This repository contains my end-to-end Machine Learning Fundamentals project, completed as part of my ML lecture. The project covers both Regression and Classification problems using real-world datasets from Kaggle and follows the complete machine learning lifecycle: understanding the problem, exploratory data analysis (EDA), data preparation, model building, evaluation, and interpretation.
The objective of this project was not only to build machine learning models, but also to understand the logic behind each step, interpret model behavior, and connect technical implementation with real-world decision-making.
The project is divided into two main parts:
In this section, the California Housing dataset was used to predict median house prices based on features such as:
- Median income
- Number of rooms
- Number of bedrooms
- Population
- Households
- Latitude and longitude
- Ocean proximity
- Housing median age
- Simple Linear Regression
- Multiple Linear Regression
- Ridge Regression
- Lasso Regression
-
Dataset loading and understanding
-
Exploratory Data Analysis (EDA)
-
Missing value handling using median imputation
-
One-hot encoding for categorical features
-
Train-test split (80:20)
-
Feature scaling using StandardScaler
-
Model training and prediction
-
Performance evaluation using:
- R² Score
- RMSE (Root Mean Squared Error)
-
Regularization using Ridge and Lasso
-
Alpha tuning for model optimization
-
Final comparison of all regression models
Among all regression models, Ridge Regression performed slightly better on unseen test data with the best balance between prediction accuracy and generalization.
In this section, the Heart Disease dataset was used to predict whether a patient has heart disease (1) or not (0) using medical attributes such as:
- Age
- Sex
- Chest pain type
- Blood pressure
- Cholesterol
- Maximum heart rate achieved
- Exercise-induced angina
- Oldpeak
- Number of major vessels
- Thalassemia and others
- Logistic Regression
-
Dataset loading and understanding
-
Exploratory Data Analysis (EDA)
-
Class balance analysis
-
Correlation heatmap and feature relationship analysis
-
Train-test split with stratification
-
Feature scaling using StandardScaler
-
Logistic Regression model training
-
Sigmoid function understanding and interpretation
-
Model evaluation using:
- Confusion Matrix
- Accuracy
- Precision
- Recall
- F1 Score
- Classification Report
The Logistic Regression model achieved strong recall for heart disease detection, which is especially important in healthcare since minimizing false negatives is critical.
- Python
- Jupyter Notebook
- VS Code
- Pandas
- NumPy
- Matplotlib
- Seaborn
- Scikit-learn
This project helped strengthen my understanding of:
- Supervised Learning
- Regression vs Classification
- Model evaluation and interpretation
- Overfitting and regularization
- Feature importance
- Bias toward Precision vs Recall in healthcare use cases
- How machine learning decisions connect with business and product thinking
As someone interested in AI Product Management, this project also helped me understand how technical model decisions affect real-world product outcomes, stakeholder decisions, and user impact.
- Housing Regression Notebook
- Heart Disease Classification Notebook
- Dataset files
- Project outputs and visualizations
- Final comparison tables
- This README documentation
This project was built with a strong focus on understanding each line of code rather than only completing the assignment. Every model choice, metric, and interpretation was written with the goal of connecting machine learning concepts to practical applications and product decision-making.