def load_default(self, machine_list=['lasso', 'tree', 'ridge', 'random_forest']):
"""
Loads 4 different scikit-learn regressors by default.
Parameters
----------
machine_list: optional, list of strings
List of default machine names to be loaded.
"""
for machine in machine_list:
if machine == 'lasso':
self.machines_['lasso'] = linear_model.LassoCV(random_state=self.random_state).fit(self.X_k_, self.y_k_)
if machine == 'tree':
self.machines_['tree'] = DecisionTreeRegressor(random_state=self.random_state).fit(self.X_k_, self.y_k_)
if machine == 'ridge':
self.machines_['ridge'] = linear_model.RidgeCV().fit(self.X_k_, self.y_k_)
if machine == 'random_forest':
self.machines_['random_forest'] = RandomForestRegressor(random_state=self.random_state).fit(self.X_k_, self.y_k_)
def main():
train_x, train_y, test_x, test_y = (sio.loadmat(TRAIN_DIR)['trainx'],
sio.loadmat(TRAIN_DIR)['trainy'],
sio.loadmat(TEST_DIR)['testx'],
sio.loadmat(TEST_DIR)['testy'])
clf = linear_model.RidgeCV(alphas=[0.001, 0.01, 0.1, 1.0, 10.0, 100.0],
normalize=True,
cv=10,
store_cv_values=False)
print("Training......")
clf.fit(train_x, train_y)
print("Predicting.......")
years = clf.predict(test_x)
diff = 0.0
for i, j in zip(test_y, years):
diff += abs(i - j)
diff /= TEST_SIZE
print("MSE is: " + str(diff))
def fit(self, train, y):
internal_model=linear_model.RidgeCV(alphas=(0.1, 0.5, 1.0, 5.0, 10.0), fit_intercept=False)
bestscore=1e15
better=True
indextrain=train.dropna().index
limitlen=len(train)*self.limit_size_train
while better:
internal_model.fit(train.ix[indextrain], y.ix[indextrain])
score=metrics.mean_squared_error(internal_model.predict(train.ix[indextrain]), y.ix[indextrain])
if score < bestscore:
bestscore=score
self.bestmodel=internal_model
residual=y.ix[indextrain]-internal_model.predict(train.ix[indextrain])
indextrain=residual[abs(residual)<=abs(residual).quantile(self.quant)].index
if len(indextrain)<limitlen:
better=False
else:
better=False
self.bestmodel=internal_model
def Models(self, model_names):
self.X_train, self.X_val, self.y_train, self.y_val = train_test_split(
self.poly_features, self.y, test_size=0.3)
#model_lasso = linear_model.LassoCV(alphas=[0.0001,0.001,0.01,0.05,0.1,0.2,0.5,1,10],cv=10)
model_linear = linear_model.LinearRegression()
model_ridge = linear_model.RidgeCV(
alphas=[0.0001, 0.001, 0.01, 0.05, 0.1, 0.2, 0.5, 1, 10], cv=10)
model_en = linear_model.ElasticNetCV(
alphas=[0.0001, 0.0005, 0.001, 0.01, 0.1, 1, 10],
l1_ratio=[.01, .1, .5, .9, .99],
max_iter=5000)
model_br = linear_model.BayesianRidge(alpha_1=1e-06,
alpha_2=1e-06,
compute_score=False,
copy_X=True,
fit_intercept=True,
lambda_1=1e-06,
lambda_2=1e-06,
n_iter=300,
normalize=False,
tol=0.001,
verbose=False)
#model_svr = model_selection.GridSearchCV(LinearSVR(random_state=0, tol=1e-5),
# param_grid={"epsilon":[0,0.2],"C": [0,1]},cv = 5)
model_svr = make_pipeline(StandardScaler(),
svm.LinearSVR(random_state=0, tol=1e-5))
model_sgdr = make_pipeline(StandardScaler(),
SGDRegressor(max_iter=1000, tol=1e-3))
model_list = [
model_linear, model_ridge, model_en, model_br, model_svr,
model_sgdr
]
#print('岭回归最优alpha值:', model_ridge.alpha_, '\n', '-'*50)
for i in range(6):
model_predict = model_list[i].fit(self.X_train, self.y_train)
self.pre_y_list = model_predict.predict(self.X_val)
joblib.dump(
model_predict,
f'C:\\Users\\Administrator\\Desktop\\7.21\\models-saved\\1期 device6 模型{model_names[i]}.pkl'
)
def single_ridge_regression(X_in, y_in):
'''
Performs a single variable ridge regression given X feature and y outcome
variable
Input: Series of X feature and Series of y outcome variables
