Logistic Regression Model – Airbnb Superhost Classification

This project uses logistic regression to predict whether an Airbnb host is a superhost using preprocessed listing data. The process includes model training, evaluation, feature selection, and persistence.

Dataset

• data_LR/airbnbData_train.csv: Preprocessed dataset with one-hot encoding, scaling, and imputed values.

Techniques Used:

• Logistic Regression (scikit-learn)
• GridSearchCV for hyperparameter tuning
• Confusion matrix, ROC curve, AUC, and precision-recall curve evaluation
• Feature selection using SelectKBest
• Model persistence using pickle

Model Goals

• Classify hosts as superhosts (True or False)
• Evaluate default vs tuned logistic regression models
• Analyze feature importance and compare AUC scores

Files used:

• ModelSelectionForLogisticRegression.ipynb: The full notebook
• model_best.pkl: Pickled best model
• data_LR/airbnbData_train.csv: Cleaned Airbnb dataset

Libraries

• Python 3
• pandas, numpy
• scikit-learn
• matplotlib, seaborn

---

Developed as part of a machine learning lab exercise!