We are pleased to announce that the publication “Data Preprocessing as a Service – Outsourcing der Datenvorverarbeitung für KI-Modelle mithilfe einer digitalen Plattform” written by Erik Karger and Marko Kureljusic has been accepted and published in the journal Informatik Spektrum.

Paper Abstract:

Both in practice and in science, there has been an increasing interest in data-intensive methods, such as artificial intelligence, in recent years. The majority of these data science projects focused on the explanatory content and robustness of the models. Often neglected here was the process of data preprocessing, even though this takes up about 80% of the time of a data science project. During data preprocessing, which is also referred to as data preprocessing, data is acquired, cleaned, transformed, and reduced. The goal of this procedure is to generate a data set that is suitable for training and testing purposes of the data science models.

Thus, data preprocessing is a necessary process step for machine learning of correct patterns and relationships. However, data science projects often fail due to poor data preprocessing. For example, erroneous data is not identified in advance, potentially leading to incorrect correlations being learned. As a result, the explanatory power of data science models is significantly reduced. One way to solve this problem is to outsource data preprocessing to specialized professionals. With the help of a platform, secure and automated data exchange between customers and service providers can be ensured. This article addresses how a platform for data preprocessing can be used to enable more efficient and faster provision of data.

The article can be downloaded at the link below: