Research on quality risk early warning model of blood diagnostic reagents based on big data analysis

In this study, we constructed a quality risk early warning model for blood diagnostic reagents based on big data, integrating Random Forest, Support Vector Machine and Long and Short Term Memory Network algorithms, which achieved an early warning accuracy of 87.6%, and identified the potential risks 15.3 days in advance on average. By analysing data from 12,456 batches of reagents, the model identified key risk factors and their interactions, such as the coefficient of variation between batches of raw materials (0.187) and fluctuations in environmental temperature and humidity (0.165). The validation results show that the accuracy of the model in predicting the risks of raw materials and production environment reaches 92.1% and 89.3%, respectively, and 17 quality problems were successfully avoided in 5 enterprises in 6 months, which provides an accurate and intelligent solution for the quality risk management of medical devices.