Python Pipeline.name方法代码示例

# 需要导入模块: from sklearn.pipeline import Pipeline [as 别名]
# 或者: from sklearn.pipeline.Pipeline import name [as 别名]
def get_estimator_pipe(name, model, vec_only, num_alch_cat, lsa_comp, reducer, stem):
    ''' Concatenate a transform chain and a classifier. '''
    chain = get_trf_chain(vec_only, num_alch_cat, lsa_comp, reducer, stem)
    chain.append((name, model))
    pipe = Pipeline(chain)
    pipe.name = name
    return pipe

# 需要导入模块: from sklearn.pipeline import Pipeline [as 别名]
# 或者: from sklearn.pipeline.Pipeline import name [as 别名]
def build_simple_pipes():
    ''' Create classifier-only pipes (without prior transforms, if this is done upfront e.g.). '''
    clfs = get_all_classifiers()
    pipes = []
    for clf in clfs:
        pipe = Pipeline([(clf.name, clf)])
        pipe.name = clf.name
        pipes.append(pipe)
    return pipes

# 需要导入模块: from sklearn.pipeline import Pipeline [as 别名]
# 或者: from sklearn.pipeline.Pipeline import name [as 别名]
        optimizer = keras.optimizers.Nadam(lr=0.002)
        model.compile(loss='mse', optimizer=optimizer)

        model.fit(x_seq, y_seq, batch_size=self.batch_size, verbose=1, nb_epoch=self.n_epochs, shuffle=False)
        self.model = model
        return self


    def predict(self, x):
        # merge the train and the test
        x_merged = pd.concat([self.x_train, x])

        start = len(x_merged.index) % (self.batch_size * self.sequence_length)

        x_seq = self.sliding_window(x_merged.iloc[start:])
        pred = self.model.predict(x_seq, batch_size=self.batch_size, verbose=1)

        pred = np.vstack(pred)
        res = pred[-len(x):, :]
        return res


lstm = Pipeline([
      ("drop", FeatureRemover([c for c in x_all.columns if c[-2] == "-"])),
      ("scaleandnorm", ScaleAndNorm()),
      ("lstm", MyLSTM(batch_size=32, sequence_length=20, n_epochs=10))
      ])
lstm.name = "lstm"

lstm = SomeLinearWrapper(lstm)

# 需要导入模块: from sklearn.pipeline import Pipeline [as 别名]
# 或者: from sklearn.pipeline.Pipeline import name [as 别名]
                del x[col]
        return x

    def fit(self, x, y=None):
        return self


model_ridge = Pipeline([
      ("drop", FeatureRemover(["ATTB"])),
      ("dropna", preprocessing.Imputer()),
      ("scale", preprocessing.StandardScaler()),
      #("norm", preprocessing.Normalizer()),
      ("ridge", linear_model.Ridge(normalize=True, fit_intercept=True, alpha = 0.4)),
      ])

model_ridge.name = "ridge"


# subclassed to play with eta decay and dart booster
class MyXGBRegressor(xgboost.XGBRegressor):


    # overwriting to get desired behaviour
    def fit(self, X, y, eval_set=None, eval_metric=None,
            early_stopping_rounds=None, verbose=True):
        # pylint: disable=missing-docstring,invalid-name,attribute-defined-outside-init
        """
        Fit the gradient boosting model

        Parameters
        ----------

# 需要导入模块: from sklearn.pipeline import Pipeline [as 别名]
# 或者: from sklearn.pipeline.Pipeline import name [as 别名]
        model.add(Dense(700, input_dim=input_dim))
        model.add(Activation('tanh'))
        model.add(Dense(700))
        model.add(Activation('tanh'))
        model.add(Dense(300))
        model.add(Activation('tanh'))
        #model.add(Dropout(0.1))
        model.add(Dense(output_dim))
        model.add(Activation('relu'))

        optimizer = keras.optimizers.Adam()
        model.compile(loss='mse', optimizer=optimizer)

        model.fit(x.as_matrix(), y.as_matrix(), batch_size=self.batch_size, verbose=1, nb_epoch=self.n_epochs, shuffle=True)
        self.model = model
        return self

    def predict(self, x):
        pred = self.model.predict(x.as_matrix(), batch_size=self.batch_size, verbose=1)
        pred = np.vstack(pred)
        return pred


nn = Pipeline([
      ("scaleandnorm", ScaleAndNorm()),
      ("nn", NN(batch_size=32, n_epochs=10))
      ])
nn.name = "nn"

nn = SomeLinearWrapper(nn)