Open AccessA comparison of extreme gradient boosting, SARIMA, exponential smoothing, and neural network models for forecasting rainfall data
Author(s) -
Ryoichiro Agata,
I Gede Nyoman Mindra Jaya
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1397/1/012073
Subject(s) - exponential smoothing , autoregressive integrated moving average , gradient boosting , artificial neural network , mean squared error , gradient descent , mathematics , boosting (machine learning) , deviance (statistics) , extreme value theory , statistics , meteorology , climatology , computer science , time series , geography , machine learning , geology , random forest
Extreme gradient boosting, is a combination of gradient descent and boosting that can be used to build an optimal model for time series data. This method was used to forecast the rainfall data in city of Bandung for period 2018-2019 and compared to Seasonal Autoregressive Integrated Moving Average (SARIMA) exponential smoothing, and artificial neural network which were used as benchmarks. Data used in this study were monthly rainfall from 2000 through 2017. The extreme gradient boosting had the lowest mean absolute deviance, root mean squared error deviance, and mean absolute percentage error. This indicates the extreme gradient boosting model performed better than the SARIMA, exponential smoothing, and neural network. Based on the extreme gradient boosting model, it is concluded that the highest rainfall will occur between September 2018 and May as a rainy season in Bandung, and the lowest rainfall will occur between June and August as a dry season.