Predictive Modeling question on Weka I would like to predict the number of flu cases in the future using predictive modeling. I am very new to statistics, so I'm not sure which classifier to use in this case.
For the attributes, I'm using historical data (2 years) as well as geographical location and age groups.
Are these attributes enough?
What classifier should I use? 
Can I do this on Weka? If not, suggest alternatives please.
Thank you for your help. Any help is appreciated.
 A: You really need to spend more time learning statistics or consult with a professional statistician.
Your question is similar to asking: "I have had a biology class and now want to perform surgery, can I do this using a rib spreader?"
The amount of information that you need to proceed is far beyond this medium, a few more good statistics classes and/or some meetings with a statistical consultant will help much more than anything that we could include in an answer here.
A: Chances are this is not enough by itself, but you definitely have a start! To give you a bit of direction, you should look at the size of your data set, and consider whether you have enough observations to make create a convincing model. It's pretty rare that anyone knows the answer to this question the first time they want to answer a question like yours in a new field, so don't worry if you don't know! My first step is usually finding studies that have been published which you address similar questions as mine, and, perhaps, using similar independent variables as the ones I'm considering. For example, you could look into the typical size of data sets in other epidemiological studies that have either looked at flu incidence in terms of geography and at-risk populations (I imagine there are several!), or a study using a similarly contagious outcome. You should pay particular attention to their methods sections and the way they went about constructing their model. For many of these, they probably describe the way they determined that they had enough data to answer their question at all (i.e., power analysis), and the approach they used to determine which predictor variables were actually important for their model (i.e., model selection and comparison). Hopefully that gives you a place to start, but let me know if you have a more specific question!
