E-Commerce Spending Prediction Dashboard

This project predicts yearly customer spending based on their activity on an e-commerce platform. It includes data exploration, regression modeling (Linear, Ridge, and Lasso), evaluation, and a deployment-ready Streamlit dashboard.

---

Tech Stack & Libraries

• Languages: Python 3

• Libraries:

  • Data Handling: pandas, numpy
  • Visualization: matplotlib, seaborn
  • Modeling: scikit-learn, statsmodels, joblib
  • App Interface: streamlit

---

Dataset

• Source: Ecommerce Customers.csv

• Features Used:

  • Avg. Session Length
  • Time on App
  • Time on Website
  • Length of Membership
  • Yearly Amount Spent (target)

---

Exploratory Data Analysis (EDA)

• Scatterplots to explore relationships (e.g. Time on App vs Yearly Spending)
• Barplots to analyze binned behavior
• Pairplots & Jointplots for pairwise feature insights
• Distribution plots for residual analysis

---

Modeling

Linear Regression (OLS)

• Performed using both scikit-learn and statsmodels
• Extracted coefficients and evaluated performance (R² score)

Ridge & Lasso Regression

• Added regularization to improve generalization
• Used pipelines with standard scaling
• Compared metrics: MAE, MSE, RMSE, R2

Multicollinearity Check

• Used Variance Inflation Factor (VIF) to ensure no multicollinearity among features

---

Model Evaluation

• Metrics Used:

  • Mean Absolute Error (MAE)
  • Mean Squared Error (MSE)
  • Root Mean Squared Error (RMSE)
  • R² Score

• Plots:

  • Actual vs Predicted
  • Residual distribution
  • Normal Q-Q plot

---

This project builds and deploys a machine learning model to predict yearly e-commerce customer spending based on usage metrics like session length, time on app, and membership duration.

Key highlights:

Model Used: Linear Regression (with comparisons to Ridge & Lasso)

Best Model R² Score: 0.98, indicating a strong predictive fit

Mean Absolute Error (MAE): 7.54

Root Mean Squared Error (RMSE): 9.36