Selling Laptops: Smart Marketing

Table of Contents

1. Overview
   • Key Features
2. Learning Objectives
3. Setup and Installation
   • Prerequisites
   • Setup Instructions
4. Project Components
   • Dataset Overview
   • The UserPredictor Class
   • Performance Metrics
5. Accomplishments
6. Hints and Suggestions
7. Future Enhancements
8. Acknowledgments
9. License

---

Overview

This project focuses on using machine learning to predict whether users will click on a promotional email for laptops based on historical user data and browsing logs. The goal is to target marketing efforts effectively while minimizing unnecessary emails.

Key Features

• High Prediction Accuracy: Achieved 75%+ accuracy in predicting email clicks.
• Efficient Data Processing: Reduced data processing time by 30% through optimized feature engineering.
• Robust Classification: Developed a reliable classifier using Python libraries like scikit-learn, pandas, and NumPy.
• Comprehensive Evaluation: Used cross-validation and confusion matrices for better model interpretability and validation.

---

Learning Objectives

This project demonstrates:

• The integration of purchase histories and browsing logs to build predictive models.
• Advanced feature engineering techniques to improve data processing and model performance.
• Evaluation of machine learning models with metrics like accuracy, cross-validation scores, and confusion matrices.

---

Setup and Installation

Prerequisites

• Python 3.x installed
• Required libraries: pandas, numpy, scikit-learn

Setup Instructions

1. Clone the repository:

  git clone https://github.com/SrujayReddy/Selling-Laptops.git
  cd Selling-Laptops

2. Install dependencies:

   pip install pandas numpy scikit-learn
   

3. Ensure datasets (train_users.csv, train_logs.csv, train_y.csv) are available in the data/ directory.

---

Project Components

Dataset Overview

The project uses three datasets for training (train) and testing (test1, test2):

1. Users Dataset (*_users.csv): Contains demographic and account-related information.
2. Logs Dataset (*_logs.csv): Records user browsing history, including pages visited and time spent.
3. Target Dataset (*_y.csv): Indicates whether users clicked on a promotional email (1 for yes, 0 for no).

The UserPredictor Class

The classifier is implemented in main.py as the UserPredictor class with two key methods:

1. fit(train_users, train_logs, train_y):
   • Combines user and log data into a unified feature set.
   • Trains a scikit-learn pipeline, leveraging LogisticRegression or other classifiers.
2. predict(test_users, test_logs):
   • Predicts email click outcomes for the test dataset.
   • Returns predictions as a numpy array of Booleans.

Performance Metrics

• Accuracy: Primary metric for evaluation.
• Cross-Validation: Used to assess model robustness with metrics like mean and standard deviation.
• Confusion Matrix: Provides insights into false positives, false negatives, and overall prediction quality.

---

Accomplishments

• Achieved 75%+ Accuracy: Developed a robust classifier that consistently performs above the threshold for full credit.
• Optimized Data Processing: Engineered features that reduced data processing time by 30%.
• Enhanced Interpretability: Evaluated models using cross-validation and confusion matrices for better insights.

---

Hints and Suggestions

1. Start Simple: Begin with features from the *_users.csv dataset for a one-to-one mapping with predictions.
2. Feature Engineering: Create log-based features (e.g., total time spent, unique pages visited) to enhance model performance.
3. Cross-Validation: Use cross_val_score to evaluate model stability across different data splits.
4. Model Pipelines:
   • Combine StandardScaler with LogisticRegression for efficient processing and classification.
5. Handle Missing Data: Address cases where users lack log entries by imputing or creating default values.

---

Future Enhancements

• Explore advanced models like Random Forests or Gradient Boosting for higher accuracy.
• Automate hyperparameter tuning with tools like GridSearchCV or Optuna.
• Visualize feature importance to better understand model decisions.

---

Acknowledgments

This project was developed as part of the CS 320 course at the University of Wisconsin–Madison. Special thanks to the teaching staff for guidance and support.

---

License

This project was developed as part of the CS 320 course. It is shared strictly for educational and learning purposes only.

Important Notes:

• Redistribution or reuse of this code for academic submissions is prohibited and may violate academic integrity policies.
• The project is licensed under the MIT License. Any usage outside academic purposes must include proper attribution.