Meta-Model Structure Selection (MetaMSS) algorithm for building Polynomial NARX models
Meta-Model Structure Selection (MetaMSS) algorithm for building Polynomial NARX modelsΒΆ
Example created by Wilson Rocha Lacerda Junior
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sysidentpy.model_structure_selection import MetaMSS, FROLS
from sysidentpy.metrics import root_relative_squared_error
from sysidentpy.basis_function._basis_function import Polynomial
from sysidentpy.utils.display_results import results
from sysidentpy.utils.plotting import plot_residues_correlation, plot_results
from sysidentpy.residues.residues_correlation import compute_residues_autocorrelation, compute_cross_correlation
df1 = pd.read_csv('examples/datasets/x_cc.csv')
df2 = pd.read_csv('examples/datasets/y_cc.csv')
df2[5000:80000].plot(figsize=(10, 4))
<AxesSubplot:>
# we will decimate the data using d=500 in this example
x_train, x_test = np.split(df1.iloc[::500].values, 2)
y_train, y_test = np.split(df2.iloc[::500].values, 2)
basis_function = Polynomial(degree=2)
model = MetaMSS(
norm=-2,
xlag=3,
ylag=3,
estimator="recursive_least_squares",
k_agents_percent=10,
estimate_parameter=True,
maxiter=30,
n_agents=10,
loss_func='metamss_loss',
basis_function=basis_function
)
model.fit(X_train=x_train, y_train=y_train, X_test=x_test, y_test=y_test)
c:\Users\wilso\Desktop\projects\GitHub\sysidentpy\sysidentpy\narmax_base.py:824: RuntimeWarning: overflow encountered in power
np.power(raw_regressor, model_exponents[j])
c:\Users\wilso\miniconda3\envs\sysidentpy\lib\site-packages\numpy\core\fromnumeric.py:87: RuntimeWarning: invalid value encountered in reduce
return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
c:\Users\wilso\Desktop\projects\GitHub\sysidentpy\sysidentpy\model_structure_selection\meta_model_structure_selection.py:435: RuntimeWarning: overflow encountered in square
sum_of_squared_residues = np.sum(residues**2)
c:\Users\wilso\Desktop\projects\GitHub\sysidentpy\sysidentpy\metrics\_regression.py:215: RuntimeWarning: overflow encountered in square
numerator = np.sum(np.square((y_predicted - y)))
c:\Users\wilso\Desktop\projects\GitHub\sysidentpy\sysidentpy\metrics\_regression.py:216: RuntimeWarning: overflow encountered in square
denominator = np.sum(np.square((y_predicted - np.mean(y, axis=0))))
c:\Users\wilso\Desktop\projects\GitHub\sysidentpy\sysidentpy\metrics\_regression.py:217: RuntimeWarning: invalid value encountered in true_divide
return np.sqrt(np.divide(numerator, denominator))
c:\Users\wilso\miniconda3\envs\sysidentpy\lib\site-packages\numpy\core\fromnumeric.py:87: RuntimeWarning: overflow encountered in reduce
return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
c:\Users\wilso\miniconda3\envs\sysidentpy\lib\site-packages\numpy\linalg\linalg.py:2567: RuntimeWarning: divide by zero encountered in power
absx **= ord
<sysidentpy.model_structure_selection.meta_model_structure_selection.MetaMSS at 0x27b3c21e490>
yhat = model.predict(X_test=x_test, y_test=y_test)
rrse = root_relative_squared_error(y_test, yhat)
print(rrse)
r = pd.DataFrame(
results(
model.final_model, model.theta, model.err,
model.n_terms, err_precision=8, dtype='sci'
),
columns=['Regressors', 'Parameters', 'ERR'])
print(r)
plot_results(y=y_test, yhat = yhat, n=1000)
ee = compute_residues_autocorrelation(y_test, yhat)
plot_residues_correlation(data=ee, title="Residues", ylabel="$e^2$")
x1e = compute_cross_correlation(y_test, yhat, x_test)
plot_residues_correlation(data=x1e, title="Residues", ylabel="$x_1e$")
0.038462417177693324
Regressors Parameters ERR
0 1 -3.2204E+02 0.00000000E+00
1 y(k-1) 1.3930E+00 0.00000000E+00
2 y(k-2) -5.4673E-01 0.00000000E+00
3 y(k-3) 6.3161E-02 0.00000000E+00
4 x1(k-1) 2.2296E+01 0.00000000E+00
5 x1(k-2) 1.7402E+01 0.00000000E+00
6 x1(k-3) -4.2603E+00 0.00000000E+00
7 y(k-3)y(k-1) 2.8055E-05 0.00000000E+00
8 x1(k-1)y(k-1) -1.5244E-01 0.00000000E+00
9 x1(k-2)y(k-1) -1.3620E-01 0.00000000E+00
10 x1(k-3)y(k-1) 3.9445E-02 0.00000000E+00
11 y(k-3)y(k-2) -1.6138E-05 0.00000000E+00
12 x1(k-1)y(k-2) 9.2426E-02 0.00000000E+00
13 x1(k-2)y(k-2) 8.1630E-02 0.00000000E+00
14 x1(k-3)y(k-2) -2.0450E-02 0.00000000E+00
15 x1(k-1)y(k-3) -2.3994E-02 0.00000000E+00
16 x1(k-2)y(k-3) -2.1218E-02 0.00000000E+00
17 x1(k-1)^2 1.1148E+02 0.00000000E+00
18 x1(k-2)x1(k-1) -3.6302E+00 0.00000000E+00
19 x1(k-2)^2 8.7010E+01 0.00000000E+00
20 x1(k-3)^2 -2.1302E+01 0.00000000E+00
# Ploting the evolution of the agents
plt.plot(model.best_by_iter)
