About NeuroTrade

An end-to-end ML pipeline for stock signal prediction.

Project Overview

NeuroTrade is a full-stack machine learning portfolio project that predicts Buy/Sell/Hold signals for 20 US stocks using historical OHLCV data and technical indicators.

The pipeline trains three base models — XGBoost, LightGBM, and LSTM — on 2019-2023 data, then combines them using a Logistic Regression stacking ensemble.

All predictions are from models trained strictly on 2019-2023 data. The 2024 validation set was opened only once for final evaluation.

ML Pipeline

01
Data Collection
Downloaded 20 stocks from Yahoo Finance (2019-2024), daily OHLCV
02
EDA
Exploratory analysis of price distributions, correlations, volume patterns
03
Feature Engineering
17 technical indicators: RSI, MACD, Bollinger Bands, ATR, volume features
04
XGBoost
Gradient boosted trees with 5-fold purged TimeSeriesSplit CV
05
LightGBM
Leaf-wise growth trees, faster training, comparable performance
06
LSTM
20-day sequence model capturing temporal momentum patterns
07
Stacking Ensemble
Logistic Regression meta-model combining OOF predictions from all 3 models

Tech Stack

ML Models
  • XGBoost
  • LightGBM
  • TensorFlow/Keras
  • scikit-learn
Data
  • yfinance
  • pandas
  • numpy
  • Yahoo Finance API
Backend
  • FastAPI
  • Python 3.11
  • uvicorn
  • joblib
Frontend
  • Next.js 16
  • TypeScript
  • Tailwind CSS
  • Axios
LLM
  • Groq API
  • Llama 3.3 70B
  • Natural language explanations
Deployment
  • Vercel (frontend)
  • Hugging Face Spaces (backend)

ML Methodology

No Data Leakage
5-day purge gap at fold boundaries prevents target leakage from 5-day forward returns
Date-based CV Splits
TimeSeriesSplit on unique dates not row indices ensures correct chronological ordering
Per-ticker Sequences
LSTM sequences built inside groupby loop, no cross-ticker boundary contamination
Honest Evaluation
Validation set opened exactly once. No retuning after seeing results.

Developers

O
Om Naik
ML Engineer
LinkedInGitHub
M
Murali Manohara
ML Engineer
LinkedInGitHub

Disclaimer

NeuroTrade is a portfolio demonstration project built for educational purposes. All predictions are generated by machine learning models trained on historical data. Past performance does not guarantee future results. This is not financial advice. Do not make investment decisions based on these signals.