Links

• Dataset

Write-up

Introduction

• Describe the dataset.
• Explain the problem you aim to solve.

Unsupervised Analysis

• Explore patterns or structure using clustering and dimensionality reduction (e.g., PCA).
• Visualize training data:
  • Plot individual feature distributions (e.g., histograms, density plots).
  • Plot relationships between features and the target variable.
  • Create a correlation matrix.
• Discuss any interesting structures or explain attempts to find them.

Supervised Analysis

• Train >=3 different models: (1) Logistic Regression, (2) SVM, (3) Neural Networks
• Use the following for implementation:
  • Custom implementation or existing libraries (e.g., Keras, scikit-learn, TensorFlow).
• Experiment with different feature transformations (at least three, e.g., polynomial, PCA, radial-basis function kernel).
• Apply different regularization techniques (at least six values per model).
• Document all transformations and regularization results.

Table of Results

• Include training accuracy and validation metrics for every model.
• Provide results for varying parameter settings:
  • Classification metrics (e.g., precision, recall).
  • Regression metrics (e.g., MSE, R²).
• Plot and analyze performance metrics (e.g., accuracy, precision, recall, MSE) with different transformations and hyperparameters.

Analytical Discussion

• Analyze and explain experimental results.
• Include a chart of key findings.
• Discuss the impact of:
  • Feature transformations.
  • Regularization techniques.
  • Other hyperparameters on model performance.
• Interpret:
  • Overfitting and underfitting observations.
  • Bias-variance trade-offs.
  • Parameter choices improving generalization.

Submission Requirements

• Upload the following to Gradescope:
  • Presentation slides.
  • Project write-up (PDF format).
  • Project code as a Jupyter Notebook (or GitHub link, if necessary).
  • Custom dataset (if used, either upload or provide a GitHub link).