sherlocked27/Assignment_IR_Recommender

Recommendor System

Below is the list of the algorithms used to implement recommendor systems which have been implemented which include Collaborative filtering, Singular Value Decomposition and CUR.

Functionality Implemented

1. Collaborative Filtering (calculating similarity by users and predicting missing ratings)

2. Collaborative Filtering using global baseline approach.

3. SVD

4. SVD with 90% retained energy

5. CUR with sampling of rows and columns with replacement

6. CUR with sampling of rows and columns without replacement

The Dataset

• Movie Rating Corpus - dataset contains users and their ratings.

Software/frameworks used:

• python-Numpy, Pandas

1. Install pip: sudo apt-get install python-pip

2. Install Numpy : sudo pip install -U numpy

3. Install pandas : sudo pip install pandas

*Check for istallation by Opening up a Python prompt by running the following:

	python

​ At the prompt, type the following: ​ ​

import pandas

import numpy

print numpy.version

• python-xlrd module -

1. Install xlrd: pip install xlrd

2. Install xlwt: pip install xlwt

Platforms:

• To run recommender system make sure you got python installed.

Run:

• nevigate to IR_AS3 directory.
• change path to the dataset in files to run
• use python3 filename.

RESULT

TECHNIQUE	RMSE	PRECISION AT TOP 50	SPEARMAN CORREALTION	TIME TAKEN (in sec)
Collaborative	0.08	1.389	0.999	300
Collaborative with baseline	0.1581	1.440	0.999	180
SVD	0.60	1.53	0.9999	600
SVD with 90%	0.58	1.55	0.999	450
CUR (with repetition)	495.81	14.82	0.998	300
CUR (without repetition)	2260	568	0978	200