I am planning to create a predictive model. However my approach uses a dataset which is not similar to prior works. So I can't apply my model on the previous datasets and those models will not work with my dataset. How can I validate my research in such cases?