README.md

---
title: DataSynthis ML Job Task
emoji: 📈
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: "4.44.0"
app_file: app.py
pinned: false
license: mit
---
# DataSynthis_ML_JobTask – Gradio Space

## Models
- **ARIMA** → classical statistical model.  
- **Prophet** → trend + seasonality decomposition.  
- **LSTM (Multivariate)** → deep learning with multiple features.  
 
**Data:** AAPL daily prices (2015–2025)

## How it works
- **Forecast tab:** Choose horizon (7–90d). See future projections overlayed on history.
- **Backtest tab:** Evaluate models on the last N days (RMSE & MAPE table).

## Files
- `data/AAPL_stock_2015_2025.csv`
- `models/arima_model.pkl`, `models/prophet_model.pkl`, `models/lstm_model.h5`, `models/lstm_scaler.pkl`
- `app.py`, `requirements.txt`

> LSTM multi-step uses naive future covariates for demo. For production, predict covariates or use a univariate target model.