In this session, we will talk about our company pilots and real-world projects that use various popular and "rare" machine learning algorithms: from recommendation systems to deep neural networks. We’ll consider the technical realization on Java (Deeplearning4j), PHP, Python (keras/tf) platforms using the Apache Mahout (Taste), Apache Lucene, Jetty, Apache Spark (incl. Streaming) open source libraries, Amazon Web Services tool spectrum. We'll underline the importance of certain algorithms and libraries, the relevance of their use and demand on the market.
Let's consider the following implemented projects:
- Clustering Bitrix24 users with Apache Spark
- Calculating the probability of user leaving (churn rate), future profit (CLV) and other business metrics in big data and highload context
- Collaborative recommender system for more than 20 000 online stores
- LSH commodity catalog clustering method
- Content-based recommender service for more than 100 million Russian internet users
- Bitrix24 technical support calls classifier based on a neural network (anything but n-gramm models)
- Chatbot answering to questions on the basis of a neural network that connects semantic spaces of Q and A.