Table 2: Models obtained by applying the different data mining methods.
CART decision tree IF (Dyspnoea when climbing one flight of steps = No) THEN COPD = No (338/3) IF (Dyspnoea when climbing one flight of steps = Yes) IF (Age group = ( < 50)) IF (Asthma family history = Yes) THEN COPD = No (6/0) IF (Asthma family history = No) IF (Phlegm in the morning = Yes) IF (Phlegm day =Yes) THEN COPD = No (3/0) IF (Phlegm day = No) THEN COPD = Yes (3/2) IF (Phlegm in the morning = No) THEN COPD = Yes (2/0) IF (Age group = ( ≥ 50)) THEN COPD = Yes (37/8) |
J48 decision tree IF (Dyspnoea when climbing one flight of steps = No) THEN COPD = No (341/3) IF (Dyspnoea when climbing one flight of steps = Yes) IF (Age group = ( < 50)) THEN COPD = No (16/5) IF (Age group = ( 50-60)) IF (Phlegm daily for three months = Yes) THEN COPD = No (4/1) IF (Phlegm daily for three months = No) THEN COPD = Yes (17/1) IF (Age group = ( > 60)) THEN COPD = Yes (24/4) |
JRip rule IF ((Dyspnoea when climbing one flight of steps = Yes) AND (Dyspnoea when walking = Yes)) THEN COPD = Yes (28/5) IF ((Dyspnoea when climbing one flight of steps = Yes) AND (Phlegm daily for three months = No) AND (Smoking intensity = ( ≥ 30)) THEN COPD = Yes (20/4) IF ((Cough daily for three months = Yes) AND (Phlegm daily for three months =No)) THEN COPD = Yes (2/0) COPD = No (352/4) |
PART decision list IF (Dyspnoea when climbing one flight of steps = No) THEN COPD = No (341/3) IF ((Heart disease = No) AND (Age group = ( ≥ 60))) THEN COPD = Yes (20/4) IF ((Heart disease = No) AND (Age group = (50 - 60)) AND (Phlegm daily for three months = No)) THEN COPD = Yes (14/1) IF (Heart disease = Yes) THEN COPD = Yes (8/0) IF (Asthma family history = Yes) THEN COPD = No (8/0) IF (Cough in the morning = Yes) THEN COPD = No (8/2) COPD = Yes (3/0) |
Logistic regression Logit pCOPD = ln pCOPD/(1-pCOPD) = -5.06 + 5.15* (Dyspnoea when climbing one flight of steps = Yes) + 1.86* (Dyspnoea when walking = Yes) + 1.89* (Heart disease = Yes) -2.16*(Waking up drowning = Yes) + 1.08* (Smoking intensity = ( > 30)) -1.70* (Age group = ( < 50)) pCOPD = probability of COPD, ln: natural logarithm (IF Logit pCOPD > 0 THEN COPD = Yes) |
Numbers in brackets, say (a/b), represent that there is a individuals of the sample with 10-fold cross validation well classified and b misclassified when the indicated decision is made.