A Data Science Central Community
The Comprehensive Analysis of Time Series (CATS) is an increasingly important use case in the field of Big Data analytics. Cat videos on the Internet notwithstanding, the prevalence of time series is perhaps even more universally ubiquitous in big data applications: customer purchase histories, web click logs, social events, human behaviors, speech patterns, weather reports, climate science, numerical simulation science, spread of infectious diseases, market quotes, device monitors, biosensors, video surveillance cameras, basketball play-by-play histories, etc. Time series analysis is now a playground with many diversions.
Analytics on time series has historically focused on monitoring, pattern detection, description, characterization, and classification, with roughly equal attention to prediction. But the latter has become the dominant area of interest, given the vast quantities of historical training data (Big Data) on which we can now build remarkably accurate predictive models.
Predictive Analytics is one of two major categories of Data Mining that is particularly well suited to comprehensive analysis of time series - the other is Descriptive Analytics. In descriptive analytics, the focus is on hindsight and oversight: monitor, detect, describe, characterize, and classify! In predictive analytics, the focus is on foresight and insight: what will happen, what are the important causal factors, and what is the sensitivity of various outcomes to changes in the input parameters? A new kid on the block is Prescriptive Analytics, which addresses: what can be done to change or prevent a potential outcome?
One of the major Data Science modeling methods for these CATS applications has been Markov Models. You can read more about Markov Models (specifically related to Predictive Analytics for CATS) in the article "Markov Models & Predictive Analytics with Cats," published at http://www.bigdatarepublic.com/author.asp?section_id=3146&doc_i....
Follow Kirk on Twitter at @KirkDBorne