Research And Application Of Technologies In Personnel Examination

With the development of education and examination informationization, forms and contents of all kinds of examination are improved continually. Personnel examination is the main way to assess and select the talents. Because the qualities of examinees improve gradually and the contents of examination increase year by year, the methods used now are difficult to meet the demand of personnel assessment, especially the problem about test difficulty prediction and intelligent test paper composition are more outstanding. It is necessary to use new technologies to resolve them for making personnel exam scientific, reasonable and justice.In research on predicting difficulty of test paper, a new method based classification in Data Mining was proposed. Prediction of a qualified score was resolved first in realizing the method, and the ratio of the qualified score and total score made a standard as establishing the difficulty of test paper. The prediction about qualified score need use classification to analysis the data information on examinees in former years. In this paper, we realized the whole classification process about predicting the qualified score by C4.5 and BP networks with the economist qualification exam as an example, and predicted approximately the difficulty scope of 2005's test paper. On the basis of the realized model, the paper compared the efficiency of decision tree with artificial neural networks, and the model about predicting difficulty of test paper realized by decision tree was proved more reasonable and feasible.In research on composing test paper, the test model structure which difficulty coefficient is suitable for the demand of test controlled objectivity by analyzing the created test aim was built and realized by Particle Swarm Optimization (PSO). Using particle swarm optimization for composing test paper is a discrete solution problem, in which some particles were initiated, and then iterated them until an optimization solution was obtained in limited evolution. It is a new kind of extension for PSO application in composing test paper. On the basis of the realized model, the rapid convergence and high efficiency of PSO were proved.