期刊
JOURNAL OF COMPUTATIONAL SCIENCE
卷 60, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.jocs.2022.101570
关键词
Breast cancer survival prediction; Stacked ensemble method; Artificial Bee Colony; Parallel Bayesian hyperparameter optimization; Multi-omics data
In this paper, a parallel Bayesian hyperparameter optimized Stacked ensemble (BSense) model is proposed for breast cancer prediction. The model utilizes multi-omics data and stacking of machine learning models, and incorporates Bayesian optimization and parallel computing to enhance performance. Experimental results demonstrate that the BSense model outperforms other models in breast cancer survival prediction.
Breast Cancer is a disease with high risk and mortality rate associated with female health. The multi-omics data having genome, proteome, transcriptome, metabolome data, and pathological images are available for analysis. Artificial intelligence has huge scope in developing models to predict the occurrence, reoccurrence, and survival in breast cancer. In this paper, we proposed a parallel Bayesian hyperparameter optimized Stacked ensemble (BSense) model. The model is developed using stacking of machine learning models, i.e., Deep Neural Network (DNN), Gradient Boosting Machine (GBM), and Distributed Random Forest (DRF). Bayesian optimization with Gaussian Processes is used to find the best hyperparameters for the machine learning models. The parallelism is coupled with Bayesian optimization to address the high computational time. Artificial Bee Colony (ABC) algorithm is used to select the most effective features from a dataset. The proposed model is evaluated on breast cancer dataset available on the TCGA (The Cancer Genome Atlas) portal and is validated on the METABRIC (Molecular Taxonomy of Breast Cancer International Consortium), Metabolomics, and RNA-seq datasets. It is evident from the experimental results that proposed BSense model outperforms DNN, GBM, and DRF models for breast cancer survival prediction. The BSense model achieved 83.9%, 87.3%, 91.1%, and 80.1% Area Under Curve (AUC) for TCGA, METABRIC, Metabolomics, and RNA-seq dataset, respectively, which is more in comparison to the existing models taken into consideration.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据