Output: dataframe of beta coefficients, and the ridge regressor
'''
# Ridge regression using only cumulative number of COVID cases
# standardizes X values for the ridge regression
scaler = StandardScaler()
X = X_in.values.reshape(-1, 1)
y = y_in.values.reshape(-1, 1)
X_std = scaler.fit_transform(X)
# formulates the training and tests set for each variable
X_train, X_test, y_train, y_test = train_test_split(X_std,
y,
test_size=.2,
random_state=0)
# ridge regression algorithm
regressor = linear_model.RidgeCV(alphas=[.1, 1, 10])
regressor = regressor.fit(X_std, y)
regressor.alpha_
# coefficient dataframe and y prediction results of the ridge regression
y_pred = regressor.predict(X_test)
df = pd.DataFrame({
"Actual": y_test.flatten(),
"Predicted": y_pred.flatten()
})
# shows model performance
print("Ridge regression score: " + str(regressor.score(X_std, y)))
print("Ridge regression explained_variance: " +
str(metrics.explained_variance_score(y_test, y_pred)))
print("Ridge regression MSE: " +
str(metrics.mean_squared_error(y_test, y_pred)))
return regressor, df
def multi_ridge_regression(X_in, y_in):
'''
Performs a multivariate ridge regression given X features and y outcome
variable
Input: dataframe of X features and Series of y outcome variables
Output: dataframe of beta coefficients, and the ridge regressor
'''
# Ridge regression
# standardizes X values for the ridge regression
scaler = StandardScaler()
X = X_in.values
y = y_in.values
X_std = scaler.fit_transform(X)
# formulates the training and tests set for each variable
X_train, X_test, y_train, y_test = train_test_split(X_std,
y,
test_size=.2,
random_state=0)
# ridge regression algorithm
regressor = linear_model.RidgeCV(alphas=[.1, 1, 10])
regressor = regressor.fit(X_std, y)
regressor.alpha_
# coefficient dataframe and y prediction results of the ridge regression
coeff_df = pd.DataFrame(regressor.coef_,
X_in.columns,
columns=["Coefficients"])
y_pred = regressor.predict(X_test)
df = pd.DataFrame({"Actual": y_test, "Predicted": y_pred})
print("Ridge regression score: " + str(regressor.score(X_std, y)))
print("Ridge regression: " +
str(metrics.explained_variance_score(y_test, y_pred)))
print("Ridge regression: " +
str(metrics.mean_squared_error(y_test, y_pred)))
return coeff_df, df, regressor
def ridgeCV_model(X_train,X_valid,y_train,y_test,y_name, y_train_mean,y_train_std):
print('head items to fit are: ', y_name)
# In[ ]:
for head_item in range(len(y_name)):
y_train_item = y_train[:,head_item]
y_train_item = np.reshape(y_train_item,[y_train.shape[0],1])
y_test_item = y_test[:,head_item]
y_test_item = np.reshape(y_test_item,[y_test_item.shape[0],1])
print('********************************** Fitting RidgeCV on %s Data **********************************' % y_name[head_item])
#Declare model
model = linear_model.RidgeCV(alphas=[0.1, 1.0, 10.0],normalize=True,fit_intercept=True)
#Fit model
model.fit(X_train,y_train_item)
#Get predictions
y_valid_predicted=model.predict(X_valid)
training_prediction=model.predict(X_train)
R2s_training=get_R2(y_train_item,training_prediction)
print('R2 on training set = ', R2s_training)
#Get metric of fit
R2s=get_R2(y_test_item,y_valid_predicted)
print('R2s:', R2s)
print('saving prediction ...')
np.savez(y_name[head_item] + '_RidgeCV_ypredicted.npz',y_test=y_test_item,y_prediction=y_valid_predicted,
y_train_=y_train_item,training_prediction=training_prediction,
y_train_mean=y_train_mean[head_item],y_train_std=y_train_std[head_item])
#print 'saving model ...'
joblib.dump(model, y_name[head_item] + '_Ridge.pkl')
print('plotting results...')
