In this thesis, I present my complete research work in the field of action rules, more precisely object-driven and temporal action rules.The drive behind the introduction of object-driven and temporally based action rules is to bring forth an adapted approach to extract action rules from a subclass of systems that have a specific nature, in which instances are observed from assumingly different distributions (defined by an object attribute), and in which each instance is coupled with a time-stamp. In previous publications, we proposed an object-independency assumption that suggests extracting patterns from subsystems defined by unique objects, and then aggregating similar patterns amongst all objects. The motivation behind this approach is based on the fact that same-object observations share similar features that are not shared with other objects, and these features are possibly not explicitly included in our dataset. Therefore, by individualizing objects prior to calculating action rules, variance is reduced, and over-fitting is potentially avoided. In addition to the object-independency assumption, temporal information is exploited by taking into account only the state transitions that occurred in the valid direction.The common nature of object-driven and temporal action rules made us believe that this work is general enough to solve a diverse fields of areas where it is highly needed. In our case study, we show how our approach was applied to an information system of hypernasality patients; our results were further investigated by physicians collaborators to confirm them.