June 3, 2019 at 12:44Gazprom Neft invites you to visit: GPN Data Science Meetup
At first glance, it seems that only effective managers work at Gazprom Neft, and in IT there are a maximum of system administrators and seconded implementers of all sorts of Oracle and SAPs. From the second - you can find your own development (web fronts, mobile applications, software for managing gas stations), and, for example, the date of Scientists. Under the cut - about the internal mitap DS, which we decided to make external.
First, recall the business. Gazprom Neft is a large company, it has exploration and production, there are oil refineries, there is logistics and sales - we are engaged in them. In this part of the company are responsible for the delivery and sale of motor fuels - from factories to end customers. The most noticeable part for motorists is our gas station network. This is a large retail (more than 1800 stations in Russia, the CIS and Europe) and we sell not only gasoline (for example, we sell coffee in large quantities).
The tasks in DS are like any retail: sales forecasting, pricing, assortment optimization, logistics, marketing and geo-analytics, client profiles, outflow management, video and audio analytics, a bit of credit scoring. The work is carried out in different departments in small groups, while the tasks, methods and tools often overlap. To exchange experience, align competencies and discuss common problems and interests, we organized the Scientists Date Guild.
Free and not very Scientist companies gather for guild meetings every 3-4 weeks, listen and discuss colleagues' reports on their projects, topics, tools, and best practices. Recently, we began to invite external speakers. And now we are ready to invite interested listeners.
In Friday, June 7,, we invite practicing Scientists and everyone who is interested in the topics below to a meeting in our Lecture Hall (St. Petersburg, 14 Vilensky lane). You will hear three reports, and you can also announce your topic for the next meeting.
Registration here
Presentations this time:
“Interpreted ML: the task of estimating the demand elasticity function by price”
Ruslan Ermagambetov, Data Scientist, Gazpromneft
Ruslan will talk about what tools exist for interpreting models - “black boxes”, that is, extracting the nature of the relationship between a target and features, and will consider the most popular of them (Feature Importance in “wooden” models, LIME and SHAP). Then he will tell how these methods are used in the pricing project to evaluate the price elasticity function of customer demand.
“Architecture of Generative-Competitive Neural Networks (GAN), Examples of Their Application”
Maxim Balashov, a student at ITMO University, a specialist at the Gazpromneft Advanced Analytics Center
Report on the popular architecture of neural networks: basic architectural details, there will be examples of applications of architecture modifications.
“License plate recognition at Gazpromneft gas stations
Alexander Popkov, specialist at the Competence Center for Advanced Analytics (Gazpromneft), master's degree at SPbSU in the field of Information Business Analytics
Alexander will talk about the problem as a whole, will analyze in detail the application of the classical Viola-Jones method for detecting an object in an image (license plate number) .
Link to the broadcast here .
First, recall the business. Gazprom Neft is a large company, it has exploration and production, there are oil refineries, there is logistics and sales - we are engaged in them. In this part of the company are responsible for the delivery and sale of motor fuels - from factories to end customers. The most noticeable part for motorists is our gas station network. This is a large retail (more than 1800 stations in Russia, the CIS and Europe) and we sell not only gasoline (for example, we sell coffee in large quantities).
The tasks in DS are like any retail: sales forecasting, pricing, assortment optimization, logistics, marketing and geo-analytics, client profiles, outflow management, video and audio analytics, a bit of credit scoring. The work is carried out in different departments in small groups, while the tasks, methods and tools often overlap. To exchange experience, align competencies and discuss common problems and interests, we organized the Scientists Date Guild.
Free and not very Scientist companies gather for guild meetings every 3-4 weeks, listen and discuss colleagues' reports on their projects, topics, tools, and best practices. Recently, we began to invite external speakers. And now we are ready to invite interested listeners.
In Friday, June 7,, we invite practicing Scientists and everyone who is interested in the topics below to a meeting in our Lecture Hall (St. Petersburg, 14 Vilensky lane). You will hear three reports, and you can also announce your topic for the next meeting.
Registration here
Presentations this time:
“Interpreted ML: the task of estimating the demand elasticity function by price”
Ruslan Ermagambetov, Data Scientist, Gazpromneft
Ruslan will talk about what tools exist for interpreting models - “black boxes”, that is, extracting the nature of the relationship between a target and features, and will consider the most popular of them (Feature Importance in “wooden” models, LIME and SHAP). Then he will tell how these methods are used in the pricing project to evaluate the price elasticity function of customer demand.
“Architecture of Generative-Competitive Neural Networks (GAN), Examples of Their Application”
Maxim Balashov, a student at ITMO University, a specialist at the Gazpromneft Advanced Analytics Center
Report on the popular architecture of neural networks: basic architectural details, there will be examples of applications of architecture modifications.
“License plate recognition at Gazpromneft gas stations
Alexander Popkov, specialist at the Competence Center for Advanced Analytics (Gazpromneft), master's degree at SPbSU in the field of Information Business Analytics
Alexander will talk about the problem as a whole, will analyze in detail the application of the classical Viola-Jones method for detecting an object in an image (license plate number) .
Link to the broadcast here .