plot_results(y_test_item,y_valid_predicted,y_name[head_item],R2s,model_name='RidgeCV')
return model
def __init__(self, X, y, kind):
self.X = X
self.y = y
kind = kind.upper()
if kind == 'SVM':
from sklearn import svm
self.regressor = svm.SVR()
elif kind == 'RIDGECV':
from sklearn import linear_model
self.regressor = linear_model.RidgeCV(
alphas=[
x * y for x in [0.01, 0.1, 1, 10] for y in [1, 5]
]
)
elif kind == 'SVM_GRID':
from sklearn import svm
from sklearn.model_selection import GridSearchCV
self.regressor = GridSearchCV(
svm.SVR(),
{
'C': [1e0, 5e0, 1e1, 5e1, 1e2, 5e2, 1e3],
'epsilon': [1e-3, 1e-2],
'kernel': ['linear', 'rbf', 'poly'],
'degree': [2, 3, 4]
},
scoring='neg_mean_squared_log_error'
)
elif kind == 'SVM_GRID_SIMPLE':
from sklearn import svm
from sklearn.model_selection import GridSearchCV
self.regressor = GridSearchCV(
svm.SVR(),
{
'C': [1e0, 5e0, 1e1, 5e1, 1e2],
'gamma': [1e-3, 1e-2, 1e-1],
'kernel': ['linear'],
'degree': [2]
},
scoring='neg_mean_squared_log_error',
n_jobs=-1,
cv=7
)
def evaluate_lcc_model(en_train_matrix, en_test_matrix, fr_train_matrix,
fr_test_matrix, dimensions, evaluation_function):
scores = []
for dimension in tqdm(dimensions):
en = en_train_matrix[:, :dimension] - np.mean(
en_train_matrix[:, :dimension], axis=0)
fr = fr_train_matrix[:, :dimension] - np.mean(
fr_train_matrix[:, :dimension], axis=0)
sample_size = en.shape[0]
zero_matrix = np.zeros((sample_size, dimension))
X1 = np.concatenate((en, zero_matrix), axis=1)
X2 = np.concatenate((zero_matrix, fr), axis=1)
X = np.concatenate((X1, X2), axis=0)
Y1 = np.concatenate((en, fr), axis=1)
Y2 = np.concatenate((en, fr), axis=1)
Y = np.concatenate((Y1, Y2), axis=0)
reg = linear_model.RidgeCV(alphas=[1e-10, 1e-3, 1e-2, 1e-1, 1, 10])
reg.fit(X, Y)
pca = PCA(n_components=int(dimension))
pca.fit(reg.predict(X))
rrr = lambda X: np.matmul(pca.transform(reg.predict(X)), pca.
components_)
#sample_size = len(en_docs_test)
en = en_test_matrix[:, :dimension] - np.mean(
en_train_matrix[:, :dimension], axis=0)
fr = fr_test_matrix[:, :dimension] - np.mean(
fr_train_matrix[:, :dimension], axis=0)
zero_matrix = np.zeros((en_test_matrix.shape[0], dimension))
X1 = np.concatenate((en, zero_matrix), axis=1)
X2 = np.concatenate((zero_matrix, fr), axis=1)
X = np.concatenate((X1, X2), axis=0)
english_encodings_lcc = rrr(X1)
french_encodings_lcc = rrr(X2)
score = evaluation_function(english_encodings_lcc,
french_encodings_lcc)
scores.append(score)
return scores
def fit_and_predict(self, TEST_YEAR, regularization=True):
X, Y, xTrain, yTrain, xTest, yTest, names = self.build_data_arrays(TEST_YEAR)
if regularization:
predictor = linear_model.RidgeCV(alphas=[0.1, 1.0, 10], fit_intercept=True, normalize=self.shouldNormalize)
else:
predictor = linear_model.LinearRegression(fit_intercept=True, normalize=self.shouldNormalize, copy_X=True, n_jobs=1)
scores = {}
relativeError = {}
coefficients = {}
output = {}
for p in self.positions:
if len(xTrain[p]) > 1 and len(xTest[p]) > 1:
predictor.fit(np.array(xTrain[p]), np.array(yTrain[p]))
coefficients[p] = pd.DataFrame(zip(self.features, predictor.coef_), columns = ['feature', 'coefficient']).sort_values(by=['coefficient'], ascending=False)
prediction = predictor.predict(np.array(xTest[p]))
output[p] = pd.DataFrame(zip(names[p], prediction), columns = ['name', 'value']).sort_values(by=['value'], ascending=False)
scores[p] = (mean_squared_error(np.array(yTest[p]), np.array(prediction)),
r2_score(np.array(yTest[p]), np.array(prediction)))
relativeError[p] = self.get_relative_error(output[p], TEST_YEAR)
return output
def ridge_regress(self, cv = 20, alphas = None ):
"""perform k-folds cross-validated ridge regression on the design_matrix. To be used when the design matrix contains very collinear regressors. For cross-validation and ridge fitting, we use sklearn's RidgeCV functionality. Note: intercept is not fit, and data are not prenormalized.
