Customer Personality Analysis: Campaign Response Prediction

Goal/Objective of the Project

The primary objective of this project is to build a supervised machine learning classification model capable of predicting whether a customer will respond positively to a marketing campaign. By analyzing customer demographics, purchasing behavior, and historical interaction with previous campaigns, the model helps businesses optimize their marketing strategies. Instead of spending resources targeting all customers, the business can accurately identify the segment most likely to accept the offer, thereby increasing conversion rates and maximizing return on investment (ROI).

Dataset

Dataset URL: Customer Personality Analysis on Kaggle

The dataset contains detailed information regarding the company's customer base, structured into various attributes:

Target Variable

• Response: 1 if the customer accepted the offer in the last campaign, 0 otherwise.

Features/Columns Description

People (Demographics & History)

• ID: Customer's unique identifier.
• Year_Birth: Customer's birth year.
• Education: Customer's highest education level.
• Marital_Status: Customer's marital status.
• Income: Customer's yearly household income.
• Kidhome: Number of children in the customer's household.
• Teenhome: Number of teenagers in the customer's household.
• Dt_Customer: Date of the customer's enrollment with the company.
• Recency: Number of days since the customer's last purchase.
• Complain: 1 if the customer complained in the last 2 years, 0 otherwise.

Products (Spending Habits over the last 2 years)

• MntWines: Amount spent on wine.
• MntFruits: Amount spent on fruits.
• MntMeatProducts: Amount spent on meat products.
• MntFishProducts: Amount spent on fish products.
• MntSweetProducts: Amount spent on sweets.
• MntGoldProds: Amount spent on gold products.

Promotion (Campaign History & Deal Usage)

• NumDealsPurchases: Number of purchases made with a discount.
• AcceptedCmp1: 1 if the customer accepted the offer in the 1st campaign, 0 otherwise.
• AcceptedCmp2: 1 if the customer accepted the offer in the 2nd campaign, 0 otherwise.
• AcceptedCmp3: 1 if the customer accepted the offer in the 3rd campaign, 0 otherwise.
• AcceptedCmp4: 1 if the customer accepted the offer in the 4th campaign, 0 otherwise.
• AcceptedCmp5: 1 if the customer accepted the offer in the 5th campaign, 0 otherwise.

Place (Channels of Purchase)

• NumWebPurchases: Number of purchases made through the company's website.
• NumCatalogPurchases: Number of purchases made using a catalog.
• NumStorePurchases: Number of purchases made directly in stores.
• NumWebVisitsMonth: Number of visits to the company's website in the last month.

(Note: Features like Z_CostContact and Z_Revenue have constant values across the dataset and are typically removed during preprocessing).

Data Cleaning and Preparation

To ensure the model receives high-quality data, the following techniques were utilized for data preprocessing:

• Handling Missing Values: Missing values (primarily found in the Income column) were imputed using the median to avoid skewing the distribution with extreme values.
• Date/Time Formatting: The Dt_Customer feature was converted to a standard datetime format.
• Feature Engineering: * Derived an Age feature by subtracting Year_Birth from the current data year.
  • Created an Enrollment_Duration (or Tenure) feature to measure customer loyalty based on Dt_Customer.
  • Aggregated spending columns to create a Total_Spend feature.
  • Combined Kidhome and Teenhome to create a Total_Dependents feature.
• Dropping Irrelevant Features: Removed identifier columns like ID and constant variables like Z_CostContact and Z_Revenue as they hold no predictive power.
• Categorical Encoding: Applied One-Hot Encoding and Label Encoding to convert categorical text data (e.g., Education, Marital_Status) into numeric formats. Marital_Status anomalies (like 'YOLO' or 'Absurd') were merged into broader logical categories (e.g., 'Single').
• Feature Scaling: Standardized numerical variables using StandardScaler to bring all features to a similar scale, optimizing algorithm convergence.
• Class Imbalance Treatment: Applied techniques like SMOTE (Synthetic Minority Over-sampling Technique) to address class imbalances in the target Response variable, preventing the model from becoming biased toward the majority class.

Algorithms and Model Evaluation

Algorithms Used

Various classification algorithms were trained and tested to find the best performing model for predicting campaign response:

• Logistic Regression: Serving as a strong, interpretable baseline model.
• Decision Tree Classifier: To capture non-linear relationships and segment boundaries.
• Random Forest Classifier: An ensemble method used to reduce overfitting and improve predictive accuracy.
• Gradient Boosting Classifiers (XGBoost / LightGBM): Advanced ensemble techniques that sequentially correct errors from previous decision trees, highly effective on tabular data.
• Support Vector Machines (SVM): Used to find the optimal hyper-plane separating customers who accept offers from those who do not.

Evaluation Metrics

To robustly assess model performance beyond simple accuracy, the following metrics were evaluated:

• Accuracy: The overall percentage of correctly predicted responses.
• Precision: The proportion of predicted positive responses that were actually positive (minimizing false positives to save marketing costs).
• Recall (Sensitivity): The proportion of actual positive responses correctly identified by the model (minimizing false negatives to avoid missing out on potential buyers).
• F1-Score: The harmonic mean of Precision and Recall, providing a balanced metric—especially useful due to target class imbalances.
• ROC-AUC (Receiver Operating Characteristic - Area Under Curve): Measures the model's ability to distinguish between the two classes (responders vs. non-responders) across various threshold levels.
• Confusion Matrix: Utilized to visualize the breakdown of True Positives, True Negatives, False Positives, and False Negatives.

Tech Stack

Backend:

• Django (Python Web Framework)
• SQLite (Default Database)

Frontend:

• HTML / CSS / JavaScript
• Bootstrap 5 (via django-crispy-forms and crispy-bootstrap5)

Machine Learning & Data Processing:

• XGBoost
• Scikit-Learn
• Pandas
• Joblib

Screenshots

🏠 Homepage

📊 Analytics Dashboard

🔮 Single Prediction Wizard

📁 Bulk CSV Prediction

🧠 Model Analytics & Feature Importance

📈 Data Preparation Strategies

📋 Inference Logs / History

⚙️ Retrain Model

Installation & Setup

Follow these steps to set up and run the system locally:

1. Clone the repository:

   git clone <repository-url>
   cd Customer-Personality-Analysis

2. Create a virtual environment:

   python -m venv venv
   
   # On Windows:
   venv\Scripts\activate
   
   # On macOS/Linux:
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Apply database migrations:

   python manage.py migrate

5. Run the development server:

   python manage.py runserver

6. Access the application: Open your web browser and navigate to http://127.0.0.1:8000/

License

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