A public dataset on long-distance running training in 2019 and 2020

This dataset contains 10,703,690 records of running training during 2019 and 2020, from 36,412 athletes from around the world. The records were obtained through web scraping of a large social network for athletes on the internet.<br>The data with the athletes' activities are contained in dataframe objects (tabular data) and saved in the Parquet file format using the Pandas library, part of the Python ecosystem for data science. Each Pandas dataframe contains the following data (as different columns) for each athlete (as different rows), the first word identifies the name of the column in the dataframe:- datetime: date of the running activity;- athlete: a computer-generated ID for the athlete (integer);- distance: distance of running (floating-point number, in kilometers);- duration: duration of running (floating-point number, in minutes);- gender: gender (string 'M' of 'F');- age_group: age interval (one of the strings '18 - 34', '35 - 54', or '55 +');- country: country of origin of the athlete (string);- major: marathon(s) and year(s) the athlete ran (comma-separated list of strings).<br>For convenience, we created files with the athletes' activities data sampled at different frequencies: day 'd', week 'w', month 'm', and quarter 'q' (i.e., there are files with the distance and duration of running accumulated at each day, week, month, and quarter) for each year, 2019 and 2020. Accordingly, the files are named 'run_ww_yyyy_f.parquet', where 'yyyy' is '2019' or '2020' and 'f' is 'd', 'w', 'm' or 'q' (without quotes). The dataset also contains data with different government’s stringency indexes for the COVID-19 pandemic. These data are saved as text files and were obtained from https://ourworldindata.org/covid-stringency-index. The Jupyter notebooks that we created and made available in the https://github.com/BMClab/covid19 repository exemplify the use of the data.本数据集包含2019至2020年间来自全球36412名运动员的10703690条跑步训练记录，所有数据通过网络爬虫抓取互联网上某大型运动员社交平台获取。 运动员活动数据以数据框对象（表格型数据）形式存储，采用Python数据科学生态中的Pandas库保存为Parquet文件格式。每个Pandas数据框以不同行对应每名运动员，各列则包含如下字段（列名以首项标注，即首项为数据框内的列名）： - datetime：跑步活动的日期 - athlete：运动员的计算机生成ID（整数类型） - distance：跑步距离（浮点数，单位：千米） - duration：跑步时长（浮点数，单位：分钟） - gender：性别（字符串类型，取值为'M'或'F'） - age_group：年龄区间，可选字符串为'18 - 34'、'35 - 54'或'55 +' - country：运动员的原籍国家（字符串类型） - major：运动员参与过的马拉松赛事及对应年份（逗号分隔的字符串列表） 为便于使用，我们针对2019年与2020年，分别按日（'d'）、周（'w'）、月（'m'）、季度（'q'）四种聚合频率对运动员活动数据进行采样，生成包含各时间周期内累计跑步距离与时长的对应文件。此类文件的命名格式为`run_ww_yyyy_f.parquet`，其中`yyyy`代表年份（取值为'2019'或'2020'），`f`代表聚合频率（取值为'd'、'w'、'm'或'q'，均不含引号）。 本数据集还包含新冠疫情防控严格指数（COVID-19 Stringency Index）数据，此类数据以文本文件形式存储，获取自https://ourworldindata.org/covid-stringency-index。 我们在https://github.com/BMClab/covid19仓库中公开了自编的Jupyter Notebook示例代码，用于演示该数据集的使用方法。