Fields of research LWS
The Laboratory for Web Science actively researches in the research fields of GeoHealth Analytics and Data Science for Energy, Environment and Materials. Results from both research fields flow into industry through projects.
The term data science covers a wide range of individual aspects and tasks, including data selection and linking, analytics, technical implementation, corporate relevance and ethical issues.
Information and knowledge are the raw materials in every company which, if used correctly, can offer a decisive economic advantage. We distinguish between structured data, which are stored in a defined and fixed form (e.g. in databases and tables), unstructured data (data, which are available in all possible formats and platforms, e.g. e-mails, audio and video files etc.), and semi-structured data, e.g. HTML and XML.
Due to the development of information technology, ever larger amounts of data are available. These often contain implicit knowledge which, if known, would have great economic or scientific significance. Data mining is an area of research that deals with the search for potentially useful knowledge in large amounts of data, and machine learning is one of the key technologies within this area.
The aim of machine learning is to develop methods for the implementation of adaptive technical systems. Machine learning can be divided into two large classes: In supervised learning, systems for the classification and modelling of functional dependencies are trained based on existing example data. The goal of unsupervised learning is to autonomously find relevant structures in data.
The LWS evaluates internal and external data using machine learning methods.
Deep learning as a subset of machine learning opens up further undreamt-of possibilities in dealing with data. These are neural networks with many layers between data input and output. These deep networks enable the mapping of classification functions with high non-linearity, i.e. very complex tasks can be solved. The best example are convolutional neural nets, which as a disruptive technology have led to enormous progress in the field of image and signal processing.
The LWS has built up a lot of know-how and uses this technology for projects.
Recommender systems can be divided into three large classes:
- Content-Based: By analyzing already consumed content, services, etc. of a user, the system calculates the probability that a not yet consumed object is interesting for the corresponding user.
- Collaborative filtering: By analyzing the consumption behavior of all participants in a system, the system generates a user-user or object-object network by projecting a bipartite graph. Based on this network, recommendations are given to the users.
- Network based: In contrast to collaborative filtering, the bipartite graph is analyzed directly. This has the advantage that no information is lost through the projection. (see: B-Rank: A top N Recommendation Algorithm).
The field of application of Recommender Systems is almost inexhaustible. From online platforms to expert systems in medical diagnostics. The methods for researching and developing such systems come primarily from the field of statistical learning.
The Laboratory for Web Science develops new algorithms and system architectures, followed by a subsequent evaluation of applicability to industry projects. In addition to application-oriented developments, theoretical models are also developed. These models should provide a deeper understanding of which algorithm is suitable for a given database.
GeoHealth analyses the relationship between people, space, time and health. In the scientific sense, we interpret "GeoHealth Analytics" as an intersection of the fields of geoinformatics, health informatics and data science.
The scientific tasks range from epidemiological questions (spatial reference object: earth) to local tasks in health documentation (spatial reference object: human). We use the tools of data science for storage (Big Data), acquisition (Internet of Things) and analysis (Machine Learning) to transform data into valuable information. In order to fulfill these tasks, we always pay attention to the optimization of process-related and economic aspects.
The research field "GeoHealth Analytics" as part of the Laboratory for Web Science (LWS) deals with spatial and/or health-related tasks in research and teaching and examines the applicability of the associated solutions or prototypes via third-party funded projects or services in cooperation with industry and public institutions. Thus the research of the LWS generates knowledge, which is transferred to practice, provided that this also creates an economic benefit.
Data Science for Energy, Environment and Materials
Climate change and the energy transition are challenges that engage thousands of people around the world and where research can be a driving force for the transition to sustainable energy consumption. Materials science integrates into this field through the application of Data Science methods in sustainable energy management. Research questions from the energy and materials fields can benefit from artificial intelligence methods. The goal of the Data Science for Energy, Environment and Materials research field is to provide an interface between these research directions. The Laboratory for Web Science carries out application-oriented projects and services in cooperation with companies as well as national and international research institutions.