L.K-prototypes: An improved K-prototypes algorithm used for hybrid data clustering

In the era of big data, more effective hybrid data clustering methods are expected in data mining and data preprocessing. As one of the typical clustering algorithms, K-Prototypes can deal with mixed type dataset. However, there remains two limitations: (1) the randomly selected initial points lack representativeness; (2) the dissimilarity definition is still rough. In this paper, we propose an improved algorithm called L.K-Prototypes to remedy above two drawbacks. To evaluate the performance of this algorithm, we implemented and conducted experiments with a real dataset. The results showed that L.K-Prototypes can improve both the accuracy and efficiency of clustering.