Adaptive SOMMI (Self Organizing Map Multiple Imputation) base on Variation Weight for Incomplete Data

Incomplete data occurs with the missing data repeatedly causing problems in data processing. The Self Organizing Map Multiple Imputation (SOMMI) method is proposed to fill the data repeatedly by using the appropriate weight when learning. SOMMI had been handle data complexity that is difficult to handle appropriately (for example, mixed data). This paper proposes the Method of Self Organizing Maps (SOMMI) in a non-linear approach to overcome continuous and categorical attributes. The recursive learning procedure is stopped when the SOM algorithm clustering has converged. The results showed that learning use alpha > 0.5 and particularly in higher missing rates caused the longest time with RMSSTD is also large, and vice versa.