Implementation of ANN for Predicting the Percentage of Illiteracy in Indonesia by Age Group

Abstract

Illiteracy is one of the serious problems experienced by Indonesia. The lack of care from the government and the private sector towards illiterate people makes the illiteracy rate quite high. Because of this, this problem must be one of the government’s targets going forward, because directly or indirectly illiteracy plays a role in increasing the number of poverty in Indonesia. The purpose of this study is to predict the percentage of the illiterate population in Indonesia according to the age group 15-44 years because this age group category is a productive age. The results of this prediction are expected to be a reference and benchmark for the government in determining and making policies to reduce illiteracy rates. The data that will be predicted is the illiteracy rate data for each province in Indonesia sourced from the Indonesian Central Bureau of Statistics from 2011 to 2017. The method used for prediction is Backpropagation Artificial Neural Network. Data analysis and calculation were carried out with the help of Matlab and Microsoft Excel software. This study uses 5 architectures, 4-5-1, 4-6-1, 4-9-1, 4-14-1 and 4-18-1. Of the five models, the best network architecture is 4-14-1 with an accuracy rate of 91% and the Mean Squared Error 0.00274166. Using this 4-14-1 network model, a prediction on the percentage of illiteracy in Indonesia will be calculated for 2018 until 2020.