基于随机森林分析广西地区20~39岁青年人超重及肥胖的影响因素

彭迎春; 莫海娟; 赖银娟; 黄志碧; 罗茜倩; 韦瑜群

doi:10.16462/j.cnki.zhjbkz.2021.06.020

基于随机森林分析广西地区20~39岁青年人超重及肥胖的影响因素

doi: 10.16462/j.cnki.zhjbkz.2021.06.020

1.
530012 南宁，广西体育高等专科学校运动科学与健康学院
2.
530021 南宁，广西医科大学公共卫生学院
3.
530021 南宁，广西医科大学附属肿瘤医院

基金项目:

广西中青年教师科研基础能力提升项目 2020KY26008

详细信息

通讯作者:
莫海娟，E-mail：mo18729026426@163.com

中图分类号: R723.14
计量
- 文章访问数: 395
- HTML全文浏览量: 178
- PDF下载量: 49
- 被引次数: 0
出版历程
- 收稿日期: 2020-11-23
- 修回日期: 2021-02-04
- 刊出日期: 2021-06-10

Influential factors of overweight and obesity in young adults aged 20 to 39 in Guangxi based on random forest algorithm

1.
School of Sports Science and Health, Guangxi College of Physical Education, Nanning 530012, China
2.
School of Public Health, Guangxi Medical University, Nanning 530021, China
3.
Guangxi Medical University Affiliated Tumor Hospital, Nanning 530021, China

Funds:

The Project of Improving the Basic Scientific Research Ability of Young and Middle-aged Teachers in Guangxi 2020KY26008

More Information

Corresponding author: MO Hai-juan, E-mail: mo18729026426@163.com

摘要

摘要: 目的探讨广西地区20~39岁青年人的交通方式、工作状态、闲暇时间的活动对超重和肥胖的影响，为提高全民身体素质水平提供参考意见。方法对广西14个地市的国民体质监测中心监测点的20~39岁青年人进行随机整群抽样，共纳入7 534例研究对象。将其分成20~＜30岁、30~39岁两组资料过采样处理后建立随机森林分类模型。结果根据基尼指数的下降幅度对变量重要性排序结果显示：20~＜30岁青年超重和肥胖的影响因素主要为骑摩托车、电动车时长，读书、用电脑等时长，看电视、听广播时长，以静坐伏案为主时长，轻家务劳动时长；30~39岁青年超重和肥胖的影响因素主要为骑摩托车或电动车时长，看电视、听广播时长，读书、用电脑等时长，轻家务劳动时长，以静坐伏案为主时长。结论人们要尽量缩短静坐行为的持续时间。短距离路程尽量选择骑自行车和步行，工作方式与闲暇活动方面避免久坐，增加科学规律的体育锻炼。
- 青年 /
- 超重 /
- 随机森林
Abstract: Objective To explore the influence of transportation mode, working status, and leisure time activity on overweight and obesity among young adults aged 20 to 39 in Guangxi, so as to provide a reference for improving the physical fitness level of the whole population. Methods A randomized whole-group sample of young adults aged 20-39 years from institutions in cities of Guangxi was conducted, with a total of 7 534 cases. The data were divided into 20- < 30 and 30-39 ages groups, and a random forest classification model were constructed after over-sampling a random forest classification model. Results According to the decline of the Gini index, we ranked the importance of the variables. The main influencing factors of overweight and obesity for young people aged 20- < 30 were the time spent riding motorcycles, electric bicycles, reading and using computers, the time spent watching TV and listening to the radio, mainly sitting at a desk, and light housework. As for 30-39 age group, the influencing factors of overweight and obesity were, in order, the length of time riding a motorcycle or electric vehicle; the length of watching TV and listening to the radio; the length of reading and using the computer; the length of light housework; the length of sitting mainly at the desk. Conclusions People should try to shorten the duration of sedentary behavior. In a short distance, try to choose cycling and walking, avoid sedentary work style and leisure activities, and increase scientific and regular physical exercise will be good choices.
- Youth /
- Overweight /
- Random forest algorithm

HTML全文

图 1 20~＜30岁的变量个数、树的数量调整的袋外错误率

Figure 1. OOB error of mtry and ntree adjustment in 20- < 30 years of age

下载: 全尺寸图片幻灯片

图 2 30~39岁变量个数、树的数量调整的袋外错误率

Figure 2. OOB error of mtry and ntree adjustment in 30-39 years of age

下载: 全尺寸图片幻灯片

图 3 广西地区20~＜30岁青年超重与肥胖影响因素重要性的排序

Figure 3. Ranking of influencing factors of overweight and obesity in 20- < 30 years in Guangxi

