Yury Svirydovich

Developed and implemented new pipelines for CLV and churn probability prediction, achieving up to 30 percentage points improvement in prediction accuracy. Tasks included: -Data preprocessing and normalization -Data resampling -Outliers removal -Dimensionality reduction -Identifying latent subgroups using cluster analysis and data distribution-based heuristics -Implemented models: Bayesian hierarchical, tree-based, and deep learning for regression -and classification Built Tableau visualisations and reports

-Developed and deployed a series of bed disposal prediction models based on LSTM, CNN-LSTM, probabilistic deep learning and XGBoost -Designed and tested statistical hypotheses related to rehabilitation data -Developed a series of classifiers and regressors for various clinical outcomes prediction, based on classical models, tree methods and deep architectures -Implemented a BERT-based model for keyword extraction from freetext fields of anamneses and questionnaires

-Implemented players' subscriptions and churn prediction using time series and deep learning methods -Performed segmentation of players into purchase groups (Recency- Frequency-Monetization) using heuristic approaches and methods of cluster analysis -Carried out first time user experience analysis -Researched various aspects of players' behavior using statistical hypotheses testing and regression modelling -Determined prices for new bundle offers, reaching 35% more conversion compared to existing bundles