Bio: PhD in Operational research from Tel Aviv University
MBA from Recanati Business School - Tel Aviv University
MSc in Condensed Matter Physics from Tel Aviv University
BSc in Physics from Tel Aviv University
CTO and founder of DMWay Analytics
15 years of practical experience in Advance Analytics including Actuary, behavioral targeting, Credit risk, Foresting, customer retention, fraud and more.
Verity of positions in the Israeli startup industry.
The Automation of the Feature Selection Process
Abstract: Feature selection, the process of finding the most influential predictors in a regression model, from amongst a much larger set of potential predictors, is definitely the most complex problem in building large scale regression models. It is a hard combinatorial problem which does not have an analytic solution, often requiring a resort to heuristic solution.
In this talk we describe an expert system approach for addressing the feature selection problem that mimics the way an experienced data scientist goes about in building large scale regression models. This approach also address such complications as non-linearity, multi-collinearity, noise, missing values, outliers and others, on the fly, yielding models which, by and large, are better than models built by human experts but at a fraction of the cost and much less time.