An interactive web-based machine-learning suite

Abstract

We propose a comprehensive and interactive web-based machine learning suite that will allow researchers and practitioners to use a wide collection of basic and experimental learning algorithms and sophisticated visualization and analysis tools. Component based frameworks incorporating data input/output, pre-processing, classi cation, clustering, regression and visualization schemes have been implemented before in various programming languages, for use on di erent platforms, to operate using a variety of data formats. ML-Lab includes a large variety of machine learning algorithms for resampling, feature selection and extraction, classi cation and ensemble methods, as well as tools to visualize the experimental results of statistical comparison and testing. It provides a sophisticated and easy-to-use interface for creating work ows and a component-based framework intended for both experienced users and also those who are just entering the eld. The collection of machine learning algorithms are implemented in Python, a modern easy-to-use scripting language with clear but powerful syntax and extensive set of additional libraries. ML-Lab has an extensible architecture and allows adding new capabilities to the system infrastructure easily.