【摘要】 目的:基于公共数据库的前列腺癌数据，通过机器学习的方法构建模型来预测前列腺癌复发。方法:下载前列腺癌RNA测序数据以及临床数据，对前列腺基因以及临床数据进行处理，筛选前列腺复发相关特征基因，建立相关模型，并对模型效能进行验证。在默认参数下比较随机森林、支持向量机(径向核、线性核、二项式核、sigmoid核)、和梯度下降树这5个模型预测效果并选取具有较高效能的模型进行进一步验证。结果:总共获得148个复发差异基因，根据重要性筛选5个基因构建预测模型。基于这些基因使用不同的方法构建的模型均具有较好的精确度和准确度，其中随机森林法构建的模型最佳，其预测前列腺癌复发的准确度为87%,受试者工作特征曲线下面积为0.84。结论:通过基因表达数据构建机器学习模型能够较好地来预测前列腺癌的复发。更多还原

【Abstract】 Objective: Based on the prostate cancer data from public databases, a model was constructed to predict prostate cancer recurrence by machine learning methods. Methods: Prostate cancer RNA sequencing data as well as clinical data were downloaded, and prostate genes as well as clinical data were processed to screen prostate recurrence-related feature genes. Relevant models were constructed, and the model efficacy was validated. Random forests, support vector machines(radial kernel, linear kernel, binomial kernel, sigmoid kernel), and gradient descent trees were compared with the default parameters, and the models with higher performance were selected for further validation. Results: A total of 148 recurrence difference genes were obtained, and 5 genes were screened according to their importance to construct prediction models. The models constructed based on these genes using different methods had good precision and accuracy, among which the model constructed by the random forest method was the best, with an accuracy of 87% in predicting the recurrence of prostate cancer, and the area under the working curve of the subjects was 0.84. Conclusion: Constructing a machine learning model from gene expression data can be used to predict the recurrence of prostate cancer in a better way.更多还原