There are two subsamples in the dataset - on one the target is real(valid), and on the other it is approximate (I don't know how it differs yet, on one sample the real price of an apartment, and on the other the price from ads, you need to predict the real one, of course). Any ideas what to do about it? I have two ideas - to normalize the target from ads (to bring the expectation and variance to an real target), and also to modify the loss so that it punishes more for an error on an real target. There are no more ideas. Therefore, I ask for help.
upd: Sorry for not giving enough details. The the problem is to predict apartment price, which is made by professional realtors. The dataset is plenty features (like amount of shops in some radius, distant to closest school, etc.), and we have two subsets in this dataset: first is dataset with prices developed by realtors, and second is subset with prices from advertisings. The goal is to predict price the way realtors would do it, but of course realtor predictions are expensive, so we don't have enough data, and we use data from advertisings as well. So i'm asking how is better to treat subset with target values from advertisings.
