Data_projects_TripleTen

Model Comparison (Classification)

Goal:
Compare the performance of multiple classification models to predict user churn and identify the most effective algorithm.

Process:

• Prepared the dataset and split into train/test sets
• Trained logistic regression, decision tree, and random forest models
• Evaluated results using metrics: accuracy, precision, recall, F1, and ROC-AUC
• Interpreted feature importance to explain model behavior

Result:
The Random Forest model achieved the highest overall performance, offering strong predictive accuracy and balanced precision-recall trade-offs.

Skills Used: scikit-learn, pandas, machine learning, model evaluation, visualization

Files:

• model_comparison.ipynb

View Project: Notebook

Future Improvements

• Apply hyperparameter tuning (GridSearchCV or RandomizedSearchCV) to further optimize model accuracy and reduce overfitting.
• Add feature selection or dimensionality reduction (e.g., PCA or SelectKBest) to evaluate which features contribute most to model performance.
• Deploy the trained model via a simple API endpoint or Streamlit app to make predictions interactively.
• Incorporate cross-validation and ROC curve comparison for a more robust evaluation across multiple models.