Thesis On Students Performance Predictor using Multi Channel Classifier

Abstract

Abstract newlineData mining refers to the set of techniques to derive hidden patterns from the large existing data. These patterns can be useful for the analysis and prediction purpose. Education data mining refers to the set of data mining applications in the education field. These applications refer to the analysis of students and teachers data. The analysis could be for the purpose of classification or prediction. In today s competitive world, it is essential for an institute to predict the performance of students, classify according to their skills and try to improve their performances in future examinations. Students could be informed well in advance to focus in a particular direction for the betterment of their academic performances. Such analysis helps an institute to reduce failure rates. This research work predicts students performances in a course, based on their previous performances in related courses. Association rule mining is used to find out a set of related subjects which are correlated with each other Performance of one subject may depend on the performance of other subjects. Along with Association rule mining, feedback of students and expertise of teachers helped in finalizing a list of correlated subjects. Students database is designed to cover assessment based and skill based information. Based on this information, Students performances are predicted using a multi channel classifier which uses classification algorithms like decision tree and naive bayes. As students results are extremely difficult to predict because of a lot of parameters and as no classification could be accurate all the time, Classification smoothing algorithm is introduced to select one of the most appropriate classified performances from the set of available predictions. This research work has been tested for a wide database of students of Bachelor of Computer Applications Saurashtra University. The prediction results are tested with the available data to measure accuracy. The prediction results have received wide accept