:param cv: cross-validated folds, inherits RidgeCV cv argument's functionality.
:type cv: int, standard = 20
:param alphas: values of penalization parameter to be traversed by the procedure, inherits RidgeCV cv argument's functionality. Standard value, when parameter is None, is np.logspace(7, 0, 20)
:type alphas: numpy array, from >0 to 1.
:returns: instance variables 'betas' (nr_betas x nr_signals) and 'residuals' (nr_signals x nr_samples) are created.
"""
if alphas == None:
alphas = np.logspace(7, 0, 20)
self.rcv = linear_model.RidgeCV(alphas=alphas,
fit_intercept=False,
cv=cv)
self.rcv.fit(self.design_matrix.T, self.resampled_signal.T)
self.betas = self.rcv.coef_.T
self.residuals = self.resampled_signal - self.rcv.predict(self.design_matrix.T)
self.logger.debug('performed ridge regression on %s design_matrix and %s signal, resulting alpha value is %f' % (str(self.design_matrix.shape), str(self.resampled_signal.shape), self.rcv.alpha_))
def fit_ridge(oe,mw): #### oe = x; mw = y variable.
# fit line:
oe = np.array(oe)
mw = np.array(mw)
clf = linear_model.RidgeCV(alphas=[0.001, 0.01, 0.1, 1.0, 10.0, 100.0, 1000.0])
num_codes = min(len(oe),len(mw))
print 'oe.shape = ', oe.shape
oe = oe.reshape(-1, 1)
print 'oe.shape = ', oe.shape
mw = mw.reshape(-1, 1)
clf.fit(oe[0:num_codes], mw[0:num_codes])
print '****************************************************************************'
print 'RIDGE REGRESSION SCORE = ', clf.score(oe[0:num_codes],mw[0:num_codes])
m = clf.coef_[0][0]
c = clf.intercept_[0]
print 'm,c = ', m,c
return m,c
def poly_ridge(self,deg=2): ''' Polynomial Ridge Regression ''' from sklearn import linear_model # Training t0 = time.time() phi = cm.naivePolyFeature(self.X,deg=deg,norm=True) lm = linear_model.RidgeCV(alphas=np.logspace(-10,-1,10)) lm.fit(phi,self.y) print lm.alpha_ t_tr = time.time() - t0 # Predicting t0 = time.time() phi_pred = cm.naivePolyFeature(self.X_pred,deg=deg,norm=True) y_lr = lm.predict(phi_pred) t_pr = time.time() - t0 eel = np.mean(np.maximum(self.Value0-self.c-np.sum(y_lr,axis=1),0)) return (eel, t_tr, t_pr)
def RidgeRegression(x, y, z, degree=5, alpha=10**(-6), verbose=False):
# Split into training and test
x_train = np.random.rand(100, 1)
y_train = np.random.rand(100, 1)
z = FrankeFunction(x_train, y_train)
# training and finding design matrix X_
X = np.c_[x_train, y_train]
poly = PolynomialFeatures(degree)
X_ = poly.fit_transform(X)
ridge = linear_model.RidgeCV(alphas=np.array([alpha]))
ridge.fit(X_, z)
beta = ridge.coef_
#intercept = ridge.intercept_
# predict data and prepare for plotting
x_, y_ = np.meshgrid(x, y)
x = x_.reshape(-1, 1)
y = y_.reshape(-1, 1)
M = np.c_[x, y]
M_ = poly.fit_transform(M)
predict = M_.dot(beta.T)
if verbose:
print("X_: ", np.shape(X_))
print("M: ", np.shape(M))
print("M_: ", np.shape(M_))
print("predict: ", np.shape(predict))
# show figure
fig = plt.figure()
ax = fig.gca(projection='3d')
ax.plot_surface(x_,
y_,
predict.reshape(20, 20),
cmap=cm.coolwarm,
linewidth=0,
antialiased=False)
plt.show()
return beta
def __initialize_model(model_name, lamda=0, hyper_parameters={}):
"""
initialize machine learning model.
