An enterprise-friendly book recommendation system for very sparse data

Desai, Tejash; Gandhi, Sahil M.; Murlidhar, Pranav; Gupta, Sankalp; Vijayalakshmi, M.; Bhole, Girish P.

doi:10.1109/cast.2016.7914968

Cited by 3 publications

(1 citation statement)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This method effectively reduces the sparseness of user scoring data. In [13], the author used biclustering to handle the existing duality between users and items. The proposed algorithm consists of a hybrid approach containing an initial cluster phase which is taken as input for a biclustering phase and proves to be scalable dealing with large amounts of sparsity.…”

Section: Introductionmentioning

confidence: 99%

A Novel Personalized Recommendation Algorithm for the Metrology Industry with Massive Sparse Data

Zheng¹

2019

JCP

View full text Add to dashboard Cite

Sparsity of source data sets is one major reason causing the poor recommendation quality. In order to solve this problem in the recommendation system of metrology industry with limited an unordered data, this paper proposes a novel personalized recommendation algorithm incorporating industry information and service category information to alleviate the influence of source data sparsity. First, the user's industry information and service category information are added to existing user-service preference data. Then, the K-means clustering algorithm is used to calculate the different user clusters. And then, the user-service preference matrix and the user-service category preference matrix are constructed separately from the user data in each cluster. And then, the nearest neighbor set of target user is calculated by the measure of cosine similarity. Finally, we use the user-based collaborative filtering algorithm to implement personalized recommendations for each user. Experimental results show that the proposed method can improve the recommendation accuracy rate in the metrology industry with sparse data set. The time to calculate for the nearest neighbor is shortened and the recommended speed is improved by reducing the nearest neighbor search range using clustering.

show abstract

Section: Introductionmentioning

confidence: 99%