International Core Journal of Engineering 2020-26 | Page 113

Changsha City is smaller than that of Beijing, and the stations
are densely distributed. The correlation factors between the
stations in the city are strong, so the prediction accuracy is
high. To improve the prediction accuracy, the air quality
monitoring station can be added in the actual situation, or the
air quality monitoring data can be collected for a longer time
as a training set to train the model.
2, 2019. ). Figure 3 and 4 shows the predictive AQ grade and
real AQ grade on the four whole prediction datasets. As for
Beijing and Changsha, the prediction AQ grade is consistent
with the overall trend of the real grade. In general, there is a
problem that the forecast value lags behind the true value. It
can be found from the figures that in Beijing and Changsha,
the forecast model is sensitive to the situation that the air
quality begins to turn better.
V. C ONCLUSION
This paper proposes a new forecasting method for air
quality grade prediction. The main contribution is the use of
the ensemble learning algorithm to predict air quality grade.
We collect real-world data sets through web crawlers and train
the models. Experimental results show that the proposed
method has good prediction effect and good forecasting ability
on the real forecast dataset.
R EFERENCES
[1]
[2]
Fig. 4. Forecasting results of LB prediction model of Changsha [3]
Since the air quality changes are relatively frequent on a
small scale, it can be clearly seen that the prediction accuracy
of the station scale is lower than the city scale. From the
perspective of the overall trend of AQ grades, the forecast
results are basically consistent with the real situation. When
the AQ grades of some stations changes, the prediction model
does not necessarily predict the moment when the change
occurs. From 17:00 to 19:00 pm on March 29, the actual AQ
grade of Beijing Olympic Center increased from grade 1 to 4
and then back to 1, while the model only predicted that the
peak value of the change was 2. Similar to the prediction of
the city scale, the prediction of the forecast model on the
station scale is also more sensitive to the change of the grade. [4]
[5]
[6]
[7]
[8]
From the above results, it can be found that the prediction
model has a good effect on both the city scale and station scale
air quality grade prediction. The accuracy of Changsha are
higher than Beijing. This may be because the urban area of
[9]
91
S.K. Grange, D.C. Carslaw, A. Lewis, E. Boleti, C. Heuglin, “Random
forest meteorological normalisation models for swiss pm10 trend
analysis“, Atmospheric Chemistry and Physics Discussions, 2018.
I.. Bougoudis, K. Demertzis, L.S. Iliadis, "Hisycol a hybrid
computational intelligence system for combined machine learning: The
case of air pollution modeling in athens", Neural Computing and
Applications, vol. 27, pp. 1191-1206, 2016.
X. Zhao, R. Zhang, J.L. Wu, P.C. Chang, "A deep recurrent neural
network for air quality classification", vol. 9, pp. 346-354, 2018.
C.J. Huang, P.H. Kuo, "A deep cnn-lstm model for particulate matter
(pm2.5) forecasting in smart cities", vol. 18, pp. 2220, 2018.
B.C. Liu, A. Binaykia, P.C. Chang, M.K. Tiwari, C.C. Tsao, "Urban
air quality forecasting based on multi-dimensional collaborative
support vector regression (svr): A case study of beijing-tianjin-
shijiazhuang", Plos One, vol. 12, e0179763-, 2017.
C.M. Vong, W.F. Ip, P.-K. Wong, J.Y. Yang, "Short-term prediction
of air pollution in macau using support vector machines", vol. 2012,
2012.
A. Bifet, G. Holmes, B. Pfahringer, "In Leveraging bagging for
evolving data streams", European Conference on Machine Learning &
Knowledge Discovery in Databases, 2010.
N.C. Oza, "In Online bagging and boosting", 2005 IEEE International
Conference on Systems, Man and Cybernetics, 12-12 Oct. 2005, vol.
2343, pp. 2340-2345, 2005.
A. Bifet, G. Holmes, R. Kirkby, B. Pfahringer, "Moa: Massive online
analysis", J. Mach. Learn. Res. vol. 11, pp. 1601-1604, 2010.