Abstract
Channelizing product sales with the aid of Recommender Systems is ubiquitous in e-commerce firms. Recommender systems help consumers by reducing their search cost by directing them to interesting and useful products. It also helps e -commerce firms by pushing the range of products a user may purchase on their e-commerce platform. The emergence of marketplace model provides platform for large fragmented buyers and sellers, where shelf space is not a constraint. Owing to unlimited shelf space, it is in the interest of e-commerce platforms to push niche products to idiosyncratic users. However, the current recommender systems, in general, recommends popular and obvious products leading to a few Long-Tail items. In this paper, our focus is on matching the niche products to idiosyncratic users such that the needs of users are satiated. We propose an innovative and robust model of matrix factorization that engenders recommendations based on a user’s optimal liking of the long-tail items. We also propose an adaptive model that pursues to promote the long tail items in the recommendation list. Comprehensive empirical evaluations consistently show the gains of the proposed techniques for handling the long tail on real world data sets like Amazon dataset over different algorithms.
Highlights
Improving customer experience in the digital world is the prime focus of most e-commerce firms
The number of latent factors (I) used in Regularized singular value decomposition (RSVD), proposed model-1 and proposed model-2 are kept constant at the value for which RSVD performs best on cross-validated Root Mean Square Error (RMSE) accuracy measure for respective datasets in order to compare with the proposed models on long-tail measures
The experimentation of the proposed models (PM-1 and PM-2) on the publicly available benchmark datasets demonstrates their suitability i) to provide personalized recommendations to a user by taking into account the respective taste for long-tail items, and ii) to promote long-tail items to idiosyncratic users.PM-1 is an additive model of matrix factorization and novelty measure which ensures that user’s taste towards long-tail item is captured during personalized recommendation
Summary
Improving customer experience in the digital world is the prime focus of most e-commerce firms. The trend is to provide a unique shopping experience in the digital arena to every user so that overall customer satisfaction increases [31]. Recommender systems are a specific type of personalized web-based decision support systems that analyse data about customers and products to help customers find items of interest [16]. Online retailers take the advantages of unlimited shelf space over brick and mortar retailers, with the aid of recommender systems (RS), by pushing the niche products to idiosyncratic users. These niche products are mainly non-hit or miss products that account for significant sales in online platforms. Miss (non-hit) or insignificant number of hits for most of items are often referred as the long tail (TLT) phenomenon in the context of recommender systems [18]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
