Case 2. Regression: apartment price estimation

Implementation in pure PHP

In this case, we apply the same k-nearest neighbors idea to a regression task. Instead of predicting a class label, we estimate a numeric value: the price of an apartment. We compute distances to known examples, take the k closest ones, and average their prices.

Example of use


                
<?php

// k-Nearest Neighbors (k-NN) for regression: estimate apartment price.
function euclideanDistance(array $a, array $b): float {
    // Euclidean distance between two vectors (points).
    $sum = 0.0;

    foreach ($a as $i => $value) {
        $sum += ($value - $b[$i]) ** 2;
    }

    return sqrt($sum);
}

// Dataset:
// Each item is [features, price].
// Here features are [area_m2, distance_to_center_km] and price is a numeric value.
$dataset = [
    [[40, 12], 120000],
    [[50, 10], 150000],
    [[60, 8], 190000],
    [[70, 6], 240000],
    [[80, 5], 300000],
];

// Apartment we want to price.
$query = [65, 7];

// How many nearest neighbors to use.
$k = 3;

$distances = [];

foreach ($dataset as [$features, $price]) {
    $distances[] = [
        'distance' => num_format(euclideanDistance($features, $query), 7),
        'price' => $price,
    ];
}

// Sort distances.
usort($distances, fn ($a, $b) => $a['distance'] <=> $b['distance']);
$neighbors = array_slice($distances, 0, $k);

echo 'Neighbors: ' . PHP_EOL . array_to_matrix($neighbors) . PHP_EOL . PHP_EOL;

// Calculate average price of neighbors.
$sum = 0;

foreach ($neighbors as $neighbor) {
    $sum += $neighbor['price'];
}

$prediction = num_format($sum / $k, 0);

echo "Estimated price: $prediction";

Result: Memory: 0.007 Mb Time running: < 0.001 sec.


                Neighbors: 
[5.0990195135928, 190000]
[5.0990195135928, 240000]
[15.132745950422, 300000]

Estimated price: 243333.33333333