Predicting the future academic performance of undergraduate students with artificial neural networks

Thesis Overview

<p> </p><div><p><strong>INTRODUCTION</strong></p><p><strong>1.1 &nbsp; &nbsp; &nbsp; BACKGROUND TO THE STUDY</strong></p><p>Most institutions of higher learning today are concerned with predicting the paths of undergraduate students. By analysing student performances as they move from one level to the next, it is possible to determine which students will join particular course programs and the various fields the students will excel at. Today, one of the biggest challenges that educational institutions face is the explosive growth of educational data and how to use this data to improve the quality of managerial decisions. This research is aimed at using Neural Networks to see how educational data can be made more useful.</p><p>Artificial neural network is defined by Dr Robert (1989) an inventor of one of the first neurocomputers as “a computing system made up of a number of simple, highly interconnected processing elements, which process information by their dynamic state response to external inputs”. ANNs are processing devices that are modelled based in the neuronal structure of the mammalian cerebral cortex but on rather much smaller scales. A large neural network might be made of thousands of processor units whereas the human brain contains billions of neurons with corresponding increase in magnitude of their overall interaction and emergent behaviour (Maureen, 1989). The motivation of using a neural network approach is its learning algorithm that learns the relationships between variables in sets of data and then builds models to explain these relationships (Radi and Samy, 2013).</p><p></p><p>Artificial Neural Networks can be used to perform different tasks depending on the training received. With the correct training, neural networks should be able to generalise and should be able to recognize similarities in different input patterns. When designing a neural network, certain parameters must be decided upon, these include:</p><p>(a) &nbsp; The number of layers</p><p>(b) &nbsp; The number of neurons per layer</p><p>(c) &nbsp; The number of training iterations per layer</p><p>(d) &nbsp; The number of hidden neurons</p><p>(e) &nbsp; The learning rate and</p><p>(f) &nbsp; The momentum parameter.</p><p>Neural Networks can be applied in the following areas:</p><p>§ Capturing associations or discovering regularities in within a set of patterns;</p><p>§ Where the volume, number of variables or diversity of the data is very great;</p><p>§ Where the relationships between variables are vaguely understood; or</p><p>§ When the relationships are difficult to describe adequately with conventional approaches.</p><p>Their advantage over other networks is that they can capture various kinds of relationships therefore allowing the user to quickly and relatively easily model phenomena which may have originally been difficult to explain. Some ANNs are adaptive systems and can for instance be used to model populations and environments which constantly change.</p><p>Artificial Neural Networks possess a remarkable ability to derive meaning form complicated or imprecise data and can be used to extract patterns and detect trends that are too complex to be noticed by humans or other computer techniques. A well trained neural network can be considered and expert in the range of information it has been given to analyse. Other advantages are;</p><p>(a) &nbsp; Adaptive learning:</p><p>The ability to learn how to perform tasks based on the data given for training or initial experience.</p><p>(a) &nbsp; Self-organisation:</p><p>The ability to create its own organisation or representation of the information it receives during learning time.</p><p>(b) &nbsp; Real Time Operation:</p><p>The computations of an ANN may be carried out in parallel, and special hardware devices are being designed and manufactured which take advantage of this capability.</p><p>(c) &nbsp; Fault tolerance through redundant information:</p><p></p><p>The partial destruction of a network leads to corresponding degradation of performance, however, some of the network capabilities may be retained even with major network damage.</p><p>The analysis of educational data is not particularly a new practise. However, latest advances in educational technology as well as increase in computing power and the ability to enter fine-grained data about students use of computer based learning atmosphere have led to increased interest in developing techniques for analysing large amounts of data that are often generated in the educational field.</p><p>In this project, artificial neural networks will be explored so as to find meaningful ways to evaluate academic data to aid future decision making so as to improve on academic performances of students</p><p></p></div><div><p><strong>1.2 &nbsp; &nbsp; &nbsp; STATEMENT OF THE PROBLEM</strong></p><p>There is a need for improvement in the service and quality of education in tertiary institutions of learning, the ne1ed to take into consideration the performance of students in each course so as to see how the students can improve in them and in their overall performances. Most times, institutions does not take a look at students’ in the previous session before promoting them to the next level or allowing them to continue in a particular course of study. This has caused a great decline in the quality of their final results (CGPA) thereby producing undergraduates with poor grades that also do not meet the necessary requirements for graduation and employment therefore lacking the ability to effectively compete with their counterparts in the wider world.</p><p>This research is centred on using Artificial Neural Networks (ANN) to predict students’ performances so as to avoid the above mentioned problems thereby making the school counsellors to effectively guide students towards academic excellence.</p><p><strong>1.3 &nbsp; &nbsp; &nbsp; AIM AND OBJECTIVES</strong></p><p>The aim of this project is to predict the feature academic performance of undergraduate students using Artificial Neural Networks. This is to be achieved by the following objectives:</p><p>(i) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; To collect data of undergraduate students.</p><p>(ii) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; To transform the raw data into the suitable format for the prediction tool to be used.</p><p>(iii) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; To train the Neural Network with the transformed data using a suitable neural network model.</p><p><strong>1.4 &nbsp; &nbsp; &nbsp; SIGNIFICANCE OF THE STUDY</strong></p><p>The ability to predict students’ feature performances will create a more customised student experience as students will be better advised to enable them improve on their academic performances. It will also enable the lecturers and counsellors determine the strengths and weaknesses of these students as this will enable them understand the students better and know what areas they would fit best in. With predictive ability, failure rates will be greatly reduced.</p><p>The groups of people to gain from this are the students, parents and counsellors.</p><p>With predictive ability each group involved will have an insight to the students’ capabilities as well as changes they need to make to ensure the smooth running of the institution. By so doing, the institution, the lecturers and parents can better advice the students on what career paths to pursue and the student as well are able to make better choices for themselves.</p><p><strong>1.5 &nbsp; &nbsp; &nbsp; SCOPE OF THE STUDY</strong></p><p>This project is centred on evaluating and predicting students’ performances using Artificial Neural Networks (ANN). The grades of the undergraduate students of computer science department, school of information and communication technology will be used for this research. </p><p><strong>1.6 &nbsp; &nbsp; &nbsp; LIMITATIONS</strong></p><p>§ This research work covers only the Computer Science Department of the Federal University of Technology Minna, and thus may not be generalizable by other institutions.</p><p>§ The network cannot be used if there is only very little data available</p></div> <br><p></p>