def predict(self, model_conf, data_sources, data_sinks, model, args_dict):
if "test_key" in args_dict:
# URI param example (an uri param is part in the args_dict just like any other input)
key = args_dict["test_key"]
logger.info(
"Got the uri parameter (ID) %s. Looking up input data and predicting...",
key,
)
df = data_sources["batch_input"].get_dataframe()
df["myid"] = df["myid"].apply(str)
X = order_columns(df.loc[df["myid"] == key])
my_tree = model
y = my_tree.predict(X.drop("myid", axis=1))[0]
return {"iris_variety": y}
elif "sepal.length" not in args_dict or args_dict[
"sepal.length"] is None:
# Batch prediction example
logger.info("Doing batch prediction")
X = order_columns(data_sources["batch_input"].get_dataframe())
my_tree = model
y = my_tree.predict(X.drop("myid", axis=1))
X["pred"] = y
data_sinks["predictions"].put_dataframe(X)
return {"status": "ok"}
# "Normal" prediction example
logger.info('Doing "normal" prediction')
X = order_columns(pd.DataFrame(args_dict, index=[0]))
my_tree = model
y = my_tree.predict(X)[0]
return {"iris_variety": y}
def create_trained_model(
self, model_conf, data_sources, data_sinks, old_model=None
):
df = data_sources["petals"].get_dataframe()
X = df.drop("variety", axis=1)
y = df["variety"]
X_ordered = order_columns(X)
my_tree = tree.DecisionTreeClassifier()
my_tree.fit(X_ordered, y)
return my_tree
def test_trained_model(self, model_conf, data_sources, data_sinks, model):
df = data_sources["petals_test"].get_dataframe()
X_test = order_columns(df.drop("variety", axis=1))
y_test = df["variety"]
my_tree = model
y_predict = my_tree.predict(X_test)
acc = accuracy_score(y_test, y_predict)
conf = confusion_matrix(y_test, y_predict).tolist()
metrics = {"accuracy": acc, "confusion_matrix": conf}
return metrics
def create_trained_model(self,
model_conf,
data_sources,
data_sinks,
old_model=None):
df = data_sources["petals"].get_dataframe()
X = df.drop("variety", axis=1)
y = df["variety"]
X_ordered = order_columns(X)
my_tree = tree.DecisionTreeClassifier()
my_tree.fit(X_ordered, y)
# We can add info (numbers, strings, dicts, lists) to be saved in the training report (JSON metadata):
report("meaning_of_life", 42)
report(
"train_data_used", df
) # If we pass dataframes, their summary will be added to the training report
return my_tree