Forecasting air pollution PM2.5 in Beijing using weather data and multiple kernel learning

• Date: 01 March 2020
• Published date: 01 March 2020
• Author: Xiang Xu
• DOI: http://doi.org/10.1002/for.2599

Received: 15 November 2018 Revised: 13 March 2019 Accepted: 11 April 2019

DOI: 10.1002/for.2599

RESEARCH ARTICLE

Forecasting air pollution PM2.5 in Beijing using weather

data and multiple kernel learning

Xiang Xu

Antai College of Economics and

Management, Shanghai Jiao Tong

University, Shanghai, PR China

Correspondence

Xiang Xu, Antai College of Economics and

Management, Shanghai Jiao Tong

University, 1954 Huashan Rd.,Shanghai,

200030, PR China.

Email: xuxiang2_2009@foxmail.com

Funding information

National Natural Science Foundation of

China, Grant/AwardNumber: 71871140

Abstract

PM2.5 mass concentration prediction is an important research issue because of

the increasing impact of air pollution on the urban environment. In this paper,

aPM

2.5 forecasting framework incorporating meteorological factors based on

multiple kernel learning (MKL) is proposed to forecast the near future PM2.5.

In addition, we develop a novel two-step algorithm for solving the primal MKL

problem. Compared with most existing MKL 2-step algorithms, the proposed

algorithm does not require the optimal step size for updating kernelcombination

coefficients by linear search. To demonstrate the performance of the proposed

forecasting framework, its performance is compared to single kernel-based sup-

port vector regression (SVR). Data sets of an inland city Beijing acquired from

UCI are used to train and validate both of two methods. Experiments show that

our proposed method outperforms the SVR.

KEYWORDS

first-order primal–dual algorithm, meteorological factors, multiple kernel learning, PM2.5

forecasting

1INTRODUCTION

With the rapid growth of industrialization and urbaniza-

tion in China, air pollution has become one of the most

important environmental problems and has exerted a neg-

ative influence on human health, such as respiratory ill-

nesses and cardiovascular diseases (Kampa & Castanas,

2008; Xing, Xu, Shi, & Lian, 2016). Among the various pol-

lutants, PM2.5 (particulate matter with a diameter equal

to or less than 2.5 μm) is one of the critical ones during

the atmospheric complex pollution process. As reported

in previous research (Orru et al., 2011), among all Euro-

pean Union (EU) citizens, fine particles have decreased

the average life expectancy at birth by 8.6 months and

led to the premature death of 348,000 people in Europe

in 2000. Investigating in 29 European countries, Analitis

et al. (2006) found that the daily number of people who

died of respiratory diseases increased by 0.58% when PM10

increased by 10 𝜇gm

−3. As for the USA, after analyzing

approximately 500,000 air pollution deaths among adults

across the USA, Pope et al. (2002) concluded that each

10 𝜇gm

3elevation in PM2.5 mass concentration was asso-

ciated with approximately a 4%, 6%, and 8% increased risk

of all-cause, cardiopulmonary and lung cancer mortality,

respectively. In addition, a study by the American Cancer

Society, which tracks 1.2 million US adults (1982–2008),

found that mortality from lung cancer increased by 15–27%

when the concentration of PM2.5 increased by 10 𝜇gm

3

(Turner et al., 2011). A study of 63,520 people from six

districts in three prefectures in Japan showed that higher

mortality from lung cancer and respiratory diseases in

Japan were closely associated with long-term exposure to

ambient air pollution (Katanoda et al., 2011). Compared

with Western countries, research into the hazard of PM2.5

in China began just a decade ago. Using meta-analysis,

Qian (2005) studied the epidemiological literature pub-

Journal of Forecasting. 2020;39:117–125. wileyonlinelibrary.com/journal/for © 2019 John Wiley & Sons, Ltd. 117