House Valuation AI (v0.1.0-pre.alpha.1)

Note

This is a machine learning-powered real estate valuation tool built in Rust. It uses Elastic Net Regression to predict property prices based on historical data from real estate listings. As of v0.1.0-pre.alpha.1, Spain + Idealista API is the only country available for scraping.

This AI model processes location-based property features (size, floor, amenities, etc.) to estimate real estate values. It includes scraping, training, and prediction capabilities.

---

Table of Contents

• Features
• Installation
• Usage
  • Scrape Data
  • Train the Model
  • Predict Property Prices
• Technical Overview
• Dependencies
• License

---

Features

Warning

This model is experimental. It is not a financial advisory tool and should not be used for critical investment decisions.

• Property Price Estimation - Uses Elastic Net Regression to predict real estate prices.
• Automated Data Scraping - Fetches property data from Idealista.
• Machine Learning Training - Builds a model using historical property listings.
• Feature Extraction - Uses location, size, rooms, bathrooms, and more as predictive factors.
• Evaluation Metrics - Computes R² score to assess model performance.
• Prebuilt Database - The package comes with 2500+ homes in spain with 13 data fields per house.

---

In action

What do you want to do?: Predict Property Price
Enter the size (m²) of the property: 196
Enter the floor number (leave empty if not applicable): 6
Enter the latitude: 32
Enter the longitude: 0.43
Does the property have a lift? yes
Enter price per m² (if not known, leave empty): 3500
Number of bedrooms: 4
Number of bathrooms: 3
Does the property have a swimming pool? yes
Does the property have a garden? yes
Does the property have a garage? yes
Loaded model from output/cervo_model.bin
✅ Loaded existing trained model.
💰 Predicted price: €1122271.88
What do you want to do?:
> Scrape Data
  Predict Property Price
  Train Model
  Exit

---

Installation

Prerequisites

• Rust (latest stable)
• Cargo package manager
• CSV dataset (automatically generated from scraper)

Clone the Repository

git clone https://github.com/NEBYTE/HouseValuation.git
cd HouseValuation

Build the Project

cargo build --release

---

Usage

Scrape Data

cargo run --release

This will scrape Idealista real estate listings and save them to data/idealista_homes_spain.csv.

---

Train the Model

cargo run --release

The model will process the dataset and generate a trained Elastic Net Regression model saved to output/cervo_model.bin.

---

Predict Property Prices

To estimate the price of a custom property:

cargo run --release -- predict

It will prompt the user for property details (size, rooms, location, etc.) and return a predicted price.

---

Technical Overview

Machine Learning Model

• Algorithm: Elastic Net Regression
• Input Features:
  • Geolocation (latitude, longitude)
  • Size (square meters)
  • Floor level
  • Number of rooms & bathrooms
  • Amenities (pool, garden, garage, lift)
• Evaluation Metric: R² Score

Dataset Handling

• CSV Format
• Auto-generated via scraper
• Stores property attributes & prices

Risk & Error Handling

• Missing values handling
• Data normalization
• Cross-validation (K=5)

---

Dependencies

[dependencies]
csv = "1.2"
ndarray = "0.15"
linfa = "0.7.0"
linfa-elasticnet = { version = "0.7.0", features = ["serde"] }
serde = { version = "1.0", features = ["derive"] }
serde_json = "1.0"
dotenv = "0.15.0"
dialoguer = "0.11.0"

---

License

Distributed under the GNU AGPLv3 license.