Variable selection for collaborative filtering with market basket data

• DOI: http://doi.org/10.1111/itor.12518
• Author: Wook‐Yeon Hwang
• Date: 01 November 2020
• Published date: 01 November 2020

Intl. Trans. in Op. Res. 27 (2020) 3167–3177

DOI: 10.1111/itor.12518

INTERNATIONAL

TRANSACTIONS

IN OPERATIONAL

RESEARCH

Variable selection for collaborative filtering with market basket

data

Wook-Yeon Hwang,∗

College of Global Business, Dong-A University, Busan, South Korea

E-mail: wyhwang@dau.ac.kr

Received 15 December 2016; receivedin revised form 25 October 2017; accepted 20 January 2018

Abstract

The marketbasket data in the form of a binary user–item matrix or a binary item–user matrixcan be modeled as

a binary classification problem, whichactually tackles collaborative filtering (CF) as well as target marketing.

Effective variable selection (VS) can increase the prediction accuracy as well as identify important users or

items in CF as well as target marketing.Therefore, we propose twonew VS approaches: a Pearson correlation-

based approach and a forward random forests regression-based approach, comparing the performance in a

variety of experimental settings. The experimentalresults show that the proposed VS approaches outperform

the conventional approaches in the examples. Furthermore, the experimental results are more reasonable

and informative than the previous experimental results because the binary misclassification error and Top-N

accuracy for the user CF, the item CF, the user modeling, and the item modeling are all considered in this

paper.

Keywords: market basket data; Pearson correlation; random forests; supervised learning-based collaborative filtering;

target marketing; variable selection

1. Introduction

Recommender systems can suggest to users which items can be more valuable for them. The main

drivers of recommender systems are collaborative filtering (CF) techniques. Although there have

been main challenges for CF, such as data sparsity, shilling attack, and privacy protection, re-

searchers have striven to overcome the difficulties. There are three categories of CF techniques:

memory-based, model-based, and hybrid CF techniques (Su and Khoshgoftaar, 2009). The appli-

cation areas of CF cover movies, books, images, and so on (Park et al., 2012). We mainly focus on

data sparsity, model-based CF techniques, and movies in this paper.

In general, voting scores aremissing, which is called data sparsity.The model-based CF techniques

struggle with data sparsity because they should depend on the information of votingscores. In order

∗Corresponding author.

C

2018 The Authors.

International Transactionsin Operational Research C

2018 International Federation ofOperational Research Societies

Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main St, Malden, MA02148,

USA.