Classification of Underdeveloped Regency using Probabilistic Neural Networks

Abstract

One of the supporting data in the government work plan in order to realize national priorities of
development of underdeveloped regency is data determining the status of regional backwardness. Determination
of the status regency is a process of classifying the area as an underdeveloped regency or not. Classification cases
can be resolved very quickly using computational algorithm's Neural Network. Probabilistic Neural Network
(PNN) is one method of neural network that can be used to resolve cases of classification with very good
performance. PNN only requires one iteration of training so that the process is faster when compared with a Back
Propagation neural network that requires several iterations of training. PNN has a structure consisting of four
layers, namely input layer, pattern layer, summation layer and decision layer. This study applied the PNN method
in the classification of underdeveloped regency in Indonesia. Classification results show that as many as 458 of
the 491 regencies predicted in accordance with the actual status. The classification method with PNN use leads to
good performance with the accuracy value of 93,29 percent, sensitivity of 91,42 percent and specificity of 94,10
percent.
Keywords: Neural Network, Probabilistic Neural Network, classification, underdeveloped regency