Self-Organising Networks (SON) was introduced by the next Generation Mobile Networks (NGMN) in 4G/LTE networks to automate network management and reduce operational costs. However, SON functions (SF) are provided by different vendors as black boxes making it harder to predict their performance in untested network environments. Also, deploying many SFs leads to conflicting objectives. These conflicts are expected to be more apparent with more tunable parameters in future networks. In this article, we propose empowering SON with distributed intelligence by using machine learning and association rules from historical and temporal network conditions. These rules can be applied to the network any-time similar conditions are experienced.