machine-learning-challenge

Compared the various model's and the SVM model and Random Forest Model, seemed to be giving the best results among all. And on comparing the SVM Model and Random Forest Model, the Random Forest Model gives a better accuracy and seems like the best for the Exoplanet dataset. Saving grid2 which is of Random Forest as it has a better accuracy of .90

SVM Model

SVC model Train Acc: 0.846 SVC model Test Acc: 0.842

Classification Report precision recall f1-score support

FALSE POSITIVE 0.70 0.62 0.66 411 CANDIDATE 0.71 0.76 0.73 484 CONFIRMED 0.98 1.00 0.99 853

  accuracy                           0.84      1748
 macro avg       0.80      0.79      0.79      1748

weighted avg 0.84 0.84 0.84 1748

After Parameter Tuning

{'C': 1000, 'gamma': 0.0001, 'kernel': 'linear'} Best Score : 0.887659736791913 Test Score : 0.8890160183066361 precision recall f1-score support

FALSE POSITIVE 0.81 0.72 0.76 411 CANDIDATE 0.79 0.84 0.81 484 CONFIRMED 0.98 1.00 0.99 853

  accuracy                           0.89      1748
 macro avg       0.86      0.85      0.86      1748

weighted avg 0.89 0.89 0.89 1748

Random Forest Model

RandomForestClassifier model Train Acc: 1.000 RandomForestClassifier model Test Acc: 0.901 Feature of highest Importance is -- koi_fpflag_co Classification Report precision recall f1-score support

FALSE POSITIVE 0.83 0.76 0.80 411 CANDIDATE 0.84 0.85 0.85 484 CONFIRMED 0.96 1.00 0.98 853

  accuracy                           0.90      1748
 macro avg       0.88      0.87      0.87      1748

weighted avg 0.90 0.90 0.90 1748

After Parameter Tuning

{'max_depth': 175, 'n_estimators': 300} Best Score - 0.8933816517261111 Test Score - 0.9038901601830663 Classification Report

            precision    recall  f1-score   support

FALSE POSITIVE 0.84 0.77 0.80 411 CANDIDATE 0.84 0.86 0.85 484 CONFIRMED 0.97 1.00 0.98 853

  accuracy                           0.90      1748
 macro avg       0.88      0.87      0.88      1748

weighted avg 0.90 0.90 0.90 1748