ALZ predictor ·

ALZ predictor is a JavaScript library for symbols sequence prediction, based on Gopalratnam and Cook Active LeZi algorithm.

• AI: ALZ predictor makes it possible to create proactive and smart UIs. Improve your application UX by adding a touch of artificial intelligence. Guess and prefetch required data to enhance performance.

• Simple: Model each user action as a char, then train ALZ by simply adding them as they come (one by one). Predict next user action at any time.

• Lightweight: No dependencies.

FAQ

• Why JavaScript and Frontend AI at all?

  • No server-client latency
  • Offline
  • Privacy
  • Cost
  • TensorFlow.js and ONNX.JS

• How to save learning data?

  Serialize and save to browser localStorage.

• What is the difference with Guess.js?

  Google Analytics is not required for training.

Installation

# Yarn
yarn add alz-predictor

# NPM
npm install --save alz-predictor

Usage

import Predictor from 'alz-predictor'

const predictor = new Predictor()
predictor.add('a')
predictor.add('b')
predictor.add('a')

const predictions = predictor.predict()

Examples

• Next-page data predective prefetching:

  • Online demo: https://mudafar.github.io/alz-predictor-prefetch-example/

  • Source code

Documentation

Constructor

const predictor = new Predictor()

Methods

• add: add char to the sequence.
• predict: get predictions object.
• loadJSON: recover state from JSON.

Serialization

const predictor = new Predictor()
predictor.add('a')
predictor.add('b')

const json = JSON.stringify(predictor)

License

ALZ predictor is MIT licensed.