Reflection 3

Feature	Accuracy	Precision	Recall	F1
Gender	0.6532	0.6480	0.6532	0.6500
Daytime/evening attendance	0.6134	0.5074	0.6134	0.5200
Scholarship holder	0.6122	0.3749	0.6122	0.4650
Tuition fees up to date	0.7221	0.7886	0.7221	0.6758
International	0.6070	0.3684	0.6070	0.4585
Debtor	0.6838	0.7070	0.6838	0.6243
Educational special needs	0.6088	0.3709	0.6088	0.4609

One feature contingency table for “tuition fees up to date”

Feature (w/tuition)	Accuracy	Precision	Recall	F1
Gender	0.7254	0.7920	0.7254	0.6816
Daytime/evening attendance	0.7270	0.7868	0.7270	0.6844
Scholarship holder	0.7208	0.7868	0.7208	0.6757
International	0.7460	0.7995	0.7460	0.7073
Debtor	0.7404	0.7618	0.7404	0.7124
Educational special needs	0.7193	0.7814	0.7193	0.6754

Two feature contingency table for tuition and debtor

Feature (w/tuition and debtor)	Accuracy	Precision	Recall	F1
Gender	0.7300	0.7560	0.7300	0.7022
Daytime/evening attendance	0.7212	0.7517	0.7212	0.6900
Scholarship holder	0.7257	0.7679	0.7257	0.6919
International	0.7254	0.7569	0.7254	0.6955
Educational special needs	0.7156	0.7356	0.7156	0.6858

Three feature contingency table for tuition debtor and gender

Feature (w/tuition, debtor and gender)	Accuracy	Precision	Recall	F1
Daytime/evening attendance	0.7267	0.7424	0.7267	0.7040
Scholarship holder	0.7337	0.7342	0.7337	0.7339
International	0.7300	0.7570	0.7300	0.7011
Educational special needs	0.7212	0.7453	0.7212	0.6920

Four feature contingency table for tuition, debtor, gender, and scholarship holder

Feature (w/tuition, debtor, gender and scholarship)	Accuracy	Precision	Recall	F1
Daytime/evening attendance	0.7337	0.7401	0.7337	0.7251
International	0.7325	0.7309	0.7325	0.7315
Educational special needs	0.7331	0.7446	0.7331	0.7210

Five feature contingency table for tuition, debtor, gender, scholarship holder, and international

Feature (w/tuition, debtor, gender, scholarship and international)	Accuracy	Precision	Recall	F1
Daytime/evening attendance	0.7368	0.7535	0.7368	0.7117
Educational special needs	0.7420	0.7418	0.7420	0.7427

Six feature contingency table for tuition, debtor, gender, scholarship holder, international, and educational special needs

Feature (w/tuition, debtor, gender, scholarship, international and special needs)	Accuracy	Precision	Recall	F1
Daytime/evening attendance	0.7282	0.7447	0.7282	0.7089

Seven feature contingency table for tuition, debtor, gender, scholarship holder, international, educational special needs and daytime/evening attendance

After conducting this experiment, I concluded that the best Naive Bayes classifier for this dataset to predict if someone graduates or drops out uses the six features tuition, debtor, gender, scholarship holder, international, and educational special needs. These six features created a classifier with the highest tested F1 score of 0.7427