基于机器学习的南京地区献血者精准招募方法的建立和应用

Machine Learning-based Development and Application of a Precise Blood Donor Recruitment Strategy in Nanjing

目的通过基于机器学习的方法,建立南京地区献血者精准招募模型并开展应用,以提高献血招募的效率和质量,招募更多的献血者,确保血液供应的安全性和充足性。方法对江苏省血液中心2017—2022年的献血和短信招募数据进行回顾性研究,利用极端梯度提升、支持向量机、K近邻算法、逻辑回归、决策树、随机森林、多层感知机等模型,并使用合成少数类过采样技术、下采样等多种采样技术,结合代价敏感方法(MFE、MSFE损失函数)进行训练,通过网格搜索选择性能较佳的机器学习模型。结果研究发现,机器学习模型对高意愿献血者招募成功率提升57.79%,机器学习模型可减少40.05%的短信发送数量,每条短信招募效率较常规方法平均提升12.26%,减少了短信发送数量,提高了每条短信的招募效率。结论机器学习算法可以对献血者进行精准识别,提高招募效率,减少不必要的短信发送,降低招募成本,为确保血液供应的安全性和充足性提供了有效的手段。

Objective To establish and apply a precise blood donor recruitment model based on machine learning in Nanjing,thus enhancing the efficiency and quality of blood donation recruitment,incresing the number of blood donors,and ensuring the safety and sufficiency of blood supply.Methods This study retrospectively investigated blood donation and SMS recruitment data from Jiangsu Blood Center from 2017 to 2022.Various machine learning models,including eXtreme Gradient Boosting,Support Vector Machine,K-Nearest Neighbors,Logistic Regression,Decision Tree,Random Forest,and multi-layer perceptron models were used.These models were trained via techniques such as synthetic minority oversampling,under-sampling and cost-sensitive methods(mean false error and mean squared false error).The grid search method was used to select the machine learning models with better performance.Results The implemented machine learning model demonstrated a 57.79%improvement in the success rate of recruiting high willingness blood donors.Additionally,it reduced the number of SMS sending by 40.05%,and increased the recruitment efficiency of each SMS by an average of 12.26%compared with the conventional method.Conclusion Machine learning algorithms could accurately identify potential blood donors,thereby improving recruitment efficiency,reducing unnecessary SMS messages,decreasing recruitment costs,and providing an effective means to ensure the safety and adequacy of blood supply.