/COVID-19-outcome-classification

This is my analysis of different classifiers on WHO organization's dataset using scikit-learn.

Primary LanguageJupyter Notebook

COVID 19 Outcome Classifier

This is my analysis of different classifiers on WHO organization's dataset using scikit-learn.

KNeighborsClassifier, LogisticRegression, and GaussianNB: COVID19_outcome_classifier.ipynb

DecisionTreeClassifier and SVC: COVID19_outcome_classifier2.ipynb

View the notebook using nbviewer from here:

Best hyperparameters for each model

K-Nearest Neighbors (KNN)

Parameter value
n_neighbors 3
weights uniform
algorithm auto
leaf_size 10
p 2
metric minkowski
metric_params None
n_jobs None

Logistic Regression (LR)

Parameter value
penalty l2
dual False
tol 0.0001
C 112.9
fit_intercept True
intercept_scaling 1
class_weight None
random_state 42
solver lbfgs
max_iter 100
multi_class auto
verbose 0
warm_start False
n_jobs None
l1_ratio None

Naïve Bayes (NB)

Parameter value
priors None
var_smoothing 1000.0

Decision Trees (DT)

---

Support Vector Machines (SVM)

---

Each model's performance

--- Precision Recall f1 Cross Validation ROC AUC
K-Nearest Neighbors (KNN) 0.96 0.96 0.96 0.7069 --- 0.90
Logistic Regression (LR) 1.00 0.95 0.97 0.7055 --- 0.93
Naïve Bayes (NB) 1.00 0.86 0.92 0.9375 --- 0.88
Decision Trees (DT) --- --- --- --- --- ---
Support Vector Machines (SVM) --- --- --- --- --- ---

Conclusion

References

  1. Understanding The Confusion Matrix From Scikit Learn