For the ride prediction task, we collect open NYC transportation data ranging from 2016-01-01 to 2016-06-30.
The schema of the corresponding table is presented below, consisting of a singular table with a total of eleven columns, comprising two timestamp columns, two string columns, and seven numeric columns.
For the flu forecast task, we have collected covid19 data ranging from 2020-01-22 to 2020-04-07.
The corresponding table schema is presented below, comprising a singular base table with a total of six columns, consisting of one timestamp column, two string columns, and three numeric columns.
For the power forecast task, we have collected the open wind-power data ranging from 2009-07-01 to 2012-06-26.
The table schema is shown below. It consists of 10 tables containing a total of 61 columns. Among them, seven farm tables record power changes separately for each farm, while the remaining three training tables are utilized for the prediction model training process.
For the sales-prediction task, we have collected real data in Uniqlo ranging from 2017-12-31 to 2021-05-30.
The data schema comprises a total of 7 tables with 85 columns, incorporating both steam tables, such as canceled orders storage, and attribute tables, such as product information.
The task aims to predict whether customers will pay back their loans on time for a credit card company. Q4 owns 9 tables together with 1GB data and 245 columns. The schema is shown below.
Q5 owns 10 tables with over 13GB origin data and 773 columns. The schema is shown below.
