Published June 14, 2020 | Version 4
Dataset Open

KDD Cup Dataset (without Missing Values)

  • 1. PhD Student at Monash University
  • 2. Lecturer at Monash University
  • 3. Professor at Monash University
  • 4. Lecturer at University of Sydney


This dataset was used in the KDD Cup 2018 forecasting competition. It contains long hourly time series representing the air quality levels in 59 stations in 2 cities: Beijing (35 stations) and London (24 stations) from 01/01/2017 to 31/03/2018. The air quality level is represented in multiple measurements such as PM2.5, PM10, NO2, CO, O3 and SO2.

The dataset uploaded here contains 270 hourly time series which have been categorized using city, station name and air quality measurement. 

The original dataset contains missing values. The leading missing values of a given series were replaced by zeros and the remaining missing values were replaced by carrying forward the corresponding last observations (LOCF method). 


Files (2.4 MB)

Additional details

Related works