Args:
model_name: learning algorithm name
lamda: coefficient of standardization item
hyper_parameter: other parameters for algorithms
See parameters for RandomForest Regression in sci-kit-learn
Returns:
an initialized classifier
"""
if model_name == constants.MODEL_NAME_LASSO:
# note: alpha in scikit-learn reprsents lamda which is the constant that
# multiplies the regularization term
clf_lasso = linear_model.Lasso(alpha=lamda)
return clf_lasso
elif model_name == constants.MODEL_NAME_ELASTICNET:
clf_elasticnet = ElasticNet(alpha=lamda)
return clf_elasticnet
elif model_name == constants.MODEL_NAME_RIDGE:
clf_ridge = linear_model.Ridge(alpha=lamda)
return clf_ridge
elif model_name == constants.MODEL_NAME_RIDGECV:
clf_ridgecv = linear_model.RidgeCV(alphas=constants.lamdaArray)
return clf_ridgecv
elif model_name == constants.MODEL_NAME_LARS:
clf_lars = linear_model.Lars(n_nonzero_coefs=1)
return clf_lars
elif model_name == constants.MODEL_NAME_BAYESIAN:
clf_bayesian = linear_model.BayesianRidge()
return clf_bayesian
elif model_name == constants.MODEL_NAME_SGD:
clf_sgd = linear_model.SGDRegressor(alpha=lamda)
return clf_sgd
elif model_name == constants.MODEL_NAME_RANDOM_FOREST:
clf_random_forest = RandomForestRegressor(**hyper_parameters,
random_state=0,
n_jobs=-1)
return clf_random_forest
def cross_validate_model(X_train, Y_train):
"""
Here we perform cross validation of models to choose the best one.
"""
# Divide the training and testing data
train, test, y_actual, y_predict = train_test_split(X_train,
Y_train,
test_size=0.5,
random_state=42)
# List the regression methods to use.
clf_random_forest = ensemble.RandomForestRegressor(n_estimators=50)
clf_adaboost_reg = ensemble.AdaBoostRegressor(n_estimators=50)
clf_lasso_larscv = sklinear.LassoLarsCV(cv=9)
clf_ridge = sklinear.RidgeCV()
clf_elastic_net = sklinear.ElasticNet()
clf_extra_tree = ensemble.ExtraTreesRegressor(n_estimators=50)
clf_mlpr = neural_network.MLPRegressor(solver='adam')
# Add the above methods in an array
# More ameable for looping
methods = [
clf_random_forest, clf_adaboost_reg, clf_lasso_larscv, clf_elastic_net,
clf_extra_tree, clf_mlpr
]
methods_label = [
'clf_random_forest', 'clf_adaboost_reg', 'clf_lasso_larscv',
'clf_elastic_net', 'clf_extra_tree', 'clf_mlpr'
]
method_mse = np.zeros((len(methods), 1))
# Fit and predict for each method
for i in range(len(methods)):
methods[i].fit(train, y_actual)
method_predict = methods[i].predict(test)
method_mse[i] = metrics.mean_squared_error(y_predict, method_predict)
print('MSE for %s while cross validation : %f' %
(methods_label[i], method_mse[i]))
# We return the method which has the minimum mse
return np.argmin(method_mse)
def __init__(self, train=True, train_set=[], clf='ridge'):
self.identity = None
self.nb_victories = 0
self.nb_games = 0
self.next_action = {}
self.gamma = 1
self.initial_alpha = 0.05
self.min_alpha = 0.005
self.alpha_update_rate = 1 # pow(self.initial_alpha/self.min_alpha, 1/500)
self.alpha = self.initial_alpha
self.epsilon = 0.1
self.state_space_x = [
'low_prestige', 'high_prestige', 'can_buy_prestige',
'can_buy_card', 'can_reserve', 'can_take_2', 'can_take_3'
]
self.training_phase = train
self.is_trained = not train
self.classifier = {
'ridge': linear_model.RidgeCV(),
'mlp': neural_network.MLPRegressor()
}[clf]
self.state_space_y = [0, 1] # 0: few, 1: some, 2: a lot
self.state_space, self.state_space_inverted_index = self.build_state_space(
)
# self.state_space_inverted_index = self.build_inverted_index(self.state_space)
self.action_space = [
'buy_prestige', 'buy_card', 'reserve', 'take_3', 'take_2',
'do_nothing'
]
self.action_space_inverted_index = self.build_inverted_index(
self.action_space)
self.nb_actions = len(self.action_space)
self.nb_states = len(self.state_space)
self.state = {}
self.actions = []
self.init_features()
if len(train_set) > 0:
self.load_training(train_set)
def poly_regression(x_tr,
x_ts,
y_tr,
y_ts,
degree,
filename='Poly_regression'):
print("Polynomial Regression")
x_train = x_tr
x_test = x_ts
y_train = y_tr
y_test = y_ts
poly_features = PolynomialFeatures(degree=degree)
x_train_poly = poly_features.fit_transform(x_train)
poly_model = linear_model.RidgeCV(alphas=np.logspace(-9, 9, 19),
normalize=True)
#poly_model = linear_model.LinearRegression(normalize=True)
print("Fitting...")
