A dataset of Korean weather with anomaly score from 2010 to 2020

As the harmful effects of climate change on human society increase, the analysis of abnormal weather is becoming an important issue. Therefore, this work provides the Korean weather dataset, including the anomaly score measurements by using seven different methods. In this dataset, seven types of weather data for each day in 64 Korean cities from 2010 to 2020 are provided by Weather Radar Center in Korea Meteorological Administration. For measuring the weather anomaly scores, this work first constructs a graph by extracting the relationship features between the given weather data and maps the graph into low-dimensional vector space to obtain the embedding vector of weather data for each day. In the embedding space, seven anomaly detection methods are applied to measure the anomaly scores for each day by calculating the degree of an anomaly for each embedding vector.

随着气候变化对人类社会造成的危害日益加剧，异常天气的分析已成为一项重要议题。因此，本研究提供了一项韩国气象数据集，其中包括利用七种不同方法进行的异常分数测量。该数据集由韩国气象厅气象雷达中心提供，涵盖了2010年至2020年间64个韩国城市的每日七种类型天气数据。为了测量天气异常分数，本研究首先通过提取给定天气数据之间的关联特征构建图，并将该图映射到低维向量空间中，从而获得每日天气数据的嵌入向量。在嵌入空间中，应用了七种异常检测方法，通过对每个嵌入向量计算异常程度来测量每日的异常分数。