Abstarct: Nowadays, with the increasing rate of data generation and the emergence of the data stream concept, we can extract useful information from data by data mining techniques. This information can be used in a variety of areas such as social network analysis, patient medical information grouping, and network intrusion detection. Among the information extraction strategies, we can point to clustering, which is a convenient way of organizing data. In this thesis, a new approach is introduced baxsed on accordant clustering. Accordant clustering attempts to create more interpretable clusters by considering the background knowledge in order to discover possible group relationships among data and obtain more useful information from clusters. If the data already has a predefined grouping and we want to use this information in clustering, such that a relationship exists between the clusters and the prior groups, accordant clustering can be used. Accordant clustering, despite adopting a new approach to clustering, is not applicable on data streams and executes centrally. In the proposed method of this thesis, the accordant clustering process is presented in an online and incremental manner in order to apply the concept of accordant clustering on the data streams and discover the group relationships in such environments. Conventional clustering methods attempt to put similar data in a cluster; while the proposed method is using the background knowledge with a new approach in the concept of clustering. One of the important advantages of the proposed method is the possibility of finding the existing group relationships in a data stream, reducing the memory required for storing data, reducing the time and computational complexity, as well as the lack of dependence on the number of clusters in order to create the final clustering. The results of the experiments on artificial and real data sets confirm the proper performance of the proposed method in comparison with the basic accordant clustering and K-Means clustering.