An Interactive Visualisation Tool for Interpreting Predictive Machine Learning Models of Event Organisation Data (DAAV)

In recent years, the usage of machine learning models (ML models) and especially deep neural networks (DNN) outside of the world of research has increased rapidly. One of the major challenges when working with ML models is to correctly and efficiently interpret the results given by a model. Additionally, understanding how the model came to its conclusions is a very complicated task even for domain experts in the field of machine learning. For laypeople, ML models are often just black-boxes giving some results. The lack of understanding of a model and its reasoning leads to users not trusting the model predictions. Especially in what-if analysis, being able to interpret a model prediction is of utmost importance. In this thesis, we work with an ML model trained on event organisation data. The goal is to create an interactive visualisation tool that enables event organisers to efficiently work with the mentioned model. The main user goals in this scenario are to maximise profits and being able to prepare for the predicted number of visitors. Tasks to achieve this goal include: finding optimal event and marketing parameters and performing what-if analysis to prepare for unexpected parameter changes e.g. change of weather. The proposed visualisation tool incorporates adapted versions of multiple state-of-the-art model-agnostic interpretation methods like partial dependence plots (PDP) and feature importance. Since model-agnostic methods are independent of the ML model, they provide high flexibility. Many state-of-the-art approaches to explain ML models struggle with being simple enough to be understood by laypeople. Since our target group of event organisers cannot be expected to have a vast amount of technical knowledge, especially in the field of machine learning, one focus of this thesis is to overcome this problem. Therefore, our visualisation tool is created using a human-centred design approach performing multiple case studies with potential users during the whole development circle.