A New Approach for Mobile Advertising Click-Through Rate Estimation Based on Deep Belief Nets

Table 1

Description of dataset.

Title	Description	Type
ID	The ID of this presentation	Category, unique
click	0: unclicked, 1: clicked	Category, 0,1
hour	Time	Continuous, 10 days, 24 hours a day
C1	Anonymous features	Continuous, 7 different values
banner_pos	The position of the ads	Category, 7 different values
site_id	The ID of the site	Category, 2,865 different values
site_domain	The domain of the site	Category, 3,394 different values
site_category	The category of the site	Category, 2 different values
app_id	The ID of the app	Category, 4,154 different values
app_domain	The domain of the app	Category, 287 different values
app_category	The category of the app	Category, 31 different values
device_id	The ID of the device	Category, 368,962 different values
device_ip	The ip of the device	Category, 1,078,153 different values
device_model	The model of the device	Category, 6,098 different values
device_type	The type of the device	Category, 4 different values
device_conn_type	The connection type of the device	Category, 4 different values
C14–C21	Anonymous features	Most categories, few continuous