An Uncertain Temporal Similarity Proposal for Temporal CBR . ?

Domains in which time plays a relevant role, CBR systems require temporal similarity measures to compare cases. Temporal cases are traditionally represented by a set of temporal features, describing time series and temporal sequences. In the particular situation where these features are not homogeneous (i.e. combination of qualitative and quantitative information such as the patient evolution in medical domains), systems find difficulties to perform the CBR cycle. At the retrieval step, temporal similarity functions cannot directly apply the efficient time series techniques, and the temporal management needs a flexible representation for this set of heterogeneous features. Temporal constraint networks have demonstrated to be useful tools for temporal representation and reasoning. An approach to solve some retrieval and adaption problems could be the transformation of these heterogeneous sequences on uncertain temporal relations, obtaining a temporal constraint network. In this work we propose a temporal similarity measure based on: 1) the estimation of the temporal possibilistic constraint network proposed by HadjAli, Dubois, and Prade [1]; and 2) a similarity criteria based on the quantification of the implied uncertainty. This paper also motivates the proposal with a practical example in a concrete medical domain.