Figure 4
Feature importance of donors for recruitment models. The trained XGBoost model automatically calculated feature importance with a f(eature) score on 13 features associated with willingness to donate blood, we could see that the plot shown F7 (donation interval) had the highest importance and F5 (blood test) had the lowest importance. The features such as donation interval, age, and donation frequency having a strong correlation with blood donors’ intention to donate blood, were considered more important than other features. Besides, total donation volume, education level and professional occupation also played important roles in our recruitment model. f7: Donation interval, f0: Age, f11: Donation frequency, f3: Total donation volume, f8: Education level, f6: Professional occupation, f1: Gender, f2: Recent blood donation, f4: Blood donation times, f9: Living status, f5: Blood test.
