Analyzing students behavioral Patterns in learning process using Random swap expectation maximization Algorithm

Abstract

Extracting necessary patterns from huge data source for analyzing newlineknowledge and making it useful to the society are most difficult task in newlineknowledge extraction. The progression carried out due to the factor of data newlinemining influencing makes the knowledge extraction process in simplified newlinemanner. The knowledge extracted should be most likely to solve some of issues newlinein feasibility, scalability and effectiveness in usage of data. The basic stages newlinefollowed in data mining are preprocessing, feature extraction, enhancement, newlinetransformation and analyzing. The process of preprocessing techniques used to newlineclean the irrelevant information from collected data. The irrelevancy in data newlineusually compared with the outcome of the research work for the efficiency of newlinethe techniques and algorithms. The feature extraction process usually carried newlineout after cleaning stage of data mining process. The features mostly taken for newlinefine-tuning the research work mostly avoided in many researches. The newlinenecessary features selected in feature selection process with the use of relevant newlinerules in implementation process. newlineThe arrangement of extracted features are very essential for analyzing newlinestage in data mining technique. The clustering and classification algorithms newlineused for cataloging the proposed algorithm in this work. The decision-making newlineconsidered the most important part in every research analysis. The used newlinealgorithms might solve some of serious issues faced while analyzing applied newlinetechnique into the system through research strategy. newlineThe educational data mining (EDM) is an application of data mining newlinefollows in understanding behavioral patterns followed by the students in newlinelearning process. Educational data mining uses learners and instructors process newlinepatterns for understanding behavior. The EDM general idea is to follow the newlinexi newlinevarious stages of data mining starting from preprocessing to analyzing stage in newlinestudents records. This research work carried out by using collected data set newlinefrom various colleges located in different geographical locations. The data newlinecollected from different locations purely made with the help of questionnaires newlineframed based on leaners and instructor s behavior in learning and teaching newlineaspects. Learning systematic research (LSR) and Educational data mining are newlinebasic areas helps in finding behavioral patterns in learning and teaching as well newlineas those areas used for academicians to understand the capability of newlinemethodology implemented in learning and teaching. The predictions made after newlinethe EDM research may be useful for institutions to convert the teaching newlinemethodology as well as to understand the difficulties faced by the learners in newlineunderstanding the concept. newlineThe modern educational system faces practical difficulties in analyzing newlinethe students learning aspects based on the background of the students and newlinelearning methodologies followed by the students. This research work makes an newlineinitiative process in evaluating the learning strategy carried out by the student newlineand analyzing the impact of locality of the student while considering the newlinelearning process. The work focuses on developing an evolutionary Genetic newlineAlgorithm (GA) based approach and proposed an algorithm named as Random newlineSwap Expectation Maximization (RSEM) for understanding the behavioral newlinepatterns followed in learning process based on the locality of the learner. newlineThe results of the proposed Random Swap Expectation Maximization newlinetechnique is compared with the existing clustering algorithms such as k Means, newlinek Mediods, Fuzzy C Means and Expectation Maximization for finding the newlineaccuracy in analyzing the student s behavioral patterns in learning process. The newlinevarious algorithms measured with various evolutionary process such as newlinePrinciple component, Chi Squared Test, Filtered attribute Evaluation, newlinexii newlineCorrelation-Based Feature, Gain attribute Evaluation and Relief attribute newlineEvaluation along with proposed Random Swap Expectation Maximization. The newlinefindings and various testing process carried out in this research work clearly newlineshows that the Mean and Variance for proposed Random Swap Expectation newlineMaximization technique is better compared with other clustering algorithms. newlineFinally, this work suggests the method which suites for analyzing student data newlineset by means of its efficiency and accuracy. newline