基于改进狮群算法优化神经网络的糖尿病风险预测
An Optimized Neural Network Prediction for Diabetes Risk Based on Improved Lion Swarm Algorithm

糖尿病风险评估预测有助于早期发现糖尿病，降低发病率和并发症。针对糖尿病风险预测问题，提出一种基于改进狮群算法优化神经网络的糖尿病风险预测模型。引入非线性扰动因子改进狮群算法，使得算法既能加强全局优化能力，避免陷入局部最优，又能保证局部优化能力，提高算法的收敛速度。利用改进狮群算法(ILSO)的寻优能力优化神经网络的权重和偏置参数，建立基于ILSO-BP神经网络的预测模型。同时，采用少类样本合成过采样技术和递归特征消除方法对糖尿病数据进行预处理，提高模型预测能力。在真实糖尿病数据集PIMA上的实验结果表明，基于ILSO-BP神经网络的糖尿病风险预测模型，其预测性能优于基线模型，也优于基于遗传算法、鲸鱼优化、粒子群优化等算法优化的神经网络预测模型，对糖尿病风险具有良好预测能力，能够对糖尿病早期筛查起到辅助作用。
Diabetes risk prediction helps to detect diabetes early and reduce the incidence and complications. To address the problem of diabetes risk prediction, a diabetes risk prediction of the neural network based on improved lion swarm optimization (ILSO) algorithm is proposed. A nonlinear perturbation factor is introduced to improve the lion swarm algorithm, so that the algorithm enhances the global search capability to avoid falling into local optimum, and enhances the local search capability to provide convergence speed. The weights and bias parameters of the neural network are optimized by using the optimization ability of the Improved Lion Swarm optimization algorithm (ILSO), and a prediction model based on the ILSO-BP neural network is developed. Meanwhile, a synthetic minority oversampling technique and recursive feature elimination are employed for pre-processing diabetes data to enhance the model prediction capability. The experimental results on the real diabetes dataset PIMA show that the diabetes risk prediction based on ILSO-BP neural network has better prediction performance than the baseline models, and better than the neural network model based on genetic algorithm, whale optimization and particle swarm optimization. The proposed model has good prediction ability for diabetes risk and can play an auxiliary role in early diabetes screening.