Diversification-oriented accuracy prediction in recommender systems

Tremendous amount of data generated by e-commerce users on items (e.g., purchase or rating history), sets some key challenges for the online knowledge discovery scheme. Recommendation systems are an important element of the digital marketplace such as e-stores and service providers that use the generated information to discover preferred products of the consumers. Developing an effective recommender system that produces diverse suggestions without compromising the precision of the customised list is challenging for online systems. This paper aims at diversifying recommendation by integrating graph-based algorithm supported with significant nearest neighbour strategy for enhancing recommendation precision. The experimental efficacy on the 100K dataset of MovieLens shows that the proposed hybrid model has a strong coverage and superior efficiency in product recommendations. © 2022 Inderscience Enterprises Ltd.