Given temperature histories for NYC and many small towns, find the most volatile town and the town most like NYC, then predict NYC from the towns with greedy feature selection.

You have historical temperature series for New York City and for many small towns. (1) Which town's temperature fluctuates the most? (2) Which town is most similar to NYC? (3) Predict NYC's temperature from the towns' data and score the prediction by mean squared error. (4) Select features greedily: at each step add the town that reduces the error the most, until a given number of towns is in the model.