Sales Forecasting & Demand Intelligence System

An end-to-end data science project built on 4 years of US retail sales data (Superstore dataset).

What this project covers

• Time series decomposition and stationarity testing (ADF)
• 4 forecasting models: SARIMA, Prophet, XGBoost, LSTM
• Anomaly detection using Isolation Forest and Z-Score methods
• Product demand segmentation using K-Means clustering + PCA
• Interactive Streamlit dashboard with 5 pages including an AI-powered Q&A page

Tech Stack

Python, Pandas, Statsmodels, Prophet, XGBoost, TensorFlow, Scikit-learn, Plotly, Streamlit

Dataset

Superstore Sales Dataset (2015–2018) — 9,800 rows, 18 features