下载: 全尺寸图片幻灯片

图 4 广西地区30~39岁青年超重与肥胖影响因素重要性的排序

Figure 4. Ranking of influencing factors of overweight and obesity in 30-39 years in Guangxi

下载: 全尺寸图片幻灯片

表 1 研究对象样本构成[n(%)]

Table 1. The composition of the research object sample [n(%)]

年龄(岁)	体重正常	超重、肥胖	χ²值	P值
20~＜30	2 851(55.36)	846(35.49)	257.519	< 0.001
30~39	2 299(44.64)	1 538(64.51)

下载: 导出CSV

表 2 体重影响指标的代码

Table 2. Codes for the weight impact indicators

类别	代码
交通方式
乘车(船)	BUS
自驾私家车	DRIVE
骑摩托车、电动车	MOTO
骑自行车	BIKE
步行	WALK
工作时状态
以静坐伏案为主	SED
工作中静坐并伴有上肢活动或以站为主	ULM
以走为主	AMB
搬运、举重物或挖掘	CARRY
闲暇时活动
看电视、听广播	TV
下棋、打牌等	CHESS
读书、用电脑等	READ
散步	STROLL
轻家务劳动	LHW
重家务劳动	HHW

下载: 导出CSV

表 3 各个年龄段训练集和测试集的模型评估结果

Table 3. Model evaluation results of training sets and test sets for each age group

年龄(岁)	数据集	实际情况	预测体重正常(人)	预测超重、肥胖(人)	准确率(%)
20~＜30	训练集	体重正常	1 746	242	92.59
		超重、肥胖	53	1 940
	测试集	体重正常	727	136	68.65
		超重、肥胖	217	46
30~39	训练集	体重正常	1 169	409	78.89
		超重、肥胖	257	1 319
	测试集	体重正常	547	174	58.02
		超重、肥胖	315	129
全年龄段	训练集	体重正常	2 850	718	85.32
		超重、肥胖	330	3 240
	测试集	体重正常	1 243	339	60.68
		超重、肥胖	578	172

下载: 导出CSV

参考文献(20)

