# Neural Network
#nn = MLPRegressor(hidden_layer_sizes=(30), activation='tanh',max_iter=300)
nn = MLPRegressor(hidden_layer_sizes=(5), activation='tanh', solver='lbfgs')
#solver='lbfgs'
n = nn.fit(X_train, y_train)
print('NN', nn.score(X_train, y_train))
print('NN', nn.score(X_test, y_test))
#y_rbf = svr_rbf.fit(X, y).predict(XX)
#y_knn = KNeighborsRegressor(n_neighbors=20).fit(X,y).predict(XX)
#linreg = linear_model.LinearRegression().fit(X, y).predict(XX)
#lasso =linear_model.Lasso().fit(X, y).predict(XX)
#nn = MLPRegressor(hidden_layer_sizes=(30), activation='tanh',max_iter=300).fit(X, y).predict(XX)
nn = MLPRegressor(hidden_layer_sizes=(5), activation='tanh',
solver='lbfgs').fit(X, y).predict(XX)
lw = 2
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
ax.scatter(df_final['Calls Offered'],
df_final['Overflow Calls'],
df_final['number_agents'],
c=df_final['number_agents'],
marker='o',
s=30)
ax.plot(x_line, y_line, nn.ravel(), color='navy', lw=lw, label='RBF model')
plt.title('Agent Forecasting')
ax.set_xlabel('Calls Offered')
ax.set_ylabel('Overflow Calls')
ax.set_zlabel('Number of Agents')
