Perbandingan Kriteria Decision Tree Pada Pengetahuan Masyarakat Pada Pemilihan Umum Presiden Indonesia

Abstract

In the Presidential election, gen z as a new voter, must know in advance who the presidential candidate will be in the 2024 election as well as the election process, because if voters do not know and do not understand, it will cause the wrong choice will result in their votes cannot be used even to abstain. What factors cause milennials and gen z generations to not know about elections can be determined using a decision tree. Therefore, in this study, a questionnaire was given to millennials and gen z generation to find out whether voters know the presidential candidate to be elected. The data from the questionnaire is processed to become training data and testing data with a ratio of 70:30. Then measure the accuracy level using the C4.5 algorithm with a comparison of splitting criteria, namely gain ratio, information gain and gini index. By knowing the right splitting criteria, the decision tree model can help overcome the problem of overfitting in the data. Overfitting occurs when the model is too complex in memorizing training data, thus failing to generalize well to read new data. The calculation results show the difference in accuracy values between Gain ratio, Information gain and Gini index, namely 81.67%, 83.33% and 83.33%. It can be concluded that for the use of Algorithm C4.5 splitting criteria Gain ratio and Gini index have the same accuracy value for accuracy measurement in this study.