Can we transform our data into events and mine association rules from these?

Can we "scan" our time series data for a single test looking for correlations, or maybe just similarity if time does not allow complexity?

Let's label sequences as "Sustain" or "Not Sustain". We can then use an existing system such as ARMiner on the resulting data set. Before it was said that we aim to use time series attributes directly for mining association rules. It may appear as though we are simply categorizing or simplifying the time series attributes into single valued attributes. It can thought of as picking one characteristic of a time series attribute to focus on, simplifying our initial implementation and proving a framework in which to conduct further study. The resulting rules would explain or concern that characteristic of the time series. We have simply chosen a characteristic that applies to the whole time series rather than a portion at a time.

Let's not forget the idea that the single valued attributes might be used to determine the range and number of values for time series attributes in an instance.