Data Mining, Neural Network Algorithm to Predict Student's Grade Point Average: Backpropagation Algorithm

Peneliti

Selvia Lorena Br Ginting, S.Si., M.T

Deskripsi/Abstrak

The Grade Point Average (GPA) for a tertiary institution is one of the elements used to measure the quality of teaching and learning. In the Department of Computer Systems, Universitas Komputer Indonesia, there are many students who have bachelor's degrees, but there are still many students who do not graduate on time with unsatisfactory GPA. By utilizing the pile of academic data from students who have graduated, modelling the data mining technique using the Artificial Neural Network (ANN) method with the backpropagation algorithm can be done. The purpose of this study is to test the ANN modelling with backpropagation algorithm in making a student's Grade Point Average (GPA) prediction information system. The student GPA prediction system is used as a supporting material for the academic advisor to provide evaluation guidance to students who are considered to have academic problems. The parameter attributes used in this study included Grade Point Semester (GPS) one to four as well as several course scores as data input and GPA as an output data. The test was carried out in four scenarios, where the best results of ANN were obtained in scenario-IV with the parameters of the number of input layer neurons 18, the number of hidden layer neurons 24, the number of output layer neurons 1, learning rate (?) 0.15, 2000 iterations (epoch) and 591 data sets. The data is divided into 85% data set for training data as well as 15% data set for test data. The average accuracy obtained from this information system is 97.2%. It is expected that this study can be used as a reference for designing a data mining architecture model using the ANN Backpropagation algorithm in creating values prediction information system.