model-performance-investigator 1.0.2

Package give idea of a models performance based on given data

Model Performance Investigator

Short Description

    Model Performance Investigator is used to analyse the performance Machine Learning and
Deep learning models. It gives the user basic idea how the model performance on given data.
So that user can start work on the right model.

Installation

pip install model-performance-analyzer

How to use ?

* Import the package.
    from model_analyzer import ml_models as ml

* Call the predector function, assign it to a variable and run.
    data = [train_X, train_y, text_x]
    new_out = ml.predector('regression', data, alg_type=['linear','lasso'], 
              score_type=['r2'], tune_param='default')

Parameters ?

1. 'regression' -> Current Version supports Regression and Classification 
                   type problems. It is mandatory.

2. data -> data must contain "train_X, train_y, text_x". train_X, train_y is for fitting
           the model and test_x is for predicting. It is mandatory.

3. alg_type -> Type of the algorithm. It is optional parameter. Provide the parameter
               values in list.  
               Default is "Linear Regression" algorithm.

               Currently it Supports:

                Algorithm Name             | Parameter 
                ---------------------------|----------
                Linear Regression        | linear
                Polynomial Regression    | polynomial
                Redig Regression         | ridge
                Lasso Regression         | lasso
                ElasticNet Regression    | elasticnet
                Random Forest Regression | random_forest_regressor
                XG Boost Regressor       | xgb_regressor

4. score_type -> Type of the score. It is optional parameter. Provide the parameter
                 values in list.   
                 Default is "r2" algorithm.
                1. Regression:
                   It Supports:
                    * r2
                    * explained_variance
                    * max_error
                    * neg_mean_absolute_error
                    * neg_mean_squared_error
                    * neg_mean_squared_log_error
                    * neg_median_absolute_error
                    * neg_mean_poisson_deviance
                    * neg_mean_gamma_deviance

                2. Classification:
                    It Supports:
                    * jaccard
                    * f1
                    * neg_log_loss
                    * roc_auc
                    * accuracy
                    * balanced_accuracy
                    * average_precision

5. tune_param -> Tuning Parameter for the model. It is optional. By Default parameter 
                 this package uses some basic parameters for each models.
                 User can provide own Tuning Parameter.

                 Template:
                     tune_param = {Name of the model: {parameters}}
                     Name of the model -> should be similar like alg_type.

                 Example:
                     linear_tune_param = {
                          "fit_intercept"       : [True, False],
                          "normalize"           : [True, False],
                          "copy_X"              : [True, False],
                          "n_jobs"              : [ -1]
                        }
                    lasso_tune_param = {
                          "alpha":[1e-15, 1e-10, 1e-8, 1e-3, 1e-2, 
                                   1, 5, 10, 15, 20, 40, 50,
                                   85, 100, 300, 500, 1000
                                   ]
                        }

                     tune_param = {'linear': linear_tune_param, 
                                  'lasso': lasso_tune_param}

                 Note : User don't have to specify tuning parameter for all the
                        models used in 'alg_type'. If tuning parameter is not provided
                        then this package use default tuning parameter.

Project History

The project was started in 2019 by Srikandan Rajua and Sathish Anandha.