-querying routine (`fetch_real_data()`) in `backend/core/pipeline.py` that queries 2 years of daily candles for `BTC-USD` from Yahoo Finance API (saving to `backend/data/BTC-USD.csv`) and queries real-time funding rates from the Binance USDS-M Futures REST API.
* **Environment Dependency**: Installed `scikit-learn` in the local `Miniconda3` python environment using pip.
* **Pipeline Model Training**: Executed model training using standard Walk-Forward validation on actual daily closing prices. Trained 5 estimators (Random Forest, Gradient Boosting, Logistic Regression, SVM, MLP) across 3 horizons (T+1, T+5, T+10).
* **Dynamic Ensemble Export**: Exported real probabilities into `public/data/ensemble_predictions.json` with `isShieldActive: false` to allow the terminal's Walk-Forward Ensemble Radar to render live active signals.
### Active Bugs / Compile Status
* **Active Bugs**: None.
* **Type Checker Status**: Verified 100% clean type verification (`npx tsc --noEmit` returns exit code 0).