model.best_by_iter[-1]
0.004650987888650369
# You have access to all tested models
# model.tested_models
from sklearn.neighbors import KNeighborsRegressor
from sklearn.svm import SVC, LinearSVC, NuSVC
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import RandomForestRegressor, AdaBoostRegressor, GradientBoostingRegressor
from sklearn.naive_bayes import GaussianNB
from catboost import CatBoostRegressor
from sklearn.linear_model import BayesianRidge, ARDRegression
from sysidentpy.general_estimators import NARX
xlag = ylag = 10
estimators = [
('NARX_KNeighborsRegressor', NARX(
base_estimator=KNeighborsRegressor(),
xlag=xlag,
ylag=ylag,
basis_function=basis_function)),
('NARX_DecisionTreeRegressor', NARX(
base_estimator=DecisionTreeRegressor(),
xlag=xlag,
ylag=ylag,
basis_function=basis_function)),
('NARX_RandomForestRegressor', NARX(
base_estimator=RandomForestRegressor(
n_estimators=200),
xlag=xlag,
ylag=ylag,
basis_function=basis_function
)),
('NARX_Catboost', NARX(
base_estimator=CatBoostRegressor(
iterations=800,
learning_rate=0.1,
depth=8),
xlag=xlag,
ylag=ylag,
basis_function=basis_function,
fit_params={'verbose': False}
)),
('NARX_ARD', NARX(
base_estimator=ARDRegression(),
xlag=xlag,
ylag=ylag,
basis_function=basis_function
)),
('FROLS-Polynomial_NARX', FROLS(
order_selection=True,
n_info_values=50,
extended_least_squares=False,
ylag=ylag, xlag=xlag,
info_criteria='bic',
estimator='recursive_least_squares',
basis_function=basis_function
)
),
('MetaMSS', MetaMSS(norm=-2,
xlag=xlag,
ylag=ylag,
estimator="recursive_least_squares",
k_agents_percent=10,
estimate_parameter=True,
maxiter=20,
n_agents=15,
loss_func='metamss_loss',
basis_function=basis_function)
)
]
resultados = {}
for nome_do_modelo, modelo in estimators:
resultados['%s' % (nome_do_modelo)] = []
if nome_do_modelo == 'MetaMSS':
modelo.fit(X_train=x_train, y_train=y_train, X_test=x_test, y_test=y_test)
yhat = modelo.predict(X_test=x_test, y_test=y_test)
else:
modelo.fit(X=x_train, y=y_train)
yhat = modelo.predict(X=x_test, y=y_test)
if nome_do_modelo in ['MetaMSS', 'FROLS-Polynomial_NARX']:
result = root_relative_squared_error(y_test[modelo.max_lag:], yhat[modelo.max_lag:])
else:
result = root_relative_squared_error(y_test, yhat)
resultados['%s' % (nome_do_modelo)].append(result)
print(nome_do_modelo, '%.3f' % np.mean(result))
NARX_KNeighborsRegressor 1.992
NARX_DecisionTreeRegressor 0.264
NARX_RandomForestRegressor 0.223
NARX_Catboost 0.175
NARX_ARD 0.074
FROLS-Polynomial_NARX 0.047
c:\Users\wilso\Desktop\projects\GitHub\sysidentpy\sysidentpy\narmax_base.py:824: RuntimeWarning: overflow encountered in power
np.power(raw_regressor, model_exponents[j])
c:\Users\wilso\miniconda3\envs\sysidentpy\lib\site-packages\numpy\core\fromnumeric.py:87: RuntimeWarning: invalid value encountered in reduce
return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
c:\Users\wilso\Desktop\projects\GitHub\sysidentpy\sysidentpy\model_structure_selection\meta_model_structure_selection.py:435: RuntimeWarning: overflow encountered in square
sum_of_squared_residues = np.sum(residues**2)
c:\Users\wilso\Desktop\projects\GitHub\sysidentpy\sysidentpy\model_structure_selection\meta_model_structure_selection.py:445: RuntimeWarning: invalid value encountered in sqrt
se_theta = np.sqrt(var_e)
c:\Users\wilso\miniconda3\envs\sysidentpy\lib\site-packages\numpy\core\fromnumeric.py:87: RuntimeWarning: overflow encountered in reduce
return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
c:\Users\wilso\Desktop\projects\GitHub\sysidentpy\sysidentpy\metrics\_regression.py:215: RuntimeWarning: overflow encountered in square
numerator = np.sum(np.square((y_predicted - y)))
c:\Users\wilso\Desktop\projects\GitHub\sysidentpy\sysidentpy\metrics\_regression.py:216: RuntimeWarning: overflow encountered in square
denominator = np.sum(np.square((y_predicted - np.mean(y, axis=0))))
c:\Users\wilso\miniconda3\envs\sysidentpy\lib\site-packages\numpy\linalg\linalg.py:2567: RuntimeWarning: divide by zero encountered in power
absx **= ord
MetaMSS 0.025
for aux_results, results in sorted(resultados.items(), key=lambda x: np.mean(x[1]), reverse=False):
print(aux_results, np.mean(results))
MetaMSS 0.025150310881841744
FROLS-Polynomial_NARX 0.04663897799085836
NARX_ARD 0.07413356855178779
NARX_Catboost 0.17463008529882576
NARX_RandomForestRegressor 0.2230679886687048
NARX_DecisionTreeRegressor 0.2643109391717239
NARX_KNeighborsRegressor 1.9917717932592363