A former quant, my day job is now at Los Alamos National Laboratory. Like Google, I'd like to help make the world a better place, so I coupled a very powerful data correlation/prediction engine I built before coming to LANL to an Excel/VBA macro which scrapes Google Search volume from a (or your) dictionary, as well as an Excel/VBA macro for scraping pairs of phrases to compare relative volume. (Download Excel files at http://precisiondatamining.net/) I then have "the" Fortran 90 engine find all significant correlations in the Google Search volume data (plus weather data, financial data, etc.) and generate predictive models for future Google Search volume for a given search phrase, or for energy demand, etc.. It's great for macroeconomics, SEO, etc, I would be willing to share the F90 correlation/prediction engine if there is any interest. Cheers, Alex
At http://precisiondatamining.net/ you will finds lots of links to Word dcouments about applications of the technology to SEO (search engine optimization), Energy Demand Forecasting, Threat Reduction Alarming, Genomics and Proteomics for advancing the pace of development of individual molecular medicine.
If you have a worthy cause, and you can prove yourself legitimate, I would consider either giving you the Fortran 90 executable, or running it for you and sending you its manifold kinds of correlation and prediction dumps, including multivariate models in multiple function bases. Worthy APPS would include financial, macroeconomic and Google Insights data to project potential econometric states in the near term 2-4 week future.
CHECK THIS OUT FOR YOURSELF----What Geico should know:
(Who would have thought that "writing resume" or "cheap car insurance" mimicked each other? Maybe Geico would like to know. http://www.google.com/insights/search/#q=writing%20resume%2Ccheap%20car%20insurance&cmpt=q) I'm interested in threat reduction applications, energy efficiency, etc. Go out and use the data for your internal purposes--be better, do better, run better, understand better.
The Excel Macros run in C: _Main (you can change the macro)
Two files are produced:
inDeck.txt The weekly Google Search Volume data. Import it with Excel and graph it, or whatever. Use the delimeted, and space options when importing a text file.
inHead.txt, a header for the F90 executable telling it how many data streams were captured, and how many weeks of the Google Insights file are to be used starting from 4 Jan. 2004--currently 256 weeks.
Example: GoogleInsights.xls Dictionary Worksheet explained:
There are 17 phrases to try to scrape from Google Insights--some may have zero or insufficient data. There are 256 weeks worth of data--but since I write to Excel 2003 in rows, you can only pull either the first 255, or 2 thru 256 weeks. (You could modify the macro to write to colums for Excel 2003 or older) Excel 2007 doesn't have this problem. 5 is for the F90 code, telling it to use a 2^5=32 week sliding window while looking for correlation. Warning: The most stable platform seems to be XP with Excel 2007. Excel 2003 doesn't really close files when told to do so by VBA, and hence, the macros will eventually crash. Not so for Excel 2007.
17 255 5
nuclear
great satan
israel
hezbollah
uranium
plutonium
scientology
atomic weapons
nuclear weapons
bombs
anfo
atom bombs
anarchist cookbook
pipe bombs
suicide bomber
iran
iraq
By viewing this simple macro you can see how to modify it as Google Insights puts out more weeks.
By viewing the pairs macro, you can tinker to try all the way up to 5 search phrases, the Google Insights limit.
If you want, I can even show you how to modify the macros to scrape only one particular country.
Dr. Alex Alaniz, Ph.D.
If interested, I would consider either giving you the Fortran 90 executable, or running it for you and sending you its manifold kinds of correlation and prediction dumps, including multivariate models in multiple function bases.
|
