Module rml::knn

Implementation for K-Nearest Neighbors. Allows for predicting data based on a KNN search. A full, working example is contained in the examples/knn directory.

Example

// Collect and parse data to a format consistent with:
// type CSVOutput = (Vec<Vec<f64>>, Vec<i32>);

let training_data: CSVOutput = (Vec::new, Vec::new);
let testing_data: CSVOutput = (Vec::new, Vec::new);

// Create a new KNN struct.
let knn = knn::KNN::new(
5, // 5-nearest
training_data.0, // x
training_data.1, // y
None, // Default distance(euclidean)
Some(math::norm::Norm::L2), // L2 Normalization
);

// Get a prediction for each point of the testing data.
let pred: Vec<i32> = testing_data.0.iter().map(|x| knn.predict(x)).collect();

// Count the number that were predicted correctly.
let num_correct = pred
    .iter()
    .cloned()
    .zip(&testing_data.1)
    .filter(|(a, b)| *a == **b)
    .count();

println!(
    "Accuracy: {}",
    (num_correct as f64) / (pred.len() as f64)
);

Structs

KNN

KNN struct handles the computation and data for the K-Nearest Neighbors algorithm. It is highly recommended to not change values inside of this struct manually. Always create a new one using ::new.

Point

A data point.