NYC House Types Classifier

A scikit-learn classification project that predicts the type of accommodation listed in New York City Airbnb data.

The model classifies each listing into one of these room_type categories:

Entire home/apt
Private room
Shared room

This project is intended as an educational baseline for exploring tabular data, feature preprocessing, model comparison, and hyperparameter tuning.

Model Details

Task: Multiclass tabular classification
Target variable: room_type
Framework: scikit-learn
Saved format: .pkl via joblib
Saved model: House_Types_Classifier.pkl

The workflow compares Logistic Regression, Decision Tree, Random Forest, and Gradient Boosting classifiers. Random Forest is also evaluated with Grid Search and Randomized Search hyperparameter tuning.

Dataset

The dataset contains NYC Airbnb listing records with information about listing location, price, minimum stay, reviews, host listings, and availability.

Raw columns

id
name
host_id
host_name
neighbourhood_group
neighbourhood
latitude
longitude
room_type
price
minimum_nights
number_of_reviews
last_review
reviews_per_month
calculated_host_listings_count
availability_365

The target column is room_type.

Preprocessing

The notebook applies the following preparation steps:

Removes identifier, name, host, and date fields that are not used as model features.
Fills missing reviews_per_month values with zero.
Caps price and minimum_nights at their 99th percentiles to reduce the effect of extreme values.
Uses median imputation and standard scaling for numeric features.
Uses most-frequent imputation and one-hot encoding for categorical features.
Splits the data into training and test sets using an 80/20 stratified split.

The preprocessing steps are included in the saved scikit-learn pipeline.

Training and Evaluation

Models are compared using cross-validated accuracy and macro F1 score on the training data, followed by accuracy and macro F1 evaluation on the held-out test set.

The notebook also evaluates tuned Random Forest pipelines using:

Grid Search CV
Randomized Search CV

Run the notebook to reproduce the model comparison and generate the evaluation results for the current dataset.

How to Use

import joblib
import pandas as pd

model = joblib.load("House_Types_Classifier.pkl")

sample = pd.DataFrame([{
    "neighbourhood_group": "Manhattan",
    "latitude": 40.7536,
    "longitude": -73.9838,
    "price": 225,
    "minimum_nights": 1,
    "number_of_reviews": 45,
    "reviews_per_month": 0.38,
    "calculated_host_listings_count": 2,
    "availability_365": 355
}])

prediction = model.predict(sample)
print(f"Predicted room type: {prediction[0]}")

The model expects the feature columns used by the preprocessing pipeline. The pipeline handles imputation, scaling, and categorical encoding during prediction.

Limitations

The model reflects patterns in this dataset and may not generalize to other cities or time periods.
Listing prices, availability, and review behavior can change over time.
The dataset may contain sampling, reporting, and geographic bias.
Predictions should be treated as estimates for learning and experimentation, not as authoritative property classifications.

Files

notebook.ipynb - Full workflow covering data loading, EDA, cleaning, preprocessing, training, tuning, and evaluation.
data.csv - NYC Airbnb listing dataset.
House_Types_Classifier.pkl - Saved scikit-learn classification pipeline.
pyproject.toml - Project metadata and Python dependencies.

License

Released under the MIT License. Use freely for learning and experimentation.

Downloads last month: -; Downloads are not tracked for this model. How to track