I've joked in the past that the data miner is to the BI analyst what the weather person is to the news reporter. One reports on the past, and the other predicts the FUTURE! Pretty cool when you look at it that way. So let?s do something meaningful today that can help our entire nation. Let's forecast the unemployment rate using the Microsoft SQL Server 2012 Data Mining Add-ins for Office 2010!
I started by downloading and installing the FRED Excel add-in which you can find here. I pulled the unemployment rate by month into Excel and formatted it as a table. I set aside the data from 2012 to use as a gauge to see how well my model does at predicting. Clicking the forecast button in Table Tools Analyze ribbon and I was on my way.
I've got great news! The unemployment rate is going down!
Comparing my forecasted values to the actuals shows there is some room for improvement.
Next we'll add the job openings figures to the mix. All I have to do is paste these values next to unemployment rate numbers and delete the rows from my previous forecast.
Unfortunately, this doesn't improve our forecast.
Next, I add the number of Hires for each month. It didn't improve my unemployment forecast; but, it did improve my job openings forecast.
I went on to try many other data sets including gas prices, compensation, the dollar index relative to other currencies and more. Although I didn't improve my predictions I hope I did manage to demonstrate the methodology one might use to see what factors influence the unemployment rate.