The Online-test System Based On The Genetic Algorithm

With the development of computer and network technology, the methodsof test changes greatly. The computerized test system can shorten the circle ofexam workflow, lessen the repeating work of teachers, and improve theefficiency. The intelligent paper forming function it supports can improve thequality of test papers and increase the objectivity of test. At the same time, theweb-based test can be taken at any time and any place. The form of testbecomes more flexible and free.The system is based on B/S-three-tier model and it provides good securityin the fashion of three-tier authorized system including administrator, teacherand student. The administrator is responsible for the system maintenance andcan do manipulation with the system such as adding or deleting content in thetest question database. Teacher is responsible for managing test questiondatabase, producing test papers, evaluating test integratedly and so on. Studentcan use the test system to do test and query grade.The system includes 6 subsystems, test question management subsystem,paper forming subsystem, on-line test subsystem, grade subsystem, paper andgrade analysis subsystem and grade query subsystem. Among them, thefunction of the test question management subsystem is adding, deleting andmodifying the resources of the test questions database. Paper formingsubsystem includes paper forming and paper browse. The user can form paperseither by hand or by algorithm intelligently. To form paper by hand means tochoose the test questions which meet the requirement of the test to form a testpaper. The user can search the test questions either by type or by key words,and the system is able to save many pieces of papers, which the user can lookthrough all the contents by hyperlink and choose one as an examination paper.On-line test includes examination and the self-test. In the self-test, studentsmay choose the quantity of the test questions first, and in the end of the test,the system will give the students' grade and the simple paper analysisautomatically. Examination paper is the production of the paper forming. Theexamination has time restriction, and the system will display the surplus timedynamically on the screen, when time is over, the paper will be submittedautomatically. During the process of the examination, the student's status andanswer will be saved into the database, so that when accident happens, theexamination will resume according to the database information. For theobjective questions, the system will compare the students' answers to thestandard answers, and give the score of every student automatically. Thefunction of the grade subsystem is to provide the subjective questions whichhave no scores to the teachers, the teachers will give the scores manually, afterall the scores are given, the system will gather the scores, that is to gather thescores of the subjective questions and objective questions together and putthem into the database. The paper and grade analysis subsystem is based onClassics Test Theory(CTT), and evaluate the test integratedly and scientificly.Further more, the other function of it is to input the scores of a test, that is aslong as we input the scores of the students, the system can evaluate any testpapers. The function of the grade query subsystem is querying the scores ofthe student themselves. I mainly study the algorithm in intelligent paperforming and the test theories.The essential of the computerize intelligent paper forming is to manage aoptimization test paper that satisfies the parameters which the user inputs, sothat the paper forming problem is a classical Constraint SatisfactionProblem(CSP). The genetic Algorithm is in favor of solving the CSP for itswhole searching, randomicity, well coding method and high parallel, so thesystem adopt the Genetic Algorithm to manage the paper forming problem.The Genetic Algorithm is a new kind of whole optimization algorithm,and it uses the biologic genetics's point of view to raise the individul'sadaptability by the natural selection, heredity, mutation.There are 5 elements in the Genetic Algorithm—parameter coding, theinitial population forming, adaptability function design, genetic operationdesign and the control parameter enactment. Thereinto, the genetic operationincludes selection, crossover and mutation.In this dissertation, I convert part of the restricting condition of paperforming to objective function, and I adopt binary coding. In the geneticoperation, to ensure the new afterworld still satisfies the constraint, I adopt onepoint crossover, which the whole type of questions exchanges, and the twopoints mutation to implement.I have improved upon the traditional Genetic Algorithm. I adopt elitereservation strategy, which reserves the best chromosome in this generation toenter the next generation, and other individuals are chosen by the diskselection method. In this way, I can ensure the excellent individuals get morebreeding opportunity, and can also prevent the few excellent individualsfrom controling the whole population.Contraposing the so-called deceptive problems, I adopt the zoom fitnessto solve. The problems are as follows:(1) In the early period of the GA's evolution, there are usually someexceptional excellent individuals, which control the selectionprocess for the outstanding competition.(2) In the late period of the GA's evolution, because the differenceamong the individuals' fitness in the population is small, it is hardto continue to optimize, and it will get the partial optimizationsolve.Here, I adopt a linear scaling method to transform the fitness as follow..By this zoom technic, reduce(enlarge) the primarymaximal(minimal) fitness to weaken the discrete phenomenon. In the end, Ialso discuss the parameter's influence to the algorithm.The other emphasis of the research is the test theory. For the test questions,the selection of the index relates to the quality of the test paper, so when theitem bank is built, there must be theories foundation to evaluate the testquestions index. Moreover, the grade evaluating also needs the theoriesfoundation as guidance. In this dissertation, I choose the Classical TestTheory(CTT) as the basic theory.Classical Test Theory was put forward in the early years of the 20 century,and formed into an integrated system in the late 1950s. One of the traits is allthe various kinds of test's parameters must get from the sample of the students.The paper and grade analysis subsystem is the application of it. The content isas follows:(1). Grade statistic, including calculate the average score, square subtract,standard subtract, the excellent rate,the pass rate and the frequence distributedchart. The average score represents the concentrative trend of a set of data, andthe square subtract represents the discrete trend of a set of data, and alsoreflect the polarization situation in fact, furthermore, we can also learnwhether the result is the normal school.(2). Test questions analysis, calculating the difficulty and easiness of thetest questions by the difficulty expression, and put them into the database.(3). Test paper analysis, including four aspect of the test paperanalysis—difficult degree, believable degree, differentiate degree, availabledegree. The differentiate degree represents the students ability of high, low,good, bad. I adopt a high, low 27% grouping method. The believable degreerepresents the reliability of a test, that is the test result whether can reflect thereal level of the test. The available degree represents how the test achieves itspurpose.To sum up, this paper mainly introduce the online-test system based onWeb. it implements functions such as test questions management, paperforming, on-line test, grade, paper analysis and grade query. Then on the basisof investigating two education measurement theory which are ClassicalTesting Theory and Item Response Theory put forward the idea of setting theparameters of test questions and the method of the paper analysis in thedirection of CTT. In order to solve the abuse of the traditional paper forming, Iadopt the improved Genetic Algorithm, which makes the constringency speedquicken and avoid the 'precocious phenomenon'. Experiment proves that ithas good capability and practicability.