helper_methods.py

from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LassoCV, LogisticRegression, LogisticRegressionCV, MultiTaskLassoCV, RidgeCV
from sklearn.multiclass import OneVsRestClassifier
from sklearn.decomposition import PCA
from sklearn.metrics import mean_squared_error
import scipy


def pipeline_regression(X_train,y_train,X_test,regression_method,seed,n_components=None):

    scaler = StandardScaler()
    X_train = scaler.fit_transform(X_train)
    X_test = scaler.transform(X_test)
    var = 0

    if n_components is not None:
        pca = PCA(n_components=n_components)
        X_train=pca.fit_transform(X_train)
        X_test=pca.transform(X_test)

        variance_explained = pca.explained_variance_ratio_
        for i in range(20):
            var = var+variance_explained[i]

    linreg =regression_method(X_train,y_train,seed)
    # return var
    return linreg,X_test,var

def custom_logistic_regression(X,y,seed):
    clf = LogisticRegressionCV(cv=5, random_state=seed,max_iter=200,scoring='roc_auc')
    multi_clf =  OneVsRestClassifier(clf,n_jobs=-1)
    estimator= multi_clf.fit(X,y)
    return estimator


def custom_linear_regression(X, y, seed):
    # print(y.shape)
    if len(y.shape) > 1:
        linreg = MultiTaskLassoCV(max_iter=1000, n_alphas=200, random_state=seed, n_jobs=-1)
    else:
        linreg = LassoCV(max_iter=1000, n_alphas=200, random_state=seed, n_jobs=-1)
    try:
        estimator = linreg.fit(X, y)
    except ValueError:
        print("Error in fitting")
        return None
    return estimator


def metrics_per_descritor(X, y, linreg):
    predicted = linreg.predict(X)
    mseerrors = []
    correlations = []
    if len(y.shape) > 1:
        for i in range(y.shape[1]):
            mseerror = mean_squared_error(predicted[:, i], y[:, i])
            correlation = scipy.stats.pearsonr(predicted[:, i], y[:, i])
            mseerrors.append(mseerror)
            correlations.append(correlation)

    else:
        mseerror = mean_squared_error(predicted, y)
        correlation = scipy.stats.pearsonr(predicted, y)
        mseerrors.append(mseerror)
        correlations.append(correlation)

    return predicted, mseerrors, correlations