Guest blog by Jay Gendron, Associate Data Scientist at Booz Allen Hamilton; Author; Data Analytics; Speaker.
In today’s edition of “Coffeehouse Connect” we take a look at a major predictive event in the United States that occurs each year on February 2.
Today is Groundhog Day. It occurs in the chilly time of the year aroundCandlemas, a midpoint between the winter and spring. On this day we await the prediction of Punxsutawney Phil – a 20 pound prognosticator from Pennsylvania. Learn more about the history and hullabaloo from this Groundhog Day Podcast.
Prediction is a mark of the human condition. Why are there so many cases of people commonly mistaking correlation with causation? There is a burning desire for us humans to know the future. You can see some wonderfully silly examples inSpurious Correlations: 15 examples recently posted by Laetitia Van Cauwenberge on Data Science Central.
Today, Punxsutawney Phil did not see his shadow and predicts an early spring. How well can we trust his prediction? Don’t bet a Starbucks on it! Phil has predicted correctly 13 of 28 times since 1988. According to the National Centers for Environmental Information (NCEI) in Asheville, NC
There is no predictive skill for the groundhog during the most recent years of analysis
Here is a USA Today infographic showing the historical predictions over 129 years.
Simply stated, Phil’s prediction is about as good as a coin flip. Yet prediction power is growing. Global competitions like Kaggle have brought together leading predictive analytics practitioners to tackle the toughest problems in fields like commercial retail, health, and many others.
There are many things media analysts predict: oil futures, political debates, Oscar Award winners. Despite their importance on society, none of these predictive events is a national observance like Groundhog Day. Data Scientists rejoice! This is our day – National Predication Day? How will you be celebrating?