To ensure Risk Mitigation of all aspects of our operations, we implement various Risk-Mitigation Tools at key aspects of the process: we implement Data Science & Machine Learning practices in almost all our points of contact with client and all over our regular business:

•  Application process
  Our platform is easy to integrate with different data providers like credit bureau and state registers in order to get 360degrees view of customer profile and propose tailored credit offer.
  We use state-of-art models to predict the outcome of loan, our stack of technologies include, but not limited to:
  
  • Latest and most powerful boosting (i.e. XGBoost, LiteGBM, CatBoost) models for predictions based on clients’ data;
  • Big Data and Neural Networks for model based on clients session data. This models used not only for application decisions, but also gives insight for marketing and fraud prevention models;
  • Ensembling to bind together different models built with different algorithms.
    
    
• Fraud prevention
  We use a complex system of rules and cross-checks of information from different data sources to be sure that we deal with clients not with cheaters who stole identity. We do it with help of:
  • Estimation of statistical significance of used rules and checks;
  • Tracking unusual client behavior on website / mobile app help us to reveal a cheater (using Neural Networks and Big Data).
    
    
• Marketing
  We do data-driven marketing that includes:
  • Using machine learning models to target marketing campaigns;
  • Reinforcement learning and next best action models for customize clients experience with our company;
  • Implementing AI powered chat-bot to help customers with getting information quicker and more precisely.
    
    
• Debt collection
  We automatized many things in this area and as a result save operational expenses and increase collected amounts, in particular:
  • Use reinforcement learning to define whole and consistent collection strategy, that is individual in each case and for each borrower;
  • Prioritize efforts of collection staff on most promising cases (by all possible actions) to increase collection effectiveness;
  • Use AI powered chat-bot and IVR-bot (which is available thanks to Recursive Neural Networks and Voice recognition technologies) to connect with borrower and save staff time.
    
    
• Outstanding Data Mining
  We use Data Science & ML techniques to obtain the next generation of features to feed them to our models:
  • Convolutional Neural Networks and Image recognition, that allows us to:
  • Recognize data in scanned files and compare it to data from other sources of information
  • Face detection technologies that are a core part of our fraud prevention system
  • Voice recognition that allows us to convert speech to text and dig from this text bunch of useful features for different models;
  • Getting fresh insights from bank statement using classification of them with Recursive Neural Networks;
  • Scanning behavior on websites and getting different session data to predict real need of the client and prevent fraudulent cases.

Such an un-parallel industry-innovative approaches ensures accuracy for our customers in their complex decision making process