Graph Laplacian for Heterogeneous Data Clustering in Sensor-Based Internet of Things

Traditional clustering algorithms are not suited for the heterogeneous data of the Sensor-Based Internet of Things. The accuracy of real-time data processing, in such applications, is further compromised because of the noise and missing values in the data. Considering the need for accurate clustering, a graph Laplacian-based heterogeneous data clustering is proposed in this work. Exploiting the correlation structure of the data, weight graphs are used to generate a graph Laplacian matrix to obtain co-related data points. Eigenvalues are further used to obtain distance-based, accurate clusters. The proposed algorithm is validated on five different real-world data sets and is able to outperform most of the existing algorithms. A detailed mathematical analysis followed by extensive simulation on real-world data sets proves the dexterity of the proposed method, as the performance gap, with respect to the state-of-the-art methods, in terms of accuracy and purity is as high as 30%.