X = np.zeros((len(data[i][0]), 400))
y = np.zeros(len(data[i][0]))
for j in tqdm.trange(len(data[i][0])):
temp = produceWordEmbd(data[i][0][j])
X[j] = temp
y[j] = data[i][1][j]
Res = []
kf = KFold(n_splits=5)
c = CorrelationPearson()
for train_index, test_index in kf.split(X):
X_train, X_test = X[train_index], X[test_index]
y_train, y_test = y[train_index], y[test_index]
model = regModel
model.fit(X_train, y_train)
model_predicted = model.predict(X_test)
Res.append(c.result(y_test, model_predicted))
print(regMethod + "- Pearson Coefficient for " + emotions[i] + ": ",
c.result(y_test, model_predicted))
print(
regMethod + ":Avg of pearson-coefficients for the " + emotions[i] +
" : ",
sum(Res) / 5)
finRes.append(sum(Res) / 5)
print("--------------------------------------------")
print("Final PC for " + regMethod, sum(finRes) / 4)
print("--------------------------------------------")
sys.stdout = orig_stdout
f.close()
return model
finRes = []
for i in range(4):
X = np.zeros((len(data[i][0]), 400))
y = np.zeros(len(data[i][0]))
for j in tqdm.trange(len(data[i][0])):
temp = produceWordEmbd(data[i][0][j])
X[j] = temp
y[j] = data[i][1][j]
Res = []
kf = KFold(n_splits=5)
c = CorrelationPearson()
for train_index, test_index in kf.split(X):
X_train, X_test = X[train_index], X[test_index]
y_train, y_test = y[train_index], y[test_index]
model = KerasRegressor(build_fn=NN_model, epochs=100, batch_size=5, verbose=0)
model.fit(X_train, y_train)
model_predicted = model.predict(X_test)
Res.append(c.result(y_test, model_predicted))
print(regMethod +"- Pearson Coefficient for "+ emotions[i] + ": ", c.result(y_test, model_predicted))
print(regMethod + ":Avg of pearson-coefficients for the " + emotions[i] + " : ", sum(Res)/5)
finRes.append(sum(Res)/5)
print("--------------------------------------------")
print("Final PC for "+ regMethod ,sum(finRes)/4)
print("--------------------------------------------")
sys.stdout = orig_stdout
f.close()
kf = KFold(n_splits=5)
c = CorrelationPearson()
for train_index, test_index in kf.split(X):
#print("TRAIN:", train_index, "TEST:", test_index)
X_train, X_test = X[train_index], X[test_index]
y_train, y_test = y[train_index], y[test_index]
Rmodel = regModels[z]
Rmodel.fit(X_train, y_train)
Rmodel_predicted = Rmodel.predict(X_test)
"""
y_test_new = get_pos_half(y_test, y_test)
Rmodel_predicted_new = get_pos_half(Rmodel_predicted, y_test)
y_test_new = y_test_new[y_test_new != 0]
Rmodel_predicted_new = Rmodel_predicted_new[Rmodel_predicted_new != 0]
"""
Res.append(c.result(y_test, Rmodel_predicted))
print("Feature used: "+ feature_names[g] +"Regression model: "+regMethods[z] +"---- Pearson Coefficient for "+ emotions[i] + ": ", c.result(y_test, Rmodel_predicted))
finRes.append(sum(Res)/5)
print("--------------------------------------------")
print("Avg PC for "+ emotions[i] + " with "+ feature_names[g] + " and "+ regMethods[z] ,sum(finRes))
print("--------------------------------------------")
avg_pc[z][i] += sum(finRes)
"""
X_new = hstack([features[0], features[1]]).todense()
Res2 = []
print("Now working for bigram+unigram:++++++++++++++++++++")
for train_index, test_index in kf.split(X_new):
#print("TRAIN:", train_index, "TEST:", test_index)
X_train, X_test = X_new[train_index], X_new[test_index]
y_train, y_test = y[train_index], y[test_index]
Rmodel = regModels[z]
for k in tqdm.trange(sum_arr[j - 1], sum_arr[j]):
dev_arrays[k - sum_arr[j - 1]] = model[emotions[j] + str(k)]
dev_labels[k - sum_arr[j - 1]] = data[j][1][k - sum_arr[j - 1]]
print("Training a Neural Network with ")
mlp = MLPRegressor(solver='sgd',
alpha=1e-5,
hidden_layer_sizes=(11, 6, 5),
random_state=1,
activation='relu',
learning_rate='adaptive')
mlp.fit(train_arrays, train_labels)
mlp_predicted = mlp.predict(dev_arrays)
# print(dev_labels - mlp_predicted)
c = CorrelationPearson()
print("pearson-coefficient for " + emotions[i] + ": ",
c.result(dev_labels, mlp_predicted))
regMethods = [
"Neural Nets", "Decision Tree", "Random Forests", "K-NN", "ADA-Boost",
"Gradient-Boost"
]
regModels = [
MLPRegressor(solver='lbfgs',
alpha=1e-5,
hidden_layer_sizes=(9, 5, 7),
random_state=1,
activation='tanh',
learning_rate='adaptive'),
DecisionTreeRegressor(random_state=0),
RandomForestRegressor(max_depth=2, random_state=0),
KNeighborsRegressor(n_neighbors=2),
#pip install correlation-pearson
from correlation_pearson.code import CorrelationPearson
X_Speed = [0.73, 0.81, 1.53, 1.97, 2.29, 2.86]
X_Energy = [1.507, 1.235, 0.654, 0.864, 0.656, 0.490]
correlation = CorrelationPearson()
print('Correlation coefficient of speed and Energy:' +
str(correlation.result(X_Speed, X_Energy)))
Y_Power = [1.0, 1.0, 1.1, 1.4, 1.5, 1.7]
Y_Energy = [0.654, 1.235, 1.507, 0.490, 0.656, 0.864]
print('Correlation coefficient of Power and Energy:' +
str(correlation.result(Y_Power, Y_Energy)))