/CollaborativeFiltering

simple c++ implementation of collaborative filtering.

Primary LanguageC++MIT LicenseMIT

Recommender System Assignment

C++ Implementation of recommender system with collaborative filtering.

To run the implementation

  1. Keep project files in one folder.

  2. compile using command make.

To compile without using the makefile, type the following command.

g++ -std=c++11 recommender.cpp -o recommender.exe

(Note that -std=c++11 option is must be given in g++.)

  1. Run using following command.

./recommender.exe [base file name] [test file name]

Summary of the algorithm

Collaborative filtering (CF) algorithm is the most prominent approach to generate recommendations.

Advantages

  1. used by large, commercial e-commerce sites
  2. well-understood, various algorithms and variations exist
  3. applicable in many domains (book, movies, DVDs, ..)

Basic assumption and idea

  1. Users give ratings to catalog items (implicitly or explicitly)
  2. Customers who had similar tastes in the past, will have similar tastes in the future

Approach

  1. Consider user c
  2. Find set D of other users whose ratings are “similar” to c’s ratings
  3. Estimate user’s ratings based on ratings of users in D

Input

  1. A matrix of given user–item ratings

Output

  1. A (numerical) prediction indicating to what degree the current user will like or dislike a certain item.
  2. A top-N list of recommended items.

Method

There are two types of collaborative filtering.

  1. User-based
  2. Item-based

Any other specification of the implementation and testing

  • Note that I use c++11, not c++. therefore -std=c++11 option is must be given in g++.

  • self test result

u1 RSME: 0.9094779

u2 RSME: 0.8995277

u3 RSME: 0.8900281

u4 RSME: 0.8941476

u5 RSME: 0.8968278