改进鱼群算法优化的ELM在乳腺肿瘤辅助诊断中的应用研究

发布时间：2018-11-06 14:32

【摘要】：针对传统极限学习机的输入权值矩阵和隐含层偏差是随机给定进而可能会导致在乳腺肿瘤的辅助诊断应用研究中存在精度明显不足的情况,提出用改进鱼群算法优化ELM方法。在完成对乳腺肿瘤有效的辅助诊断的过程中,本研究工作充分利用ELM能快速地完成训练过程且具有很好的泛化能力的特点,并结合用改进鱼群算法对ELM的隐含层偏差进行优化,构造出了乳腺肿瘤与从乳腺肿瘤样本数据中提取的10个特征向量之间的非线性映射关系。将本文提出的乳腺肿瘤识别方法的仿真结果与AFSA-ELM方法、ELM方法、LVQ方法、BP方法的仿真结果分别从识别准确率、假阴性率、学习速度三个方面做对比分析,仿真结果表明,本文所提方法对乳腺肿瘤诊断具有较高的分类识别准确率、假阴性率以及较快的学习速率。
[Abstract]:In view of the fact that the input weight matrix and hidden layer deviations of traditional LLMs are given randomly and may lead to obvious inaccuracy in the application of breast tumor diagnosis, an improved fish swarm algorithm is proposed to optimize the ELM method. In the process of effective assistant diagnosis of breast tumor, this study makes full use of the characteristic that ELM can complete the training process quickly and has good generalization ability, and optimizes the hidden layer deviation of ELM by using improved fish swarm algorithm. The nonlinear mapping relationship between breast tumor and 10 feature vectors extracted from breast tumor sample data is constructed. The simulation results of breast tumor recognition method proposed in this paper are compared with those of AFSA-ELM method, LVQ method and BP method from three aspects of recognition accuracy, false negative rate and learning speed, respectively. The simulation results show that, The method proposed in this paper has high classification and recognition accuracy, false negative rate and fast learning rate for the diagnosis of breast tumors.
【作者单位】：安徽理工大学计算机科学与工程学院;
【基金】：国家自然科学基金(5574257) 安徽理工大学中青年骨干教师
【分类号】：R737.9;TP18

【相似文献】