The tremendous growth of customers and products in recent years poses some key challenges for recommender systems. These are producing high quality recommendations and performing many recommendations per second for millions of customers and products. New recommender system technologies are needed that can quickly produce high quality recommendations, even for very large-scale problems. The authors address the performance issues by scaling up the neighborhood formation process through the use of clustering techniques. By using association rule learning, it has been observed that customers who purchase the items t-shirt and jeans have an increasing trend to buy shoes, etc. These systems, especially the k-means clustering-based ones, are achieving widespread success in e-commerce nowadays, and the results are encouraging (i.e., the category silver is preferable as purchasing amount is concerned). Enterprises can use the model to predict the stock and customer for their business sustainability.
TopIntroduction
With the increasing popularity of Internet technology, there is a new trend of buying various products from E-commerce website which offers different products for sale. As the customers can acquire a lot of information on-line, they may get confused with information overload. Information and communication technology (ICT) infrastructures such as Personal computers, mobile phones, servers, routers etc. are needed for E-Commerce business. Extensive use of these technologies may lead to energy consumption and electronic waste. Virtual business network and digitization of E-commerce business has created an impact on transportation, land usage and productivity of resources which in turn may affect environment. E-Commerce business has also changed the lifestyle and consumption patterns of people that indirectly affect the environment (Niu, 2016). Miniaturization of this devices and its low power consumption rate, make the electronic gadgets very popular in market.
Clustering has been extensively used as an unsupervised learning method (Ester, Kriegel, Sander, & Xu, 1996, Pitsilis, Zhang & Wang, 2011) which creates a great impact on the researcher in the field of computer science. Clustering techniques has been used as a subject of research in the area of Recommender systems, still there is a need to explore this field. According to Pitsilis, Zhang & Wang (2011) to make future choice of neighbor it is needed to partition choices into smaller sets permanently. The research of Sarwar, Karypis, Konstan & Riedl (2001) is considered as the pioneer in the field of Recommendation system which uses clustering techniques. Sumathi et al. (2010) introduced the concept of recommendation systems, and is based on navigational patterns of the user’s. They used model based clustering, where recommendations was provided to needs of the user based on their requirements (Thiyagarajan, Thangavel & Rathipriya, 2014). Irvin (1994) has done the life time value analysis to identify lucrative customers and to build up strategies for new customers. Choi, Yoo, Kim & Suh (2012) discover the buying pattern of customers of different categories, as it helps for the future prediction using K-means clustering algorithm. K-means clustering algorithm uses Euclidian distance, where the distance is computed by finding the square of the distance between each data points, and then summing the squares to get the desired result (Oyelade, Oladipupo & Obagbuwa, 2010).
Recommendation system plays a significant role to address these problems where various data mining techniques are incorporated, and by using the knowledge gathered about products actions can be taken which may benefit the users. Customers’ preference rating for different products has also been successfully mined for automatic product recommendation. For the creation of recommendation rules, the association rule mining was applied based on the frequent purchase patterns obtained from different group of customers (Agrawal, Imielinski & Swami, 1993; Liu & Shih, 2005; Bandyopadhyay, Thakur, & Mandal, 2018). Using collaborative filtering, the frequent purchase patterns represents the purchasing behaviour of customers having similar taste for products (Aggarwal, Wolf, & Yu, 1999; Thakur, Kundu & Sing, 2011; Paul, Sarkar, Chelliah, Kalyan, & Nadkarni, 2017). In this work, Association Rule Learning (ARL) (Agrawal & Srikant, 1994) and traditional k- means clustering algorithm have been implemented and categorization has been done. This work uses Euclidean distance which is a measure of similarity has been chosen for categorization of customer in terms of their buying pattern.