采购订单金额异常检测算法模型

1.数据采集：从电子卖场平台采集政企采购订单信息，具体包含成交价值、商品品类、时间字段信息； 2.数据处理：1）按商品品类-月度维度对采集到的输入数据进行聚合计算，通过统计方法得出各品类对应月度的成交均价μ与标准差σ；2）基于统计模型进行异常判定，当单笔订单金额超出μ+3σ上限或低于μ-3σ下限时，将该订单标记为金额异常；3）针对同一供应商单日产生的多笔相近金额订单进行合并计算；4）整理并输出异常订单的明细ID、异常类型（高价/低价）及偏离系数； 数据应用：将模型嵌入采购平台实时数据流，实现异常订单的秒级预警，为审计工作和反舞弊监管提供技术支撑，适配政企采购领域风险防控的实际业务场景，助力提升采购过程的合规性与安全性。1. Data Collection: Procurement order information of government and enterprise entities was collected from an electronic marketplace platform, specifically including transaction value, product category, and timestamp fields. 2. Data Processing: 1) Aggregate the collected input data by product category and monthly granularity, and calculate the monthly average transaction price μ and standard deviation σ for each category via statistical methods; 2) Conduct anomaly detection using statistical models: flag an individual order as having an abnormal transaction amount if its value exceeds the upper bound of μ+3σ or falls below the lower bound of μ-3σ; 3) Consolidate and recalculate multiple orders with comparable amounts generated by the same supplier within a single day; 4) Compile and output the detailed ID of the abnormal orders, anomaly type (high-price/low-price), and deviation coefficient; 3. Data Application: Embed the model into the real-time data stream of the procurement platform to achieve second-level early warning for abnormal orders, providing technical support for audit work and anti-fraud supervision. This solution adapts to the actual business scenarios of risk prevention and control in the government and enterprise procurement domain, and helps improve the compliance and security of the procurement process.