📊 AI-Powered Sales Forecasting Dashboard

⏳ Time-Series Forecasting • 📉 SARIMA Model • 🌐 Streamlit App • 🛒 Walmart Sales Data

This project predicts future weekly retail sales using real historical data and provides an interactive dashboard built with Streamlit. It includes full data preprocessing, sales trend analysis, time-series modeling (SARIMA), forecasting, and deployment.

🔗 Live Demo: https://vidit-ml-forecast.streamlit.app

🔗 Dataset: Walmart Weekly Sales Dataset (Kaggle)

---

🚀 Project Features

✅ 1. Data Processing & Cleaning

• Convert dates to datetime format

• Sort weekly data chronologically

• Handle missing values

• Aggregate sales by date

✅ 2. Data Visualization

• Line plots of weekly sales

• Rolling 12-week average trend

• Seasonal decomposition (trend, seasonality, residuals)

✅ 3. Time-Series Forecasting (Modeling)

• SARIMA (1,1,1)(1,1,1,52)

• 12-week future forecast

• Model diagnostics & evaluation

• MSE, RMSE, MAE metrics

✅ 4. Dashboard (Streamlit)

• Interactive sections include:

📅 Weekly Sales Table (with rolling mean)

📈 Actual vs Forecasted Sales Chart

🔮 Future Predictions Table

• Responsive & clean layout

---

🗂️ Project Structure

sales-forecasting-dashboard/
│
├── app.py                # Streamlit dashboard
├── weekly_sales.csv      # Cleaned dataset (exported from Colab)
├── forecast.csv          # SARIMA model forecast (exported from Colab)
├── requirements.txt      # Required dependencies
└── README.md             # Project documentation

---

🧠 Modeling Approach

🟦 SARIMA Model

• I used a Seasonal ARIMA (SARIMAX) model with weekly seasonality (52 weeks). Model configuration:

order = (1, 1, 1)
seasonal_order = (1, 1, 1, 52)

📈 Forecast Output

• The model predicts 12 weeks ahead, generating a clean forecast used in the Streamlit dashboard.

---

🛠️ Tech Stack

Category Tools:

• Programming Python

• Data Handling Pandas, NumPy

• Visualization Matplotlib

• Time-Series Modeling Statsmodels (SARIMAX)

• Cloud Notebook Google Colab

• Dashboard Streamlit

• Deployment Streamlit Cloud

• Version Control GitHub

---

▶️ How to Run the Project Locally

1️⃣ Clone the repository

git clone https://github.com/your-username/sales-forecasting-dashboard.git
cd sales-forecasting-dashboard

2️⃣ Install dependencies

pip install -r requirements.txt

3️⃣ Run Streamlit

streamlit run app.py

---

🌐 Deployment (Streamlit Cloud)

The project is deployed on Streamlit Cloud.

Key steps:

1. Upload project folder to GitHub

2. Go to https://streamlit.io/cloud

3. Deploy → Select GitHub repo

4. Choose app.py as the entry point

5. Add environment file (requirements.txt)

6. Deploy & get a public URL

---

📸 Dashboard Preview

---

📉 Results

• The SARIMA model successfully forecasts 12 weeks of sales

• Rolling averages reveal seasonal patterns

• Dashboard provides intuitive visual insights

• RMSE and MAE validate model performance

---

⭐ Future Improvements

• Add Facebook Prophet / NeuralProphet models

• Add advanced EDA with store-wise or department-wise charts

• Add interactive filters in Streamlit

• Add anomaly detection on unusual sales spikes

---

🤝 Contributing

• Pull requests, suggestions, and forks are welcome!

• Feel free to open an issue if you'd like to improve the dashboard.

---

📬 Contact

Vidit Kumar

📧 Email: vidit.kumar624@gmail.com

🌐 LinkedIn: linkedin.com/in/viditkumar-in

---