Statistical Analysis of Massive Data Streams: Proceedings of a Workshop (2004)
Chapter: 10. SUMMARY
Figure 10. Certain types of traditionally tedious computations can be performed very quickly once the underlying empirical distribution of the data is known. This tool takes advantage of that fact for performing interactive financial analysis of the value (in currency) of a stream based on pricing models input by the user. The empirical distributions can either be extracted in real-time or archived for later comparison and analysis.
Because of the compactness of the models this kind of computation can be performed by the client in a few milliseconds, which enables “what-if” modeling based on actual or forecast-extended distribution models of subscriber usage behavior.
Other tools currently in development include network analysis and forecasting for capacity planning as well as a suite of security analysis tools. A more complete discussion of how these more advanced tools take advantage of streaming analysis will be the subject of follow-on papers.
10. SUMMARY
High-momentum data streams and rivers can be expensive to store and require long processing times to analyze using the traditional store first, then analyze later techniques. Although some types of analysis will always require this time-proven approach, we are discovering that a great deal of valuable insight can be extracted from these streams prior
