Student Evaluation For E-learning

With the rapid development of computer technology, network communicationtechnology, multimedia technology and virtual reality, the E-learning education technologyhas developed fast. Web-based student evaluation which has become the challenges inE-learning system has drawn attention from many researchers. It becomes a new researchpoint. Testing which is an important method of student evaluation has a close relationshipwith student’s real situation evaluation. Tests assembling algorithm that is a corecomponent of the testing system, which has a directly related to the authenticity andaccuracy of student evaluation has attracted researchers’ attention.This dissertation concentrates on studying testing in web-based student evaluation ofE-learning systems, focuses on studying the problem of comprehensive tests andpersonalized assembling algorithm.The main research contents and innovative contributions of this dissertation are asfollows:1) A student evaluation framework which considers object and subject conditions ispresented for solving the problem of the limited flexibility and expansion in web-basedstudent evaluation framework of current E-learning systems. Contents of hard and softevaluation are studied. This framework could be easily expanded and maintains flexibility.For constructing a general learning strategy recommendation system which could be usedin multiple areas, testing results information is used in constructing process. Learningstrategies model, learning strategy-oriented services model and a general learning strategyrecommendation algorithm (GLSR) are proposed. And a learning strategy recommendationsystem (BIT-LSS) is constructed. Experimental results show that BIT-LSS can effectivelyrecommend suitable learning strategy for students.2) For solving multi-subject test sheet generation problem, test sheet generationconstraints in comprehensive tests which include multiple subjects are analyzed. A testsheet generation algorithm based on intelligent genetic algorithm and hierarchical planning(GAHP) is presented. The proposed approach utilizes hierarchical planning to simplify themulti-subject testing problem and adopts genetic algorithm to process the layered criteria,enabling the construction of good test sheets according to multiple test item requirements.GAHP is proved to be more rapid and more convenient in assembling tests for multiple subjects.3) To meet the needs of testing students with more non-mastered knowledge and thesimulation tests could support suitable students’ level, a personalized genetic algorithm(PGA) is proposed to assemble appropriate test sheets for individual students according totheir different personalized information. The proposed approach incorporates personalizedinformation as preference gene bit into the crossover operator of genetic algorithm andfitness function, so as to select more non-mastered questions in the final test. Experimentalresults show that PGA can effectively assemble personalized test sheets for each student.4) For keeping the balance of items in test sheets, the definition of good itemdistribution in knowledge hierarchy is presented. Considering item distribution, animproved algorithm incorporated item distribution (IGAID) is presented to assemblesimulation test sheets which have good item distribution in knowledge hierarchy for eachstudent. Experimental results show that IGAID can is effective in assembling test sheetswith good item distribution and students’learning level.