特征选择方法预测胃癌的性能比较

IF 0.4 Q4 ONCOLOGY
Hamed Mazreati, Reza Radfar, Mohammad-Reza Sohrabi, Babak Sabet Divshali, Mohammad Ali Afshar Kazemi
{"title":"特征选择方法预测胃癌的性能比较","authors":"Hamed Mazreati, Reza Radfar, Mohammad-Reza Sohrabi, Babak Sabet Divshali, Mohammad Ali Afshar Kazemi","doi":"10.5812/ijcm-138653","DOIUrl":null,"url":null,"abstract":"Background: Gastric cancer (GC) is a leading cause of cancer-related deaths, emphasizing the importance of timely diagnosis for effective treatment. Machine learning models have shown promise in assisting with GC diagnosis. Objectives: This study aimed at comparing the performance of various feature selection methods in identifying influential factors related to GC based on lifestyle using machine learning models. The ultimate goal was to enhance early detection and treatment of the disease. Methods: The data of patients from Shahid Ayatollah Modarres Hospital and Shohadaye Tajrish Hospital between 2013 and 2021 were utilized. Three feature selection methods (filter, wrapper, and filter-wrapper) were employed. The k-fold method validated each model. Four classifiers k Nearest Neighbor (kNN), Decision Tree (DT), Random Forest (RF), and Gradient-Boosted Decision Trees (GBDT) compared their outputs based on feature selection methods. Results: The filter-wrapper method outperformed others, achieving an area under the ROC curve and F1 score of 95.8% and 94.7%, respectively. GBDT also performed well. The wrapper and RF classifiers achieved an area under the ROC curve and F1 scores of 95.7% and 93.6%, respectively, after the filter-wrapper method. Without feature selection methods, the RF classifier had an area under the ROC curve and F1 scores of 95.6% and 91.7%, respectively, surpassing other classifiers. Conclusions: This study suggests that appropriate feature selection methods for identifying influential factors related to GC based on lifestyle can facilitate early diagnosis and treatment. The filter-wrapper method demonstrated the best performance in this regard.","PeriodicalId":44764,"journal":{"name":"International Journal of Cancer Management","volume":null,"pages":null},"PeriodicalIF":0.4000,"publicationDate":"2023-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Comparing the Performance of Feature Selection Methods for Predicting Gastric Cancer\",\"authors\":\"Hamed Mazreati, Reza Radfar, Mohammad-Reza Sohrabi, Babak Sabet Divshali, Mohammad Ali Afshar Kazemi\",\"doi\":\"10.5812/ijcm-138653\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Background: Gastric cancer (GC) is a leading cause of cancer-related deaths, emphasizing the importance of timely diagnosis for effective treatment. Machine learning models have shown promise in assisting with GC diagnosis. Objectives: This study aimed at comparing the performance of various feature selection methods in identifying influential factors related to GC based on lifestyle using machine learning models. The ultimate goal was to enhance early detection and treatment of the disease. Methods: The data of patients from Shahid Ayatollah Modarres Hospital and Shohadaye Tajrish Hospital between 2013 and 2021 were utilized. Three feature selection methods (filter, wrapper, and filter-wrapper) were employed. The k-fold method validated each model. Four classifiers k Nearest Neighbor (kNN), Decision Tree (DT), Random Forest (RF), and Gradient-Boosted Decision Trees (GBDT) compared their outputs based on feature selection methods. Results: The filter-wrapper method outperformed others, achieving an area under the ROC curve and F1 score of 95.8% and 94.7%, respectively. GBDT also performed well. The wrapper and RF classifiers achieved an area under the ROC curve and F1 scores of 95.7% and 93.6%, respectively, after the filter-wrapper method. Without feature selection methods, the RF classifier had an area under the ROC curve and F1 scores of 95.6% and 91.7%, respectively, surpassing other classifiers. Conclusions: This study suggests that appropriate feature selection methods for identifying influential factors related to GC based on lifestyle can facilitate early diagnosis and treatment. The filter-wrapper method demonstrated the best performance in this regard.\",\"PeriodicalId\":44764,\"journal\":{\"name\":\"International Journal of Cancer Management\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.4000,\"publicationDate\":\"2023-09-30\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"International Journal of Cancer Management\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.5812/ijcm-138653\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q4\",\"JCRName\":\"ONCOLOGY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"International Journal of Cancer Management","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.5812/ijcm-138653","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"ONCOLOGY","Score":null,"Total":0}
引用次数: 0

