DESIGN AND IMPLEMENTATION OF A SYSTEM FOR PREDICTING STUDENT PERFORMANCE USING ARTIFICIAL NEURAL NETWORK

DESIGN AND IMPLEMENTATION OF A SYSTEM FOR PREDICTING STUDENT PERFORMANCE USING ARTIFICIAL NEURAL NETWORK

ABSTRACT

Artificial intelligence has made it possible to create more sophisticated and efficient student models that depict and identify a greater variety of student behavior than was previously possible. The construction of a user-friendly software tool based on a neural network classifier for forecasting students’ performance in course will be made in this project. This tool has a simple interface and can be used by educators to classify and differentiate pupils with low achievement or weak kids who are likely to have low achievement. The observed low quality of graduates from this institution in recent years has been partly attributed to the shortcomings of some or most of the lecturers at the University, which relates to their capacity to manage students. An Artificial Neural Network (ANN) model for predicting student performance will be built and evaluated in this study. The system will be constructed and trained using data from five generations of graduates from one of the school’s departments. The application of artificial intelligence has enabled the creation of more sophisticated and efficient student models capable of representing and detecting a greater variety of student behavior than was previously achievable.

1.0 INTRODUCTION TO CHAPITRE ONE

The use of artificial intelligence in education has risen tremendously in recent years, owing to the fact that it allows us to discover fresh, intriguing, and useful information about pupils. Educational data mining (EDM) is a new discipline concerned with developing ways for examining the distinct types of data generated by educational contexts. While typical database queries can only answer questions like “find the students who failed the exams,” data mining can answer more abstract questions like “find the students who may pass the exams.” One of the most important areas of EDM use is the creation of student models that predict student traits or performance in their educational institutions. As a result, researchers have begun to look into various data mining tools to assist educators in evaluating and improving the structure of their course setting.

The primary goal of the admissions procedure is to identify individuals who are likely to do well at university or who can perform well during the academic year, or to produce students of high grade and intelligence. The quality of candidates admitted to any higher education institution affects the level of research and training within the institution, and thus has an overall effect on the development of the country itself, as these candidates eventually become key players in the country’s affairs in all sectors of the economy.

However, there has recently been a dramatic decline in the quality of graduates from various Nigerian universities. This drop has been attributed to, among other things, the shortcomings of the current university admissions system. The growing discrepancy between the number of students seeking admission and the total number of available admission slots has put further strain on the admissions process. This pressure has resulted in an increase in incidents of admission fraud and other difficulties.

In Nigeria, pupils must join secondary school after completing at least six years of primary school and passing a stipulated National Common Entrance Examination. A student then attends Secondary School for a minimum of six years before taking the General Certificate of Education Examination (GCE), often known as the Senior Secondary Certificate Examination (SSCE) or the Ordinary Level Exams. The test allows for a maximum of nine and a minimum of seven subjects, with Mathematics and English Language being required. Each topic has nine possible grades: A1, A2, A3 (distinctions grades), C4, C5, C6, (credit grades), P7, P8 (pass grades), and F9 (failure).

As a result, this study takes an engineering approach to addressing the admissions problem, looking for ways to make the process more successful and efficient. The study specifically attempts to investigate the feasibility of employing an Artificial Neural Network model to forecast a student’s performance prior to enrolling the student.

Intuitively, one would anticipate a student’s performance to be a function of a variety of elements (parameters) related to the student’s history and intelligence. It is evident, however, that constructing an analytical (or mathematical) model that can adequately explain this performance/factors relationship will be difficult. However, one practical technique to predicting a student’s success may be ‘extrapolating’ from historical data of previous students’ backgrounds and associated performances.

The disadvantage here is the difficulty in selecting an appropriate function capable of capturing all types of data relationships as well as automatically modifying output in the event of additional information, because a candidate’s performance is influenced by a number of factors, and this influence/relationship is unlikely to be any simple known regression model.

A more generic approach that can handle this type of problem is an artificial neural network, which mimics the human brain in problem solving. As a result, we attempted to create an adaptive system, such as an Artificial Neural Network, to forecast a candidate’s performance based on the effect of these elements.

The prediction results can also be utilized by instructors to designate the most appropriate instructional actions for each set of students and offer them with further support suited to their needs. Furthermore, the prediction results may assist students in developing a good understanding of how well or poorly they will perform and then developing an appropriate learning strategy. Accurate prediction of student progress is one method for improving educational quality and providing better educational services (Romero and Ventura, 2007). Traditional mathematical models as well as newer data mining techniques have been used to predict student academic success. A series of mathematical formulas was utilized in these approaches to define the quantitative relationships between outputs and inputs (i.e., predictor variables). If the difference between the projected and actual numbers is within a short range, the forecast is correct.

