Skip to main content

Semantic Reasoning for Predictive Data Analytics

Jiaoyan Chen ( University of Heidelberg )

Semantic reasoning has shown its advances in many stages of the predictive data analytics process. The report first presents our study on solving the concept drift problem in stream learning. Data stream learning has been largely studied for extracting knowledge structures from continuous and rapid data records. In the Semantic Web, data is interpreted in ontologies and its ordered sequence is represented as an ontology stream. Our work exploits the semantics of such streams to tackle the problem of concept drift i.e., unexpected changes in data distribution, causing most of models to be less accurate as time passes. Second, the report introduces our ongoing work about utilizing semantic reasoning for improving transfer learning, which aims at applying a model learnt from one data domain to another different data domain or from one task to another different task. We try to extend the first study by learning the weights of the entailments so that a knowledge base can be automatically embedded into a vector. In both studies, we use two real world problems, air quality forecasting in Beijing and bus delay forecasting in Dublin for experiments.

Share this: