Finding temporal classification data sets can be difficult. As previously mentioned (Section 6.2), although many data sets are inherently temporal, many people are eager to convert them to propositional form immediately. Although time series data is easy come by, what is hard to obtain from experts is the classification - this exclude many of the dynamic-systems or financial time series data sets.
The first data set, the Auslan data set, was selected because the author had collected the data himself and was one of the motivating problems behind this research; there did not seem to be many different algorithms for recognising time series. Secondly, unlike other data sets, rules produced by the system could be directly compared with the definitions found in the Auslan dictionary [Joh89].
The second data set, the ECG data set, was selected for a number of reasons. Firstly, work had recently concluded on a ``propositionalisation'' of the data set [dC98] that converted an ECG into a list of attribute values. This would allow easy comparisons between results using domain-specific information and a temporal classification system.