LightRec - lr-ibcf Documentation

Model: lr-ibcf

Purpose

lr-ibcf (Item-Based Collaborative Filtering) is designed to recommend items similar to those a user likes. It analyzes the interactions between users and items to compute item similarity and generate relevant suggestions.

Key Functions

act(itemId, userId, points, modify=true): Records or updates a user's interaction with an item. `points` can be any numeric value. `modify` updates the similarity matrix immediately.
feed(data): Feed a batch of interactions (array of {itemId, userId, points}) to the engine. Automatically builds the similarity matrix.
modifyItemSimilarity(itemId): Updates similarity scores for a single item. Used internally by act.
buildItemSimilarity(): Builds the full item similarity matrix for all items in memory.
recommendForUser(userId, topN=10): Generates top-N recommended items for a specific user based on their history.
printItemUserMatrix(): Prints the internal user-item interaction matrix for debugging purposes.

Installation & Usage

Install via npm:

npm install @lightrec/lr-ibcf

Basic usage with comments:

import { RecEngine } from '@lightrec/lr-ibcf'; // 1. Create a new engine instance const engine = new RecEngine(); // 2. Feed interaction data engine.feed([ { userId: 'u1', itemId: 'i1', points: 5 }, { userId: 'u1', itemId: 'i2', points: 4 }, { userId: 'u2', itemId: 'i1', points: 3 }, ]); // 3. Add a single interaction dynamically engine.act('i3', 'u1', 4); // 4. Generate recommendations for a user (top 5) const recommended = engine.recommendForUser('u1', 5); console.log(recommended); // Output: Array of recommended itemIds // 5. Optional: Debug internal matrix engine.printItemUserMatrix();

About the Author

MD Muktadirul Islam Mahi, Full-stack developer and open-source enthusiast. Passionate about AI, web dev, and lightweight recommendation engines.