rummanadib023/DataSynthis_ML_JobTask

DataSynthis ML JobTask - Stock Forecasting Models

Model Overview

This repository contains multiple time series forecasting models for stock price prediction:

📊 Models Included:

1. ARIMA - Traditional statistical model
2. LSTM - Long Short-Term Memory neural network
3. GRU - Gated Recurrent Unit neural network

📈 Performance Metrics:

ARIMA Model:

• RMSE: $7.9293
• MAE: $5.3609
• MAPE: 5.70%

Deep Learning Models:

LSTM:

• RMSE: $7.8890
• MAE: $5.4603
• MAPE: 5.58%
• Direction Accuracy: 49.51%

GRU:

• RMSE: $8.1412
• MAE: $5.6259
• MAPE: 5.69%
• Direction Accuracy: 49.33%

📋 Dataset Information:

• Stock: AAPL (Apple Inc.)
• Date Range: 2015-01-02 00:00:00 to 2024-12-31 00:00:00
• Observations: 2516
• Current Price: $249.53

🚀 Quick Start

from huggingface_hub import hf_hub_download
import joblib
import tensorflow as tf

# Download ARIMA model
arima_path = hf_hub_download(repo_id="rummanadib023/DataSynthis_ML_JobTask", filename="arima_model.joblib")
arima_data = joblib.load(arima_path)
arima_model = arima_data['model']

# Download LSTM model
lstm_path = hf_hub_download(repo_id="rummanadib023/DataSynthis_ML_JobTask", filename="lstm_model.keras")
lstm_model = tf.keras.models.load_model(lstm_path)

# Download GRU model
gru_path = hf_hub_download(repo_id="rummanadib023/DataSynthis_ML_JobTask", filename="gru_model.keras")
gru_model = tf.keras.models.load_model(gru_path)

# Download metadata
metadata_path = hf_hub_download(repo_id="rummanadib023/DataSynthis_ML_JobTask", filename="dl_metadata.joblib")
dl_metadata = joblib.load(metadata_path)

📝 Model Details

ARIMA Model

• Order: (2, 2, 3)
• Type: Statistical time series model
• Best for: Medium-term trend analysis

Deep Learning Models

• Architecture: Multi-layer LSTM/GRU with dropout and batch normalization
• Input: Normalized returns with 60-day lookback
• Output: Next-day return prediction
• Best for: Short-term directional forecasts

⚠️ Disclaimer

This model is for educational and research purposes only. Stock market predictions are inherently uncertain and past performance does not guarantee future results.

---

Model created on 2025-10-03 18:57:18