Abstract: Objectives 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.

Key words: Blood donor recruitment, SMS recruitment, Machine Learning, Artificial Intelligence