31 July 2009

Non-predictable analytics

Each of us builds a matrix of predicted outcomes based on specific incomes. We use these to guess how early we should leave for an event, what time we should go to bed, when we should eat lunch, what the weather will be like. Key to build a good predictive matrix is getting good data, reliable data through a process called indexing. I have noticed that some things that should be simple to index and prove nearly impossible to index.

First example, the traffic patterns for the section of the city my parents live in. One morning I can leave at 9:45am and face nothing but solid traffic. The next morning, still a weekday, I can leave at the exact same time and face nothing but open roads. This varying of traffic seems to have no tie in to the day of the week or even hour of the day. Though I am sure that if one set far enough back to look at enough data one could find enough of a pattern to build an acceptable matrix. Such is, however, beyond my skill.

Second example, perhaps even more unpredictable is my sisters’ wakeup time. They seem to go to bed at nearly the same time each night. But come morning, the wake up times vary by hours, which I am grateful for. Sometimes my sisters wake up at 7:30am and sometimes they don’t wake up until after 9:30am. Sometimes they wake up at the same for a whole week. Other times they wake up at different times each morning.

As a whole I have realized that some things are really easy index and thus predict, some things are not.

No comments:

Post a Comment