Comparison of Multi Layer Perceptron, Random Forest & Logistic Regression on Students Performance Test

The test is one thing that can be taken to measure a person's ability to understand a material or a competency. In general, there is a final test taken by students at the school level, before reaching the final test, usually students will take a series of preparatory tests. In reality, of course, not all students can take the test preparation well. therefore of course the school has data related to test preparation. From this test preparation data, a classification technique can be used to classify the data of students who have completed the preparatory test and students who have not completed the preparatory test, so that schools can prepare the best strategy. To assist in classifying data, data classification techniques are needed, in this study the Multi-Layer Perceptron, Random Forest and Simple Logistics algorithms were used. These three methods produce different accuracy when used for the data classification process. For testing the data, scenarios are used using cross-validation. The results of this test scenario show that the Logistic Regression method is superior to the Random Forest and Multi-Layer Perceptron methods with an accuracy of 73.9%. The best Root Mean Square Error results are in the Multi-Layer Perceptron method with the smallest value of 0.363.