In Strava, not all users are honest. Some cheats, paying for example, other people to run instead. Others believe they are running while they move by bicycle or by car. The social network of athletes, which claims more than 150 million athletes, is aware of them and tries to put these bad practices.
For this purpose, he shared his efforts this Wednesday, May 21 to “guarantee ranking righteous and legitimate.” Strava in fact offers classifications according to the segments, that is, several parts of a route. Therefore, users are put in competition, which, for example, can obtain the title “King of the Mountain” (Kom) or “Mountain Queen” (Qom) if they are faster in the general classification in a segment.
For several months, Strava has been betting on artificial intelligence to detect these cheats.
4.45 million incorrect activities eliminated
Strava was able to detect these activities with the “classification integrity” function. Released in May 2024. It is able to “automatically inform irregular, unlikely or impossible activities registered on the platform,” the social network said in a press release.
This function benefited from a first improvement last February, with each activity placed in Strava analyzed by its automatic learning model. “This last analysis 57 different factors, such as speed and acceleration, to determine if part of the activity has been recorded in a vehicle.”
Therefore, when such activity is detected, the user is invited to make it private or to “readjust.” This same technology is also used to prevent bicycle outings from appearing in electric classifications and bicycles in hiking classifications.
An initiative that allowed him to eliminate “4.45 million activities whose type of sport was incorrect or registered in vehicles, which helped restore Kom and Qom to reflect real performance.” And this Wednesday, Strava announced that he continues to develop the applications of his automatic learning model to “remove the first 10 classification of racing and bicycle races.”
Source: BFM TV
