Forecasting Air Quality in Taiwan by Using Machine Learning

Abstract

fang 1,2* this study proposes a gradient-boosting-based machine learning approach for predicting the pM 2.5 concentration in Taiwan. The proposed mechanism is evaluated on a large-scale database built by the Environmental Protection Administration, and Central Weather Bureau, Taiwan, which includes data from 77 air monitoring stations and 580 weather stations performing hourly measurements over 1 year. By learning from past records of PM 2.5 and neighboring weather stations' climatic information, the forecasting model works well for 24-h prediction at most air stations. This study also investigates the geographical and meteorological divergence for the forecasting results of seven regional monitoring areas. We also compare the prediction performance between Taiwan, Taipei, and London; analyze the impact of industrial pollution; and propose an enhanced version of the prediction model to improve the prediction accuracy. The results indicate that Taipei and London have similar prediction results because these two cities have similar topography (basin) and are financial centers without domestic pollution sources. The results also suggest that after considering industrial impacts by incorporating additional features from the Taichung and Thong-Siau power plants, the proposed method achieves significant improvement in the coefficient of determination (R 2) from 0.58 to 0.71. Moreover, for Taichung City the root-mean-square error decreases from 8.56 for the conventional approach to 7.06 for the proposed method.

Noise Prediction Using Machine Learning with Measurements Analysis

“…The procedure is described in the equations in detail. The testing samples are put into the prediction model F(x t ) to calculate the prediction results [8].…”

“…Moreover, β is a gradient decent step size and (x i t , y i t+24 ) is the i-th training sample pair. When the value of the loss function L(y t +24, F(x t , β )) is minimized as [8,21,25]…”

“…Successful examples include predicting real estate sales prices, which includes noise similar to that of time series data [7]. A recent study applied this method for forecasting air quality in Taiwan and also extracted meaningful time and historical features as input to the gradient lifting technology model [8].…”

“…However, the observed data information and output values are calculated error, which uses a loss function. The frequently used loss functions include square-error, absolute-error, and negative binomial log-likelihood functions [8]. Then, gradient lifting technology was applied to find the fitting function where the expected value of loss function is minimized.…”

“…The NSRRC installed 12 noise detection sensors around the work environment. As shown in Figure 2a, it showed the circle building has a 120 m circumference, and 24 straight line experimental stations that are differential function research experiments; the locations of the sensors were divided into the inner circle (1)(2)(3)(4)(5)(6) and the outer circle (7)(8)(9)(10)(11)(12), shown as green dots. In addition, there are many noise sources from vacuum pumps, cooling pumps, liquid nitrogen pressure relief, computer servers, etc.…”