Study On The Multi-objective Optimization Method And Applying It Inthe Health Decision

Fast non-dominated sorting genetic algorithm(Fast Non-dominated Sorting Genetic Algorithm, NSGA-II), is a typical multi-objective optimization method, widely used in many engineering fields. On the basis of studying the algorithm detailly, this paper mainly does the following research work to improve the NSGA-II algorithm.Co-evolution strategy based on multi population divides the target space into non overlapping and small regions. Each sub-population search in the corresponding small area. It can improve the search accuracy and accelerate the convergence speed. It can prevent the population converging toward a small region of the target space, so that it can control the whole diversity.Crossing parent selection method based on Pareto non-dominated level and congestion degree will make excellent individual to have more opportunities to be selected as crossing parent. They will pass on their excellent information to the next generation. Compared with the original selection strategy, it can improve the convergence and diversity of the algorithm.Dynamic cross strategy is based on crowding distance. It divides the area into crowded area and sparse area based on crowding distance. The strategy uses SPX crossover which makes parent center as the center for the parents of the crowded area. So that offspring is far away from parents. Does not make the parents around to be more crowded. I use SBX crossover which makes parent as the center for the parents of the sparse area. In order to make the offspring could be close to the parents. Make up the parent around empty area. Compared with the original single SBX crossover, it could ensure the algorithm diversity better and ensure the algorithm adaptability for different functions.In addition, this subject introduces L dominance based on Pareto dominance. When using the Pareto dominance cannot compare the merits of the individuals, this subject presents to compare the merits of the individuals by L dominance. So the elite individuals can be retained more effectively.Above improved algorithm is tested on a test set of functions ZDT and DTLZ, the test results show that the improved algorithm is better than NSGA-II on composite indicator IGD of all the test functions, improved algorithm improves the effectiveness of the algorithm.Improved multi-objective evolutionary algorithm which is proposed in this subject will be applied to health decisions which is a practical application case. The problem needs to be considered a number of indicators, such as diet, calorie diet, exercise, cost, user preferences, etc. These indicators act as the objective functions and constraint functions in optimization modeling. Through the contrast test of health decision system, we can get better health plans by health decision sys tem of using the improved multi-objective optimization method, all the plans are in accordance with the health optimization indicators and we can select different plans according to the actual needs.