Machine Learning: Hyperparameter Optimization
This project focuses on evaluating how the variation of hyperparameters affects a machine-learning model designed to predict whether a car will be sold based on variables such as mileage, model year, and price. The project uses a for loop to test models from scikit-learn with varying hyperparameters. Using pandas and seaborn, the results are analyzed to find the hyperparameter combination that offers the highest accuracy. Parameters such as max_depth, min_samples_leaf, and min_samples_split of the DecisionTreeClassifier model are varied and refined through several tests to achieve satisfactory parameters, aiming to reduce the risk of overfitting.
The project focuses on the following:
- Understand hyperparameters and parameter spaces.
- Explore parameter spaces deterministically.
- Optimize machine learning models.
- Avoid overfitting.
- Implement behind-the-scenes exploration.
- Use GridSearchCV for hyperparameter exploration.
- Apply nested cross-validation.