近年来，我国的天气问题成为了国家关注的新问题，围绕天气的研究成为了学界的热 点话题。天气的变化会对个体的生理、心理和行为造成影响，特别是极端天气和空气污染 会给个体带来许多消极的影响。本研究根据以往研究发现极端天气和空气污染可以通过使 个体产生消极情绪体验或导致个体的认知执行功能下降，最终导致个体的绩效表现下降。 为了探求天气相关指标对绩效的影响，本研究利用网络爬虫技术在互联网中爬取了 2015-2017 年国内气象监测站和空气质量监测站 供的的数据，以及 3 年内所有上场的 438 名 10122 个数据点的中超联赛球员的技术统计数据，并进行整理和挖掘。之后利用有监督 机器学习方法 Lasso 回归、岭回归以及支持向量回归(SVR)三种算法建立预测模型。本 研究建模时把球员位置分为了前锋和后卫，绩效类型划分为关系绩效和任务绩效，并且在 建模时还纳入了进攻和防守的关键指标“射门次数”和“抢断次数”，并且还利用“犯规次数” 进行建模，探索球员的攻击行为与天气指标的关系，最终建立了 7 个预测模型。结果发现: 本研究的假设 1:气候指标(温度、降雨量、风力等)处于极端情况时，能够负向预测球 员的关系绩效和任务绩效。以及假设 2:空气质量指标(PM2.5、PM10、CO 等)，能够负向 预测球员的关系绩效和任务绩效。对应的预测模型是模型 1 到模型 4，模型一的预测效果 最好(R2=0.421)，模型二(R2=0.237)和模型三(R2=0.206)也有较好的预测成果，模型 四的预测效果不是很好(R2=0.045)。因此假设 1 是成立的，假设 2 由于模型 4 的预测效果 不好，因此假设 2 不完全成立，只证明了天气能够比较好地预测前锋的关系绩效，却不能 预测后卫的关系绩效。本研究的假设 3:极端天气(极端气候、空气污染)能够正向预测运 动员攻击行为(犯规)的产生。通过模型七进行验证，模型七预测效果不理想(R2=0.098)， 说明天气因素对球员攻击行为的解释率较低。

In recent years, the issue of weather in China has become a new issue of concern to the nation. Research on the weather has become a hot topic in academic circles. Changes in the weather can have an impact on an individual's physical, mental, and behavioral behavior. In particular, extreme weather and air pollution can have many negative effects on individuals. Based on previous studies, this study found that extreme weather and air pollution can lead individuals to experience negative emotions or cause individuals to decline in cognitive executive function, which ultimately leads to a decline in individual performance. In order to explore the impact of weather-related indicators on performance, this study used web crawler technology to crawl the data provided by domestic weather stations and air quality monitoring stations during the period of 2015-2017, and 438 players in all 3 years. The data points of the technical statistics of the Super League players are collated and mined. Afterwards, three predictive models were built using supervised machine learning methods Lasso regression, Ridge regression, and Support Vector Regression (SVR). In this research modeling, the player position was divided into forward and rearward. The performance type was divided into relational performance and task performance. In the modeling, the key indicators of offense and defense were included in the “number of shots” and “number of steals”, and Using the “Foul Times” model to explore the relationship between the player’s attack behavior and weather indicators, he eventually established seven forecasting models. Findings: Hypothesis 1 of the study: When climate indicators (temperature, rainfall, wind, etc.) are in extreme conditions, they can be negative.