Management Forecasting in Business Data when Outliers Existing

Abstract

This academic paper discusses the appropriate business data management and forecasting when outliers existing. Business data management is administrative process by which the required data are acquired, validated, stored, protected, and analyzed, and by which its accessibility, reliability, and timeliness are ensured to satisfy the needs of the data users. Outlier data greatly diverges from the rest of other data. The current parameter estimate method cannot represent the population and explain the quality of the mixed up outliers pupation completely. The outliers have the biggest effect on the mean and variance. Outliers in time series or business data may have a moderate to significant impact on the effectiveness of the standard methodology for time series and business data analysis in the aspect of model identification, estimation, and forecasting. That creates error and unreliability. The methods in correcting and detecting outliers in business data are Graph Method, Iterative Procedure, Deletion Diagnostics and Grubbs’ test. Forecasting time series data with outliers is through Ordinary Least Square Estimation Method, Maximum Likelihood Estimation Method with Iterative for Adjusted Series, Ordinary Least Square Estimation Method with Iterative for Adjusted Series, Conditional Least Squares Method, Joint Estimation of Model Parameters and Outliers Effect Method, Bootstrap Weighted Least Squares Method, Recursive Mean Ordinary Least Squares Method, Recursive Median Ordinary Least Squares Method and Improved Recursive Median Ordinary Least Squares Method. Outliers regularly found in the business and time series data are Additive outlier (AO) and Innovational outlier (IO).