Decision theory in expert systems and artificial intelligence
International Journal of Approximate Reasoning
Approximating probabilistic inference in Bayesian belief networks is NP-hard
Artificial Intelligence
Pattern recognition using neural networks: theory and algorithms for engineers and scientists
Pattern recognition using neural networks: theory and algorithms for engineers and scientists
Mining fuzzy association rules
CIKM '97 Proceedings of the sixth international conference on Information and knowledge management
Mining fuzzy association rules in databases
ACM SIGMOD Record
Bayesian Networks and Decision Graphs
Bayesian Networks and Decision Graphs
Inference via Fuzzy Belief Petri Nets
ICTAI '03 Proceedings of the 15th IEEE International Conference on Tools with Artificial Intelligence
Data Mining for Case-Based Reasoning in High-Dimensional Biological Domains
IEEE Transactions on Knowledge and Data Engineering
Hi-index | 0.00 |
An outstanding problem is how to make decisions with uncertain and incomplete data from disparate sources without NP-hard algorithms. Here we introduce a new reasoning methodology, fuzzy-inferenced decisionmaking (FIND), to solve this problem in polynomial time. In this methodology, a fuzzy-belief-state base (FBSB) is created from historical data of the states of a system by clustering the set of values for each state variable into three clusters upon whose center fuzzy set membership functions LOW, MEDIUM and HIGH are defined. The FBSB is mined for fuzzy association rules using the fuzzy set memberships to infer values for the missing data via these rules. When given an incomplete and uncertain observation of the system state, the observed state is completed via fuzzy association rules. Then each case in the FBSB is matched against the inference-completed observation to retrieve the best matching fuzzy belief state record that contains a decision as an extra variable. The process is analogous to case-based reasoning, but it uses fuzzification to ameliorate uncertainty and to complete missing data. The test results on real world data prove the effectiveness of this methodology.