Intelligent Archive Search System Using Cuckoo Search-Enhanced K-Prototypes Clustering

Abstract

With the rapid advancement of the times, the amount of data stored in archives is increasing. Aiming at the problem of low data search accuracy of traditional search algorithms in archives, this research proposes an intelligent search method for archives that uses cuckoo search algorithm to improve K-prototypes clustering algorithm. First, the Cuckoo search algorithm is innovatively utilized to improve the K-prototype clustering algorithm and solve the problem of poor searching ability of the K prototype algorithm. Then, the CS-K-prototypes algorithm is introduced for intelligent data search in archives. Finally, experiments are conducted using 200 sets of data collected from machine learning repositories as well as real data from a large archive, and performance evaluation metrics such as Precision-Recall curve, MAE, RMSE, Jacquard Coefficient (JC), Rand Index (RI), and Fowlkes Mallows Scores are used. The experiment findings denote that the K-prototypes clustering algorithm proposed by the research combined with the cuckoo search algorithm has an accuracy and recall curve offline area of 0.9744 when conducting intelligent search of archive data. In contrast with the K-value average clustering algorithm, the accuracy and recall curve offline area increase by 0.073. Compared to the artificial bee colony algorithm, it improves by 0.2252 under conditions of abundant data. In the practical application experiment of the algorithm model, the proposed model achieves a search accuracy of 97.67%. The above results indicate that the improved K-prototypes clustering algorithm proposed by the research can improve the search accuracy of archive data.