Tuning of Abnormal Financial Data Alarm Model for Pharmaceutical Enterprises–Fusion Based on PSO-RF Algorithm

In order to more effectively identify the abnormal financial data of pharmaceutical enterprises and avoid financial risks, the study summarized the abnormal alarm signal indicators and behavior characteristics of listed pharmaceutical enterprises' financial data, and 1653 items were calculated based on valid financial data of 87 pharmaceutical enterprises from 2013 to 2022 in the CSMAR database. The real estate enterprises are divided into normal category (34) and abnormal category (53), and a financial data anomaly alarm model combining MinMax normalization and Particle Swarm Optimization(PSO) with Random Forest (RF) algorithm (PSO-RF) is proposed. By testing and analyzing the characteristics of pharmaceutical enterprises and their financial data, the alarm index system and alarm model are screened and constructed. The research shows that: ①The alarm model constructed by the integration of MinMax normalization and PSO-RF model can identify the abnormal financial data of pharmaceutical enterprises more accurately and carefully. ②The fused alarm model outperforms the standard random forest model by 12.85%, providing more dependable application value. ③Money metrics like the Operating Profit Ratio, Fixed Asset Turnover and Net Profit Ratio have a significant role in determining abnormal alarms of financial data, and pharmaceutical enterprises should focus on such indicators. The study is expected to provide important reference for pharmaceutical enterprises and auditors to identify and discover potential financial data anomalies in time.