TimeFly algorithm: a novel behavior-inspired movie recommendation paradigm
- PDF / 1,002,174 Bytes
- 8 Pages / 595.276 x 790.866 pts Page_size
- 16 Downloads / 177 Views
SHORT PAPER
TimeFly algorithm: a novel behavior‑inspired movie recommendation paradigm Bam Bahadur Sinha1 · R. Dhanalakshmi2 · Ramchandra Regmi3 Received: 24 November 2019 / Accepted: 30 March 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract This paper proposes a novel behavior-inspired recommendation algorithm named TimeFly algorithm, which works on the idea of altering behavior of the user with respect to time. The proposed model considers solving two recommendation problems (fluctuating user interest over time and high computation time when dataset shifts from scarcity to abundance) and presents a real application of the proposed method in the field of recommendation engine. It describes a system which enrolls the changing behavior of user to furnish personalization suggestions. The results obtained by TimeFly are compared with the results of other well-known algorithms. Simulation results on 100K, 1M, 10M, and 20M MovieLens dataset reveal that using TimeFly leads to high accurate predictions in less computation time. Keywords Recommendation system · Collaborative filtering · Scalability · Time series analysis · E-commerce
1 Introduction Recommendation systems (RSs) have proven to be of significant assistance for managing item/information overload and user experience through quality recommendations. It tends to predict and recommend items according to user preferences. Nowadays, it is being used in almost all knowledge-based management systems, and various top leading e-commerce websites like Amazon, Flipkart, Walmart, etc., harness the power of recommendation system by analyzing various factors connecting the user with the product. Traditional methods of recommendation have been parenthesized as collaborative filtering [1], content-based filtering [2], and hybrid filtering [3]. These methods have been developed to * Bam Bahadur Sinha [email protected] R. Dhanalakshmi [email protected] Ramchandra Regmi [email protected] 1
Department of Computer Science and Engineering, National Institute of Technology Nagaland, Dimapur, India
2
Department of Computer Science and Engineering, Indian Institute of Information Technology Tiruchirappalli, Tiruchirappalli, India
3
Societe Generale, Bangalore, Karnataka, India
deal with the massive amount of data being generated on the internet. Collaborative filtering (CF) [1] is the most widely used recommendation technique that works on the notion of a similar taste of users, i.e., users who have rated similar items in the past are more likely to have a similar interest in the future. It can be either memory based [4] or model based [4]. It has become apparent as a knowledge discovery tool that provides fruitful recommendations after analyzing the behavior and belief of other alike neighbors. Despite its wide usage, collaborative filtering [5, 6] suffers from the issue of slow computation when the dataset shifts from scarcity to abundance [7, 8]. As the dataset shifts from scarcity to abundance, the complexity of understandi
Data Loading...