poly_model.fit(x_train_poly, y_train)
# Save The Model into File
filename_model = f'Trained Models\{filename}.sav'
pickle._dump(poly_model, open(filename_model, 'wb'))
#print("Best Alpha : ", poly_model.alpha_)
print("Predicting...")
prediction = poly_model.predict(poly_features.fit_transform(x_test))
print(
"Poly Accuracy: ",
round(
poly_model.score(poly_features.fit_transform(x_test), y_test) *
100), "%")
print('Poly Mean Square Error',
metrics.mean_squared_error(y_test, prediction), "\n\n")
print("First Value of Test Samples' Actual Output: ",
np.asarray(y_test)[0])
print("First Value of Test Samples' Predicted Output: ", prediction[0])
'''fig, ax = plt.subplots()
def run_ridge_cv(x_df, y_df, analyspar, alphas=None):
"""
run_ridge_cv(x_df, y_df, analyspar)
"""
if alphas is None:
alphas = (0.1, 0.1, 2.0)
steps = [("scaler", preprocessing.StandardScaler()),
("model",
linear_model.RidgeCV(normalize=True,
alphas=alphas,
store_cv_values=True))]
pipl = pipeline.Pipeline(steps)
pipl.fit(x_df, y_df)
log_sklearn_results(pipl,
analyspar,
name="ridge_cv",
var_names=x_df.columns)
def build_models(predictors, responses, modelNo):
if(modelNo==0):
# Linear Regression
model = linear_model.LinearRegression();
modelName = "Linear Regression";
if(modelNo==1):
# Ridge Regression
model = linear_model.RidgeCV(alphas = (0.1,0.1,10));
modelName = "Ridge Regression";
if(modelNo==2):
# lasso Regression
model = linear_model.MultiTaskLassoCV(eps=0.001, n_alphas=100, alphas=(0.1,0.1,10));
modelName = "Lasso Regression";
model.fit(predictors, responses);
predictions = model.predict(predictors);
Result = {};
Result['modelName'] = modelName;
Result['predictions'] = predictions;
Result['model'] = model;
Result['Corr'] = pearsonr(predictions,responses)[0][0];
return Result;
def _regression_DKL(n_, fo_, fx_, stims_all_LM_, stims_all_S_, stims_all_LUM_):
""" regress all coordinates but only hue matters """
reg_fo_LM = linear_model.RidgeCV()
reg_fo_S = linear_model.RidgeCV()
reg_fo_LUM = linear_model.RidgeCV()
reg_fx_LM = linear_model.RidgeCV()
reg_fx_S = linear_model.RidgeCV()
reg_fx_LUM = linear_model.RidgeCV()
reg_fo_LM.fit(fo_[:n_], stims_all_LM_[:n_])
reg_fo_S.fit(fo_[:n_], stims_all_S_[:n_])
reg_fo_LUM.fit(fo_[:n_], stims_all_LUM_[:n_])
reg_fx_LM.fit(fx_[:n_], stims_all_LM_[:n_])
reg_fx_S.fit(fx_[:n_], stims_all_S_[:n_])
reg_fx_LUM.fit(fx_[:n_], stims_all_LUM_[:n_])
return reg_fo_LM, reg_fo_S, reg_fo_LUM, reg_fx_LM, reg_fx_S, reg_fx_LUM
def load_default(self,
machine_list=[
'lasso', 'tree', 'ridge', 'random_forest', 'svm'
]):
"""
Loads 4 different scikit-learn regressors by default.
Parameters
----------
machine_list: optional, list of strings
List of default machine names to be loaded.
Returns
-------
self : returns an instance of self.
