recommendersystem's Introduction

Recommender System using Collaborative Filtering

Language used : Python 2.7

Dataset used : MovieLens 100k, available at http://grouplens.org/datasets/movielens/

Features Implemented :

User Based Recommendation
Item Based Recommendation

How to use it:

Step 1: Import Recommender Class

Step 2: Initialise the Recommendation System

    Recommender function requires four arguments:
      
      1.Rmin -> minimum rating in a rating scale
      
      2.Rmax -> maximum rating in a rating scale 
      
      3.User Dataset
      
      4.Item Dataset

Step 3a: For UserBased Recommendation System, UserRecommendation method is implemented

        It requires 2 arguments:
          
          1. UserID
          
          2. Method for Similarity measures
    
    Six different similarity measures are implemented:
      
      1.Proximity-Impact-Popularity (PIP) measure, as proposed by Hyung Jun Ahn in "A new similarity measure for              collaborative filtering to alleviate the new user cold-starting problem", available at 
        http://www.sciencedirect.com/science/article/pii/S0020025507003751
      
      2.Euclidean distance
      
      3.Pearson’s Correlation (COR)
      
      4.Cosine (Cos)
      
      5.Constrained Pearson’s Correlation(CPC)
      
      6.Spearman’s Rank Correlation (SRC)

Step 3b: For ItemBased Recommendation System, ItemRecommendation method is implemented

        It takes 3 arguments:
          
          1. UserID
          
          2. Method for Similarity measures
          
          3. Top n matches required

Recommend Projects