Video from Avito Data Science meetup
Hello to all! Today we publish videotapes from the Data Science meeting for professionals in our Moscow office on June 24th. Under the cut there are reports on building recommender systems from specialists from Yandex.Dzen, OZON.ru and Avito, as well as detailed descriptions of the decisions of the finalists of our competition, which took place on the site Dataring.ru. And, of course, awarding its winners!
Evgeny Sokolov and Dmitry Ushanov told us about the main types of signals that they use in recommender algorithms, the complexity of scaling Yandex.Dzen and the nuances of working with content (classification, evaluation, metrics). How to combine explicit and implicit feedback from users, to get good scalable factors, to compare content and click information - see in the video.
Ksenia Boksha, a leading analyst at OZON.ru, talks about the recommendation system that is used in the marketplace. What technology stack (software and hardware) is used on the site, how OZON.ru cares about users and what types of recommendations provide the most sales, you will learn from the video.
')
In his report, Vasiliy Leksin, the Avito recommendations leader, talks about the tasks his team faces, the types of recommendations needed for our service, and the approaches used to build them.
Also in this video is a story about how the Avito team participated in the RecSys Challenge 2017 competition . We solved the problem of cold start of a reference job search service and tried in parallel two approaches: content-based and more general, based on matrix factorization.
On the experience of participation in the RecSys Challenge 2017, our other colleague, Andrey Ostapets, also spoke in great detail at a training session on machine learning at Yandex. His report starts at 17:50.
On June 24, we awarded the winners of the competition , which was held at the site Dataring.ru. Mikhail Kamenshchikov spoke in detail about the task of the competition, the data that was provided to the participants and the choice of quality metrics. And also summed up the statistics and results of the competition. And the finalists of the competition prepared reports on their decisions. Watch the video!
Thanks to everyone who came to the mitap! The entire playlist is available here . For those who do not like to watch the video, we have laid out the slides . Photo report published on our Facebook page , mark yourself and your friends!
We invite you to the following meetings for professionals. Look for announcements of events in Facebook AvitoTech , Twitter and Telegram .
Machine learning in Zen
Evgeny Sokolov and Dmitry Ushanov told us about the main types of signals that they use in recommender algorithms, the complexity of scaling Yandex.Dzen and the nuances of working with content (classification, evaluation, metrics). How to combine explicit and implicit feedback from users, to get good scalable factors, to compare content and click information - see in the video.
Recommendations in OZON.ru
Ksenia Boksha, a leading analyst at OZON.ru, talks about the recommendation system that is used in the marketplace. What technology stack (software and hardware) is used on the site, how OZON.ru cares about users and what types of recommendations provide the most sales, you will learn from the video.
')
What tasks solves the team recommendations in Avito
In his report, Vasiliy Leksin, the Avito recommendations leader, talks about the tasks his team faces, the types of recommendations needed for our service, and the approaches used to build them.
Also in this video is a story about how the Avito team participated in the RecSys Challenge 2017 competition . We solved the problem of cold start of a reference job search service and tried in parallel two approaches: content-based and more general, based on matrix factorization.
On the experience of participation in the RecSys Challenge 2017, our other colleague, Andrey Ostapets, also spoke in great detail at a training session on machine learning at Yandex. His report starts at 17:50.
Results of the Avito contest for building a recommendation system
On June 24, we awarded the winners of the competition , which was held at the site Dataring.ru. Mikhail Kamenshchikov spoke in detail about the task of the competition, the data that was provided to the participants and the choice of quality metrics. And also summed up the statistics and results of the competition. And the finalists of the competition prepared reports on their decisions. Watch the video!
Thanks to everyone who came to the mitap! The entire playlist is available here . For those who do not like to watch the video, we have laid out the slides . Photo report published on our Facebook page , mark yourself and your friends!
We invite you to the following meetings for professionals. Look for announcements of events in Facebook AvitoTech , Twitter and Telegram .
Source: https://habr.com/ru/post/332408/
All Articles