"""
self.estimators_ = {}
for machine in machine_list:
try:
if machine == 'lasso':
self.estimators_['lasso'] = linear_model.LassoCV(
random_state=self.random_state).fit(
self.X_k_, self.y_k_)
if machine == 'tree':
self.estimators_['tree'] = DecisionTreeRegressor(
random_state=self.random_state).fit(
self.X_k_, self.y_k_)
if machine == 'ridge':
self.estimators_['ridge'] = linear_model.RidgeCV().fit(
self.X_k_, self.y_k_)
if machine == 'random_forest':
self.estimators_['random_forest'] = RandomForestRegressor(
random_state=self.random_state).fit(
self.X_k_, self.y_k_)
if machine == 'svm':
self.estimators_['svm'] = LinearSVR(
random_state=self.random_state).fit(
self.X_k_, self.y_k_)
except ValueError:
continue
return self
def save_standard_results(model_type, year):
p_train, r_train, p_test, r_test = pipeline(year, "pixel_level",
"daily_total")
scale = preprocessing.StandardScaler()
p_train = scale.fit_transform(p_train)
p_test = scale.transform(p_test)
if model_type == "Lasso":
lr = linear_model.LassoCV()
elif model_type == "Ridge":
lr = linear_model.RidgeCV()
elif model_type == "OLS":
lr = linear_model.LinearRegression()
elif model_type == "Elastic_Net":
lr = linear_model.ElasticNetCV()
else:
raise TypeError("Invalid model_type")
model = lr.fit(p_train, r_train)
predicted_train = model.predict(p_train)
predicted_test = model.predict(p_test)
train_correlation = fast_evaluate.anti_correlation(predicted_train,
r_train)
test_correlation = fast_evaluate.anti_correlation(predicted_test, r_test)
train_rmse = fast_evaluate.root_mean_square_error(predicted_train, r_train)
test_rmse = fast_evaluate.root_mean_square_error(predicted_test, r_test)
train_mae = fast_evaluate.mean_absolute_error(predicted_train, r_train)
test_mae = fast_evaluate.mean_absolute_error(predicted_test, r_test)
coefficients = model.coef_.tolist()
filename = "Standard_{0}_year_{1}.csv".format(model_type, year)
with open(filename, 'wb') as f:
writer = csv.writer(f)
writer.writerow([
"train_correlation", "test_correlation", "train_rmse", "test_rmse",
"train_mae", "test_mae", "coefficients"
])
writer.writerow((train_correlation, test_correlation, train_rmse,
test_rmse, train_mae, test_mae, coefficients))
def calculateRAPM(units, points, weights):
u = DictVectorizer(sparse=False)
u_mat = u.fit_transform(units)
# config.debug
# print(u_mat)
# print(points[:25])
# print(weights[:100])
playerIDs = u.get_feature_names()
# print(json.dumps(u.get_feature_names()[:25], indent=4*' '))
# print(json.dumps(u.inverse_transform(u_mat)[:1], indent=4 * ' '))
clf = linear_model.RidgeCV(alphas=(np.array([0.01, 0.1, 1.0, 10, 100, 500, 1000, 2000, 5000])), cv=5)
clf.fit(u_mat, points, sample_weight=weights)
# print(clf.alpha_)
ratings = []
for playerID in playerIDs:
ratings.append((playerID, clf.coef_[playerIDs.index(playerID)]))
ratings.sort(key=lambda tup: tup[1], reverse=True)
return ratings
def train_models(mod,
save=True,
cutoff=0.999,
percent=50,
plot=True,
scale=False):
if mod == 'linear':
clf = linear_model.LinearRegression(n_jobs=-1)
elif mod == 'lasso':
clf = linear_model.Lasso(alpha=1000,
max_iter=10000,
tol=0.001,
normalize=True,
positive=True)
elif mod == 'lassolars':
clf = linear_model.LassoLars(alpha=0.001)
elif mod == 'multilasso':
clf = linear_model.MultiTaskLasso(alpha=0.1)
elif mod == 'ridgeCV':
clf = linear_model.RidgeCV(alphas=[0.01, 0.1, 1.0, 10.0])
elif mod == 'ridge':
clf = linear_model.Ridge(alpha=[1000])
elif mod == 'bayes':
clf = linear_model.BayesianRidge()
elif mod == 'huber':
clf = linear_model.HuberRegressor()
elif mod == 'poly':
#clf = poly_clf()
clf = PolynomialFeatures(degree=2)
clf, continuum = train(clf,
mod,
save=save,
cutoff=cutoff,
percent=percent,
plot=plot,
scale=scale)
return clf, continuum
def _estimate_model(self):
"""Estimates ridge regression model.
Returns
-------
model : sklearn ridge regression or ridge cv object
Fitted ridge model.
