Павел Филонов — Сложности формулирования требований в ML-проектах; как поможет пирамида метрик

Learn more about the Flow conference: https://jrg.su/CAm5kF — — Data science specialists are adept at using metrics to evaluate the quality of their models. But these metrics are often difficult to map to real business needs. And sometimes, even data science specialists are poor at explaining their meaning. Ask them, for example, to explain the Kullback-Leibler distance to the average person, and they'll realize they don't understand it themselves. Understanding how data science metrics can impact final business results is a crucial issue that needs to be clarified at the earliest stages of a project. This can be done by constructing a metrics pyramid, which will help understand how these non-trivial ROC-AUCs can be converted into real payroll savings or other important business metrics. The talk uses examples from real projects to demonstrate the steps for building such connections. To do this, participants will take a brief dive into the world of data science metrics. Download the presentation from the Flow website — https://cutt.ly/AwDshDfM #metrics #datascience