ML Based Customer Purchase Prediction Model

Project Overview

A machine learning project predicting customer purchase behavior using multiple classification algorithms.

Dataset Description

• Total Rows: 10,000
• Total Columns: 9

Columns

1. customer_id: Unique customer identifier
2. age: Customer's age
3. gender: Customer's gender
4. annual_income: Customer's annual income
5. last_visited_days_ago: Days since last website visit
6. session_duration: Time spent on website
7. pages_visited: Number of pages browsed
8. device: Device used for browsing
9. purchase: Target variable (Purchase/No Purchase)

Sample Data

Below is a sample dataset that provides customer details, session activity, and purchase behavior:

customer_id	age	gender	annual_income	last_visited_days_ago	session_duration	pages_visited	device	purchase
1	56	male		7	17	15	desktop	0
2	69	female	47617	4	35	19	mobile	0
3	46	male	94258	30		15	mobile	0
4	32	female	70075	19	4	12	mobile	0
5	60	male	146998	16	51		mobile	0
6	25	male	42631	8	31	16	desktop	1
7	38	female	143120		31	6	desktop	1
8	56	male	117158	24	9	20	tablet	0
9	36	female	158955	12	31	15	desktop	0

Model Performance

• Logistic Regression: 48% accuracy
• Random Forest: 70% accuracy
• Gradient Boosting: 70% accuracy

Final Model

• Selected Model: Random Forest
• Best Accuracy: 70%

Requirements

• Python 3.8+
• pandas
• numpy
• scikit-learn
• matplotlib
• seaborn

Installation

1. Clone the repository

git clone https://github.com/PRANAYBHUMAGOUNI/ML-Based-Customer-Purchase-Prediction-Model/

2. Create virtual environment

python -m venv venv
venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

Example Prediction

new_customer = pd.DataFrame({
    'age': [69],
    'gender': ['female'],
    'annual_income': [47617],
    'last_visited_days_ago': [4],
    'session_duration': [35],
    'pages_visited': [19],
    'device': ['mobile']
})

# Prediction
prediction = model.predict(new_customer)
probability = model.predict_proba(new_customer)

Prediction Result

• Prediction: Will not purchase
• Purchase Probability: 21.48%

Feature Engineering

1. days_since_visit_ratio: Ratio of days since last visit to session duration
2. pages_per_minute: Pages visited per minute
3. income_age_ratio: Annual income divided by age
4. engagement_score: Combination of pages visited and session duration

Preprocessing Steps

• Handled missing values
• Encoded categorical variables
• Created engineered features
• Scaled numerical features

Hyperparameter Tuning

• Used GridSearchCV for Random Forest and Gradient Boosting
• Explored parameters:
  • Number of estimators
  • Max depth
  • Min samples split
  • Learning rate

Future Improvements

• Collect more diverse data
• Experiment with advanced ensemble methods
• Incorporate more complex feature engineering
• Try deep learning approaches