Detection of Outliers in The Peruvian Fruit Production Time Series Using Arima Models

Abstract

The present applied, non-experimental, descriptive and prognostic research; was aimed at detecting outliers in the agricultural production of Mangifera indica (mango), Persea americana (avocado) and Citrus x aurantifolia (lemon) at the national level, was performed by applying an ARIMA Model. To fulfill it purposes, documentary analysis was used at the National Institute of Statistics and Informatics (In Spanish, INEI). The study sample consisted of the mango, avocado and lemon production indices 2000-2020. As a result, the models were obtained arima mango (1,0,0) (2,1,2) (AIC=5448.99, BIC=5473.35 and RMSE=19067.93), arima avocado (0,1,3) (2,1,0) (AIC=4687.05, BIC=4707.91 and RMSE=4114.35) and arima lemon (1,0,1) (0,1,1) (AIC=4484.36, BIC=4501.76 and RMSE=2551.96) with a 12 months period, the diagram of boxes and whiskers was also made with which it was identified that atypical data (Outliers) abound in the periods of greatest production.

Keywords: Outliers, Forecast, Time Series, Arima, Agricultural Crops

DOI: 10.54941/ahfe1001008