—This research aims to study the forecasting model
to predict the 24-hour average PM10 concentration in the
Northern region of Thailand. This research presents a hybrid
model that combines the autoregressive part of the
Autoregressive Integrated Moving Average (ARIMA) model
with the support vector regression technique. The data used in
this study are the 24-hour average PM10 concentration from 3
locations. Each of the data sets is the daily univariate time series
during 1st January to 31th May 2016. We evaluate predictive
performance of our hybrid model using the two measurements:
Root Mean Squared Error (RMSE) and Mean Absolute
Percentage Error (MAPE). The performance of our hybrid
model has been compared against the ARIMA model. From the
experimental results, we found that a hybrid model has lower
RMSE and MAPE than the ARIMA model for all three data
sets. Therefore, we concluded that our hybrid model can be
used to forecast the 24-hour average PM10 concentration in the
Northern region of Thailand.
—PM10, ARIMA model, support vector
regression, hybrid model.
Ronnachai Chuentawat is with the Nakhonratchasima Rajabhat
University, Thailand (e-mail: firstname.lastname@example.org).
Nittaya Kerdprasop and Kittisak Kerdprasop are with the Suranaree
University of Technology, Thailand (e-mail: email@example.com,
Cite: Ronnachai Chuentawat, Nittaya Kerdprasop, and Kittisak Kerdprasop, "The Forecast of PM10 Pollutant by Using a Hybrid Model," International Journal of Future Computer and Communication vol. 6, no. 3, pp. 128-132, 2017.