Combining Case Based Reasoning and KNearest Neighbor for Prediction System of Students Graduation

Authors

  • Henderi , Abas Sunarya , Maria Purnamasari , Untung Rahardja , Ankur Singh Bist

Abstract

The high percentage of the number of students not graduating on time always has
negative impacts on the education performance of an institution. Among them is the ratio of
lecturers to students, lecturers need in pursuing further studies. Therefore, the number of students
not graduating on time have to be suppressed and the impact anticipated. This paper proposes a
model of a predictive system of student graduation rates, and predicts whether students can
graduate on time or not. The model was developed using the Case Based Reasoning (CBR) as a
method and algorithm of K-Nearest Neighbor (KNN). The study was conducted using data of
students who have graduated as training data, and sixth semester student data as testing data. The
required research attributes consisted of age, achievement index (IPS) value in 1 semester, IPS
value in 2, IPS value in 3, IPS value in 4, and IPS value in 5. The testing of the model was carried
out with four scenarios of training data distribution and testing data. The result of the experiment
shows that the stability of the accuracy value of data in CBS using the KNN algorithm is found
in the scenario of 90% training data and 10% testing data, namely 90.24% with the nearest
neighbors K = 5, K = 7 and K = 9. This concludes that the prediction system model produced in
this study was able to predict whether students will graduate on time or not. This research when
implemented will benefit higher education institution performance in terms of achieving the
students to graduate on time.

Published

2020-10-16

Issue

Section

Articles