A comparative study on the efficient reordering methods of sparse matrix problem for large-Scale surveying network adjustment

When a large sparse matrix is calculated for a horizontal geodetic network adjustment, it needs to go through the process of matrix reordering for the efficiency of time and space. In this study, several reordering methods for sparse matrix were tested, using Sparse Matrix Manipulation System (SMMS) program, total processing time and Fill-in number produced in factorization process were measured and compared. As a result, Minimum Degree (MD) and Mutiple Minimum Degree (MMD), which are based on Minimum Degree are better than Gibbs-Poole-Stockmeyer (GPS) and Reverse Cuthill-Mckee (RCM), which are based on Minimum Bandwidth. However, the method of the best efficiency can be changed dependent on distribution of non-zero elements in a matrix. This finding could be applied to heighten the efficiency of time and storage space for national datum readjustment and other large geodetic network adjustment.