Machine Learning and Hybrid Models

In practice, model-based optimization can lead to substantial improvements in processes – whether planning and implementing supply chains, scheduling the tasks in a production chain, or for the layout and operation of a production plant. Reliable models that lead to meaningful, practice-relevant statements are an essential requirement for success in these challenges.

ITWM has many years of experience in adapting models to the current problem, in addition to training of the models by means of data obtained from the process. Modeling, simulation, and optimization (MSO) methods are employed in such a way that the result is the continuous improvement of the model parameters through continuous comparison of the model predictions with actual practice. This also enables data quality to be evaluated with increasing reliability, for example, for the early identification of outliers or system error.

In this way, data and models are given equal consideration – models learn from data (model learning), data can be compared and assessed with models (Smart Data). This approach has been implemented in a series of projects.

Example Projects

 

Grey-Box-Models

The aim of this project is to develop the methods to significantly improve both – availability and quality – of the model predictions, by interlinking the deterministic white box models with the data-driven black box approaches.

 

Optimizing Sales Promotion

The sales forecast for a trade promotion is one of many project examples in which smart data has proven itself superior to the pure, data driven, Big Data models.

 

Capacity Planning for Complex Value Streams

We developed a tool for proactive capacity planning that addresses challenges through practical modeling.