Step 7: Evaluate & Apply Model

Apr 8 – Apr 22 | Evaluate & Apply Model

Step 7 - Evaluate & Apply Model

Before proceeding to Step 7 - Evaluate & Apply Model, you will need to download the models folder (MEDomicsLab_TestingPhase_Step7.zip) into your EXPERIMENTS folder. This folder contains the models we prepared for you. One of the models corresponds to the model you created during Step 6 - Create Model. However, for consistency across all participants of the Testing Phase, we recommend using the model we provided specifically for Step 7 - Evaluate & Apply Model.

Additionally, for this step, you will also need the data we sent you for Step 6 - Create Model (MEDomicsLab_TestingPhase_Step6.zip), which contains the holdout patient set we will use to evaluate our models.

An invitation to access the MEDomicsLab_TestingPhase_Step7.zip data has been sent to you via email.

In this current Step 7 - Evaluate & Apply Model, we will explore the functionalities of the Evaluation Module by evaluating two machine learning models on our holdout set. As the models were created using only Time point 1 and Time point 2 from the learning set (see Step 6 - Create Model for more details), we are going to evaluate them on Time point 1 and Time point 2 from the holdout set.

Additionally, we will explore the functionalities of the Application Module by applying one of the models to a single patient from the holdout set.

Model 1: ExtraTrees Classifier

Documentation related to this model is available here. In our experiment, we kept the model's default values.

This model is the one we obtained at the end of Step 6 - Create Model. It is trained on Time point 1 and Time point 2 from our learning set, using the following columns (T1 and T2 suffix refers to T1 and T2 datasets):

• tslab_|_attr_MCHC__maximum_T1

• nradiology_|_attr4_T1

• nradiology_|_attr6_T1

• image_|_attr5(3)_T1

• image_|_attr7(3)_T1

• demographics_|_anchor_age_T1

• nradiology_|_attr4_T2

• nradiology_|_attr6_T2

• tslab_|_attr_Platelet_Count__mean_T2

• tslab_|_attr_MCHC__maximum_T2

• tslab_|_attr_MCH__maximum_T2

• image_|_attr5(3)_T2

• image_|_attr7(3)_T2

Model 2: Random Forest Classifier

Documentation related to this model is available here. In our experiment, we kept the model's default values.

We created this model specifically for this step. Like the previous model, it was trained on Time point 1 and Time point 2 from the learning set, using the same columns.

Apply Model

This section involves selecting a model and applying it to a new patient. You will need to enter the patient's required information, based on the columns the model was trained on, and observe the prediction the model makes for this new patient.

Please note that the Evaluation Module dashboard is a basic implementation of the ExplainerDashboard Python library. Additionally, if you are seeking information about the dashboard elements, you may find it in the ExplainerDashboard documentation.

Also, if you want to fully understand how ExplainerDashboard works in the background, it is an open-source library, and the code is available on GitHub.

Recommendations

Before proceeding with Step 7 - Evaluate & Apply Model of the MEDomicsLab Testing Phase, we recommend consulting the documentation of the Evaluation Module and the Application Module.

Evaluation Module

Application Module

Instructions for Step 7 - Evaluate & Apply Model

Content

Intro 0:00

Evaluate 1st model 2:25

Evaluate 2nd model 14:59

Apply model 18:19