[1]	Guglielmi V, Sbraccia P. Obesity phenotypes: depot-differences in adipose tissue and their clinical implications[J]. Eat Weight Disord, 2018, 23(1): 3-14. DOI: 10.1007/s40519-017-0467-9.
[2]	Yang T, Yu L, Barnett R, et al. Contextual influences affecting patterns of overweight and obesity among university students: a 50 universities population-based study in China[J]. Int J Health Geogr, 2017, 16(1): 18. DOI: 10.1186/s12942-017-0092-x.
[3]	Bray GA, Frühbeck G, Ryan DH, et al. Management of obesity[J]. Lancet, 2016, 387(10031): 1947-1956. DOI: 10.1016/s0140-6736(16)00271-3.
[4]	Alghamdi M, Al-Mallah M, Keteyian S, et al. Predicting diabetes mellitus using SMOTE and ensemble machine learning approach: The Henry Ford ExercIse Testing (FIT) project[J]. PLoS One, 2017, 12(7): e0179805. DOI: 10.1371/journal.pone.0179805.
[5]	吕红燕, 冯倩. 随机森林算法研究综述[J]. 河北省科学院学报, 2019, 36(3): 37-41. DOI: 10.16191/j.cnki.hbkx.2019.03.005. Lyu HY, Feng Q. A review of random forests algorithm[J]. Journal of the Hebei Academy of Sciences, 2019, 36(3): 37-41. DOI: 10.16191/j.cnki.hbkx.2019.03.005.
[6]	Kanerva N, Erkkola M, Nevalainen J, et al. Random forest analysis in identifying the importance of obesity risk factors[J]. European Journal of Public Health, 2013, 23(Supplement): 1. DOI: 10.1093/eurpub/ckt124.042.
[7]	中国肥胖问题工作组. 中国成人超重和肥胖症预防与控制指南(节录)[J]. 营养学报, 2004(1): 1-4. https://www.cnki.com.cn/Article/CJFDTOTAL-YYXX200401004.htm Working Group of Obesity in China. Guidelines for the prevention and control of overweight and obesity in Chinese adults (excerpt)[J]. Acta Nutr Sin, 2004(1): 1-4. https://www.cnki.com.cn/Article/CJFDTOTAL-YYXX200401004.htm
[8]	唐蓉. 基于随机森林回归的青年人体质影响因素研究[D]. 南昌: 华东交通大学, 2016. Tang R. Research on influence factors of young people physical fitness based on random forest regression[D]. Nanchang: East China Jiaotong University, 2016.
[9]	闫慈, 田翔华, 阿拉依·阿汗, 等. 基于重采样技术在医学不平衡数据分类中的应用研究[J]. 中国卫生统计, 2018, 35(2): 177-180, 185. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGWT201802004.htm Yan C, Tian XH, Alayi AH, et al. Application of the resampling technology in the classification of imbalanced medical datasets[J]. Chin J Heal Stat, 2018, 35(2): 177-180, 185. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGWT201802004.htm
[10]	Janitza S, Hornung R. On the overestimation of random forest's out-of-bag error[J]. PLoS One, 2018, 13(8): e0201904. DOI: 10.1371/journal.pone.0201904.
[11]	Biddle SJH, García Bengoechea E, Pedisic Z, et al. Erratum to: screen time, other sedentary behaviours, and obesity risk in adults: a review of reviews[J]. Curr Obes Rep, 2017, 6(3): 352. DOI: 10.1007/s13679-017-0267-6.
[12]	Tumin R, Anderson SE. Television, home-cooked meals, and family meal frequency: associations with adult obesity[J]. J Acad Nutr Diet, 2017, 117(6): 937-945. DOI: 10.1016/j.jand.2017.01.009.
[13]	Al-Hanawi MK, Chirwa GC, Pemba LA, et al. Does prolonged television viewing affect body mass index? A case of the Kingdom of Saudi Arabia[J]. PLoS One, 2020, 15(1): e0228321. DOI: 10.1371/journal.pone.0228321.
[14]	Goh EK, Kim OY, Jeon HJ. Depression is a mediator for the relationship between physical symptom and psychological well-being in obese people[J]. Clin Nutr Res, 2017, 6(2): 89-98. DOI: 10.7762/cnr.2017.6.2.89.
[15]	栗华, 张敬一, 张中朝, 等. 河北省成年居民超重肥胖流行特征及相关因素的调查分析[J]. 实用预防医学, 2007, 14(2): 283-286. DOI: 10.3969/j.issn.1006-3110.2007.02.009. Li H, Zhang JY, Zhang ZC, et al. Investigation on epidemiological characteristics and related factors of overweight and obesity among adult residents in Hebei Province[J]. Pract Prev Med, 2007, 14(2): 283-286. DOI: 10.3969/j.issn.1006-3110.2007.02.009.
[16]	Yang L, Hu L, Hipp JA, et al. Cross-sectional associations of active transport, employment status and objectively measured physical activity: analyses from the National Health and Nutrition Examination Survey[J]. J Epidemiol Community Health, 2018, 72(9): 764-769. DOI: 10.1136/jech-2017-210265.
[17]	熊荣, 袁丽凤, 杨进刚, 等. 自行车拥有状况与体力活动、静坐时间及超重或肥胖的关系研究[J]. 中国循环杂志, 2018, 33(3): 251-255. DOI: 10.3969/j.issn.1000-3614.2018.03.010. Xiong R, Yuan LF, Yang JG, et al. Correlation study between bicycle ownership status and physical activity, time of sitting, overweight or obesity[J]. Chin Circ J, 2018, 33(3): 251-255. DOI: 10.3969/j.issn.1000-3614.2018.03.010.
[18]	Nam JY, Kim J, Cho KH, et al. Associations of sitting time and occupation with metabolic syndrome in South Korean adults: a cross-sectional study[J]. BMC Public Health, 2016, 16: 943. DOI: 10.1186/s12889-016-3617-5.
[19]	Drenowatz C, Hand GA, Shook RP, et al. The association between different types of exercise and energy expenditure in young nonoverweight and overweight adults[J]. Appl Physiol Nutr Metab, 2015, 40(3): 211-217. DOI: 10.1139/apnm-2014-0310.
[20]	Jiang Y, Wang J, Wu S, et al. Association between take-out food consumption and obesity among Chinese university students: a cross-sectional study[J]. Int J Environ Res Public Health, 2019, 16(6): 1071. DOI: 10.3390/ijerph16061071.