Machine Learning Model Selection

This package aims to facilitate model selection in Machine Learning. It is a common issue that ML practitioners often struggle to decide on the most appropriate model prior to optimization, as tuning hyperparameters can be time-consuming and computationally demanding. To simplify the process, this package enables users to train several machine learning models using their default hyperparameters and compare their performance, helping them determine the most suitable model to select.

Usage

pip install mlms -U

pypi package

Then instantiate and use it like this:

from mlms.ModelSelection import Select_Regressor, Select_Classifier

Select some models to tune, this list should be the abbreviation of models as below, for example

MODELS = ['LGR', 'AB', 'CART', 'GBC', 'XGBC', 'RFC', 'ETC', 'KNN', 'NB', 'SVC', 'MLP', 'SGDC', 'GPC', 'PAC']

df_performance, fitted_classifiers = Select_Classifier('accuracy', 10, X_train, X_test, y_train, y_test, MODELS)

df_performance, fitted_regressors = Select_Classifier('neg_mean_squared_erro', 10, X_train, X_test, y_train, y_test)

For classifiers, the performance can set as accuracy , 'f1_score , precision, recall, roc_auc , balanced_accuracy_score and so on. Available classifiers are below

• ('LGR', LogisticRegression(n_jobs=-1)),
• ('AB', AdaBoostClassifier()),
• ('CART', DecisionTreeClassifier()),
• ('GBC', GradientBoostingClassifier()),
• ('XGBC', XGBClassifier()),
• ('RFC', RandomForestClassifier()),
• ('ETC', ExtraTreeClassifier()),
• ('KNN', KNeighborsClassifier(n_jobs=-1)),
• ('NB', GaussianNB()),
• ('SVC', SVC()),
• ('MLP', MLPClassifier()),
• ('SGDC', SGDClassifier(n_jobs=-1)),
• ('GPC', GaussianProcessClassifier(n_jobs=-1)),
• ('PAC', PassiveAggressiveClassifier(n_jobs=-1))

(The charts is an classifier selection example using Iris dataset)

For regressors, the performance can set as r2_score, neg_mean_squared_error and so on. Available regressors are below:

• ('KNN', KNeighborsRegressor()),
• ('CART', DecisionTreeRegressor()),
• ('SVR', SVR()),
• ('MLP', MLPRegressor()),
• ('ABR', AdaBoostRegressor()),
• ('GBR', GradientBoostingRegressor()),
• ('XGB', XGBRegressor()),
• ('RFR', RandomForestRegressor()),
• ('ETR', ExtraTreesRegressor())

Additonally, this package also alow users to plot ROC_Curve

from mlms.plot_roc_curve import Multiclass_ROC_Curve, Binary_ROC_Curve

Multiclass_ROC_Curve(X_test, y_test, fitted_model, chart_title:str)

Binary_ROC_Curve(y_true, y_pred,chart_name:str)

GitHub