Getting started with KNN and C#

Tag: machine-learning

Ricardo Mauro

• 2 min read

K-Nearest Neighbors (k-NN) regression is a simple yet effective technique for predicting numeric values. This post demonstrates how to implement k-NN regression in C# from scratch.

For simplicity, we use a small synthetic dataset:

double[][] trainData = new double[][]
{
    new double[] {1.0, 1.1, 100.0},
    new double[] {2.0, 2.2, 200.0},
    new double[] {3.0, 3.3, 300.0},
    new double[] {4.0, 4.4, 400.0}
};


// Test data
double[] testInput = { 2.5, 2.6 };

Each row represents an instance with two input features and one output value.

K-NN Regression Implementation

The algorithm calculates the Euclidean distance between the test point and training instances, selects the k nearest neighbors, and averages their output values.

using System;
using System.Linq;

class KNNRegressor(double[][] data, int kValue)
{
    int k = kValue

    public double Predict(double[] input)
    {
        var neighbors = data
            .Select(row => new 
              { 
                Distance = EuclideanDistance(input, row), 
                Value = row[^1] 
              })
            .OrderBy(x => x.Distance)
            .Take(k)
            .Select(x => x.Value);

        return neighbors.Average();
    }

    static double EuclideanDistance(double[] a, double[] b)
    {
        double sum = 0;
        for (int i = 0; i < a.Length; i++)
            sum += Math.Pow(a[i] - b[i], 2);
        return Math.Sqrt(sum);
    }
}


Example usage

var knn = new KNNRegressor(trainData, k: 3);
double prediction = knn.Predict(testInput);
Console.WriteLine($"Predicted value: {prediction}");

Evaluation

To assess the model, compare predictions with actual values and compute the Mean Absolute Error (MAE).

double[] actualValues = { 250.0 };  // Ground truth for test data
double mae = Math.Abs(actualValues[0] - prediction);
Console.WriteLine($"Mean Absolute Error: {mae}");

Conclusion

This implementation demonstrates how to apply K-NN regression in C# with minimal dependencies.

While simple, k-NN is effective for small datasets and serves as a good baseline for more complex models.

References

K-NN Algorithm: Wikipedia

Love Discord?