1.1 THE RESEARCH’S BACKGROUND

Artificial neural networks (ANNs) are a family of statistical learning models inspired by biological neural networks (the central nervous systems of animals, particularly the brain) that are used to estimate or approximate functions that can depend on a large number of inputs and are generally unknown in machine learning and cognitive science. Artificial neural networks are commonly depicted as systems of interconnected “neurons” that exchange messages. The connections include numeric weights that may be adjusted based on experience, allowing neural nets to adapt to inputs and learn. A neural network for handwriting identification, for example, is defined by a set of input neurons that can be activated by pixels in an input image. The activations of these neurons are then passed on to other neurons after being weighted and altered by a function (specified by the network’s designer). This cycle is repeated until an output neuron is engaged. Which character was read is determined by this.

The artificial neural network (ANN), a soft computing technology, has been used successfully in a variety of scientific domains, including pattern recognition, defect diagnosis, forecasting, and prediction. However, as far as we are aware, little research on forecasting student academic success employs artificial neural networks. Kanakana and Olanrewaju (2001) predicted student performance using a multilayer perceptual neural network. They used average grade 12 point scores as inputs and first-year college results as outputs. The study found that an artificial neural network-based model can accurately predict student performance in the first semester. To anticipate the students’ final achievement and classify them into two groups, a multiple feed-forward neural network was devised. A student achievement prediction approach was applied to a 10-week course in their research. The findings demonstrated that accurate prediction is attainable at an early stage, notably during the third week of the 10-week course.

Advising students on their class performance and motivating them to improve is an essential component of every instruction. To attain the aforementioned goal, techniques were required that could properly predict student achievement as early as possible and cluster them for better academic aid. According to Lykourentzou et al. (2009), student-achievement prediction can assist in identifying weak learners and adequately assisting them in their academic quest. For the aforementioned purpose, several approaches and systems have been developed, the majority of them are artificial intelligence-based. Lykourentzou et al. (2009), for example, calculated the final grades of students in e-learning courses using multiple feed-forward neural networks using multiple-choice test data from students at the National Technical University of Athens, Greece. According to the results, ANN is 91.2% efficient. Junemann, Lagos, and Arriagada (2007) employed neural networks to forecast future student academic achievement based on students’ family, social, and economic variables. The previous work focused on forecasting the achievement of 15-year-old secondary pupils in Berlin in reading, maths, and science disciplines.

In the Nigerian context, Oladokun, Adebanjo, and Charles-Owaba (2008) used a multilayer perceptron neural network to predict the likely performance of candidates being considered for admission into the University of Ibadan’s Engineering Course, using various influencing factors as input variables such as ordinary level subject scores, matriculation exam scores, age on admission, parental background, and so on. The results demonstrated that the ANN model correctly predicted the performance of more than 70% of prospective students.

However, Abass et al. (2011) used another Artificial Intelligence (AI) technique, case-base reasoning (CBR), to predict student academic performance based on past datasets, with 20 students from the Department of Computer Science at TASUED as the study domain. The excellent correlation coefficient discovered between the actual graduating CGPA and the CBR projected ones further validates the use and effectiveness of AI approaches in this type of assignment.

In this study, an Artificial Neural Network is utilized to predict students’ final university grades with a 92% accuracy.

STATEMENT OF THE RESEARCH PROBLEM

Looking into the institution these days, you will discover that 48% of the students are actually performing very poorly on their academic level, whom if asked to defend his admission status cannot (i.e. sitting for the attitude test), when proper investigation is carried out, findings show that most of them got into the school through bribe or the so-called upper hand. Another issue or problem for this research work is that some of the applied candidates are actually sound and capable of performing well when admitted, but due to some factors in the student’s immediate or surrounding environment, prevent the student from obtaining or securing his admission into the school. With this small problem, I hope to create an artificial neural network system that will solve it. Coupled with the stress of admittance and the delay in the procedure, the process ends up not being completed perfectly to the requisite standard.

Poor academic performance of some Nigerian students (tertiary and secondary) in recent years has been attributed in part to shortcomings in the National University Admission Examination System. It has become clear that the current procedure is insufficient for picking potentially good pupils. As a result, there is a need to improve the overall sophistication of the system in order to maintain the high integrity and quality. It should be underlined that stakeholders’ apprehension about the traditional admission method is not unique to Nigeria. It is a long-standing and worldwide issue. Kenneth Mellamby (1956) found that institutions around the world are dissatisfied with the procedures used to select undergraduates. While many industrialized countries’ admissions processes have benefited and been improved by numerous breakthroughs in information science and technology, the Nigerian system has failed to fully utilize these new tools and technology. As a result, this study takes a scientific approach to addressing the admissions problem by looking for ways to make the process more successful and efficient. The study specifically attempts to investigate the feasibility of employing an Artificial Neural Network model to forecast a student’s performance prior to enrolling the student.

1.3 THE STUDY’S OBJECTIVES

My primary goal in this research is to create an artificial neural network system capable of forecasting student success.

Other aims that I will address in this study work are as follows:

A mechanism that will improve this institution’s admissions procedure in terms of admitting the right student.

ANNs with a simple and friendly user interface that will allow for quick operation and examination of student output.

To identify some appropriate aspects that influence a student’s performance.

To create an artificial neural network that can be used to forecast a candidate’s performance based on pre-requisite data.