A Meta-Heuristic Algorithm Approximating Optimized Recommendations for E-Commerce Business Promotions

A Meta-Heuristic Algorithm Approximating Optimized Recommendations for E-Commerce Business Promotions

Shalini Gupta (Department of Computer Science, Atma Ram Sanatan Dharma College, University of Delhi, Delhi, India) and Veer Sain Dixit (Department of Computer Science, Atma Ram Sanatan Dharma College, University of Delhi, Delhi, India)
Copyright: © 2020 |Pages: 27
DOI: 10.4018/IJITPM.2020040103

Abstract

To provide personalized services such as online-product recommendations, it is usually necessary to model clickstream behavior of users if implicit preferences are taken into account. To accomplish this, web log mining is a promising approach that mines clickstream sessions and depicts frequent sequential paths that a customer follows while browsing e-commerce websites. Strong attributes are identified from the navigation behavior of users. These attributes reflect absolute preference (AP) of the customer towards a product viewed. The preferences are obtained only for the products clicked. These preferences are further refined by calculating the sequential preference (SP) of the user for the products. This paper proposes an intelligent recommender system known as SAPRS (sequential absolute preference-based recommender system) that embed these two approaches that are integrated to improve the quality of recommendation. The performance is evaluated using information retrieval methods. Extensive experiments were carried out to evaluate the proposed approach against state-of-the-art methods.
Article Preview
Top

Introduction

The massive growth of web and expansion of e-commerce websites generates tremendous amount of data. With the ease of access to the internet, online shopping has gained popularity. There are uncountable options that are available to the users. Recommender Systems (RSs) (Ricci, 2011; Hamidi & Mousavi, 2018) are the tools that assist users in making the choice to select the most preferable options. On the other hand these knowledge based systems maintains a good relationship with a customer which is an essential part of business strategy to set the company apart from other competitors. RS are utilized in a variety of areas such as movies, news, books, websites, research articles and online shopping to maintain a better customer relationship management (CRM) strategy. Traditional RS makes use of explicit data (Jeong et al., 2009) obtained from user in the form of ratings and queries to suggest the items (Cleger et al., 2012; Zhao et al., 2013). However these recommendations results in low accuracy of the preference level for a particular product as the obtained data is affected by user’s mood, demographic and contextual situations.

Traditional recommender approaches are not directly applicable to suggest infrequently purchased products. The work carried out explores a novel hybrid collaborative filtering and sequential clustering based approach for generating recommendations. The click stream data is generated when a user browse products on an e-commerce website. This data shows the products that are of interest to the users, based on features such as time spent, number of visits, cart placement status and purchase status. From these clicked products, a user’s preference for a particular product can be generated. The following questions were raised while studying the related work:

  • What type of information resources can be identified to extract knowledge that can be utilized by recommender systems for recommending infrequently purchased products?

  • What type of strong attributes can be extracted from those information resources and how should the knowledge be presented for use in recommending products?

  • How can the extracted knowledge be used by collaborative-filtering and sequential pattern approach to recommend products?

  • How collaborative filtering and data-mining based approaches for recommending infrequently purchased products be integrated?

The present work analyses the navigational behavior of user while browsing the commercial website. The data is extracted from various click stream sessions (Li & Tan, 2011; Kim &Yum, 2011). The data obtained is implicit data as there is no interaction from user side. Strong attributes also termed as indicators are identified such as time spent on reading a particular product, number of times a product webpage is opened, basket placement status and purchase status of a product etc. User’s absolute and relative preferences are predicted based on these attributes. Further, an effort has been made to recommend products to users, based on the sequential path (Chen et al., 2009) a user follows while navigating the website. Similarity among users is calculated based on the common sequences they follow while browsing and products of interest (Yue et al., 2012) are recommended. This is further discussed in following sections. The main contributions can be summarized as below:

  • This paper proposes an efficient method to refine recommendations based on sequential path a user follow while navigating e-commerce website by extracting implicit details. The algorithm provides ranking to the products that are clicked and viewed by the user. However, a low ranking is provided to the products that prefixes the sequence.

  • In order to enhance the accuracy of these products, absolute preference is calculated that is based on implicit features extracted from path a user follows such as time stamps, number of visits and basket and purchase status.

  • The mentioned algorithms are then merged accordingly by providing weights to each technique. Compared with previous works, the proposed scheme enhances the accuracy of recommendations and provides appropriate ranking to the products.

Complete Article List

Search this Journal:
Reset
Open Access Articles
Volume 12: 4 Issues (2021): Forthcoming, Available for Pre-Order
Volume 11: 4 Issues (2020): 3 Released, 1 Forthcoming
Volume 10: 4 Issues (2019)
Volume 9: 4 Issues (2018)
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing