Research On Adopting Data Mining And Machine Learning In E-Learning Towards Personalization And Security

Recent progress in technology has altered the learning behavior of students.In the field of elearning,a significant amount of data is generated and available on the web continuously.More sophisticated methods are therefore required to benchmark the educational data to obtain intrinsic insights.Personalized e-learning schemes are effective tools for individual learning.Students can benefit from digital resources through the wireless network to access educational data.In this thesis,we explore the literature for personalized,smart,and secure e-learning mechanism.We identified the valid problems in existing dominating solutions and then presented the proposed solutions to mitigate the issues while developing a personalized and secure E-learning system.Various clustering methods have been explored including Mean-shift clustering,K-means,K-medoids,Density Based Spatial Clustering of Applications with Noise(DBSCAN)and Hierarchical clustering.However,these approaches are not robust in identifying significant clusters in ambiguous and noisy datasets.Existing clustering approaches are dependent on the number of clusters that should be known prior to clustering the educational data.The educational data is also exchanged among students,teachers and examiners that should be secured to avoid cheating and forgery.We have identified that more robust results can be achieved by the replacement of existing methods with CFSFDP-HD.To resolve this problem,a clustering approach to data mining is suggested and incorporated with the personalized e-learning framework.The integration of data mining approaches makes the learning system more interesting.We present a clustering approach to partition students into different groups or clusters based on their learning behavior.Furthermore,personalized e-learning system architecture is presented which detects and responds to teaching contents according to the students' learning capabilities.Moreover,the administration can find essential hidden patterns to bring the effective reforms in the existing system.Our results show that the recommended approach has formed accurate clusters and took less execution time as compared with some of the existing approaches.To standardize education,the quality of teachers is to be enhanced but in Pakistan quantity of the teachers is also enumerated along with quality.The owners of schools have an excellent chance to remain competitive in their sectors if we accurately predict working teachers.We used machine learning to improve e-learning.We predict the future annual enrollment of teachers using the SVM model(support vector machine)and employ convolutional neural network(CNN)to predict the future need of total working teachers in schools and colleges of Pakistan.Time series data related to different types of institutes,like primary,middle,and high schools,inter and degree colleges were collected from National Bureau of Statistics of Pakistan.The results have shown good accuracy and outcomes suggested to increase the working staff.Our proposed method is applicable for developing countries.Finally,we propose secure E-learning system for secure data exchange for exam materials,aptitude tests and quizzes.It involves the multi-party credentials from the students to generate a secret code at trusted server.The code is shared with all students to generate session keys.The system also presents a novel protocol for session key establishment.Finally,we perform a security analysis to verify the advantages of proposed scheme with secure E-learning using cloud and Fo G servers.For the validation of work,a testbed is developed to evaluate the performance of SES in terms of fraction of untrusted students,fraction of exam exposed,and number of students from controlled and un-controlled locations,student interaction time per hour,and reputation and trust levels for students.