in this code i write code and trained data on flood_probability and evaluate my model using r2_score and then predict on test.csv and store the result in res.csv
in this code, i resized train data from 1117967 to 420000 and test from 800000 to 420000 without loss performance