【中文摘要】 在论文即将完成之际,2009年4月始发于墨西哥的甲型H1N1流感正在全球蔓延,再次为人类敲响了预防流感的警钟。流感是第一个实行全球监测的呼吸道传染病,它不仅造成大量人口发病及死亡,还会给一个国家造成巨大的经济损失和社会负担。中国是全球流感高发国家之一,1957年、1968年和1977年三次世界性流感大流行的毒株均首发于中国,1999年、2003年、2008年又在香港爆发了由新亚型流感病毒引发的流感流行,深圳毗邻港澳,人群流动频繁,常年有流感病毒活动,人群流感发病率高,危害大,在1994年深圳地区被国家流感中心选为中国重点流感监测点之一,因此对深圳市流感进行研究具有重要的现实意义。本文首次系统的收集整理深圳市2003～2007年流感就诊率资料和相应的大气环境资料,详细分析了深圳市流感的流行病学特征,重点探讨了流感发病的季节性特征与气象条件、空气质量的关系,建立了流感高峰期大气环境要素的临界值体系,并利用2008年香港流感流行特殊时段的数据资料进行了个例诊断分析,对临界值体系进行了验证修订,最终用统计方法和支持向量机方法建立了深圳市流感的预测模型,并进行独立样本预报准确率的检验与比较。获得...更多了一些重要研究成果,主要有:(1)深圳地区的流感一年四季均有发生,流感流行期为3月～9月,非流行期为10月～来年2月,夏季流感就诊率远远高于其他季节;受2003年、2004年亚洲H7N1流感流行,2003年、2004年ILI%明显比其后三年高;同时发现深圳市流感高峰期尤其是第一次高峰期有逐渐向夏季移动的趋势,有从往年的多个小高峰逐渐演变为近年来一个主高峰的变化趋势。(2)通过相关分析得知:深圳市ILI%与水汽压、绝对湿度、相对湿度、气温、气压、降水量紧密相关,同时发现流感传播受空气中水汽的绝对含量影响最大。(3)气象要素对流感的影响存在明显的超前效应。平均相对湿度、最小相对湿度超前ILI%两周时对ILI%影响最为显著,绝对湿度、平均风速超前ILI%三周时对ILI%影响最为显著,周变温在超前ILI%四周时对ILI%影响最为显著。气象要素对流感的这种超前效应具有很好的预报意义。(4)利用时间序列小波分析发现深圳市流感除存在季节性变化主周期外,还另外存在两个分别为70周、30周的局部流行爆发次周期。流感的季节性变化主周期是多种气象要素及酸雨等综合作用的结果,而30周的流感局部爆发次周期主要受相对湿度、周变温、降水量的影响,70周局部爆发次周期主要受气压的影响。(5)SO2、PM10与ILI%呈现显著反相关,即SO2、PM10有利于抑制流感流行,SO2、PM10均在超前一周时对ILI%的抑制作用最为显著,具有较好的预报意义。酸沉降可以有效的杀灭流感病毒,阻止流感病毒在人群中的传播,深圳市酸雨的季节特征对ILI%的季节流行特征起着一定的影响作用。(6)建立了流感高峰发生的大气环境要素临界值体系,并利用个例分析进行了验证修订。结果表明当水汽压处于23～29hPa、绝对湿度处于2.0 mg/m3～2.3mg/m3、平均相对湿度在60%～80%之间、平均气温在25～30℃之间、平均气压低于1008 hPa时,有发生流感高峰的可能性;当最高气温≤30℃、最低气温≤25℃时,最小相对湿度≤60%左右,ILI%随温度与湿度的上升而上升;当最高气温超过30℃、最低气温超过25℃、最小相对湿度超过60%时,LI%随温度与湿度的上升而下降;而且先干后湿、温度、气压、降水等气象要素变化起伏剧烈的天气形势更有利于流感高峰的发生。(7)利用支持向量机方法建立了深圳市ILI%的预测模型,并与统计预测模型进行预测效果比较,发现SVM预测模型的预测效果要好于统计方法预测模型,说明支持向量机方法用于疾病发病率预报是完全可行的,并且预报效果令人满意。 

【英文摘要】 When the paper is nearing completion,an influenza pandemic is global spreading,which started in Mexico in April of 2009,and sounded a warning for humanity to fence out influenza illness again.As the first kind of global monitoring respiratory infectious disease,influenza not only cause a large number of population morbidity and mortality but also a tremendous economic loss and social burden.China is one of the countries with high incidence rate of influenza,the strains which caused 1957,1968 and 1977,three world-wide pandemic influenza were first discovered in China,In 1999,2003 and 2008,an influenza pandemic caused by a new subtype of influenza virus broke out in Hong Kong,Shenzhen City lies besides Hong Kong and Macao,people from each side move frequently and the influenza virus are vitality year-round,so there is a high incidence of influenza in the crowd.Shenzhen was chosen as one of the focus of the monitoring of influenza by the National Influenza Center of China in 1994,so it is... of great practical significance to study the influenza in Shenzhen City.This article have collected Influenza-Like-Illness Rates material in Shenzhen from 2003 to 2007 year and the corresponding atmospheric environment material systematically for the first time,made a detailed analysis of the epidemiological characteristics of influenza in Shenzhen, focused on the relation of seasonal incidence of influenza characteristics with meteorological conditions and air quality,carried on the example diagnostic analysis to Shenzhen influenza at the very time when the Hong Kong influenza pandemic broke out in 2008,and established Shenzhen influenza's forecasting model using the statistical method and the Support Vector machines method finally,made the prediction accuracy test and comparison of independent samples.Accessed to a number of important research results,as the follow:(1)The influenza illness in Shenzhen happens throughout the year,its pandemic time is in March to September,the non-popular time is October to February of next year, Influenza-Like-Illness Rates in summer is much higher than other seasons;By the influence of Asian H7N1 influenza pandemic,ILI%of 2003 and 2004 is higher than the following three years; At the same time,it appears a trend that the seasonal peaks of influenza in Shenzhen,especially the first peak is moving to the summer,it also appears that a number of small peaks gradually gathering into a main peak.