Feature importance do dataset integrado

[ryukinix]

Usar decision tree ou regressão logística, extrair as métricas e feature e importance

[helen0l]

Decision Tree Feature Importance:			Logistic Regression Feature Importance:
SEVERITY	26%		COUGHING	29%
OTHER_SYMPTOMS	8%		CHEST_PAIN	12%
GENDER_FEMALE	7%		FATIGUE	6%
CHEST_PAIN	7%		RESPIRATORY_SYMPTOMNS	6%
SMOKING	6%		SNORING	5%
RESPIRATORY_SYMPTOMNS	6%		SMOKING	5%
SWALLOWING_DIFFICULTY	5%		SWALLOWING_DIFFICULTY	4%
COUGHING	5%		SEVERITY	3%
COLD_SYMPTOMNS	5%		OTHER_SYMPTOMS	2%
GENDER_MALE	4%		GENDER_FEMALE	1%
AGE_25_59	3%		GENDER_MALE	1%
SNORING	3%		COLD_SYMPTOMNS	-14%
AGE60	3%		SHORTNESS_OF_BREATH	-21%
AGE_0_9	3%		AGE_25_59	-44%
AGE_20_24	2%		AGE60	-45%
AGE_10_19	2%		AGE_0_9	-46%
FATIGUE	2%		AGE_20_24	-48%
SHORTNESS_OF_BREATH	2%		AGE_10_19	-48%

[helen0l]

Random Forest Test Accuracy: 0.9386218603627676 Random Forest Test ROC AUC Score: 0.5148739674606172

Confusion Matrix - Random Forest: [[59595 60] [ 3845 122]]

[ryukinix]

E o F1 score da classe target (1), @helen0l ? Com exceção da matriz de confusão, essas métricas são irrelevantes para esse problema com classes extremamente desbalanceadas.

[helen0l]

Random Forest Model Evaluation Metrics: Test Accuracy: 0.9386218603627676 ROC AUC Score: 0.5148739674606172 F1 Score: 0.058809351651000236 Confusion Matrix:

[[59595    60]
 [ 3845   122]]

Logistic Regression Model Evaluation Metrics: Test Accuracy: 0.9382603501933293 ROC AUC Score: 0.5062097946310351 F1 Score: 0.024826216484607744 Confusion Matrix:

[[59644    11]
 [ 3917    50]]

[ryukinix]

Obrigado por refazer as métricas! @helen0l