def evaluate(self, test, y_test, print_output=False):
pred = _prepare_prediction(test, self._ests_store, self.threshold)
return evaluate_prediction(self.method,
self.labels,
y_test,
pred,
print_output=print_output)
def train(self, train, y_train):
thresholds = np.linspace(0, 1, 10, endpoint=False)
lambdas = np.linspace(0, 1, 10, endpoint=False)
ngrams = range(2, 10) if self.use_ngram else [0]
for _lambda in lambdas:
for n in ngrams:
pred = self._prepare_prediction(train, _lambda, n)
for t in thresholds:
p = pred.apply(lambda x: _set_label(x, t), axis=1)
f1 = get_aggregate_evaluation(
self.labels,
evaluate_prediction(
"{}-{}-{}-{}".format(self.method, t, _lambda, n),
self.labels,
y_train,
p,
),
)
if f1 > self.max_f1:
self.max_f1 = f1
self.threshold = t
self._lambda = _lambda
self.ngram = n
logging.debug(
f"training found an optimum f1 of {self.max_f1} with threshold {self.threshold} and lambda {self._lambda} and ngram {self.ngram}"
)
def evaluate(self, test, y_test, print_output):
return evaluate_prediction(
self.method,
self.labels,
y_test,
self.predict(test),
print_output=print_output,
)
def evaluate(self, test, y_test, print_output=False):
pred = self._prepare_prediction(test, self._lambda, self.ngram)
pred = pred.apply(lambda x: _set_label(x, self.threshold), axis=1)
return evaluate_prediction(self.method,
self.labels,
y_test,
pred,
print_output=print_output)
def train(self, train, y_train):
thresholds = np.linspace(0, 1, 10, endpoint=False)
for t in thresholds:
p = _prepare_prediction(train, self._ests_store, t)
f1 = get_aggregate_evaluation(
self.labels,
evaluate_prediction(self.method, self.labels, y_train, p))
if f1 > self.max_f1:
self.max_f1 = f1
self.threshold = t
logging.debug(
"training found an optimum f1 of {} with threshold {}".format(
self.max_f1, self.threshold))
def train(self, train, y_train):
thresholds = np.linspace(0, 1, 10, endpoint=False)
pred = _prepare_prediction(train)
for t in thresholds:
p = pred.apply(lambda x: _set_label(x, t), axis=1)
f1 = get_aggregate_evaluation(
self.labels,
evaluate_prediction("{}-{}".format(self.method, t),
self.labels, y_train, p),
)
if f1 > self.max_f1:
self.max_f1 = f1
self.threshold = t
logging.debug(
"training found an optimum f1 of {} with threshold {}".format(
self.max_f1, self.threshold))