Research On Intelligent Test Paper Generation Based On Improved Genetic Algorithm

With the information technology and computer network rapid development and increasing popularity in the 21st century, its application field also expand, especially for the application of education. It provide a good platform for the development of education, accelerate the education reform. In the present education, examination is still an important means to measure education quality and the student's grade. It is a problem must be considered that how to adopt an effective examination method, realize the separation, ensure exam fair, achieve the purpose of a test. In this context, intelligent-arranging test paper system research becomes an important research direction, and attracts attention of more and more educators.Test paper construction is a multi-object restrict optimization problem. The core of test paper system is composing test paper algorithm. It's directly related to the composing paper efficency and test paper quality. The traditional composing test paper algorithm such as priority strategy, random strategy, parallel strategy, retrospective strategy and so on, they're time consuming and low successful rate for composing test paper. Generated papers cannot satisfy the needs of practical test. However, the genetic algorithm has the featurea to adaptive global optimization, intellectual search and nice concergence, etc, which can meet the requirement of intellectual paper test composition better. Grouping efficiency and quality of Intelligent-arranging test paper system depends entirely on the design of test-questions-database and composing test paper algorithm. In this thesis, it combine composing test paper algorithm and test-questions-database to extract test questions and form a test bank. It implements teaching-examination separation and makes teachers liberate from cumbersome examination, also can reflect exam fair, reaction was tried in real ability level. While it can reflect exam fair and the rael ability of testers. According to the problems of the prematuriy and the low speed of convergence in classical genetic algorithm are made. It use the group natural number coding strategies in this paper, overcome the weakness of binary code search space too large and length coding too long, avoid the individual decoding oftime, raises the solution speed. Using this algorithm can solve the problem of intelligent-generating test paper and improve speed. While it can avoid prematurity of standard genetic algorithm. The system has good robustness.The development environment of this system is Windows 2003 server+IIS+ PHP+MS SQL server 2005. It adopt the browser/server(B/S) mode framework. This system has friendly interface, easy operation and strong practicability. The experimental results show that the improved genetic algorithm has better ability to avoid prematurity and system speed. It can form high quality test paper to satisfy user's requirement.