B
B
boss_lexa2017-09-19 20:37:21
Algorithms
boss_lexa, 2017-09-19 20:37:21

How to detect invalid clicks, click fraud, clicks and boost impressions?

The marketplace website hosts a catalog of products from different companies.
The site sells clicks from the catalog to company product cards through an auction model with pay per click (CPC) or per impression (CPM). Also on the site you can order goods and pay for it.
All actions take place on the marketplace website and there is access to all visitor logs.
The question is how to detect invalid clicks and impressions?
What would you recommend worthy algorithms and solutions to this problem on their own?
Which paid services are better able to detect click fraud and impressions boost?
In the Russian Federation, it seems that adriver is engaged, but what systems are known abroad?

Answer the question

In order to leave comments, you need to log in

1 answer(s)
B
boss_lexa, 2017-09-22
@boss_lexa

found a study
As I understand it, this task is a classification: fraud or not.
fraud Signals
Short sessions
Low variance in observed behavior, eg time between sessions High click-through rate
Very high session count but almost no click-throughs (in a short period of time)
Machine learning algorithms with better accuracy and speed

Analyzed data
Total number of sessions
Total number of click-throughs
Distribution of time between sessions
Number of sessions marked as spurious
User IP address(es)
Distribution of session, engagement and visibility times
Distribution of time to rst click-through and mouse-over
Number of invalid sessions

@RELATION user
@ATTRIBUTE session_count NUMERIC
@ATTRIBUTE tot_session_time NUMERIC
@ATTRIBUTE avg_session_time NUMERIC
@ATTRIBUTE session_dev NUMERIC
@ATTRIBUTE tot_visibility_time NUMERIC
@ATTRIBUTE avg_visibility_time NUMERIC
@ATTRIBUTE visibility_dev NUMERIC
@ATTRIBUTE avg_time_to_first_mouseover NUMERIC
@ATTRIBUTE first_mouseover_dev NUMERIC
@ATTRIBUTE tot_engagement_time NUMERIC
@ATTRIBUTE avg_engagement_time NUMERIC
@ATTRIBUTE engagement_dev NUMERIC
@ATTRIBUTE spurious_count NUMERIC
@ATTRIBUTE click_thru_count NUMERIC
@ATTRIBUTE invalid_count NUMERIC
@ATTRIBUTE avg_time_between_sessions NUMERIC
@ATTRIBUTE time_between_sessions_dev NUMERIC
@ATTRIBUTE avg_time_to_first_click_thru NUMERIC
@ATTRIBUTE time_to_first_click_thru_dev NUMERIC
@ATTRIBUTE click_count NUMERIC
@ATTRIBUTE class {legitimate,fraudulent}

user_id STRING,
session_count INT,
total_session_time INT,
average_session_time INT,
session_time_deviation INT,
total_visibility_time INT,
average_visibility_time INT,
average_time_to_first_mouseover INT,
total_engagement_time INT,
average_engagement_time INT,
spurious_sessions INT,
click_throughs INT,
invalid_sessions INT,
average_time_between_sessions INT,
average_time_to_first_clickthrough INT

Didn't find what you were looking for?

Ask your question

Ask a Question

731 491 924 answers to any question