ML - model «Look-ALike»

There are about 7.5 million legal entities in Russia, and each of them has a unique set of needs. And now add to this all the variety of products and services that Sberbank Group offers to its customers. Under such conditions, we realized that we needed to solve this problem in a systematic way using machine learning models. The approach is based on: 1. Systematization of knowledge about customers that has already been accumulated by Sberbank. 2. Search for similar features, clustering of clients similar by certain features. Such analysis is not possible even for several dozens of analysts, that's why ML-model was developed, as which the bagging of three boosting models (LightGBM) was used.