AQI-based-property-value

This project investigates how Air Quality Index (AQI) impacts property values and health risks.
It introduces a methodology to compute property devaluation per square foot using AQI, disease burden, and relative population, and applies XGBoost regression to improve prediction accuracy.
Additionally, it provides a framework to calculate health risk scores on a 0–10 scale based on devaluation results.

---

Overview

• Analyzes property devaluation due to air quality.
• Uses AQI, health burden, and population frequency as key features.
• Employs both formula-based computation and XGBoost regression.
• Provides standardized health risk scores alongside devaluation values.
• Evaluates model accuracy using Mean Absolute Error (MAE).

---

Dataset processing

The model requires three main inputs:

1. Air Quality Index (AQI)

   • Mean AQI is computed for each location.

2. Relative Population

   • Estimated using frequency counts:

Relative Population = Total Frequency at Location / max(Total Frequency)

3. Disease Burden Score

• Combines disease frequency with average treatment costs:

Disease Burden = Frequency × log(1 + Avg Treatment Cost)

4. Aggregated features
   For each location, the dataset includes:

• Mean AQI
• Sum of Disease Burden
• Mean Relative Population

---

Devaluation calculation

Base formula

The property devaluation per square foot is estimated as:

Devaluation = (0.7 × AQI + 0.03 × Disease Burden) / (Relative Population + ϵ)

where ϵ = 10⁻⁵ is added to prevent division by zero.

Scaling factor

A scaling factor adjusts values for realism, depending on AQI levels and the Disease Burden Factor (DBF):

Scaling Factor =
0 × √DBF,         if 0 ≤ AQI ≤ 50
0.351 × √DBF,     if 51 ≤ AQI ≤ 100
0.712 × √DBF,     if 101 ≤ AQI ≤ 150
0.913 × √DBF,     if 151 ≤ AQI ≤ 200
1 × √DBF,         if AQI > 200

with

DBF = log(1 + Disease Burden) / log(1 + max(Disease Burden))

Final devaluation

The scaled devaluation is given by:

Scaled Devaluation = Predicted Devaluation × Scaling Factor

---

Predictive modeling with XGBoost

To refine predictions, an XGBoost regression model is trained.

• Features:

  • AQI
  • Disease Burden
  • Relative Population

• Target:

  • Devaluation per square foot

• Process:

  • Dataset split into training and testing sets (80/20).
  • Features standardized with StandardScaler.
  • Model trained with squared error loss.

• Evaluation metric:

MAE = (1/n) Σ |ŷᵢ – yᵢ|

Model performance

The model achieved:

Mean Absolute Error (MAE) = 7.02

This indicates strong accuracy in estimating property devaluation due to AQI.

---

Health risk calculation

Devaluation values are normalized to a 0–10 scale to indicate health risks:

Health Risk =
(Scaled Devaluation – min(Scaled Devaluation)) × 10
\---------------------------------------------------
max(Scaled Devaluation) – min(Scaled Devaluation)

This provides an interpretable risk score where 0 = lowest risk and 10 = highest risk.

---

Sample results

Location	AQI	Scaled Devaluation (per sqft)	Health Risk
Banashankari	210	293.43	9.80
Electronic City	243	299.49	10.00
HSR Layout	176	258.75	8.64
Indiranagar	182	254.77	8.51
Jayanagar	126	166.31	5.55

---

Installation

Clone the repository:

git clone https://github.com/P-Mahi10/AQI-based-property-value.git
cd AQI-based-property-value

Install dependencies (as listed in requirements.txt):

pip install -r requirements.txt

---

Usage

1. Prepare the data

   • Gather AQI, population, and medical cost datasets.
   • Compute relative population and disease burden scores.

2. Compute devaluation

   • Apply the base formula and scaling factor.

3. Train the XGBoost model

   • Use AQI, disease burden, and relative population as input features.

4. Evaluate results

   • Measure accuracy using MAE.

5. Calculate health risk

   • Normalize scaled devaluation values into 0–10 risk scores.

---

Results and insights

• AQI has a direct, quantifiable effect on property devaluation.
• Locations with higher AQI levels show significantly greater loss in value per square foot.
• Health burden magnifies this effect, as higher treatment costs increase overall devaluation.
• The XGBoost regression model achieves MAE = 7.02, confirming reliable predictions.
• Health risk scores provide a clear, interpretable scale for decision-making.

---

Conclusion

This methodology provides a systematic way to:

• Measure property devaluation caused by air pollution.
• Incorporate health and population data for more accurate results.
• Train and validate machine learning models to refine predictions.
• Quantify health risks in a standardized, actionable format.

The approach demonstrates how environmental data and health metrics can be combined to better understand real estate market impacts.