Predicting Students Academic Performance Using Artificial Neural Network
Artificial intelligence has enabled the development of more sophisticated and more efficient student models which represent and detect a broader range of student behavior than was previously possible. In this research, the implementation of a user-friendly software tool for predicting the students’ performance in course which is based on a neural network classifier will be made. This tool has a simple interface and can be used by an educator for classifying students and distinguishing students with low achievements or weak students who are likely to have low achievements. The observed poor quality of graduates of students of this institution in recent times has been partly traced to inadequacies of some or most of the lecturer in the University which goes down to ability to handle the students. In this study an Artificial Neural Network (ANN) model, for predicting the likely performance of student will be developed and tested. The system will be developed and trained using data spanning five generations of graduates from one of the department in the school. The use of artificial intelligence has enabled the development of more sophisticated and more efficient student models which represent and detect a broader range of student behaviour than was previously possible.
1.1 Background to the Study
Predicting student academic performance has long been an important research topic. Among the issues of education system, questions concerning admissions into academic institutions (secondary and tertiary level) remain important (Ting, 2008). The main objective of the admission system is to determine the candidates who would likely perform well after being accepted into the school. The quality of admitted students has a great influence on the level of academic performance, research and training within the institution. The failure to perform an accurate admission decision may result in an unsuitable student being admitted to the program. Hence, admission officers want to know more about the academic potential of each student. Accurate predictions help admission officers to distinguish between suitable and unsuitable candidates for an academic program, and identify candidates who would likely do well in the school (Ayan and Garcia, 2013). The results obtained from the prediction of academic performance may be used for classifying students, which enables educational managers to offer them additional support, such as customized assistance and tutoring resources.
The results of this prediction can also be used by instructors to specify the most suitable teaching actions for each group of students, and provide them with further assistance tailored to their needs. In addition, the prediction results may help students develop a good understanding of how well or how poorly they would perform, and then develop a suitable learning strategy. Accurate prediction of student achievement is one way to enhance the quality of education and provide better educational services (Romero and Ventura, 2007). Different approaches have been applied to predicting student academic performance, including traditional mathematical models and modern data mining techniques. In these approaches, a set of mathematical formulas was used to describe the quantitative relationships between outputs and inputs (i.e., predictor variables). The prediction is accurate if the error between the predicted and actual values is within a small range.
In machine learning and cognitive science, artificial neural networks (ARTIFICIAL NEURAL NETWORKs) are a family of statistical learning models inspired by biological neural networks (the central nervous systems of animals, in particular the brain) and are used to estimate or approximate functions that can depend on a large number of inputs and are generally unknown. Artificial neural networks are generally presented as systems of interconnected “neurons” which exchange messages between each other. The connections have numeric weights that can be tuned based on experience, making neural nets adaptive to inputs and capable of learning. For example, a neural network for handwriting recognition is defined by a set of input neurons which may be activated by the pixels of an input image. After being weighted and transformed by a function (determined by the network’s designer), the activations of these neurons are then passed on to other neurons. This process is repeated until finally, an output neuron is activated. This determines which character was read.
The artificial neural network (ARTIFICIAL NEURAL NETWORK), a soft computing technique, has been successfully applied in different fields of science, such as pattern recognition, fault diagnosis, forecasting and prediction. However, as far as we are aware, not much research on predicting student academic performance takes advantage of artificial neural network. Kanakana and Olanrewaju (2001) utilized a multilayer perception neural network to predict student performance. They used the average point scores of grade 12 students as inputs and the first year college results as output. The research showed that an artificial neural network based model is able to predict student performance in the first semester with high accuracy. A multiple feed-forward neural network was proposed to predict the students’ final achievement and to classify them into two groups. In their work, a student achievement prediction method was applied to a 10-week course. The results showed that accurate prediction is possible at an early stage, and more specifically at the third week of the 10-week course.
1.2 Statement of the Problem
The observed poor academic performance of some Nigerian students (tertiary and secondary) in recent times has been partly traced to inadequacies of the National University Admission Examination System. It has become obvious that the present process is not adequate for selecting potentially good students. Hence there is the need to improve on the sophistication of the entire system in order to preserve the high integrity and quality. It should be noted that this feeling of uneasiness of stakeholders about the traditional admission system, which is not peculiar to Nigeria, has been an age long and global problem. Kenneth Mellamby (1956) observed that universities worldwide are not really satisfied by the methods used for selecting undergraduates. While admission processes in many developed countries has benefited from, and has been enhanced by, various advances in information science and technology, the Nigerian system has yet to take full advantage of these new tools and technology. Hence this study takes an scientific approach to tackling the problem of admissions by seeking ways to make the process more effective and efficient. Specifically the study seeks to explore the possibility of using an Artificial Neural Network model to predict the performance of a student before admitting the student.
1.3 Objectives of the Study
The following are the objectives of this study:
- To examine the use of Artificial Neural Network in predicting students academic performance.
- To examine the mode of operation of Artificial Neural Network.
- To identify other approaches of predicting students academic performance.
