by: Idris Mootee
We all know how bad the real estate market is. Although the pace of the housing slow down is now decelerating, but there is no bottom in sight. And seasonally, we usually see an uptick in activity in Aug, as last minute closings take place before the new school year begins. No such bump happened in Aug 08, as sales slipped 2.2% for the month. Is this the bottom? Who has the answer?
Earlier on, Hitwise (a company measures what people are searching for on sites like Google, Yahoo and MSN), took a look at the search phrase "homes for sale" and did a statistical correlation with the NAR data. Interestingly enough, this approach predicted the recent decline in real estate. Over the past year, the Hitwise system has been wrong only once. The logic seems to make sense, imagine tracking certain words to see people sentiments can help predict consumer spend. It needs a little more work to figure how to make these information make sense.
The technology industry particularly loves prediction, and keeps legions of forecasters and futurists in business. But many predictions are wrong, technologies often arrive late, and very few live up to the hype. Why, then, ... not ... harness the collective brainpower of employees by giving them virtual trading accounts and virtual money, and letting them buy and sell “shares” in such things as project schedules or next quarter's sales. What are, in effect, elaborate computer games might help tech firms spot trends and make more accurate forecasts. Yet, oddly, hardly anyone is using them in this way. Hewlett-Packard and Intel pioneered the corporate use of prediction markets, but neither seems to be using them other than experimentally. There is a lot potential applications I can think of.
This is an example of how we can leverage the power of the enormous amount of online data to foresee trends. Google this week unveiled a new site to track the progress of the common cold. Using the same keyword tracking technology found on Google Trends, it keeps an eye on people searching for queries involving the word "flu" and tracks them both by date and location.
The fascinating part is that its data set goes back to 03, and has been cross-referenced with the last several years of survey data from the Centers for Disease Control and Prevention (CDC). Google says that because its own system is based on a constant flow of searches as opposed to surveying techniques it's able to provide results one to two weeks faster than the CDC. The same trending technique could be used in tandem with other organizations to track contagious viruses or threats besides the common cold, including AIDS, bird flu, and Africanized honey bees. We can actually check every hour seeing how the flu is spreading, a wonderful distraction at work indeed
Original Post: http://mootee.typepad.com/innovation_playground/2008/11/can-search-engines-become-prediction-engines.html
Leave a Comment