def pls_crossval(X, y, n_comp, **kwargs):
# fits a pls model with a given number of components
model = pls_regression(X, y, n_comp)
# calculated score and mse from given model
pls_scores(X, y, model)
return model
def pls_crossval(X, y, n_comp, **kwargs):
opt_comp = optimal_n_comp(X, y, n_comp)
opt_model = pls_regression(X, y, opt_comp)
pls_scores(X, y, opt_model)
return opt_model
def pls_crossval(X, y, n_comp = 10, plot=False, **kargs):
"""returns a model with the optimsed number of pls components"""
# returns n_comp with lowest loss
opt_comp = optimal_n_comp(X, y, n_comp, plot=plot)
# performs regression with n_comp
opt_model = pls_regression(X,y, opt_comp, plot=plot)
# returns regression scores
pls_scores(X,y, opt_model)
return opt_model
# pip_dev0["variable_selection"].plot(wave_number, X_0)
# # %%
# pip_dev0["variable_selection"].plot_feature_importance(wave_number)
# %%
def pls_crossval(X, y, n_comp, **kwargs):
opt_comp = optimal_n_comp(X, y, n_comp)
variance_explained(X, y, n_comp)
# %%
pls_crossval(**data_en0, n_comp=15)
# %%
model_0 = pls_regression(**data_en0, n_comp=3, plot=False)
cv_benchmark_model(**data_en0, model=model_0, y_unscaled=y, ref=ref, plot=True)
val_regression_plot(**data_en0, model = model_0)
# %%
pls_opt_0 = PLSOptimizer()
pls_opt_0.fit(X_train_0, y_train, max_comp=2)
# transformation pipeline
# %%
pip_dev1 = Pipeline(
pls_opt.fit(X_train_0, y_train, max_comp=20)
pls_opt.plot(wave_number, X_train_0)
# %%
X_train_sel = pls_opt.transform(X_train_0)
X_test_sel = pls_opt.transform(X_test_0)
data_sel = {
"X": X_train_sel,
"y": y_train,
"X_test": X_test_sel,
"y_test": y_test
}
# %%
def pls_crossval(X, y, n_comp, **kwargs):
opt_comp = optimal_n_comp(X, y, n_comp)
opt_model = pls_regression(X, y, opt_comp)
pls_scores(X, y, opt_model)
return opt_model
pls_crossval(X_train_sel, y_train, 10)
model = pls_regression(**data_sel, n_comp=2)
cv_benchmark_model(**data_sel, model=model, y_unscaled=y, ref=feat_ref)
mase_2, comp_2 = mse_minimum(X_test_0, y_test, plot=False) extra_plot_mse(mse, comp, mase_2, comp_2) # %% # variance explained and MSECV for train and test set for each component var, comp = variance_explained(X_train_0_sel, y_train, plot=False) var_2, comp_2 = variance_explained(X_test_0_sel, y_test, plot=False) extra_plot_variance_explained(var, comp, var_2, comp_2) mse, comp = mse_minimum(X_train_0, y_train, plot=False) mase_2, comp_2 = mse_minimum(X_test_0, y_test, plot=False) extra_plot_mse(mse, comp, mase_2, comp_2) # %% model_0 = pls_regression(**data_en0, n_comp=5, plot=False) print_cv_table(**data_en0, model=model_0) # %% cv_benchmark_model(**data_en0, model=model_0, y_unscaled=y, ref=ref, plot=True) #val_regression_plot(**data_en0, model=model_0) #val_regression_plot(**data_en_sel, model=model_sel) # %% model_sel = pls_regression(**data_en_sel, n_comp=2, plot=False) print_cv_table(**data_en_sel, model=model_sel) cv_benchmark_model(**data_en_sel, y_unscaled=y, ref=ref, model=model_sel)
return opt_model
# %%
variance_explained(X_train_en, y_train, n_comp=20, plot=True)
variance_explained(X_train_pip, y_train, n_comp=20, plot=True)
# %%
# %%
model_en = pls_regression(**data_en, n_comp =3)
model_pip = pls_regression(**data_pip, n_comp =2)
# %%
def print_regression_table(X, y, X_test, y_test, model, y_unscaled, ref):
print_nir_metrics(X, y, X_test, y_test, model, y_unscaled, ref)
print_regression_benchmark(X, y, X_test, y_test, model)
print_cv_table(X, y, X_test, y_test, model)
print_regression_table(**data_en, model=model_en, y_unscaled=y, ref=ref)
# %%
