Applied Binary Particle Swarm Optimize Algorithm To Test-Sheet Composition

Test-sheet quality is the key to Online Examination System (OES), thus test-sheetcomposition algorithm plays an important role in OES. Whether the test-sheet generatedby test-sheet composition algorithms is reasonable or scientific directly impact onstudents’ learning and schools’ teaching quality evaluation, further influence nextsemester‘s teaching plans and even affect the entire teaching reforms. So the test-sheetcomposition algorithm has a very important practical significance.According to specified parameters, such as item types, difficulty degree, concepts,answer time, score, etc. test-sheet composition algorithm extracts certain items fromitem bank to generate a test-sheet. It is a typical multiple-objective optimizationproblem. There is no algorithm that can be a mature perfect solution for test-sheetcomposition problem. Some algorithms have strong global search capability but poorlocal search ability and some have poor local search ability but strong global searchcapability, while some algorithms steps are more complex, have more iterations; somealgorithms’ principles and mechanisms are simpler, but there is certain subjectivity inthe parameter settings of these algorithms on the current status of research. We analyzethe advantages and disadvantages of existing test-sheet composition algorithms. On thebasic of that, we adopt a BPSO based on Bayes formula to solve the test-sheetcomposition problem. Comparing to other computational intelligence algorithms, thismethod has no crossover and mutation operator, thus its mechanism is simpler and theprogramming is easier. While comparing to standard PSO, this method gets the nextgeneration’s population with the help of Bayes formula more directly, avoids the defectsin standard PSO of calculating speed formula first, then transforming calculation resultsaccording to related rules. The amount of computation is reduced. The actual timecomplexity of algorithm mentioned above is O(n). Comparing with other CI algorithmbased on crossover, mutation and selection operator, its time complexity is improved.This plays a very important role in practical application of promoting. On the basis of the research above, we implement an Intelligent Online ExamSystem (IOES) using the improved algorithm as a test-sheet composition strategy. Then,overall system framework and UML modeling diagrams are given; data table structureis designed with PowerDesigner. In addition, we give a.NET-based browser, which hasa deep customization for students’ desktop browser. There are some special features,such as hiding the address bar completely, shielding right click and various keyboardshortcuts, etc. As a strong complement to IOES, the browser gives an effective solutionfor solving cheating problems to ensure exam’s real fair. At the end, we give anevaluation of IOES. Pbestand gbesthava high credibility in particle swarm optimizationalgorithm. The particles make decision depends on the value of pbestand gbestfor eachiteration. Pbestand gbestcontain the genetic information of the particles in the previousiteration. It directly calculates the value of xiin next generation by pbestand gbest. Usingthe Bayes formula, it can constitute a basis for decision-making in the particles’ positionvalue of next generation.