基于机器学习的前列腺癌患者分类研究

Research on Prostate Patients Classification Based on Machine Learning

基于机器学习算法,本文对前列腺癌患者构建分类模型,目的是区分出不同患者,从而采取不同的治疗措施。基于随机森林、Adaboost、GradienBoosting、XGBoost、LightBGM和Stacking模型融合算法构建前列腺癌分类预测模型,调整模型参数,并验证模型效能。在单一的机器学习算法中,每一种机器学习算法都能对第一类(前列腺增生)患者进行识别,一部分模型在第二类(前列腺癌)患者和第三类(同时有前列腺癌和前列腺增生)患者中预测错误率较高;Adaboost算法的性能最优,对每一类都能够进行有效识别;Stacking融合算法优于所有单一的机器学习算法,在测试集上的准确率达到了96%。在前列腺癌分类预测模型中,Stacking融合算法效果明显优于单一的机器学习算法。

This paper aims to distinguish different kinds of prostate cancer patients. We constructed a classification model for patients based on machine learning algorithms, so that we can take different treatment measures. Based on Random Forest, Adaboost, GradienBoosting, XGBoost, LightGBM and Stacking fusion algorithm, we constructed a prostate cancer classification model. The parameters were adjusted, and the effectiveness of models was evaluated. In all single machine learning algorithms, they can identify patients in the first category (prostatic hyperplasia). Some models have a high error rate of classification in the second category (prostate cancer) and the third category (both prostatic hyperplasia and prostate cancer);Adaboost algorithm has the best performance and it can identify each category effectively. The accuracy of Stacking fusion algorithm is up to 96%, better than all single models. Stacking fusion algorithm is better than each single machine learning algorithm significantly in the prostate cancer classification.