1.4 Significance of the Study
This study will educate on the design and implementation of Artificial Neural Network. It will also educate on how Artificial Neural Network can be used in predicting students academic performance.
This research will also serve as a resource base to other scholars and researchers interested in carrying out further research in this field subsequently, if applied will go to an extent to provide new explanation to the topic
1.6 Scope / Limitations of the Study
This study will cover the mode of operation of Artificial Neural Network and how it can be used to predict student academic performance.
Limitation of Study
Insufficient fund tends to impede the efficiency of the researcher in sourcing for the relevant materials, literature or information and in the process of data collection (internet, questionnaire and interview).
The researcher will simultaneously engage in this study with other academic work. This consequently will cut down on the time devoted for the research work.
Summary of Findings and Conclusion
5.1 Summary of Findings
The purpose of this study was to show the applicability and the effectiveness of the ANN approach to the predictive classification of students in the full range of academic performance (GPA), as well as to identify and understand the importance of the variables for each level (low, middle and high) of expected GPA. This methodology, using a predictive system, was chosen as it is very effective under conditions of very complex and great amount of data, in which a large number of variables interact in various complex and not very well understood patterns.
The results attained in this study have allowed the identification of the specific influence of each input set of variables on different levels of academic performance (high and low performance), on one hand, and common processes across all students, on the other hand. One important contribution of this predictive approach is the finding that the same variables have different effects in each group of students, defining specific patterns for each performance level. Although the contribution of each variable in a particular pattern carries a relatively small predictive weight, it is the combined effect of the pattern of variables which explains a lower or higher academic performance model.
Among the student group with the lowest 33% of academic performance, two main predictors are learning strategies components (cognitive resources/cognitive processing and time management). The importance of learning strategies as a mediating factor in a model predicting academic performance has been shown in different studies (Dupeyrat & Marine, 2005; Fenollar, et al., 2007; Simons et al., 2004; Weinstein & Mayer, 1986; Weinstein et al., 1987; Weinstein et al., 1982). However, this study added the contribution of a complex pattern of variables for a particular group of students, identifying specific learning strategies that help the classification of students in a low performance group (i.e., thoughts or behaviours that help to use imagery, verbal elaboration, organization strategies, and reasoning skills). Included in this set are learning strategies that help build bridges between what they already know, and what they are trying to learn and remember (i.e., knowledge acquisition, retention, and future application). In addition, variables related to speed of processing involved in WMC functioning have an important predictive weight for the determination and modelling of the low performance group. Other studies that have used ANN have also found that basic cognitive processing variables such as WMC and Executive Attention carried the most predictive weight in the low performance group of students (Kyndt et al., 2012, submitted; Musso & Cascallar, 2009a; Musso et al., 2012). Moreover, the literature has indicated the positive association between WMC and academic achievement (Gathercole, Pickering, Knight, & Stegmann, 2004; Riding, Grimley, Dahraei, & Banner, 2003). Regarding the relative importance of each variable, if we compare the relative role of WMC and other cognitive resources between the low and high performance groups, WMC and cognitive resources were far more important for lower GPA students. The fact that their importance for the prediction is much greater for the lower performing group is greatly due to the fact that all members of the high group had higher levels of WMC and cognitive resources, therefore not providing the necessary information to the network. On the other hand, it was an identifying characteristic of the low performing group which had consistently lower values of WMC and cognitive resources. Remediation programmes, tutorial systems and instruction methods should consider these specific learning strategies, cognitive processing characteristics and WMC resources, in order to provide basic support to students at risk. Such informed interventions would improve the possibilities of successful academic achievement for the at-risk groups, including those with particular learning difficulties.
Background variables together with reaction time measures and attentional executive control are the most important predictors for the highest academic performance group, as indicators of both efficiency in the processing and of adequate selection of information. Social background variables, such as educational level of the parents, have been found to be significant in a previous ANN study (Pinninghoff, Junemann et al., 2007), and these results have been replicated in this study. The executive control mechanism is responsible for resolving conflicts among responses (Fan et al., 2002). This attentional system has been closely related to working memory capacity (Redick & Engle, 2006), and was found to mediate and compensate WMC deficits for certain tasks (Musso et al., 2012). Other attentional networks seem to be much less discriminating among students who reach certain threshold levels needed for high academic performance. These findings have
significant implications in the way that the learning process can be addressed for students identified as potential high achievers. For this group, promoting learning through the use of metacognitive strategies, complex processing, and targeted teacher feedback would be an important way of maximizing their potential performance.
Regarding methodological implications, these results demonstrate the greater accuracy of the ANN approach compared to other traditional methods such as DA. Other studies have also made use of multilayer perceptron artificial neural networks, with positive results for the analysis of educational data (Abu Naser, 2012; Croy et al., 2008; Fong, et al., 2009; Kanakana, & Olanrewaju, 2011; Mukta & Usha, 2009; Ramaswami & Bhaskaran, 2010; Zambrano Matamala, et al., 2011). However, the present study has been able to maximize the precision obtained in the predictive classification of overall academic performance through the careful adjustment of network parameters and algorithms, producing highly accurate results with minimal misclassifications.
