| 摘要: |
| 针对现有基因微阵列数据分类中存在的数据维度高、容易发生过拟合的问题，提出了基于BTLBOGSA(Binary TLBOGSA)与卷积神经网络(Convolutional Neural Network，CNN)的基因微阵列数据分类模型(BTLBOGSA-CNN)。该模型首先针对基因微阵列数据分类时存在的数据维度高的问题，利用新的编码策略，将连续搜索空间转换为二元搜索空间，结合教与学优化(Teaching-Learning-Based Optimization，TLBO)算法的二元变体与引力搜索算法(Gravitational Search Algorithm，GSA)的各自特点，基于BTLBOGSA方法从基因微阵列数据集中选择具有高鉴别性的基因；然后针对基因微阵列数据分类易发生过拟合问题的现象，利用卷积神经网络进行基因微阵列数据的分类。利用公开的基因微阵列数据集进行仿真实验，从TLBO算法与GSA结合的有效性、BTLBOGSA与CNN结合的有效性、BTLBOGSA-CNN与其他已有分类模型相比的有效性3个方面进行对比分析，结果表明，BTLBOGSA-CNN模型可以在较少的特征基因下取得较高的分类精度，具有较高的可行性和有效性。 |
| 关键词: 基因微阵列数据分类 教与学优化算法 引力搜索算法 特征选择 卷积神经网络 |
| DOI：10.13656/j.cnki.gxkx.20220526.005 |
| 投稿时间：2021-11-19 |
| 基金项目:广西自然科学基金项目(2021GXNSFBA220080)，广西高等教育本科教学改革工程项目(2021JGA425)，广西中青年教师科研基础能力提升项目(2021KY1736)和国家自然科学基金企业创新发展联合基金(U19B2021)资助。 |
|
| Classification Model of Gene Microarray Data Based on BTLBOGSA and CNN |
|
ZHAO Wanzhao, XIE Cong
|
| (Guangxi Agricultural Vocational and Technical University, Nanning, Guangxi, 530007, China) |
| Abstract: |
| Aiming at the problems of high data dimension and easy overfitting in the existing gene microarray data classification,a gene microarray data classification model (BTLBOGSA-CNN) based on BTLBOGSA (Binary TLBOGSA) and Convolutional Neural Network (CNN) is proposed.Firstly,this model aims at the problem of high data dimension in the classification of gene microarray data and uses a new coding strategy to convert the continuous search space into a binary search space.Combined with the respective characteristics of the binary variants of Teaching Learning Based Optimization (TLBO) and Gravitational Search Algorithm (GSA),the BTLBOGSA method is used to select genes with high discrimination from gene microarray data sets.Then,in view of the phenomenon that the classification of gene microarray data is prone to over-fitting,the convolutional neural network is used to classify the gene microarray data.The simulation experiment was carried out using the public gene microarray data set,and the comparative analysis was carried out from three aspects:The effectiveness of TLBO combined with GSA,the effectiveness of BTLBOGSA combined with CNN,and the effectiveness of BTLBOGSA-CNN compared with other existing classification models.The results showed that the BTLBOGSA-CNN model could achieve higher classification accuracy with fewer characteristic genes,and had high feasibility and effectiveness. |
| Key words: gene microarray data classification teaching-learning-based optimization algorithm gravitational search algorithm feature selection convolutional neural network |
