What is the level of accuracy can be achieve in predicting student performances ???

With the recent developments in the Internet allowed many leading educational institutions to offer online teaching and learning through Learning Management Systems. Systems with different capabilities and approaches have been developed to deliver online education which makes the communication channel between student and teacher into a much more virtual one.
 

The most important consideration would be to identify the student’s performance accurately and provide the necessary support to the students to improve their knowledge levels. But the main problem of measuring the students in accurately and to classify them correctly in to group to predict their performance levels is a huge challenge.

Behrouz Minaei-Bidgoli, Deborah A. Kashy, Gerd Kortemeyer, William F. Punch [2003] have researched on applying data mining methodologies to classify the students in to different groups and try to predict their performance achievement for the future.

For this research they used Quadratic Bayesian classifier, 1-nearest neighbor (1-NN), k-nearest neighbor (k-NN), Parzen-window, multilayer perceptron (MLP), and Decision Tree as the data mining mechanisms and by combining multiple classifiers they hoped to improve classifier performance. For the decision trees they have used C5.0, CART, QUEST, CRUISE algorithms.

In this paper they focused on using a Genetic Algorithm to optimize a combination of classifiers. They used GAToolBox for MATLAB to implement a Genetic Algorithm to optimize classification performance and to find a population of best weights for every feature vector which minimize the classification error rate.

Finally they concluded that using Genetic Algorithm more than a 10% performance improvement can be achieved and having the information generated the instructor would be able to identify students at risk early.

Reference : B. Minaei-Bidgoli, D. A. Kashy, G. Kortmeyer, and W. F. Punch, “Predicting student performance: an application of data mining methods with an educational web-based system,” in Frontiers in Education, 2003. FIE 2003 33rd Annual, 2003, vol. 1, p. T2A–13.

How we can enhance the performance of the students???

Web based online learning systems are widely used around the world to provide various types of educational programs. With the increase in the networking and the use of Internet resources enables these educational programs to be offered through web based learning management systems and several web based educational systems with various functionalities have been developed to deliver online education more effectively to the students.

 

Behrouz Minaei-Bidgoli, Gerd Kortemeyer, William F. Punch [2004] have tried to research on classifying the students in to different group and predict the final grade that they will achieve in the end of the course modules. This research is based on the web based learning management systems which is known as Learning Online Network with Computer-Assisted Personalized Approach (LON-CAPA) developed at Michigan State University.

For this research they concerned on the three kinds of large data sets available in the LON-CAPA which can be categorized as educational resources, information about users and the activity log databases which log actions taken by students.  

For this study they restricted four different classifiers using the LON-CAPA dataset which were Quadratic Bayesian classifier, 1-nearest neighbor (1-NN), k-nearest neighbor (k-NN), Parzen-window. The approach they used in this research is to combine the classifiers together to increase the performance of the classifiers and this is known as online classification fusion.

Another important aspect they highlighted in this research is the use of Genetic Algorithms as an optimization tool for resetting the parameters in other classifiers. The main idea was to use the Genetic Algorithms to find a population of best weights for every feature vector, which minimize the classification error rate.

As the conclusions they mentioned that a combination of multiple classifiers leads to a significant accuracy improvement in the given data sets. Weighing the features and using a genetic algorithm to minimize the error rate improves the prediction accuracy by at least 10% test cases as well. Finally they suggested that use of these algorithms as tools can be used to identify those students who are at risk in very large classes and it will help the instructors to provide appropriate advising in more effective manner.

Reference: B. Minaei-Bidgoli, G. Kortemeyer, and W. F. Punch, “Enhancing online learning performance: an application of data mining methods,” in The 7th IASTED International Conference on Computers and Advanced Technology in Education (CATE 2004), 2004, pp. 173–178.

Using K-Nearest Neighbor Algorithm on Student Data

The k-nearest neighbor data mining method is a classical prediction method among the machine learning techniques available in data mining. It has been widely used due to its simplicity and adaptability in predicting many different types of data. The main advantage of using KNN in prediction processes is that the KNN is a lazy method which does not require a model to represent the statistics and distribution of the original training data rather it can be apply on the actual instances of the training data. Even though the KNN is a simple predictive algorithm which we can rely on and it does not make any assumption about the prior probabilities of the training dat. Also the KNN is satisfactorily used on the situations when the data set is included with noisy and incomplete data.

Due to the advantages and the simplicity of the algorithm T. Tanner and H. Toivonen [2010] has tried to implement a model using K nearest neighbor data mining algorithm to identify the student who are at high risk of failing in a specific course. By this research they suggested that good results in predicting final scores indicate that students with learning problems can be found reliably. What they have been using on the student data is to make prediction on the performance of a given student based on the similarity to all instances in the training set and find the k most similar objects in the data set. This similarity is calculated by using a Euclidean distance between the features of the test subject and the corresponding features of each instant in the training set.

In their research they showed that KNN can produce predictions accurately for the final scores even after the first lesson. Another interesting result they found is that in any skill based courses early tests on skills can be used as the predictors for the final scores and they suggested that predicting final scores for the courses can be used to identify the students with learning problems and can be used directly to implement as a early warning features for the teachers so that the students can be altered if they are likely to fail the final tests. Based on the information or features they used for the experiment with the KNN algorithm they suggest that KNN method could be just as effective in other learning management systems (LMSs) such as Moodle where only a single lesson score is available for student assessment and especially other skill-based courses could be a good fit for the KNN method.

Reference : T. Tanner and H. Toivonen, “Predicting and preventing student failure–using the k-nearest neighbour method to predict student performance in an online course environment,” International Journal of Learning Technology, vol. 5, no. 4, pp. 356–377, 2010.