def determine_errors(df_val, tree, ml_task):
predictions = decision_tree_predictions(df_val, tree)
actual_values = df_val.label
if ml_task == "regression":
# mean squared error
return ((predictions - actual_values) **2).mean()
else:
# number of errors
return sum(predictions != actual_values)
def random_forest_predictions(test_df, forest):
df_predictions = {}
for i in range(len(forest)):
column_name = "tree_{}".format(i)
predictions = decision_tree_predictions(test_df, tree=forest[i])
df_predictions[column_name] = predictions
df_predictions = pd.DataFrame(df_predictions)
random_forest_predictions = df_predictions.mode(axis=1)[0]
return random_forest_predictions
def random_forest_predictions(test_df, forest):
df_predictions = {}
for i in range(len(forest)):
column_name = "tree_{}".format(i)
predictions = decision_tree_predictions(test_df, tree=forest[i])
# keys and values
df_predictions[column_name] = predictions
# transform the dictionary into dataframe.
# rows:index of datapts; columns: prediction of each trees
df_predictions = pd.DataFrame(df_predictions)
# vote for most. Type of predicion is pandas.core.series.Series
random_forest_predictions = df_predictions.mode(axis=1)[0]
return random_forest_predictions
def random_forest_predictions(test_df, forest, ml_task):
df_predictions = {}
for i in range(len(forest)):
column_name = "tree_{}".format(i)
predictions = decision_tree_predictions(test_df, tree=forest[i])
df_predictions[column_name] = predictions
df_predictions = pd.DataFrame(df_predictions)
if ml_task == "regression":
random_forest_predictions = df_predictions.mean(axis=1)
else:
random_forest_predictions = df_predictions.mode(axis=1)[0]
return random_forest_predictions