"""
self.underlying = linear_model.Ridge(fit_intercept=self.intercept)
if (self.cv_folds is not None) or (self.solver in ['svd', 'eigen']):
#Ridge CV by default tests a very limited set of alphas, we expand on this
alphas = np.logspace(-10, 5, 100)
model = linear_model.RidgeCV(alphas=alphas,
cv=self.cv_folds,
fit_intercept=self.intercept,
gcv_mode=self.solver,
**self.kwargs)
else:
model = linear_model.Ridge(fit_intercept=self.intercept,
**self.kwargs)
model.fit(self.x_train, self.y_train)
return model
def _get_generalizer(cls, gnrl):
generalizers = dict(
MEAN = MeanClassifier(),
RFC = RandomForestClassifier(n_estimators=500, max_depth=32, n_jobs=-1, random_state=random_state),
RCV = lm.RidgeCV(alphas=np.linspace(0, 200), cv=100),
RCVp = RidgeCV_proba(alphas=np.linspace(0, 200), cv=100),
LCV = lm.LassoCV(),
LCVp = LassoCV_proba(),
LSVC = svm.LinearSVC(penalty='l2', loss='l2', dual=True, tol=0.0001, C=1.0, multi_class='ovr', fit_intercept=True, intercept_scaling=1, class_weight=None, verbose=0, random_state=random_state),
SVC = svm.SVC(C=1.0, kernel='linear', degree=3, gamma=0.0, coef0=0.0, shrinking=True, probability=True, tol=0.001, cache_size=200, class_weight=None, verbose=False, max_iter=-1, random_state=random_state),
LR = lm.LogisticRegression(penalty='l2', dual=True, tol=0.00001, C=1, fit_intercept=True, intercept_scaling=1.0, class_weight=None, random_state=random_state),
KNCuniform = KNeighborsClassifier(n_neighbors=1024, weights='uniform'),
KNC = KNeighborsClassifier(n_neighbors=1024, weights='distance'),
AUCR = AUCRegressor(),
ABC_DTC = AdaBoostClassifier(
algorithm='SAMME.R',
base_estimator=DecisionTreeClassifier(compute_importances=None, criterion='gini', max_depth=1, max_features=1.0, min_density=None, min_samples_leaf=1, min_samples_split=2, random_state=random_state, splitter='best'),
learning_rate=0.1,
n_estimators=200,
random_state=random_state),
)
return generalizers[gnrl]
def __init__(self,
data,
classifier='linear',
save=True,
load=False,
fname='FASMA_ML.pkl'):
self.classifier = classifier
self.data = data
self.save = save
self.load = load
self.fname = fname
self.X_train, self.y_train = data.X, data.y
if self.classifier == 'linear':
self.clf = linear_model.LinearRegression(n_jobs=-1)
elif self.classifier == 'lasso':
self.clf = linear_model.Lasso(alpha=0.00001)
elif self.classifier == 'lassolars':
self.clf = linear_model.LassoLars(alpha=1000)
elif self.classifier == 'multilasso':
self.clf = linear_model.MultiTaskLasso(alpha=1000)
elif self.classifier == 'ridgeCV':
self.clf = linear_model.RidgeCV(alphas=[0.1, 1.0, 10.0, 100])
elif self.classifier == 'ridge':
self.clf = linear_model.Ridge(alpha=10)
elif self.classifier == 'bayes':
self.clf = linear_model.BayesianRidge()
elif self.classifier == 'huber':
self.clf = linear_model.HuberRegressor()
# Train the classifier
if not self.load:
t = time()
self.train_classifier()
print('Trained classifier in {}s'.format(round(time() - t, 2)))
else:
with open(self.fname, 'rb') as f:
self.clf = cPickle.load(f)
def fit(self, train, y):
internal_model = linear_model.RidgeCV(
fit_intercept=True, cv=model_selection.TimeSeriesSplit(n_splits=2))
bestscore = 1e15
better = True
indextrain = train.dropna().index
limitlen = len(train) * self.limit_size_train
while better:
internal_model.fit(train.ix[indextrain], y.ix[indextrain])
score = metrics.mean_squared_error(
internal_model.predict(train.ix[indextrain]), y.ix[indextrain])
if score < bestscore:
bestscore = score
self.bestmodel = internal_model
residual = y.ix[indextrain] - internal_model.predict(
train.ix[indextrain])
indextrain = residual[
abs(residual) <= abs(residual).quantile(self.quant)].index
if len(indextrain) < limitlen:
better = False
else:
better = False
self.bestmodel = internal_model
def __remodel__(self, model_type, regr, __X_train, __Y_train):
"""
Function to retrain certain models based on optimal alphas and/or ratios
"""
if model_type == "ridge":
alpha = regr.alpha_
regr = linear_model.RidgeCV(alphas=self.__realpha__(alpha), cv=10)
elif model_type == "lasso":
alpha = regr.alpha_
regr = linear_model.LassoCV(alphas=self.__realpha__(alpha),
max_iter=5000,
cv=10)
elif model_type == "elasticnet":
alpha = regr.alpha_
ratio = regr.l1_ratio_
regr = linear_model.ElasticNetCV(
l1_ratio=self.__reratio__(ratio),
alphas=self.__elasticnet_init["alpha"],
max_iter=1000,
cv=3)
regr.fit(__X_train, __Y_train)
return regr
