Social Emotions Optimization Algorithm Based On Small World Model And Applied Research

Social emotional optimization algorithm (SEOA) is a novel population-based stochasticalgorithm inspired by human decision behavior, in which the emotion of each individual isintroduced to guide the decision. However, in the standard version, all individuals areconnected, and the complex network characters are omitted. Therefore, in this paper, thecomplex characters are employed in SEOA to improve the performance, and applied totraining artificial neural network. The main contributions are listed as follows:(1) Human society is a classical “small-world”, therefore, incorporating this characterinto the SEOA may increase the performance. With this method, we incorporate two type of“small-world”: Watts and Strogtz model (WS model) and Newman and Watts model (NWmodel) into the methodology of SEOA. In both of them, the search process is divided into twoperiods: the firstly one is with the corresponding small-world model, e.g. WS modal and NWmodel, as well as in the late, the original topology is employed to accelerate the convergencespeed. Simulation results show NW model provides a better performance than WS model.(2) Emotion index is one important parameter, however, in the standard version, theemotion index is only dominated by the changes of fitness value of itself, while other changesoccurred in its neighbors are omitted. Therefore, in this paper, we design a new emotionselection strategy in which all changes among its neighbors are considered, and apply it intoWS model and NW model. Simulation results show this strategy increase the performancesignificantly especially for NW model.(3) To further testify the performance, we apply the above mentioned algorithms to trainartificial neural networks, and two famous benchmarks: Cleveland heart disease classificationproblem and forecasting of the sunspot number are selected to compare, simulation resultsshow our modified versions are effective.