AI & Machine Learning for Real Estate

The Problem

Data scientists building real estate ML models need large, structured datasets. Publicly available datasets are often outdated, limited in scope, or missing key features. To build accurate price prediction models, recommendation engines, or market analysis tools for the UAE, you need fresh data with consistent structure and rich feature sets — location, bedrooms, area, amenities, pricing, and more.

The Solution with BayutAPI

BayutAPI provides a clean, structured data source for ML training and inference. Each property listing comes with dozens of features that are useful for machine learning: numerical features like price, area, bedrooms, and bathrooms; categorical features like location, property type, and purpose; and additional data like amenity availability and agent information. The API’s consistent JSON schema makes data preprocessing straightforward.

How It Works

Step 1: Build a Training Dataset

Collect property data across multiple areas and property types to build a diverse training set:

import requests
import pandas as pd

headers = {
    "x-rapidapi-key": "YOUR_API_KEY",
    "x-rapidapi-host": "bayut14.p.rapidapi.com"
}

def collect_listings(location_id, purpose="for-sale", max_pages=5):
    """Collect listings for ML training."""
    records = []

    for page in range(1, max_pages + 1):
        resp = requests.get(
            "https://bayut14.p.rapidapi.com/search-property",
            headers=headers,
            params={
                "purpose": purpose,
                "location_ids": location_id,
                "page": str(page)
            }
        ).json()

        for hit in resp["data"]["properties"]:
            records.append({
                "price": hit.get("price"),
                "bedrooms": hit.get("bedrooms"),
                "bathrooms": hit.get("bathrooms"),
                "area_sqft": hit.get("area"),
                "location": hit.get("location"),
                "purpose": hit.get("purpose"),
                "title": hit.get("title"),
                "currency": hit.get("currency", "AED"),
            })

        if page >= resp["data"]["totalPages"]:
            break

    return records

# Collect data from multiple areas
areas = {"5001": "Downtown", "5002": "Marina", "5548": "JVC", "5003": "Business Bay"}
all_records = []

for loc_id, area_name in areas.items():
    records = collect_listings(loc_id)
    for r in records:
        r["area_name"] = area_name
    all_records.extend(records)

df = pd.DataFrame(all_records)
print(f"Collected {len(df)} records")
print(df.describe())

Step 2: Feature Engineering

Transform raw listing data into ML-ready features:

# Calculate derived features
df["price_per_sqft"] = df["price"] / df["area_sqft"]
df["bed_bath_ratio"] = df["bedrooms"] / df["bathrooms"].replace(0, 1)

# Encode categorical variables
df["area_encoded"] = pd.Categorical(df["area_name"]).codes

# Remove outliers
q1 = df["price_per_sqft"].quantile(0.05)
q3 = df["price_per_sqft"].quantile(0.95)
df_clean = df[(df["price_per_sqft"] >= q1) & (df["price_per_sqft"] <= q3)]

print(f"Clean dataset: {len(df_clean)} records")

Step 3: Train a Price Prediction Model

Use the collected data to train a model:

from sklearn.model_selection import train_test_split
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.metrics import mean_absolute_error

features = ["bedrooms", "bathrooms", "area_sqft", "area_encoded"]
target = "price"

X = df_clean[features].dropna()
y = df_clean.loc[X.index, target]

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)

model = GradientBoostingRegressor(n_estimators=100, max_depth=5)
model.fit(X_train, y_train)

predictions = model.predict(X_test)
mae = mean_absolute_error(y_test, predictions)
print(f"Mean Absolute Error: AED {mae:,.0f}")

Step 4: Enrich with Amenity Data

Use /amenities-search to add amenity features that can improve model accuracy — properties with pools, gyms, and covered parking often command premium pricing.

Relevant Endpoints

• /search-property — Core data source with rich feature sets for ML training
• /autocomplete — Resolve locations and build area-level features
• /amenities-search — Add amenity features for improved model accuracy
• /search-new-projects — Off-plan data for market prediction models

Benefits

• Rich feature sets: Each listing provides 20+ features suitable for ML models.
• Consistent schema: Clean JSON responses mean less time on data preprocessing.
• Fresh data: Models can be retrained regularly with current market data.
• Scale: Access hundreds of thousands of listings across the entire UAE market.
• Multiple applications: Use the same data for price prediction, recommendations, market segmentation, anomaly detection, and more.