Application of Facebook’s Prophet Model for Forecasting Meteorological Data

Abstract

The wildfire risk index was calculated based on current meteorological information, for example, temperature, humidity, and wind speed. Thus, meteorological data forecasting could help estimate the probability of fire occurrence or spreading speed to prevent large wildfires. This study predicts meteorological data (e.g., temperature, humidity, and wind speed) using Facebook's Prophet library. We trained the Prophet model using meteorological data between 2016 and 2018 in Goseong, Gangwon-do (where the wildfire occurred in 2019) and predicted meteorological data for the first four months in 2019. We obtained that Facebook's Prophet model was effective in computing speed and predicting the overall trend. However, it could not predict sudden irregular changes satisfactorily. Considering its rapidity, these results could play an important role in future research, especially as a basic research for time-series forecasting.

Short- and Medium-Term Electricity Consumption Forecasting Using Prophet and GRU

“…LSTM is ideal for long-term forecasting, given its capacity to learn long-term dependencies in time-series data. However, it is sensitive to data quality and quantity and demands substantial computational resources for training and prediction [13][14][15].…”