Intelligent Ensemble T-S Fuzzy Neural Networks with RCDPSO_DM Optimization for Effective Handling of Complex Clinical Pathway Variances.
Gang Du,Zhibin Jiang,Xiaodi Diao,Yang Yao
DOI: https://doi.org/10.1016/j.compbiomed.2013.02.007
IF: 7.7
2013-01-01
Computers in Biology and Medicine
Abstract:Takagi–Sugeno (T–S) fuzzy neural networks (FNNs) can be used to handle complex, fuzzy, uncertain clinical pathway (CP) variances. However, there are many drawbacks, such as slow training rate, propensity to become trapped in a local minimum and poor ability to perform a global search. In order to improve overall performance of variance handling by T–S FNNs, a new CP variance handling method is proposed in this study. It is based on random cooperative decomposing particle swarm optimization with double mutation mechanism (RCDPSO_DM) for T–S FNNs. Moreover, the proposed integrated learning algorithm, combining the RCDPSO_DM algorithm with a Kalman filtering algorithm, is applied to optimize antecedent and consequent parameters of constructed T–S FNNs. Then, a multi-swarm cooperative immigrating particle swarm algorithm ensemble method is used for intelligent ensemble T–S FNNs with RCDPSO_DM optimization to further improve stability and accuracy of CP variance handling. Finally, two case studies on liver and kidney poisoning variances in osteosarcoma preoperative chemotherapy are used to validate the proposed method. The result demonstrates that intelligent ensemble T–S FNNs based on the RCDPSO_DM achieves superior performances, in terms of stability, efficiency, precision and generalizability, over PSO ensemble of all T–S FNNs with RCDPSO_DM optimization, single T–S FNNs with RCDPSO_DM optimization, standard T–S FNNs, standard Mamdani FNNs and T–S FNNs based on other algorithms (cooperative particle swarm optimization and particle swarm optimization) for CP variance handling. Therefore, it makes CP variance handling more effective.