| 摘 要: 在我国，恶性肿瘤死亡率最高的就是肺癌。为了提高肺癌存活性预测的准确性，本研究以随机森林(Random Forest)、LightGBM(Light Gradient Boosting Machine)和CatBoost(Categorical Boosting)三种算法为基模型，通过线性回归集成融合构建RF-LGC肺癌存活性预测模型，运用分层十折交叉验证方法进行仿真实验。实验结果显示，RF-LGC组合模型的预测精度达到了98.0242%，比单一的基模型提高了0.2%；敏感性达到了89.3957%，比单一的基模型提高了3%；特异性达到了78.4848%，比单一的基模型提高了1%。因此，该集成融合模型是一种精确、方便的肺癌存活性预测模型。 |
| 关键词: 集成学习；随机森林；十折交叉验证；癌症预后 |
|
中图分类号: TP311
文献标识码: A
|
| 基金项目: 河南省高校重点应用研究计划项目《基于Hadoop的医疗数据挖掘算法研究及应用》(19A520027). |
|
| Prediction and Analysis of Lung Cancer Survival based on Ensemble Learning |
|
LI Xiuqin, LI Lin, ZHANG Manli
|
(North China University of Water Resources and Electric Power, Zhengzhou 450046, China)
lixiuqin@ncwu.edu.cn; 1872134171@qq.com; 313364258@qq.com
|
| Abstract: In China, lung cancer has the highest mortality rates among all of the malignant tumors. In order to improve the accuracy of lung cancer survival prediction, this paper proposes to use linear regression integration and fusion to build a lung cancer survival prediction model RF-LGC, which is based on three algorithms — Random Forest, LightGBM (Light Gradient Boosting Machine) and CatBoost (Categorical Boosting). Simulation experiments are carried out by using the hierarchical ten-fold cross-validation method. Experimental results show that the prediction accuracy of the RF-LGC combined model reaches 98.0242%, which is 0.2% higher than that of a single-based model; the sensitivity has reached 89.3957%, which is 3% higher than the single-based model; the specificity has reached 78.4848%, which is 1% higher than the single-based model. Therefore, the integrated fusion model is an accurate and convenient lung cancer survival prediction model. |
| Keywords: ensemble learning; random forest; ten-fold cross validation; cancer prognosis |
