Brisbane Rental Market Analytics

Overview

A complete data-analytics product for Brisbane's rental market: a real ETL pipeline, statistical forecasting, and an eleven-tab interactive dashboard. Given a suburb, it answers the three questions renters and investors actually care about — how rent has moved over the last ~7 years versus the city, where it is likely headed over the next year, and whether the suburb is currently a premium or a bargain. It's structured to match the Queensland RTA's published rental-bond schema, so it runs on the real quarterly export in production.

What we did

• Built a real ETL pipeline — ingest raw quarterly data, standardise suburb names, drop duplicates and missing values, remove statistical outliers, and load a tidy table into SQLite via SQLAlchemy.
• OLS trend forecasting with a 95% prediction interval and R²/residual diagnostics — an honest forecast, not a straight-line guess (R² ≈ 0.94 on New Farm houses).
• KMeans suburb segmentation into Affordable / Mid-tier / Premium tiers from standardised rent-level and growth features.
• Model backtesting — OLS trend vs. Holt's exponential smoothing, trained with the last N quarters hidden and scored on MAE/RMSE/MAPE (proper train/test validation).
• Additive trend/seasonal/residual decomposition, a suburb-to-suburb correlation matrix, and Isolation Forest anomaly detection.
• An 11-tab Streamlit dashboard (overview, compare, statistics, segments, map, forecast explorer, backtesting, decomposition, correlation, anomalies, data quality) with CSV exports.

Tech stack