Implementation algorithm modification maximum standard deviation reduction in graph clustering using matrix complement as input

Clustering can be done by various methods, one of them is a graph clustering. One algorithm that can be used for graph clustering is based on the minimum spanning tree. In this algorithm, we count the value of the standard deviation weights of the vertices between nodes, which why it is called maximum standard deviation reduction (MSDR) algorithm. By using the MSDR algorithm, a cluster of optimal cluster results can be found, and the benefit is we do not have to determine the number of clusters. This method is called unsupervised learning. In MSDR algorithm, the number of clusters will be determined automatically by using polynomial regression. However, in a complex case, it will be difficult to obtain the number of clusters, so that a modification of the algorithm MSDR is proposed and called by MMSDR (Modification of Maximum Standard Deviation Reduction). The modification is done by replacing the polynomial regression to calculate the difference of standard deviation change in the value of the previous calculation. This paper will discuss the results of the implementation of the algorithm MMSDR on the data that come from Indonesia flights by airline X. While the input matrix in the MMSDR is using adjacency matrix, in this work we use matrix complement as the input.