Page 241 - Big Data Analytics for Intelligent Healthcare Management

234 CHAPTER 9 INTELLIGENCE-BASED HEALTH RECOMMENDATION SYSTEM

Table 9.1 Confusion Matrix
Total Population Condition Positive Condition Negative
Predicted condition positive True positive False positive
Predicted condition negative False negative True negative

4. False positive (FP): If the instance is negative but it is classified as positive.

• Sensitivity is the proportion of actual positives which are correctly identified as positives by the
classifier. It is described in Eq. (9.6).

TP
Sensitivity ¼ (9.6)
ð TP + FNÞ
• Specificity measures the proportion of correctly identified excluded condition when the condition is
not present. This is described in Eq. (9.7) [15].

TN
Specificity ¼ (9.7)
ð TN + FPÞ
• Predictive value positive is the proportion of positives that correspond to the presence of the
condition. This is described in Eq. (9.8).
TP
Predictive value positive ¼ (9.8)
ð TP + FPÞ
• Predictive value negative is the proportion of negatives that correspond to the absence of the
condition. It is described in Eq. (9.9).

TN
Predictivevalue negative ¼ (9.9)
ð TN + FNÞ
Precision: This is a measure of retrieved instances that are relevant, while recall can be defined as the
fraction of correctly recommended items that are also part of the collection of useful recommended
items. Precision is described in Eq. (9.10).
Correctly recommended items
ðÞ
Precision P ¼ (9.10)
Total recommended items
Recall is described in Eq. (9.11).
Correctly recommended items
Recall RðÞ ¼ (9.11)
Total useful recommended items
The F-measure aids to combine both (precision and recall) into a single score or metric. It is described
in Eq. (9.12).
2PR
F Measure ¼ (9.12)
ð P + RÞ