Similarly, the initial study of the correlation between the ANN probabilities of performance level assigned to each individual student, with the actual GPA observed, shows a significant degree of correlation between the two measures (r = .86 for the whole sample), with performance as a continuous variable. Further studies will refine the technique to maximize these individual results.
The results of the DA confirm the lack of significant linear relationships between the independent variables analysed in this study and academic performance. Neural network models have an important advantage in this respect, as they are able to model nonlinear and complex relationships among variables with greater precision and accuracy. Even though the assumptions required for traditional statistical predictive models (e.g. equality of covariance matrices) were not violated for the three stepwise discriminant analyses that were performed, the amount of variance explained was low in all three DA analyses. None of these analyses were able to discriminate with sufficient accuracy between the different levels of expected academic performance. When we compare these results with the ANNs modelled in this study, it can be concluded that ANNs are much more robust, and perform significantly better than other classical techniques, as prior studies have also indicated (Everson et al., 1994; Marquez et al., 1991).
This study has shown the power of this predictive approach using ANNs to model future overall academic performance in higher education, specifically in academic admissions and/or placement. To put the current results in perspective, if we consider one of the best known and most reliable tests currently in use, the SAT from The College Board, it has been found (Kobrin, Patterson, Shaw, Mattern, & Barbuti, 2008) that all sections of the SAT taken together, even with the more recent addition of a writing score, can predict at best 28% of the variance of the first-year college GPA for the average population of students. If we add to the SAT results the information of the GPA obtained in secondary education, the overall prediction is of only 38% of the variance of first-year college GPA (Kobrin et al, 2008). With the current ANN models, it has been possible to correctly classify 100% of student performance in the categories examined, that is, 100% of the students were correctly classified, and our research currently continues into the development of new predictive models, with much larger data sets, to classify students in much narrower bands of expected performance having already attained 98-99% accuracy in models for quintals of student performance distributions. In addition, work will also continue for the prediction of specific expected GPA results for each individual student.
In conclusion, the current predictive systems approach facilitates and maximizes the identification of those factors (or predictors) of the learning processes which participate in varying degrees in the modelling of different levels of performance in academic outcomes in higher education. If we can identify specific profiles of students, focusing on the most important variables, this opens major possibilities for the improvement of assessment procedures and the planning of pre-emptive interventions. Given that this methodology allows for the accurate prediction of actual academic performance at least one academic year in advance to it actually being measured (GPA), it has implications for the application of these methods in educational research and in the implementation of-warning” diagnostic programmes“ early settings. These results also inform cognitive theory and help in the development of improved automated tutoring and learning systems. Although some of the variables involved, such as educational level of the parents, are impossible to alter in their effects on academic performance at the time of the assessment, they do inform policy and indicate the weight that many social and environmental factors influence future academic performance. This methodological and conceptual approach allows us to consider a large number of variables simultaneously and select those which are most relevant and allow a greater degree of intervention to improve student performance, including early intervention programmes for students in need of special support.
The capacity to very accurately classify expected student performance, which is also what tests attempt to do, without the performance sampling issues of traditional testing, and using a much broader spectrum of all factors influencing a student‟s ov methodology. In fact, it also represents a more valid approach to educational assessment due to its overall accuracy and the breadth of the constructs considered to classify the expected performance. Traditional assessments are not sufficient for more complex assessments or for assessment systems that intend to serve multiple direct and indirect purposes, in complex educational situations (Mislevy, 2013; Mislevy, Steinberg, & Almond, 2003) In this respect, this new approach allows for the conceptualization and development of new modes of assessment which could facilitate breaking away from traditional forms of testing while at the same time improving the quality of the assessment process (Segers, Dochy & Cascallar, 2003).
Finally, the use of ANN together with other methods as cluster analyses and Kohonen networks could contribute to the study of the specific patterns of those variables which influence the learning process for each level of performance. In fact, a major observation resulting from the data in this study is that variables contribute to the prediction in relatively small proportions, and it is the joint effect of many contributing variables that could cause significant changes in performance. In other words, there is no “magic rather the accumulation of effects from all these various sources that produces significant changes in outcomes. These results provide an insight into learning questions from a different perspective and one that has important implications for educational policy and education at large.
The Complete Material Will Be Sent to You in Just 2 Steps
Quick & Simple…
Make Payment (Through Transfer) of ₦5,000 to Any of the Account Below
|Acc No: 0811003731
|Acc No: 1225513212
Or CLICK HERE To Pay With Debit Card
|FOR STUDENTS OUTSIDE NIGERIA
|CLICK HERE To Purchase Material ($15)
|FOR GHANIAN STUDENTS
|Make Payment of 120 GHS to 0553978005 | Douglas Cloud Osabutey | MTN MoMo
Send the Following Details on WhatsApp ( 08143831497) After Payment
- Payment Details
- TOPIC: Predicting Students Academic Performance Using Artificial Neural Network
The Complete Material Will Be Sent To You On WhatsApp After Receiving Your Details
T & C Apply