附加两个熊猫数据帧时的索引问题

Question

I'm working on dummifying a column of zipcodes in pandas so I can build a random forest model in sklearn. Here is my code:

forest_test_features = test_df[['sqft_lot', 'floors', 'waterfront', 'view', 'condition', 'grade', 'sqft_above', 'sqft_basement', 'yr_built', 'yr_renovated']] forest_test_features.append(pd.get_dummies(test_df['zipcode'])) forest_test_target = test_df['price']

I get a runtime warning, and then my model's R^2 score is much lower than when I simply leave zipcode in the model without dummifying, suggesting something went wrong. pd.get_dummies returns a dataframe, and I think the problem is in the fact that this dataframe and forest_test_features are in two different orders, but I am unsure of how to proceed. The indexes are still correct (zipcode 98144 maps to a '1' in the '98144' column of get_dummies return.)

I also get this warning: RuntimeWarning: '<' not supported between instances of 'str' and 'int', sort order is undefined for incomparable objects result = result.union(other)

Answer 1

You can specify the order. I guess you have a forest_train_features dataframe. you can do this:

feats = forest_test_features.keys()
model = RandomForestRegressor()
model.fit(forest_train_features[feats], forest_train_features['price'])
prediction = model.predict(forest_test_features[feats])

the columns should be in the same order then. You can also do the same preprocessing for the train and test in the same DataFrame and then split it