Abstract: 

The world of data is fascinating as well as confusing. We spend our days training models, evaluating their performance on validation data and crossing our fingers our model learned the right patterns. Shir presents the story of developing a machine learning model which automatically approves factoring deals, the challenges in the model training, evaluation and interpretation including a mysterious problem in the model creation and the details of its final solution. In addition she discusses data leakage, overfitting and a cutting-edge framework for model explanations (LIME).

Bio: 

Shir is a Lead Data Scientist at Bluevine which provides 100% online Invoice Factoring and Business Line of Credit. Her roles are focused around research of development of machine learning models for different aspects in risk analysis. Additionally, she organizes PyData Tel Aviv meetups and holds an MSc in electrical engineering and computers with a major in machine learning and signal processing from Ben Gurion University.