Using all features for KNN decreases RMSE


When I run the below code locally (using dc_airbnb.csv) my RMSE actually inreases when I use all features. I have followed the cleanup like in the course. However, when I run the same code in the dataquest-editor I get the results as expected.

Any idea what my cause the discrepancy?

My Code:

def get_eval(train_data, test_data, feature_columns, target_column):
    train_df = train_data.copy()
    test_df = test_data.copy()
    knn = KNeighborsRegressor(n_neighbors=5, algorithm = 'brute')[feature_columns], train_df[target_column])
    four_predictions = knn.predict(test_df[feature_columns])
    mse = mean_squared_error(test_df[target_column], four_predictions)
    return mse,  mse**(1/2)

normalized_listings = pd.read_csv('../data/normalized_listings.csv')

train_df, test_df = train_test_split(normalized_listings, random_state=1, train_size=0.760555706891855)

target_column = 'price'
features = train_df.columns.tolist()
features.remove('Unnamed: 0')

mse, rmse = get_eval(train_df, test_df, features, target_column)
print(mse, rmse)

11195.285415244598 105.80777577874227