摘要

背景:胃癌(GC)是癌症相关死亡的主要原因,强调及时诊断对有效治疗的重要性。机器学习模型在协助GC诊断方面已经显示出前景。目的:本研究旨在利用机器学习模型比较各种特征选择方法在识别基于生活方式的GC相关影响因素方面的性能。最终目标是加强对这种疾病的早期发现和治疗。方法:采用2013 - 2021年Shahid Ayatollah Modarres医院和Shohadaye Tajrish医院的患者资料。采用三种特征选择方法(滤波、包装和过滤-包装)。k-fold方法验证了每个模型。四种分类器k最近邻(kNN)、决策树(DT)、随机森林(RF)和梯度增强决策树(GBDT)基于特征选择方法比较了它们的输出。结果:过滤-包装法优于其他方法,ROC曲线下面积为95.8%,F1评分为94.7%。GBDT也表现良好。经过筛选-包装方法后,包装器和RF分类器的ROC曲线下面积和F1得分分别达到95.7%和93.6%。在不使用特征选择方法的情况下,RF分类器的ROC曲线下面积和F1得分分别达到95.6%和91.7%,优于其他分类器。结论:本研究提示,根据生活方式选择合适的特征,识别与胃癌相关的影响因素,有助于早期诊断和治疗。过滤器-包装方法在这方面表现出最好的性能。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Comparing the Performance of Feature Selection Methods for Predicting Gastric Cancer
Background: Gastric cancer (GC) is a leading cause of cancer-related deaths, emphasizing the importance of timely diagnosis for effective treatment. Machine learning models have shown promise in assisting with GC diagnosis. Objectives: This study aimed at comparing the performance of various feature selection methods in identifying influential factors related to GC based on lifestyle using machine learning models. The ultimate goal was to enhance early detection and treatment of the disease. Methods: The data of patients from Shahid Ayatollah Modarres Hospital and Shohadaye Tajrish Hospital between 2013 and 2021 were utilized. Three feature selection methods (filter, wrapper, and filter-wrapper) were employed. The k-fold method validated each model. Four classifiers k Nearest Neighbor (kNN), Decision Tree (DT), Random Forest (RF), and Gradient-Boosted Decision Trees (GBDT) compared their outputs based on feature selection methods. Results: The filter-wrapper method outperformed others, achieving an area under the ROC curve and F1 score of 95.8% and 94.7%, respectively. GBDT also performed well. The wrapper and RF classifiers achieved an area under the ROC curve and F1 scores of 95.7% and 93.6%, respectively, after the filter-wrapper method. Without feature selection methods, the RF classifier had an area under the ROC curve and F1 scores of 95.6% and 91.7%, respectively, surpassing other classifiers. Conclusions: This study suggests that appropriate feature selection methods for identifying influential factors related to GC based on lifestyle can facilitate early diagnosis and treatment. The filter-wrapper method demonstrated the best performance in this regard.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
CiteScore
1.10
自引率
0.00%
发文量
67
期刊介绍: International Journal of Cancer Management (IJCM) publishes peer-reviewed original studies and reviews on cancer etiology, epidemiology and risk factors, novel approach to cancer management including prevention, diagnosis, surgery, radiotherapy, medical oncology, and issues regarding cancer survivorship and palliative care. The scope spans the spectrum of cancer research from the laboratory to the clinic, with special emphasis on translational cancer research that bridge the laboratory and clinic. We also consider original case reports that expand clinical cancer knowledge and convey important best practice messages.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信