I am working on building a classifier to predict crime and the target variable that needs to be predicted is the category of crimes. In my dataset I have about 205321 observations these are broken down into the following categories:
Assault - 110609
Break and Enter - 43214
Autotheft - 23330
Robbery - 21500
Theftover - 6668
I split this data set into training and test sets - training 70% and test 30%
The simple decision tree that I applied only predicted the assault and break and enter categories. The other 3 categories were not predicted at all. I am leaning to believe that this is probably because of the imbalance. I need to address this in the training set. How do I go about doing this in R? I am also open to a weighted classifier.
Overall Statistics from the model:
Accuracy : 0.5626
95% CI : (0.5587, 0.5666)
No Information Rate : 0.5358
P-Value [Acc > NIR] : < 2.2e-16