Prediction of Students’ Performance Using Artificial Neural Network (ANN)


  • Fagbuagun, O. A
  • Olaiya Folorunsho
  • Obinna Nwankwo
  • Samson Akinpelu


Machine learning, TensorFlow, Students’ academic performance


Poor performance of students in institutions of higher learning is a great concern for both parents and educators. Performance is a product of factors that are both institutional, personal, and familial. Institutional factors such as facilities provided for learning, class population, lecturer availability, and others can affect students' performance. Also, personal characteristics such as students’ health index, attendance in class, and extracurricular activities can contribute to students' performance. Familial factors include family stress, parents’ educational level, and students' proper guidance from home. In this research work, home, school, and personal factors are collected from students, which forms the dataset. Cronbach's alpha was used to test for the reliability of the data. The data serves as input to an Artificial Neural Network classifier which has five (5) input layer nodes, two (2) hidden layer nodes, with one hundred (100) nodes each and eight (8) output layer nodes. The model was developed with Python programming libraries of Keras and TensorFlow, and optimized using
stochastic gradient descent optimizer. The training and testing were done with two hundred (200) epochs of learning with ten batches of data extracted from the dataset. The data was split in a ratio of 70% for training and 30% for testing. The model accuracy was 0.9193, representing 91.93%.