Journal of Digital Information Management Nadjib Mesbahi, Merouane Zoubeidi, Abdelhak Merizig, Okba Kazar 17 5 2019 https://doi.org/10.6025/jdim/2019/17/5/270-288 http://dline.info/fpaper/jdim/v17i5/jdimv17i5_2.pdf Today, enterprises use a variety of applications to manage day-by-day business activities using a large centralized database. Since a huge amount of data stored in this centralized database produced by the daily use of several systems, it is important to integrate decision- making tools to analyse and interpret these business data. For this purpose, Data Mining is a powerful technology that promote information and knowledge extraction from large databases. In this paper, we present an agent-based approach for extracting business association rules from centralized database systems. This approach combine paradigm of multi-agent system and the association rules as a data mining technique to build anefficient model. It is relying on the intelligent partitioning of data to make the execution of business association rules in a parallel and distributed way from a large centralized database. To validate our approach, we applied it during the realization of a real case study on ERP database at the National company of Well Services (ENSP)using JADE platform with machine learning WEKA toolbox for association rules mining. The developed system has been compared with the classic association rules algorithms and has proved it is more efficient and more scalable. The main objective of our work is to improve and accelerate the process of extracting association rules by business through centralized database systems. As a result, the decision process of these systems becomes more improved.