Dynamic TSK systems supported by fuzzy rule interpolation: An initial investigation

Takagi-Sugeno-Kang (TSK) Systems form one type of conventional fuzzy rule inference system, providing an effective approach for performing prediction and regression tasks. In a real-world application, the inputs are usually varying against time, thereby requiring dynamically maintaining the rule base in order to maintain and possibly improve the efficacy of such a system. Situations may become more complicated if the training data does not sufficiently cover the problem space. Fuzzy Rule Interpolation (FRI) systems may help, whilst most of which follow a static approach, tending to process a large amount of interpolated rules which are generally discarded once the results are derived. Yet, the interpolated rules may contain potentially useful information. This paper presents a dynamic TSK system by exploiting such rules to support subsequent inference and promote rule bases. The obtained intermediate rules are directly added into the sparse rule base until it reaches a certain size. Afterwards, a clustering algorithm is employed to categorise rules into different groups so that an interpolated conclusion can be computed using the closest rules selected from a small number of closest rule clusters. Through systematic experimental comparisons with the conventional static approach, it is demonstrated that the proposed dynamic TSK system not only improves the overall reasoning accuracy but also reduces the interpolation overheads by avoiding the need for interpolations of experienced similar observations.