First, the basic information
Title: "Genetic Algorithm Based Intelligent Grouping System Improvement"
Time: 2013
Source: University of Electronic Science and Technology Master Degree Thesis
Keywords: intelligent test paper, genetic algorithm, convergence, operator, fitness function
Second, research
Content: Based on the analysis of the test paper problem constraints, the establishment of a mathematical model matrix paper, paper puts forward various properties index coding method and fitness can be stored, it can adapt better to avoid double counting value, at the same time, in order to avoid the decoding process and improve operational efficiency, converting the questions into genetic algorithm. To avoid conflicts of volumes knowledge in the process, improved genetic algorithm, a multi-point strategy for the hybrid segment mutation operator of multi-point mutation strategy and a segmented for Crossover. In order to maintain the diversity of population, this paper also proposes an initial population approach paper to knowledge as the fundamental basis points, resulting answer time, questions distributed, knowledge and other basic requirements. The method uses a large ratio crossover and mutation, the genetic algorithm convergence speed can be increased, thereby increasing the speed of the test paper. In the paper the method described above, based on the design and implementation of an Intelligent Grouping System, through verification experiment, the algorithm of this paper has been good to meet the targets in terms of examination papers.
Ideas:
Three: Summary:
First, based on the modern test theory in-depth analysis of the papers as well as the constraints of the questions in-depth discussions, based on the core attributes and limitations of paper and paper targets, the establishment of a new paper matrix math model. Improved genetic algorithm, the coding method of coding a matrix into the improvement of the coding attributes may store various parameters and adaptive value consisting of individual papers, the operation speed can be improved. For improved genetic operators, proposed for segmentation mutation operator of multi-point mutation strategy, and the large variation in the use of hybrid ratio to maintain the diversity of population, effectively avoid conflict knowledge test paper process, while avoiding the search space reduced rapidly accelerate the speed of the algorithm approximation to the global optimum value, improved performance of global optimization, while, in order to ensure the best individual in the parent into the progeny, when copying, using a selection method based on sorting fitness the process of genetic selection for future generations as elitist strategy.