In this challenge, about 10,000,000 URLs are provided and only 100,000 of them are provided with label information. The task is the classification of the unlabeled URLs.

1) Data Description

The labeled samples are provided in the labeled_data.tsv (http://14.18.146.162:7180/apweb/labeled_data.tsv.gz), and unlabeled samples are provided in the unlabel_data.tsv (http://14.18.146.162:7180/apweb/unlabel_data.tsv.gz).

Each line of the labeled_data.tsv is as follows. In detail, each line contains two columns, URL and the label, and the two columns are separated by tab.

Each line of the unlabel_data.tsv is as follows. In detail, each line corresponds to an URL.

2) The descriptions of the labels are as follows:

Random selected unlabeled samples are used as the evaluation set and the normalized accuracy is used as the evaluation metric. If t_i samples are correctly classified among the c_i URLs belongs to category i, the accuracy is for the category i. Assume there are K categories, the normalized accuracy is as follows.The final board is ranked based on R.

Each submission include: 1) the source code, 2) the executable code, 3) the associated documentation with environment setting and algorithm description. The above materials should be send to apweb@189.com before the deadline, and the email title should be 'teamname+datachallenge'.

In the output file of the submitted code should be "unlabel_data.tsv", and the format of the output file must be same as that of the provided labeled_data.tsv format, i.e. each line contains two columns, URL and the label, and the two columns are separated by tab.

If you have any questions, please contact xiangy@gsta.com .

Submission Deadline: Aug 5,2015

• First Prize: 2,000 RMB
• Second Prize: 1,000 RMB
• Third Prize: 500 RMB

The top 3 teams may be invited to present their solutions in the meeting.

• APWeb2015
• previous APWeb