Phase I Analysis of Nonlinear Profiles Using Anomaly Detection Techniques

In various industries, the process or product quality is evaluated by a functional relationship between dependent variable y and one few input variables x, expressed as y=fx. This called profile in literature. Recently, monitoring has received lot of research attention. this study, we formulated an anomaly-detection problem proposed outlier-detection procedure for phase I nonlinear analysis. The developed consists three key processes. First, obtained smoothed profiles using spline smoothing method. Second, method estimating proportion outliers dataset. A distance-based decision function was to identify potential provide rough estimate contamination rate. Finally, PCA used dimensionality reduction An algorithm then employed outlying based on estimated algorithms considered study included Local Outlier Factor (LOF), Elliptic Envelope (EE), Isolation Forest (IF). that been studied researchers. We compared competing methods commonly metrics such type error, II F2 score. Based evaluation metrics, our experimental results indicate performance better than other existing methods. When considering smallest hardest-to-detect variation, LOF algorithm, with rate determined achieved errors, scores 0.049, 0.001, 0.951, respectively, while current best were 0.081, 0.015, 0.899, respectively.