Study On Immune Optimization Algorithms And Their Applications To Portfolio Selection

Biological immune system as a complex, highly parallel, distributed and adaptive information-processing system can recognize and eliminate intruding antigenic materials, while possessing the ability of learning, memory and adaptive adjustment. It can maintain inner stability of bodies. Artificial immune system, developed through using abundant resources of the immune system for reference, has become important research contents of Artificial Intelligent. It displays its prominent performance and efficiency of solving practical problems through application in many engineering fields.This dissertation, firstly, introduces some basic concepts, elements, functions and principles of the immune system while analyzing simply research contents, main achievements reported, basic theory of Artificial Immune System and basic principles of some immune algorithms. Secondly, an improved niche immune algorithm is designed after analyzing its searching performance. The focus is to improve its antibody mutation and suppression operators while introducing ideas of cell propagation and an operation of keeping excellent individuals according to immune theory. Further, It is applied to a portfolio selection model to be proposed in this dissertation to search for optimal investment decision-making projects. This model to be developed is to maximize portfolio interest, efficiency of using total investment capital, upper-jump range of stock prices while minimizing portfolio risk and exchange capital. In terms of historical data of stocks, the algorithm is applied to the model. Meanwhile, simulation shows that it can not only obtain optimal investment decision-making projects, but also correctly forecast development trend of the best stocks decided. Thus, it is efficient and robust. Finally, analyzing drawback that a reported multiobjective immune algorithm deals with practical problems, i.e., controlling population sizes and searching speed, we propose an improved multiobjective immune algorithm applied to multi-decision-making portfolio selection model. This algorithm is examined the influence of searching speed and parameters to its searching performance while being checked its efficiency and robustness. Simulation demonstrates its reasonability and efficiency for application to investment decision-making behavior for portfolio selection.