TY - JOUR
T1 - The demand effect analysis of head books and tail books in book recommendation networks
AU - Kim, Jae Kyeong
AU - Jeong, Chang Geun
AU - Li, Qinglong
AU - Choi, Il Young
N1 - Publisher Copyright:
© 2021 John Wiley & Sons, Ltd
PY - 2022/2
Y1 - 2022/2
N2 - Many existing studies related to recommendation systems have made great efforts to increase performance-oriented evaluation metrics such as accuracy, recall, F1 value, MAE, diversity, and so on. However, these metrics for measuring performance do not provide any explanation of how such recommendation systems contribute to a product provider's sales. Recently, to investigate the factors affecting product sales, researchers have started to use the product recommendation network, which is a graphical presentation of the relationship between products that are purchased simultaneously. In this research, we are going to identify factors that influence the demand for head books (top sales books) and tail books (non-top sales books) in academic and technology books, because the characteristics of specialized books are different from those of general books. To test the model, online sales data of medicine and science books of G Publishing, the leading medical book publishing and distribution company in Korea, are used. We employ social network analysis and multiple regression analysis to investigate demand effect of the book recommendation network. The former is used to measure centralities of the book recommendation network, while the latter is used to measure the demand effect. Furthermore, we inspect whether there are different attributes between head books and tail books that are related to the book demand in recommendation networks. This study is the first attempt to examine the demand effect of the recommendation network of academic and technology books. The following results are obtained general book sales of an online bookstore follow a long-tail law, while sales of academic and technology books follow Pareto's law distribution. This study investigated attributes that might be linked to demand effect of head books and tail books in the recommendation network of academic and technology books. How to increase book sales using attributes that are differently associated with the demand of head books and tail books in the recommendation network is also presented.
AB - Many existing studies related to recommendation systems have made great efforts to increase performance-oriented evaluation metrics such as accuracy, recall, F1 value, MAE, diversity, and so on. However, these metrics for measuring performance do not provide any explanation of how such recommendation systems contribute to a product provider's sales. Recently, to investigate the factors affecting product sales, researchers have started to use the product recommendation network, which is a graphical presentation of the relationship between products that are purchased simultaneously. In this research, we are going to identify factors that influence the demand for head books (top sales books) and tail books (non-top sales books) in academic and technology books, because the characteristics of specialized books are different from those of general books. To test the model, online sales data of medicine and science books of G Publishing, the leading medical book publishing and distribution company in Korea, are used. We employ social network analysis and multiple regression analysis to investigate demand effect of the book recommendation network. The former is used to measure centralities of the book recommendation network, while the latter is used to measure the demand effect. Furthermore, we inspect whether there are different attributes between head books and tail books that are related to the book demand in recommendation networks. This study is the first attempt to examine the demand effect of the recommendation network of academic and technology books. The following results are obtained general book sales of an online bookstore follow a long-tail law, while sales of academic and technology books follow Pareto's law distribution. This study investigated attributes that might be linked to demand effect of head books and tail books in the recommendation network of academic and technology books. How to increase book sales using attributes that are differently associated with the demand of head books and tail books in the recommendation network is also presented.
UR - http://www.scopus.com/inward/record.url?scp=85116959558&partnerID=8YFLogxK
U2 - 10.1111/exsy.12847
DO - 10.1111/exsy.12847
M3 - Article
AN - SCOPUS:85116959558
SN - 0266-4720
VL - 39
JO - Expert Systems
JF - Expert Systems
IS - 2
M1 - e12847
ER -