Modifikasi Data Anomali Kunjungan Wisman ke Sumatera Utara Selama Periode Covid-19 Dengan Metode Indikator Statistik dan Interpolasi Spline Time Series Untuk Kelayakan Model Time Series

Abstract

The feasibility of time series models is highly influenced by both the availability of data and the structural patterns of the series itself. The regular collection of data over time plays a crucial role in generating viable time series models for forecast ing purposes, with a minimum of fifty time series observations generally considered necessary to produce models that are sufficiently robust for predictive analysis. The feasibility of the resulting models is also affected by the temporal pattern of the data, as the presence of extreme fluctuations whether sharp increases or sharp declines can compromise the reliability of time series modeling. Based on the modifications applied to the time series data on international tourist arrivals to North Sumatra during the COVID-19 period, two statistically feasible models were identified using the statisti cal indicator method, namely UBJ-ARIMA (0,1,1) and UBJ-SARIMA (2,1,2,0,1,0). In contrast, the spline time series interpolation method produced only one feasible model, namely UBJ-ARIMA (1,1,1).