Predicting and analyzing student absenteeism using machine learning algorithm

Original article

Many developed countries use assessments tools and national surveys to assess the quality of teaching, as well as the determination of indicators that affect the motivation of students to achieve the best possible results in their studies.

S. Larabi-Marie-Sainte, R. Jan, A. Al-Matouq, S. Alabduhadi have pointed that student’s academic performance can be affected by several factors and one of them is student absences [1]. Marsh, Paulsen, and Richardson suggest that “student ratings demonstrate acceptable psychometric properties which can provide important evidence for educational research” [2–4]. Despite being aware of the harmful effects that absenteeism holds on academic performance, the absenteeism level remains high. J. Childs, R. Lofton have showed that the root causes of chronic absenteeism are complex [5]. M.H. Bahadori, A. Salari, I. Alizadeh, F. Moaddab, L. Rouhi have recommended that educational planners and policymakers pay more attention to the factors mentioned by students as the most important causes of absenteeism [6]. FTI part of UAMD (Aleksandër Moisiu University of Durrës, 2021) “the second-largest public academic institution of the Republic of Albania which enrolls about 500 students each year” is experiencing high rates of absences. If not addressed accordingly, the problem of absenteeism may reduce academic performance and have an impact on many social issues. Many factors influence student absenteeism, thus predicting it many a time proves to be very challenging. Özcan, found that “poor academic outcomes, parental involvement, school management, and

school schedules, as well as health issues and a lack of social activities, are the main factors influencing student absenteeism” [7]. Additionally, Balkis et al. observed that the major reason that was given by students for non-attendance, related to attitudes towards teacher and school, lack of motivation, level of parents education [8]. Based on the work of I. Dey and Kassarnig et al., it is understood that attendance is amongst the most crucial elements in determining a student’s academic performance and success [9; 10]. Wadesango and Machingambi found that auditor condition, socio-economic factors, and relations between students and lecturers are the main factor leading students toward nonattendance [11]. In their work, B.N. Young, W.O. Benka-Coker, Z.D. Weller, S. Oliver, J. W. Schaeffer, S. Magzamen, have shown the connection between student absenteeism and the test scores [12]. Referring to complex factors that influencing high student absenteeism, the usage of Data Mining (DM) and Machine Learning (ML) algorithms is a good method to analyze and predict student absenteeism. Helm et al. refer to ML as “an application of artificial intelligence (AI) that provides to build a model based on training data to make predictions or decisions without being programmed” [13]. DM is a process that extracts and discovers patterns with intelligent methods from a large dataset [14]. Based on the nature of the study and dataset organization, we chose classification methods to evaluate the data. Kantardzic states that “classification techniques are part of predictive methods and categorize a given dataset into classes” [15]. By using these methods, we can predict the unknown values by utilizing the known ones [15]. This dataset consists of 26 attributes and 210,000 records corresponding to the teaching hours of 500 students ranging from 18 to 25 years of age during an academic year at FTI. The attributes analyzed refer to demographic, family, university, and personal factors according to academic performance. This study is most helpful to UAMD and can be easily utilized by other universities in Albania. It is also a valuable tool for all universities in the world, serving to guide the management and provide a sense of understanding of the factors that make students not attend. The rest of the paper is structured as follows: Section 2 – an overview of the data mining classification techniques; Section 3 – the methodology; Section 4 is geared towards the findings of the study; Section 5 relates the discussion to the overall results observed and gives some recommendations.