Matrix Decomposition Methods for the Improvement of Data Mining in Telecommunications

The increasing size and complexity of telecommunication databases poses a significant challenge for conventional data mining approaches. As mobile networks mature, greater pressure is placed on network operators to make their networks more competitive. This paper proposes the use of matrix decomposition methods for the improvement of data mining in telecommunications. Using a real large mobile network dataset, the paper compares the performance of a standard data mining approach (for clustering analysis) when it is used with and without matrix decomposition. It is shown that the proposed approach decreases the computational complexity of clustering analysis and enables visualization of clusters. The proposed approach has various potential applications such as mobile subscriber classification and mobile network optimization.