Case Overview
Predicting Weekly Rental Prices in the Australian Housing Market
Dataset Source: Kaggle - Australian Rental Market Data 2026
Business Context¶
Australia’s rental market is characterized by:
High interstate migration
Urban-regional demand shifts
Affordability pressures in major metro areas
Significant variation across states and territories
A national property analytics firm wants to develop a data-driven rental pricing engine to:
Improve pricing recommendations for landlords
Identify high-demand suburbs
Support investment decision-making
Benchmark agency competitiveness
They have obtained a cleaned dataset of 6,700+ rental listings across all eight Australian states and territories, including geospatial coordinates and detailed property descriptions.
Business Problem¶
Objective¶
Develop a predictive model that estimates the weekly rental price (price_display) for residential properties using structured, geospatial, and textual data.
Target Variable¶
price_display(weekly rent in AUD)
Core Business Question¶
Given a property’s features, location, and description, what is its expected weekly rental price?
Dataset Overview¶
The dataset includes:
6,700+ unique, cleaned listings
Coverage across NSW, VIC, QLD, WA, SA, ACT, TAS, NT
Precise latitude and longitude
Sanitized text descriptions
Structured property attributes
Data Dictionary¶
| Column | Data Type | Description | Example |
|---|---|---|---|
| title | Text (String) | Marketing headline of the property listing. | “Modern 2BR Apartment with City Views” |
| price_display | Numeric (Float/Integer) | Weekly rental price in Australian Dollars (AUD), cleaned for numeric analysis. | 650 |
| description | Text (String) | Detailed property description with PII removed. | “Spacious home close to schools and transport...” |
| propertyType | Categorical | Type of dwelling (House, Apartment, Townhouse, Studio, etc.). | Apartment |
| locality | Categorical | Broader administrative region associated with the property. | Greater Sydney |
| latitude | Numeric (Float) | Geographic Y-coordinate for spatial modeling. | -33.8688 |
| longitude | Numeric (Float) | Geographic X-coordinate for spatial modeling. | 151.2093 |
| postcode | Categorical (String/Integer) | Four-digit Australian postal code. | 2000 |
| state | Categorical | State or territory abbreviation (NSW, VIC, QLD, WA, SA, ACT, TAS, NT). | NSW |
| street_address | Text (String) | Street-level property address. | 15 George St |
| suburb | Categorical | Residential suburb name. | Parramatta |
| bathrooms | Numeric (Integer) | Number of bathrooms in the property. | 2 |
| bedrooms | Numeric (Integer) | Number of bedrooms in the property. | 3 |
| parking_spaces | Numeric (Integer) | Number of dedicated parking or garage spaces. | 1 |
| agency_name | Categorical | Real estate agency managing the listing. | Ray White |
| amenities | Text / Semi-Structured | Consolidated list of key features (e.g., Air Conditioning, Balcony). | “Air Conditioning, Balcony, Dishwasher” |
Key Feature Categories¶
🏠 Property Characteristics¶
bedroomsbathroomsparking_spacespropertyTypeamenities
📍 Geographic Variables¶
latitudelongitudepostcodesuburblocalitystate
🏢 Market & Agency Information¶
agency_name
📝 Textual Data¶
titledescriptionamenities
This multi-modal structure allows for:
Regression modeling
Geospatial modeling
NLP feature extraction
Market segmentation analysis
Modeling Approach¶
Gradient Boosted Trees (XGBoost)
Why XGBoost is ideal here:
Handles nonlinear relationships
Automatically captures feature interactions
Works well with structured/tabular data
Robust to outliers
Handles missing values
Strong performance benchmark for Kaggle-style datasets
For business communication:
MAE: “Average pricing error”
RMSE: Penalizes large mispricing
R²: % of rent variation explained