Workshop on “Machine Learning with RapidMiner”
RapidMiner held an online workshop on “Machine Learning with RapidMiner” to assist and enable ARIADNE project partners on the use of Machine Learning. The workshop spanned over 2 days with a 4 hours session per day. The workshop offered information rich and highly condensed course work on applying Machine Learning in a hands-on fashion to solve predictive problems in RapidMiner. The topics presented in this workshop were task driven, i.e., the instructor explained a challenge and showed how to resolve it in RapidMiner Studio by creating ETL (Extract, Transform, Load) or ML (Machine Learning) pipelines.
The content covered in this workshop included data preparation, feature selection and engineering, dimensionality reduction, pre-processing, sampling, tips on avoiding data contamination, model validation, ensemble models, optimizing parameters of the ML pipeline process, performance and impact (gain or cost) evaluation of model, model deployment and update. Use of traditional Machine Learning methods as well Deep Learning was elaborated. The workshop served the dual purpose to familiarize the attendees with RapidMiner’s visual (coding-optional) approach to create complex Machine Learning pipelines.
The workshop also introduced RapidMiner’s automated Machine Learning tooling, namely the AutoModel, which is used to quickly generate correct ML models and allows various personas, especially those new to Machine Learning to be more productive. The generated models could be manually tuned, in respect of RapidMiner’s no black-boxing principle. The predictions were demonstrated to be optimally interpreted by automatically finding the classification threshold that maximized the performance metrics of interest. In the end, Prescriptive Analytics was introduced using RapidMiner’s model simulator, which allows to interactively inspect the influence of attributes on predictions and given an observation, can maximize prediction of a desired class, while subjecting attributes to constraints.
The workshop sessions were recorded for reuse and furthering education and enablement objectives. The video recordings and the exercise material (RapidMiner repository containing tasks, their solutions and results) are made available to project partners, so anyone interested may learn and reproduce the tasks and their solutions on self-managed pace.
