How Do Movie Preferences Correlate with e-Commerce Purchases? An Empirical Study on Amazon
The following paper presents a study on the relationship between customer movie preferences and online purchases of products from different categories. The analysis was based on the dataset of 233.1 million Amazon reviews and followed the CRISP-DM methodo
- PDF / 5,198,343 Bytes
- 13 Pages / 439.37 x 666.142 pts Page_size
- 69 Downloads / 163 Views
Abstract. The following paper presents a study on the relationship between customer movie preferences and online purchases of products from different categories. The analysis was based on the dataset of 233.1 million Amazon reviews and followed the CRISP-DM methodology. The presented findings confirm that movie preferences correlate with specific product purchase preferences. For instance, customers who watch movies from the categories documentary and drama are more likely to be interested in books purchase, whereas people who watch action movies are having higher scores in electronics. The following paper contributes especially in directly linking movie preferences and product categories purchases. Provided analysis and generalized model should be interesting for both researchers and practitioners from the e-commerce domain. Keywords: Recommender systems · Cross-domain e-commerce · Amazon · CRISP-DM
1
· Big data ·
Introduction
Recommender systems (RS) are commonly used as information search and decision support tools. They are addressing the problem of information overload by generating personalized suggestions that suit user’s needs. RS are intended to help users find products or services such as books, movies, or even people, based on a different kind of information about the user or recommended item [1]. There are three major types of recommender systems: content-based (CB), collaborative filtering (CF), and hybrid recommender systems [18]. CB exploit similarities among items, e.g., recommending music of the same genre or news articles on the same topic. In contrast, CF exploits similarity and relationships among users to provide recommendations [14]. The most successful examples of CF recommender systems are employed on Amazon1 , Netflix2 , Spotify3 , and 1 2 3
http://amazon.com. http://netflix.com. http://spotify.com.
c Springer Nature Switzerland AG 2020 W. Abramowicz and G. Klein (Eds.): BIS 2020 Workshops, LNBIP 394, pp. 184–196, 2020. https://doi.org/10.1007/978-3-030-61146-0_15
How Do Movie Preferences Correlate with e-Commerce Purchases?
185
Last.fm4 . According to the Microsoft Research report, 30% of Amazon.com’s pages views were from recommendations [17]. Similarly, RS used by Netflix is so effective that more than 80% of movies watched by Netflix users came through recommendations [11]. Therefore, RS are a very important element of current online businesses. However, despite the high effectiveness of RS, companies are still seeking to improve their algorithms since any improvements in this area can be very lucrative for them. Thus, research in this domain seems justified. Moreover, while most of the companies are focused on offering recommendations for items belonging to a single domain (e.g. Netflix, Spotify), there are large e-commerce sites like Amazon or eBay which often store customer purchasing behavior from multiple domains. It is beneficial for them to leverage the knowledge from one domain (source domain) to the other domain (target domain) to generate better recommendations. These kinds of recom
Data Loading...
