ABSTRACT
This article applied exploratory data analysis (EDA) and time series
forecasting using extreme gradient boosting (XGBoost) to bird
apprehension data from the Brazilian Institute of Environment
and Renewable Natural Resources (IBAMA) covering 2010 to 2024,
the dataset processed included 150,000 records, filtered to focus
on significant patterns across two distinct periods: a COVID-19
pre-pandemic period (2010-2020) and a period that includes the
COVID-19 pandemic (2014-2024). This analysis identified shifts in
apprehension patterns with an overall decrease of 69.17% in bird
apprehensions during the pandemic. The XGBoost model demonstrated
satisfactory root mean squared error (RMSE) on most of
the species except on Zenaida auriculata and Zenaida auriculata
noronha where the RMSE values were 37.73 and 47.80, respectively,
for the (2010–2020) period, while for the (2014–2024) period, the
RMSE values were 42.93 and 48.76, which are higher values when
compared to other results. The most apprehended bird species prepandemic
was Sicalis flaveola with 163, 437 apprehensions, while
Zenaida auriculata and Zenaida auriculata noronha remained highly
apprehended throughout both periods. The northeast regions of
Brazil, with states like Rio Grande do Norte, Ceará and Paraíba,
showed the highest apprehension rates. Overall, the model achieved
satisfactory performance in understanding the pattern of apprehension
numbers in the tested periods, with significantly low RMSE
values for the ten most apprehended species.