@bsull/augurs v0.5.4

augurs: time series forecasting

Javascript bindings to the augurs time series framework.

Usage

1. Add the package to your dependencies:

"dependencies": {
    "@bsull/augurs": "^0.4.1"
}

1. Import the default function and initialize once somewhere in your application:

import init from "@bsull/augurs";
init().then(() => console.log("Initialized augurs"));

1. Use the various ETS, changepoint, outlier, or seasonality detection algorithms. For example:

import { ets, seasonalities } from "@bsull/augurs"

const y = new Float64Array([1.0, 2.0, 3.0, 1.0, 2.0, 3.0]); // your time series data
const seasonLengths = seasonalities(y);
const model = ets(seasonLengths, { impute: true });
model.fit(y);

const predictionInterval = 0.95;
// Generate in-sample predictions for the training set.
const { point, lower, upper } = model.predictInSample(predictionInterval);
// Generate out-of-sample forecasts.
const { point: futurePoint, lower: futureLower, upper: futureUpper } = model.predict(10, predictionInterval);

Troubleshooting

Webpack

Some of the dependencies of augurs require a few changes to the Webpack configuration to work correctly. Adding this to your webpack.config.js should be enough:

{
  experiments: {
    // Required to load WASM modules.
    asyncWebAssembly: true,
  },
  module: {
    rules: [
      {
        test: /\@bsull\/augurs\/.*\.js$/,
        resolve: {
          fullySpecified: false
        }
      },
    ]
  },
}