Research on Developing a Recommender System

Added on - Sep 2019

Trusted by 2+ million users,
1000+ happy students everyday
Showing pages 1 to 1 of 3 pages
CS535Fall 2016Semester Long ProjectAssigned on 29 August 2016Final Due on 12 December 2016Total Points: 100This is a semester-long project, encouraging students to dive into the research in datamining. Specifically, this project concerns with the research on developing arecommender system. The project has three phases.The first phase begins at the beginning of the semester when you have received theassignment. You are asked to conduct the research on recommender systemsindependently by yourself.Recommender systemshave beenstudied extensively in the literature andhavealsobeenfound popularlyusefulin many real-world applications. A recommender systemis considered in the scenario where we have m different users and n different items ina specific application (e.g., in an E-commerce application we have n differentcustomerspotentiallyintending to buy m different commercial items) where a user imay give an item j a rating value k based on this users preference(i[1,... , m]; j[1,...,n]; k[1,...,K]).Essentially, the whole user preference rating data can berepresented as a matrix of n by m where each element of this matrix is the value k ifthe user in question gave a rating for the item in question, and 0 if no such rating wasgiven yet. Initially, all the elements of this matrix are 0.After we havecollected acertain amount of such user preference rating data, this matrix becomes partiallyfilled out with different non-zero values. The goal of a recommender system is that,after we have collected a certain amount of the user preference rating data, the systemis able to predict the rating value k of useri for item j if user i has not yet given such arating.In other words, given such a partially filled out matrix, a recommender systemshall be able to fill out all the predicted rating values for those elements with thecurrent value of 0.That is, a recommender system mathematically is a solution to amatrix value completion problem.In this phase, you dive into the research on recommender systems in the literature andeither develop a new recommender system by yourself or identify an existing,working recommender system. Then you implement and evaluate this implementedsystem till you are happy with the system.You may use whatever language you arecomfortable with to implement the recommender system.Now you are ready to testand report how good your system is. You are given a partially filled out rating datamatrixwith this assignment. The matrix is given in an ASCII text file where each rowis a non-zero element of this matrix with three entries: the user ID number, the item
Desklib Logo
You are reading a preview
Upload your documents to download or

Become a Desklib member to get access