Research on Stacking-Based Integrated Algorithm of Anomaly Detection in Production Process

Monitoring the production process and detecting anomalies in time is very important to ensure product quality. Given the current methods are inadequate to solve problems in the production process anomaly detection, this paper proposes a stacking-based integrated anomaly detection method SBIOD, stacking five different component learners: HBOS (Histogram-based Outlier Detection), LOF (Local Outlier Factor), iForest (Isolation Forest), DT (Decision Tree) and LR (Logistic regression), normalize the output of these learners, and then input MLP Classifier as feature values. Experiments show that the AUC of SBIOD method has been significantly improved while providing good performance and practical value.