Table 3 Stability of feature selection through resampling utilizing the Information Gain Ratio (IGR). The table presents the average IGR value and selection frequency for each feature over 100 bootstrap resampling iterations. Increased selection frequencies signify enhanced feature stability and heightened significance in stroke-risk prediction within the CHS dataset.
Attributes | Description | Mean IGR | Selection Frequency (%) |
|---|---|---|---|
AGE | AGE IN 2-YEAR CATEGORIES | 0.69451 | 98% |
GENDER | GENDER | 0.00215 | 63% |
MARIT | MARITAL STATUS | 0.09215 | 92% |
HEALTH | GENERAL HEALTH | 0.01342 | 74% |
HIBP | HIGH BLOOD PRESSURE | 0.5492 | 82% |
DIABETES | CALC. DIAB STATUS | 0.06493 | 85% |
SMOKE | SMOKING STATUS | 0.01145 | 71% |
STEPS | HAVE DIFFICULTY WALKING | 0.00514 | 65% |
EXINTEN | EXERCISE INTENSITY | 0.02648 | 75% |
OVRWT | OBESITY | 0.05941 | 83% |
HRATE | HEART RATE | 0.03841 | 78% |
ALCOH | WEEKLY ALCOHOL CONSUMPTION | 0.03248 | 78% |
Hibsug | HIGH BLOOD SUGAR STATUS | 0.06187 | 83% |
BMI | BODY MASS INDEX | 1.2457 | 94% |
HYPER | CALC. HTN STATUS | 0.02518 | 74% |
