Initial Center Weight Self Organizing Map Using Genetic Algorithm

Missing values occur in data that have incomplete or missing attribute values. Missing value can concern using the imputation method, which as applied to an appropriate value for filling in the missing value. SOMI clustering is used for clustering-based imputation but has constraints in determining the best weight and parameters specified for the artificial neural network. The problem is that choosing the optimal weight value for network node compatibility is a big deal. Furthermore, the Genetic Algorithm (GA) has proposed determining each feature's optimal weight because the output nodes grow dynamically during the input pattern grouping process. The selection of SOM weights using a genetic algorithm is called GAWSOM can be improved by determining the smallest RMSE in the winning unit of the appropriate node so that result stable cluster. The experimental results show that GA is efficient in finding the optimal initial weight value from a large search space. The results show that using the GAWSOM algorithm can find a better solution with elite chromosomes in the real data when the missing value is 10%, with a shorter time than features with mixed data. The results of GAWSOM's performance with elitist search show that the time difference is 100 seconds faster because the searching is less chromosome than GAWSOM nonelitist search.