The focus towards multi-material and lightweight assemblies, driven by legal requirements on reducing emissions and energy consumptions, reveals important drawbacks and disadvantages of established joining processes, such as welding. In this context, mechanical joining technologies, such as clinching, are becoming more and more relevant especially in the automotive industry. However, the availability of only few standards and almost none systematic design methods causes a still very time-and cost-intensive assembly development process considering mainly expert knowledge and a considerable amount of experimental studies. Motivated by this, the presented work introduces a novel approach for the methodical design and dimensioning of mechanically clinched assemblies. Therefore, the utilization of regression models, such as machine learning algorithms, combined with manufacturing knowledge ensures a reliable estimation of individual clinched joint characteristics. In addition, the implementation of an engineering workbench enables the following data-driven and knowledge-based generation of high-quality initial assembly designs already in early product development phases. In a subsequent analysis and adjustment, these designs are being improved while guaranteeing joining safety and loading conformity. The presented results indicate that the methodological approach can pave the way to a more systematic design process of mechanical joining assemblies, which can significantly shorten the required number of iteration loops and therefore the product development time.