@asymmetrik/akin

@asymmetrik/akin

Recommendation Engine Library based on Collaborative Filtering. Node.js implementation using MongoDB via Mongoose.

Provides hooks for logging user activity on items referenced by an ObjectID and an optional item type. Additional methods provide execution of a scalable, collaborative filtering algorithm that outputs recommendation results for each user into a namespaced Mongo collections. These results can then be retrieved and integrated into the application as desired.

Table of Contents

• Install
• Usage
• API
• Contribute
• License

Install

Include this module as a dependency of your application in the package.json file. It also requires that MongooseJS is available as a peer dependency. For example:

{
  ...
  dependencies: {
    "mongoose": "~4.6",
    "@asymmetrik/akin": "latest"
  }
  ...
}

Usage

Include the module via require wherever applicable:

var akin = require('@asymmetrik/akin');

The most fundamental use case of logging user activity, running the engine, and retrieving recommendations for a user is achieved via:

akin.activity.log(userId, itemId, { type: 'itemType' }, 'action');
...
akin.run();
...
akin.recommendation.getAllRecommendationsForUser(userId);

There are four classes that allow for managing activity logs and recalculating recommendations at a more granular level:

akin.model

Manages query execution for schemas

akin.activity

1. Manage (add or remove) a user's activity for any item.
2. Calculates the first phase of the recommendation engine: user's weighted scores on items

akin.similarity

1. Calculates the second phase of the recommendation engine: user's similarity to other users

akin.recommendation

1. Calculates the third and final phase of the recommendation engine: collaborative filtering to generate recommendations
2. Allows for retrieval of all or a weighted random sample of recommendations
3. Marking items as do not recommend for a user to stop them from returning in the sampling query

API

recalculate all recommendations

Execute the recommendation engine based on the data supplied to the activity.log() API

akin.run();

activity

Add a user's activity on an item

akin.activity.log(userId, itemId, { type: 'itemType' }, 'actionType');

Remove a user's activity on an item

akin.activity.removeLog(userId, itemId, 'actionType');

retrieve recommendations

Get all recommendations created for a user

akin.recommendation.getAllRecommendationsForUser(userId);

Get a sampling of recommendations created for a user based on a weighted cumulative distribution function

akin.recommendation.sampleRecommendationsForUser(userId, numberOfSamples);

Contribute

PRs accepted.

Small note: If editing the README, please conform to the standard-readme specification.

License

@asymmetrik/akin is MIT licensed.