MachineLearning.js

HomeExplorerAbout GitHub

Documentation

Algorithms, data formats, datasets, and evaluation methods.

MachineLearning.js implements a selection of classic supervised and unsupervised machine learning algorithms entirely in JavaScript — no server required. Every computation runs in your browser.

Algorithms and evaluation methods follow the descriptions in Data Mining: Practical Machine Learning Tools and Techniques by Witten, Frank, Hall, Pal & Foulds (University of Waikato). The ARFF file format is Weka's own invention.

Learning Algorithms

Each algorithm page includes the original paper citation, a pseudocode description, and a Theory → Code walkthrough mapping the paper's steps to the JavaScript implementation.

k-Nearest Neighbor

Lazy learner · Euclidean distance · majority vote

Naïve Bayes

Probabilistic · Gaussian + Laplace smoothing · log-space

Decision Tree (ID3)

Logistic Regression

Support Vector Machine

Neural Network (MLP)

Linear Regression

k-Means Clustering

Unsupervised · Lloyd's algorithm · WCSS

K2 Algorithm

Data Formats

The Explorer accepts ARFF and CSV files. Drop a file anywhere on the Explorer or use the Open file button.

ARFF FormatCSV Format

Explorer

On this page

AboutAlgorithmsData FormatsExplorer
k-Nearest Neighbor

machinelearning.js.org · open source · MIT · Marin's Web Site