基于量子粒子群智能算法的模糊逻辑系统的设计及应用

发布时间：2018-05-21 19:47

本文选题：TSK模糊逻辑系统 + 神经网络　；参考：《辽宁工业大学》2017年硕士论文

【摘要】：模糊逻辑系统由于其应用广泛,已成为学术和应用领域的研究热点。模糊系统辨识包括结构辨识和参数辨识两个方面,针对参数辨识,目前常采用最小二乘法、BP算法等,虽然方法很多,但都不是智能算法。本文将一型、A1-C1、A2-C0、A2-C1(A—规则前件,C—规则后件)区间二型TSK模糊逻辑系统融入到神经网络中,设计出四种模糊神经网络系统,采用量子粒子群(QPSO)智能算法调整模糊逻辑系统的参数,将所设计的智能系统应用于国际黄金价格和纳斯达克综合指数的预测中,并给出仿真研究。具体工作如下:(1)介绍了TSK模糊逻辑系统、神经网络以及QPSO算法的相关知识。(2)研究基于QPSO算法的一型TSK模糊逻辑系统的设计。将模糊逻辑系统融入到神经网络中形成五层模糊神经网络系统,先用QPSO算法筛选规则,再用QPSO算法优化系统参数,最后将所设计的智能系统模型应用到国际黄金价格和纳斯达克综合指数的预测问题中,并与BP算法进行比较。仿真结果表明,所设计的一型智能系统是可行的、有效的。(3)在一型TSK模糊逻辑系统的基础上,前件不变,后件变为一型模糊集,得到A1-C1型区间二型TSK模糊逻辑系统。将模糊逻辑系统融入到神经网络中形成六层模糊神经网络系统,先用QPSO算法筛选规则,再用QPSO算法优化系统参数,最后将所设计的智能系统模型应用到纳斯达克综合指数的预测问题中,并与BP算法进行比较。仿真结果表明,所设计的A1-C1型智能系统是可行的、有效的。(4)在一型TSK模糊逻辑系统的基础上,后件不变,前件变为区间二型模糊集,得到A2-C0型区间二型TSK模糊逻辑系统。将模糊逻辑系统融入到神经网络中形成五层模糊神经网络系统,先用QPSO算法筛选规则,再用QPSO算法优化系统参数,最后将所设计的智能系统模型应用到纳斯达克综合指数的预测问题中,并与BP算法进行比较。仿真结果表明,所设计的A2-C0型智能系统是可行的、有效的。(5)结合(3)和(4),前件为二型模糊集,后件为一型模糊集,得到A2-C1型区间二型TSK模糊逻辑系统。将模糊逻辑系统融入到神经网络中形成六层模糊神经网络系统,先用QPSO算法筛选规则,再用QPSO算法优化系统参数,最后将所设计的智能系统模型应用到纳斯达克综合指数的预测问题中,并与BP算法进行比较。仿真结果表明,所设计的A2-C1型智能系统是可行的、有效的。对比四种TSK模糊逻辑系统,由跟踪效果图及均方根误差可知:区间二型模糊逻辑系统比一型模糊逻辑系统的误差相对较小,具有更好的精度。
[Abstract]:Because of its wide application, fuzzy logic system has become a research hotspot in academic and application fields. Fuzzy system identification includes structure identification and parameter identification. The least square BP algorithm is often used in parameter identification. Although there are many methods, they are not intelligent algorithms. In this paper, the interval 2 TSK fuzzy logic system is integrated into the neural network. Four kinds of fuzzy neural network systems are designed, and the parameters of the fuzzy logic system are adjusted by using the Quantum Particle Swarm Optimization (QPSO) intelligent algorithm. The designed intelligent system is applied to the prediction of international gold price and NASDAQ composite index, and the simulation research is given. This paper introduces the TSK fuzzy logic system, neural network and the related knowledge of QPSO algorithm. The design of a type of TSK fuzzy logic system based on QPSO algorithm is studied. The fuzzy logic system is integrated into the neural network to form a five-layer fuzzy neural network system. QPSO algorithm is used to filter the rules, and then the QPSO algorithm is used to optimize the system parameters. Finally, the designed intelligent system model is applied to the prediction of international gold price and NASDAQ composite index, and compared with BP algorithm. The simulation results show that the first type of intelligent system is feasible and the effective. On the basis of the first type of TSK fuzzy logic system, the former part does not change, the latter part becomes a type of fuzzy set, and the A1-C1 type interval second type TSK fuzzy logic system is obtained. The fuzzy logic system is integrated into the neural network to form a six-layer fuzzy neural network system. QPSO algorithm is used to filter the rules, and then the QPSO algorithm is used to optimize the system parameters. Finally, the designed intelligent system model is applied to the prediction of NASDAQ composite index, and compared with BP algorithm. The simulation results show that the designed A1-C1 type intelligent system is feasible and effective. On the basis of the first type of TSK fuzzy logic system, the latter part remains the same, the former part becomes the interval 2 type fuzzy set, and the A2-C0 type interval 2 type TSK fuzzy logic system is obtained. The fuzzy logic system is integrated into the neural network to form a five-layer fuzzy neural network system. QPSO algorithm is used to filter the rules, and then the QPSO algorithm is used to optimize the system parameters. Finally, the designed intelligent system model is applied to the prediction of NASDAQ composite index, and compared with BP algorithm. The simulation results show that the designed A2-C0 intelligent system is feasible, and the effective TSK fuzzy logic system is composed of two types of fuzzy sets and one type of fuzzy sets. The A2-C1 type interval two type TSK fuzzy logic system is obtained. The fuzzy logic system is integrated into the neural network to form a six-layer fuzzy neural network system. QPSO algorithm is used to filter the rules, and then the QPSO algorithm is used to optimize the system parameters. Finally, the designed intelligent system model is applied to the prediction of NASDAQ composite index, and compared with BP algorithm. Simulation results show that the designed A2-C1 intelligent system is feasible and effective. Compared with four TSK fuzzy logic systems, the tracking effect diagram and root mean square error show that the error of interval type 2 fuzzy logic system is smaller than that of type 1 fuzzy logic system, and it has better precision.
【学位授予单位】：辽宁工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O159;TP18

【相似文献】