Intermediate Projects

More challenging projects that combine multiple ML concepts and techniques. Each project includes a starter.py scaffold, run it, then replace TODO sections with your implementation.

Starter usage: cd project-XX-name → download dataset per README → python starter.py

Prerequisites

Before starting these projects, you should have completed:

• All Beginner Projects
• Module 05: Model Evaluation & Optimization
• Module 06: Ensemble Methods
• Module 07: Feature Engineering
• Module 08: Unsupervised Learning (for some projects)

Bonus project ideas (not in the official 23)

These three builds extend the intermediate map. They are optional extras, not counted in the canonical 23-project list in the main README.

Deep learning curriculum map (projects)

Projects

Project 1: Handwritten Digit Recognition (MNIST)

Difficulty: Intermediate
Time: 3-5 days
Skills: Neural Networks, Image Processing, Deep Learning Basics

Build a neural network to classify handwritten digits (0-9).

What you'll learn:

• Building neural networks
• Image preprocessing
• Hyperparameter tuning
• Deep learning workflow

Dataset: MNIST (built into Keras/TensorFlow)

Extensions:

• Try different architectures
• Use data augmentation
• Achieve >99% accuracy

Project 2: Customer Churn Prediction

Difficulty: Intermediate
Time: 4-5 days
Skills: Classification, Feature Engineering, Imbalanced Data, Business Metrics

Predict which customers will leave a service (churn prediction).

What you'll learn:

• Handling imbalanced datasets
• Business-focused metrics (cost-sensitive learning)
• Feature engineering from customer data
• Ensemble methods
• Model interpretation

Dataset: Telco Customer Churn or similar

Extensions:

• Calculate cost of churn
• Recommend retention strategies
• Build a dashboard

Project 3: Movie Recommendation System

Difficulty: Intermediate
Time: 5-7 days
Skills: Collaborative Filtering, Content-Based Filtering, Matrix Factorization

Build a recommendation system to suggest movies to users.

What you'll learn:

• Collaborative filtering
• Content-based filtering
• Matrix factorization
• Evaluation metrics for recommendations
• Hybrid approaches

Dataset: MovieLens

Extensions:

• Hybrid recommendation system
• Real-time recommendations
• Cold-start problem solutions

Project 4: Credit Card Fraud Detection

Difficulty: Intermediate
Time: 4-5 days
Skills: Anomaly Detection, Imbalanced Data, Classification, Feature Engineering

Detect fraudulent credit card transactions.

What you'll learn:

• Handling highly imbalanced data
• Anomaly detection techniques
• Precision/Recall tradeoffs
• Cost-sensitive learning
• Feature engineering for fraud detection

Dataset: Credit Card Fraud Detection

Extensions:

• Real-time fraud detection
• Explainable AI for fraud cases
• Cost-benefit analysis

Project 5: Customer Segmentation

Difficulty: Intermediate
Time: 3-4 days
Skills: Unsupervised Learning, Clustering, Dimensionality Reduction

Segment customers into groups based on behavior and demographics.

What you'll learn:

• K-means clustering
• Hierarchical clustering
• Choosing number of clusters
• Interpreting clusters
• Business applications

Dataset: Mall Customer Segmentation or e-commerce data

Extensions:

• Multiple clustering algorithms
• Cluster visualization
• Actionable insights from segments

Project 6: Time Series Forecasting

Difficulty: Intermediate
Time: 4-5 days
Skills: Time Series Analysis, ARIMA, LSTM, Feature Engineering

Forecast future values in a time series (sales, stock prices, etc.).

What you'll learn:

• Time series preprocessing
• ARIMA models
• LSTM for sequences
• Feature engineering for time series
• Evaluation for time series

Dataset: Airline Passengers or stock data

Extensions:

• Multiple forecasting methods
• Confidence intervals
• Real-time forecasting

Project 7: Feature Engineering Mastery

Difficulty: Intermediate
Time: 4-5 days
Skills: Feature Engineering, Feature Selection, sklearn Pipeline, Advanced Encoding

Practice feature engineering on a complex real-world dataset.

What you'll learn:

• Apply a full feature engineering workflow
• Use sklearn Pipeline and ColumnTransformer effectively
• Handle mixed data types (numeric, categorical, text)
• Implement advanced encoding techniques (WOE, Target Encoding)
• Perform feature selection using multiple methods
• Build robust preprocessing pipelines

Dataset: House Prices, Credit Card Default, or Employee Attrition datasets

Extensions:

• Automated feature engineering
• Feature store implementation
• Feature monitoring

Project 8: Ensemble Methods Comparison

Difficulty: Intermediate
Time: 4-5 days
Skills: Ensemble Methods, Bagging, Boosting, Stacking, Hyperparameter Tuning

Compare ensemble methods: Bagging, Boosting, Stacking, and Voting.

What you'll learn:

• Understand different ensemble methods and when to use them
• Implement and compare Bagging vs Boosting
• Use advanced boosting algorithms (XGBoost, LightGBM, CatBoost)
• Build Stacking and Voting ensembles
• Tune hyperparameters for ensemble methods
• Compare model performance across the ensembles

Dataset: Titanic, Credit Card Fraud, Customer Churn, or House Prices

Extensions:

• Custom ensemble creation
• Ensemble of ensembles
• Feature importance comparison

Project Structure

Each project should include:

project-name/
 README.md              # Detailed project description
 data/                  # Dataset
 notebooks/             # Analysis notebooks
    01-data-exploration.ipynb
    02-feature-engineering.ipynb
    03-model-development.ipynb
    04-model-evaluation.ipynb
    05-results-analysis.ipynb
 src/                   # Source code modules
    data_preprocessing.py
    feature_engineering.py
    models.py
    evaluation.py
 models/                # Saved models
 results/                # Results, plots, reports
 tests/                 # Unit tests
 requirements.txt       # Dependencies
 config.yaml            # Configuration file

Tips for Success

1. Plan First: Outline your approach before coding
2. Iterate: Build baseline, then improve incrementally
3. Document: Keep detailed notes on decisions and results
4. Visualize: Create clear comparison charts
5. Compare: Try multiple approaches and compare
6. Present: Create a clear presentation of results

Learning Outcomes

After completing these projects, you should be able to:

• Handle complex, real-world datasets
• Apply advanced feature engineering
• Use ensemble methods effectively
• Handle imbalanced data
• Build end-to-end ML pipelines
• Present results to stakeholders

Additional Resources

• Kaggle Competitions: Practice with real competitions
• Papers with Code: See current published approaches and leaderboards
• Towards Data Science: Learn from others' projects

Ready for a challenge? Start with MNIST if you want to learn deep learning, or Customer Churn for a business-focused project.

Next Level: 18-projects-advanced