0.9.3 • Public • Published


automatically ensemble machine learning predictions together If you already have a bunch of classifiers trained, ensembler will go through and find the best way of ensembling them together.

NOTE: Right now this repo is designed to work with machineJS.

There's no reason it couldn't handle automatic ensembling of predictions from other sources. If somebody wants to make those modifications, that would be a wonderful PR!

What does ensembling mean?

Ensembling, sometimes referred to as SYNONYMS_HERE, probably means exactly what you think it does: taking many different machine learning algorithms, and putting them together into one super predictor.

Why is ensembling great?

Ensembling is frequently more accurate than any single algorithm can be. This is because all algorithms have their strengths and weaknesses. Some areas different classifiers might excel or struggle with:

  • outliers in the data set
  • picking out certain types of relationships in the data
  • generalization Ensembles, because they provide the consensus view of several experts (trained classifiers), will frequently avoid overfitting problems that any one classifier might face.

What this does, effectively, is minimize the risk of including inaccurate classifiers. If the data says they're not helpful in making predictions against the dataset, we will not include them.

This, then, lets you go off and train as many classifiers as you would like, over whatever time period you like, and trust that ensembler will find the best combination of them for you.

Ensembling also reduces the risk of overfitting to the data, because introducing more classifiers will bring predicted values closer to an average prediction across multiple sources, rather than the (possibly highly biased) opinion of a single classifier.


npm install ensembler ensembler is automatically installed as a dependency of machineJS.

Package Sidebar


npm i ensembler

Weekly Downloads






Last publish


  • climbsrocks