Split incremental clustering algorithm of mixed data stream

Clustering has been recognized as one of the most prominent functions in data mining. It aims to partition a given set of elements into homogeneous groups without any given knowledge about the distribution of data and according to some (dis)similarity criterion. In this paper, we propose a novel streaming algorithm, based on split technique that was introduced to avoid retaining from the scratch and to ensure the incremental clustering aspect. It intends to cluster continuously arriving chunks of data escorted with new mixed features within memory and time restrictions. Our proposed real-time clustering method clusters mixed data streams using split technique in order to tackle the incremental object, attribute, and class learning spaces at once. So, when necessary, the final distribution of the clusters has to be updated. By dint of split technique, changing the final clusters’ distribution has led to a promising clustering model. Experiments performed on real mixed data sets show that the proposal is efficient and outperforms the conventional k-prototypes method based on different evaluation measures.