Discrete looking outputs from Continous inputs

Tech-Archive recommends: Fix windows errors by optimizing your registry

Hi.

I am building a prediction model that has 6 continous inputs and a single continous output.

Using decision trees, linear regression, and neural network I constantly get the same 4 or 5 predicted output values. From investigating the decision tree model, it appears that there is a three way split at the top node and then all other parameters are input to the regression formula. Apparently, they are not weighted very heavily because varying the values of those parameters is having a negligable effect.

As a result, my scatter plot has a straight line of predicted values that are too large on one side of the ideal line and too small on the other. Is there any way I can make this output act more like a continous variable?

Thanks!
.