A Recommendation System for Visualizing Sensor Data
Project & Master Thesis (submitted in 2022)
Abstract:
IoT platforms manage technical assets and sensors and gather sensor data. Sensor data
enables an in-depth understanding of real-world industrial processes, which can lead
to data-driven decision-making. Visualizations are frequently used for analyzing data.
While domain experts are usually familiar with the context of their assets, they are
less experienced in sensor data visualization. Consequently, domain knowledge and
visualization skills should be combined to create effective visualizations.
This thesis presents a process and tools that support cooperation between domain
and visualization experts to analyze problems using visualizations. Preconfigured
visualizations will be provided in an easily accessible way. Domain experts formulate
problem statements, classify their intent, and select relevant assets and sensors. Upon
that, visualization experts can create visualizations. After the visualization has been
reviewed by domain experts, it can be published to all users of the system. In this thesis
tools and methods for common data preprocessing tasks and visualization configuration
tasks are designed and implemented. For data preprocessing, frequently occurring
preprocessing steps, such as calculating derived values or differences to previous values,
are provided.
A recommendation system supports the configuration of visualizations. The re-
commendation system uses a rule-based approach. Recommendation guidelines by the
visualization community are translated into configurations of visualizations by using
visualization grammar. The rules include pre-existing visualization configurations,
characteristics of the problem statement, and statistical properties of the data. A recom-
mendation algorithm evaluates the rules and generates recommendations so inputs
can be auto-completed when configuring a visualization. It also suggests alternative or
advanced visualizations. The prior work of domain experts and visualization experts
lowers the entrance barrier for users to analyze sensor data through visualizations.
