🏠 House Prices — End-to-End Machine Learning Project

Author: Lorena Galvan – Data Scientist

Problem Type: Supervised Regression
Objective: Predict residential property sale prices using a house prices dataset from kaggle (https://www.kaggle.com/datasets/lespin/house-prices-dataset/data)

🚀 Project Overview

This project implements a full end-to-end data science workflow, from raw data exploration to model evaluation and optimization. The goal is to build a robust regression model capable of accurately predicting house prices while demonstrating strong skills in:

• Advanced feature engineering

• Handling missing data systematically

• Statistical reasoning

• Models comparison and validation

• Regularization and generalization control

📈 Technical Overview

Exploratory Data Analysis

The target variable (SalePrice) was right-skewed → log transformation improved normality. Strong correlations found with:

• OverallQual

• GrLivArea

• GarageCars

• Outliers in living area were detected and treated to stabilize variance

• Features, Target Distribution

• Correlation matrix of main features

• Selection of the categorical variable to analyze

Model Performance

🔹 Polynomial Regression 2nd grade

Results:

• R²: 0.91
• RMSE: $23,269
• The model captured the main trend but showed sensitivity to outliers and non-linear effects.

🔹 Logistic Regression of the variable: Basement Full Bathroom

Results:

• R²: 0.88
• MAE: 0.12 (Variable takes the values: 0, 1, 2)
• The model provided interpretable decision boundaries but was limited in capturing complex relationships.
• Confusion Matrix
  

🔹 Decision Tree

Results:

• Train and Test Accuracy: 0.66, 0.62
• Train and Test F1 weighted: 0.62, 0.56
• the model achieved moderate predictive performance, with a small gap between training and testing accuracy, indicating mild overfitting.
• Feature Importance

• Snapshot of Decision Tree

🧠 Key Technical Highlights

• Exploratory Data Analysis (EDA)

• Distribution of features analysis

• Correlation analysis and multicollinearity analysis and correction

• Outliers detection and treatment

Feature Engineering

• Encoding of categorical variables

• Missing value imputation pipelines

• Log transformations for variance stabilization

• Feature scaling and normalization

⚙️ Modeling Strategy

• Regression models: Linear, Polynomial regression. Logistic Model. Decision Tree, Random Forest Classifier, Clustering Model.

• Regularized models (Ridge / Lasso / ElasticNet)

• Cross-validation

• Hyperparameter tuning

📈 Evaluation

• Evaluation metrics as RMSE, MAE, SD, Accuracy, F1 score

🛠 Tech Stack

• Python

• pandas / numpy

• matplotlib / seaborn

• scikit-learn

• Pipeline & Cross-Validation tools

• git, github

📂 Repository Structure

house-prices-regression/
│
├── README.md
├── data
│   ├── raw
├── notebooks
   └── house_prices_regression.ipynb
├── images

The notebook contains the complete workflow: data cleaning → feature engineering → modeling → evaluation.

🔎 How to Run

Clone the repository

Install dependencies

Open the notebook and run sequentially