The Use of Probabilistic Neural Network and ID3 Algorithm for Java Character Recognition

Java character is the character that is used
especially in Java island, Indonesia. Java character has a special
form of writing consist of basis characters, vowels,
complementary, etc. In this research, the authors conducted Java
character recognition using 2 methods, namely probabilistic
neural network (PNN) and ID3 algorithm. PNN is a method of
artificial neural network that can be trained supervised and
unsupervised. PNN is built based on the theory of probability
which is realized as an artificial neural network. This method is
used because PNN has a high accuracy in the classification of
data, also has high speed when performing the process. ID3
(Iterative Dichotomiser) is a decision tree algorithm. Basic ID3
algorithm using tree induction that gives attribute to the node in
the tree based on how much information increases from the node.
From experimental results, PNN method can achieve an accuracy
up to 92.35% for data that has been trained previously, and up to
61.08 % for data hasn�t been trained before. While ID3 can
achieve recognition rate of 100% for data has been trained before
but only 15.57% for data hasn�t been trained before.