Research on Personalized Book Accurate Recommendation Method Based on ISODATA Clustering Algorithm
[Purpose/significance]In order to improve the efficiency and user satisfaction of personalized book recommendation in university library,an effective online book recommendation method is proposed,which uses the clustering method to classify books,and then recommends new books according to the similarity of books.[Method/process]This paper proposes a clustering method based on ISODATA,which can adjust the fineness of clustering in real time.Ten different users were tested to calculate their TPR,FPR and F1 scores.The average value of FPR is lower than the average value of TPR,which means that the classifier is more committed to removing books that are not of interest to the readers from the reader's list.In addition,the TPR-FPR distribution map is drawn to obtain a graphical representation of the classifier accuracy.[Result/conclusion]The proposed algorithm is compared with the recommendation algorithm based on k-means clustering.The results show that the recommendation method based on ISODATA clustering algorithm is more accurate and effective than the traditional clustering algorithm.
book recommendationrecommendation systemclusteringmachine learning
万方数据
维普
