As mentioned in this Kaggle discussion, you have 12 datasets consisting of 8,000 data points, each with 28 features. The data is generated by applying one or two different smooth functions to the data points randomly. Can current data-mining techniques identify which dataset has more than one function applied to its data points?