Are there any time series techniques that are considered data mining? For instance, what about a technique that would compute auto- and cross-correlations (with various lags) between thousands of stock price time series (intraday, each time series with 10,000 observations, that is about 60MM observations total), as part of a trading system? What about another technique that would create stock taxonomies based on these cross- and auto-correlations used as similarity metrics between stocks (time series), in a hirearchical clustering framework? In other words, an algorithm for clustering thousands of time series.
Would this be considered data mining, or computational statistics? Note that there's no statistical models behind what I described, it's a totally data-driven process. Any references?