I have two databases.
One of them contains the number of this event per year. The other one contains a lot of indicators and its values per year.
I want to find those indicators that are most correlated to the numbers of this event, or maybe what indicators could explain the occurrence of this event.
Appreciate the help because I've just started to learn about data mining.
Thank you!
Hi
I suppose, using ANFIS and ANNs would be great for mentioned subject. I have done a couple of modeling and prediction similar problems by computational intelligence. You could catch a glimpse of them.
Combine the two databases over the associated area to obtain a single image. Then you can do any statistical analysis on a single table. For example; use the entropy, or correlation matrix.
To go a step further what Yasin recommended perform regression analysis as well which is tightly associated with correlations. From that analysis you can discover which indicators are significant (p-value) in describing your independent variable. Keep only the significant ones and after that get rid of collinear features. Now you have a good set of features and the regression coefficients also tell you how much any feature contributes to this event.
Thank you for the answers. It will help me a lot.
One more question, what do you think about Correlation-based Feature Subset? I was thinking about working with Weka. Is it appropriate for this kind of problem?
- Badges
- Science topic
- Similar topics
- Computer Science
- Data Mining