Resumen de Intervention analysis and automatic detection of outliers in time series analysis.: A case study
Outliers are observations that cannot be properly explained by the ARIMA model and its underlying normality assumption. Outlier analysis of time-series data comprises two key issues: outlier detection and outlier adjustment. Outlier correction is an essential one because the presence of one or more outliers in the observed series may seriously damage identi cation and estimation of the ARIMA model. We consider four types simple intervention models to represent four types of outliers that might our in practice. The standard procedures for automatic outlier detection and correction consider four types of outliers namely, additive, innovational, level shift, and transitory change outliers. Measurement and control of the dissolved aluminium levels in zinc hot-dip galvanizing baths, is of great importance to the steel industry. This study investigates the application of statistical process control and times series analysis methods for monitoring and forecasting e ective aluminium levels.
