WA Property Map Guide

Loaded rows: 14,927
Displayed map points: 3,000
Static Folium HTML uses sampled points to keep S3 hosting light.

Price View

This layer shows actual sold prices.

< $400k
$400k - $500k
$500k - $600k
$600k - $700k
$700k - $800k
$800k - $900k
$900k - $1M
$1M+

KMeans House Classification View

KMeans separates properties into a fixed number of groups. Here it uses location, price, rooms, land size, and floor area.

Type 0 — affordable or lower-profile homes
Type 1 — typical family housing
Type 2 — stronger / premium profile
Type 3 — location-sensitive high price profile
Type 4 — large land / value profile
Type 5 — compact or unusual profile

DBSCAN Density View

DBSCAN identifies natural clusters based on density without requiring a predefined number of groups. It is useful for detecting real market pockets, local patterns, and unusual outlier properties.

DBSCAN summary
Detected clusters: 29
Noise / outlier points: 2504
Grey — Noise / Outlier
Properties that do not belong to any dense group. These may be unique homes, unusual listings, or weakly connected market points.
D0 — Main Dense Local Group
The strongest and most representative cluster. Properties here share similar price, size, and location characteristics.
D1 — Secondary Dense Group
Another major cluster with a different property profile compared to D0.
D2 — Distinct Market Pocket
A clearly separated group, often representing a different suburb, pricing tier, or housing style.
D3 — Smaller Dense Segment
A smaller but meaningful cluster. Properties are similar to each other, but less common across the dataset.

If almost everything appears grey, eps is too small (increase it). If everything collapses into one group, eps is too large.

Isolation Forest Anomaly View

Isolation Forest detects local price anomalies inside a fixed nearby area. It compares each property with homes within the selected radius using price, price per sqm, suburb median gap, and nearby median gap. It is useful for spotting rare listings, unusual deals, or properties that do not match the normal market pattern.

Isolation Forest summary
Detected anomalies: 446
Normal properties: 14481
Contamination setting: 3.0%
Green — Cheap local price anomaly
Red — Expensive local price anomaly
Purple — Mixed local price anomaly
Green means cheaper than nearby/local median patterns. Red means more expensive than nearby/local median patterns. Purple means mixed signals, such as total price and price per sqm disagreeing.

Lower anomaly score means the property is more unusual inside its local price area. This is easier to interpret than global anomaly detection, but still needs manual checking.

PCA Similarity View

PCA compresses price, size, room count, land/building size, and location into two main components.

PCA summary
PC1 explains 32.4% and PC2 explains 15.2% of the scaled feature variance.
Deep Blue — very low PCA score
Blue — low PCA score
Yellow — average PCA score
Orange — high PCA score
Red — very high PCA score

Random Forest Valuation

Random Forest predicts estimated price and highlights possible undervalued candidates.

Model metrics
MAE: $72,499
R²: 0.860

Top Features

  1. cbd_dist: 0.374
  2. floor_area: 0.295
  3. longitude: 0.191
  4. latitude: 0.080
  5. nearest_stn_dist: 0.023
  6. land_area: 0.015
  7. nearest_sch_dist: 0.013
  8. bathrooms: 0.005
Green — actual price is more than 5% below predicted value
Blue — close to predicted value (±5%)
Red — actual price is more than 5% above predicted value

SHAP Model Explanation View

SHAP explains why the Random Forest predicted a certain property price. Positive values push the prediction higher. Negative values push it lower.

SHAP status: Available
The marker color is based on the largest single SHAP factor for that property.

Top SHAP Drivers

  1. cbd_dist: average impact $141,036
  2. floor_area: average impact $99,214
  3. longitude: average impact $96,674
  4. latitude: average impact $31,630
  5. nearest_stn_dist: average impact $6,516
  6. land_area: average impact $6,398
  7. bathrooms: average impact $5,567
  8. nearest_sch_dist: average impact $3,918
Strong positive — feature pushed predicted price up strongly
Positive — feature pushed predicted price up
Small effect — no single dominant factor
Negative — feature pushed predicted price down
Strong negative — feature pushed predicted price down strongly

Open a marker popup to see the top positive and negative factors, such as floor_area +$120k or cbd_dist -$80k.

Local Price Gap Zones

Nearby homes with large price differences. Useful for spotting local value boundaries.

Red line — nearby pair with a large price gap

Current rule: within 500m and at least $250k difference.