Pistologistics: S&OP Dataset for AI-Driven Supply Chain Optimization

A comprehensive Sales and Operations Planning (S&OP) dataset with synthetic data generation capabilities for supply chain AI modeling and optimization.

Overview

Pistologistics provides a rich, synthetic dataset that simulates a complete supply chain ecosystem with realistic patterns and relationships. It is designed specifically for training and evaluating AI models for supply chain optimization, demand forecasting, inventory management, and production planning.

Key Features

• Comprehensive Supply Chain Data Model: Complete schema covering master data, transaction data, and planning tables
• Realistic Data Generation: Sophisticated algorithms to create realistic patterns with trends, seasonality, external factors, and anomalies
• Advanced Supply Chain Metrics: Includes inventory optimization metrics (EOQ, safety stock), production planning, and KPIs
• Multi-horizon Forecasting Support: Historical and future data for training time-series models
• Configurable Generation Parameters: Customize data size, complexity, and characteristics
• Supply Chain Disruption Simulation: Model the impact of disruptions on the supply chain
• Promotion Impact Modeling: Simulate marketing activities and their effect on demand

Dataset Components

Master Data

• Products with categories, costs, and attributes
• Locations (DCs, stores, plants, suppliers)
• Customers with segments and regions
• Suppliers with reliability metrics
• Resources with capacity and efficiency

Transaction Data

• Demand forecasts with confidence levels
• Inventory levels with optimization metrics
• Production plans with resource requirements
• Transport lanes with costs and transit times

External Factors

• Economic indicators (GDP, inflation)
• Seasonal patterns
• Market events
• Weather impacts
• Promotional activities
• Supply disruptions

Performance Metrics

• Inventory KPIs (turns, days of supply)
• Production KPIs (efficiency, utilization)
• Supply chain KPIs (forecast accuracy, on-time delivery)

AI Model Compatibility

The dataset is designed to support various AI/ML model types:

Time Series Forecasting Models

• ARIMA/SARIMA
• Prophet
• LSTM Networks
• Transformer-based Models

Gradient Boosting Models

• XGBoost
• LightGBM
• CatBoost

Neural Network Models

• Feed-forward Neural Networks
• Recurrent Neural Networks
• Graph Neural Networks

Optimization Models

• Linear Programming
• Mixed Integer Programming
• Constraint Programming
• Reinforcement Learning

Example Use Case: Multi-horizon Demand Forecasting

import pandas as pd
from sklearn.model_selection import train_test_split
from xgboost import XGBRegressor

# Load the data
demand_df = pd.read_csv('sample_data/DemandForecast.csv')
external_factors = pd.read_csv('sample_data/ExternalFactors.csv')
promotions = pd.read_csv('sample_data/PromotionCalendar.csv')

# Merge datasets
merged_data = demand_df.merge(external_factors, on='date', how='left')

# Feature engineering
merged_data['day_of_week'] = pd.to_datetime(merged_data['date']).dt.dayofweek
merged_data['month'] = pd.to_datetime(merged_data['date']).dt.month
merged_data['is_promoted'] = merged_data['product_id'].isin(promotions['product_id'])

# Prepare features and target
features = ['day_of_week', 'month', 'gdp_factor', 'inflation_factor', 
            'seasonal_factor', 'market_event_factor', 'weather_impact_factor', 
            'is_promoted']
X = merged_data[features]
y = merged_data['forecast_quantity']

# Train-test split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Train model
model = XGBRegressor(n_estimators=100, learning_rate=0.1, max_depth=5)
model.fit(X_train, y_train)

# Evaluate
predictions = model.predict(X_test)
rmse = ((predictions - y_test) ** 2).mean() ** 0.5
print(f"RMSE: {rmse}")

Installation and Usage

1. Clone this repository:

   git clone https://github.com/yourusername/pistologistics.git
   cd pistologistics
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Generate sample data:

   python data_generator.py
   

4. Explore the generated data in the sample_data directory

Scaling Recommendations

For different use cases, adjust the data generation parameters:

• Small-scale testing: 100 products, 10 locations, 1,000 forecast records
• Medium-scale development: 500 products, 50 locations, 10,000 forecast records
• Large-scale production: 1,000+ products, 100+ locations, 50,000+ forecast records

Advanced Supply Chain Engineering Features

The dataset includes specialized features for supply chain engineers:

Inventory Optimization

• Economic Order Quantity (EOQ) calculation
• Safety stock determination based on service levels
• Reorder point calculation
• Inventory turns and days of supply metrics

Production Planning

• Resource capacity planning
• Setup time optimization
• Material requirements planning
• Production cost analysis

Supply Chain Risk Management

• Disruption impact modeling
• Supplier reliability analysis
• Risk mitigation strategies
• Resilience assessment

Performance Management

• KPI tracking and analysis
• Target setting methodology
• Variance analysis
• Performance improvement planning

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Inspired by real-world S&OP processes and supply chain management practices
• Designed for educational and research purposes in supply chain analytics