Survey on Dynamic Bayesian Network Software Tools
Hüseyin ÇAMBAŞI Özgür KURU Mehmet Fatih AMASYALI Sofiene TAHAR
AbstractBayesian networks are probabilistic graphical representations which are used to build models from data and/or expert opinion. They can be utilized for a wide range of tasks including prediction, anomaly detection, diagnostics, automated insight, reasoning, decision making, etc. Dynamic Bayesian Networks (DBN) are extensions of Bayesian networks with temporal support, which can be used to model systems that dynamically change by the time. Nowadays, DBNs are utilized in a wide range of applications including robotics, data mining, speech recognition, digital forensics, protein sequencing, and bioinformatics. Several software tools exist in the public as well as commercial domains that support modelling and simulation of DBNs. However, these DBN software tools differ in terms of features support, ease of use, documentation, user’s community, etc. Therefore, it has become important to establish various metrics for selecting the proper software tools for creating and simulating DBNs, such as cost, licensing, GUI, built-in support for inference algorithms, structural learning, data types, etc. The goal of this survey is the evaluation and comparison of existing software tools for building DBNs based on a set of users centered criteria.