This thesis aims at exploring the topic of causal discovery by learning how to build and evaluate causality inference methods particularly suited for time series data. The student should propose a generic and flexible framework to streamline causal discovery experiments, including an appropriate pipeline for causal discovery experiments and a benchmarking platform for testing multiple state-of-the-art methods. The student should also demonstrate the practical usefulness and potential of causal discovery methods by evaluating its results in different datasets and domains. Public datasets in the areas of economics, health, or climate change, may be used in this work.
This thesis will explore a new scientific area at FhP-AICOS, with potential application in many domains of our expertise. The proposed developments will streamline the process of discovering, benchmarking, and evaluating causal discovery methods for inferring causal relationships in time series data.
Author: Fernanda Almeida
Type: MSc thesis
Partner: ESS – Escola Superior de Saúde do Porto