This project analysed a dataset containing Airbnb listings in the Northern Beaches council area of Sydney, with 28 featurs including the number of beds, baths, people, cleaning & deposite fees, reviews, GPS coordinates. I then used these features to train different machine learning models in Python to predict nightly rental price of Airbnb listings.

• Exploratory Data Analysis

• Performed statistical imputation to handle significant amount of missing values
• Annalysed the distribution of continuous features, and the correlation between features and rental price.
• Performed log-normal transformation and z-score standardization of the features to handle non-linear relationships and highly-skewed distributions

• Feature Engineering

• Mapped GPS coordinates into postcodes using geo-location API to generate a new feature representing district area,
• Calculated distance from each Airbnb listing to 10 popular tourist attractions.
• Trained a XGBoost model on these distances to predict the rental price and Westfield Shopping Center and Manly Beach have the highest feature importance.

• Model Selection

• Trained XGBoost, Extremely Randomised Trees, Lasso, OLS, Ridge models and a Generalised Additive Model using Ridge with natural cubic splines.
• Combined the best models using stacking to improve their predictive accuracy: Gradient boosted Ridge model achieved the best RMSE of 57.5.