EHR Functions

# Base Metric

Metrics are used to evaluate models and print results within the Jupyter notebook. Below is the standard definition of a Metric.

# Methods

# post_train

def post_train(self, train_data, val_data):
    train_true, train_pred = train_data
    val_true, val_pred = val_data

Gets called after the model is fit.

# Utilization

Metrics can be passed in one of three ways: class type, string, and class instance.

# Class Type

In this scenario, the metric was passed with just the class type and an instance was not created. The train_model function will automatically create an instance using this class.

from ehr_functions.models import train_model
from ehr_functions.models.types import LogisticRegression
from ehr_functions.models.metrics import BaseMetric

train_model(LogisticRegression(), features, df[['PatientID', 'Over30']], metrics=[BaseMetric])

# String

In this scenario, the metric was passed using a string. The class type will automatically be found, and this reduces the need to import the class.

from ehr_functions.models import train_model
from ehr_functions.models.types import LogisticRegression

train_model(LogisticRegression(), features, df[['PatientID', 'Over30']], metrics=['BaseMetric'])

# Class Instance

In this scenario, a custom argument needs to be passed to the metric and thus an instance of it is created before being passed to the function.

from ehr_functions.models import train_model
from ehr_functions.models.types import LogisticRegression
from ehr_functions.models.metrics import BaseMetric

train_model(LogisticRegression(), features, df[['PatientID', 'Over30']], metrics=[BaseMetric(custom_param=5)])

← Binary Classification →