Title: Learning qualitative probabilistic networks with reduced ambiguous signs from complete data

Authors: Qian Wang; Wei Xu; Yali Lv; Lifang Wang

Addresses: School of Computer Science and Technology, Taiyuan University of Science and Technology, Taiyuan, 030024, China ' School of Information, Shanxi University of Finance and Economics, Taiyuan, 030006, China ' School of Information, Shanxi University of Finance and Economics, Taiyuan, 030006, China ' School of Computer Science and Technology, Taiyuan University of Science and Technology, Taiyuan, 030024, China

Abstract: Qualitative probabilistic networks (QPNs) are the qualitative abstractions of probabilistic networks or Bayesian networks, summarising probabilistic influences by qualitative signs. But the ambiguous sign is undesirable as it leads to uninformative results upon QPN inference. In this paper, we focus on learning QPN with reduced ambiguous signs from complete data. We first extend the definition of qualitative influences based on the situational signs, and prove the theorem on the range of qualitative signs. Second, we design an algorithm for learning QPN with reduced ambiguous (RAQPN Algorithm). The QPN structure can be optimised by the K2 search algorithm, and the corresponding qualitative signs can be acquired by ordering the conditional probabilities that are represented by the frequency formats. Furthermore, when the ambiguous signs appear, they can be reduced by being given the situational state of the network. Finally, experiment results verify the validity and feasibility of the proposed methods.

Keywords: qualitative probabilistic networks; ambiguous signs; Bayesian networks; K2 algorithm.

DOI: 10.1504/IJCSM.2023.135050

International Journal of Computing Science and Mathematics, 2023 Vol.18 No.4, pp.392 - 404

Received: 16 May 2022
Accepted: 06 Jun 2022
Published online: 28 Nov 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article