(2) It has been found ILI%in Shenzhen associate with aqueous vapor tension,absolute humidity,relative humidity,atmospheric temperature and atmospheric pressure closely. Simultaneously discovered that the absolute content of water vapor in air influence influenza mostly.(3) Meteorological factors' influence on the influenza exists an obvious leading effect.The average relative humidity,minimum relative humidity have strongest associations with ILI%in lag two weeks,as absolute humidity and average wind speed lag in three weeks,Week change temperature lag in four weeks.The lag effect of Meteorological factors on influenza has good meaning of forecasting.(4) By the use of wavelet analysis of influenza time series in Shenzhen found besides the main cycle of influenza seasonal changes,there are two other outbreak cycles of local influenza epidemic,which were 70 weeks and 30 weeks.The seasonal influenza cycle mainly influence by various climate elements and acid rain.At the same time,its 30 weeks cycle mainly be influenced by relative humidity,temperature and rainfall,and 70 weeks cycle by the impact of atmospheric pressure.(5) SO2,PM10 and ILI%showed a significant anti-correlation,that means SO2 and PM10 is conducive to inhibit influenza pandemic,SO2 and PM10 influence influenza most significantly one week ahead,which has better forecasting significance.Acid deposition has an effect against influenza viruses,which could prevent the spread of influenza among the crowd.The season characteristic of acid rain has a certain impact on the season pandemic feature of influenza.(6) The critical range of atmospheric environment elements has been established,which has been proved by special analysis.The results show that there is a possibility to occur the seasonal peaks of influenza in the conditions of the vapor pressure from 23hPa to 29hPa, absolute humidity from 2.0 mg/m3 to 2.3mg/m3,the average relative humidity from 60%to 80%, average temperature between 25℃to 30℃,and the average atmospheric pressure is lower than 1008hPa.At the same time,ILI%will increase as the increase of temperature and humidity when the maximum temperature lower than 30℃,minimum temperature lower than 25℃,and the minimum relative humidity less than 60%.ILI%will decrease as the increase of temperature and humidity when the maximum temperature exceeded 30℃,the minimum temperature above 25℃and the minimum relative humidity more than 60%.More than that,the influenza peak has more possibility to happen as the climate situation changes acutely,such as humidity from lower to high,and temperature,pressure,precipitation or other meteorological factors dramatic ups and downs.(7) The forecasting model of ILI%in Shenzhen has been established by Support Vector Machine method,the results has been compared with statistical forecasting model.It has been found that SVM forecasting model has a better forecast effect than statistical forecasting model, which proves SVM method could be used in the field of illness forecasting with satisfactory prediction. 

【中文关键词】 流感样病例就诊率; 气象条件; 空气质量; 预测模型; 支持向量机
【英文关键词】 Influenza-Like-Illness Rates; Meteorologic Condition; Air Quality; Forecasting Model; Support Vector Machine