Page 307 - Artificial Intellegence_v2.0_Class_11
P. 307
However, this parameter becomes useless as the actual cases where the patient has diabetes are not taken into account.
Hence, there we need to look at another parameter that considers such cases as well.
Precision
Precision is described as the percentage of positive identifications which were correct. This takes the True Positives and
False Positives into consideration.
TP
Precision = × 100%
TP+ FP
651/(651+120) × 100%
84.43%
If the Precision value is low, it means there are more false alarms than the actual ones.
Recall
Another parameter used to evaluate the performance of the model is Recall. It can be defined as the proportion of
positive cases that are correctly identified. It mainly considers the situation where the event occurred in reality and
whether the machine detects it correctly or not. In other words, it looks at true positive values.
Using the above example,
TP
Recall =
TP+ FN
651/(651+321) = 0.66
F1 Score
In some cases, you might come across high precision with low recall, or low precision but high recall. However, because
these two metrics are important, a parameter is required that takes into account both precision and recall. F1 Score gives
a measure of the balance between precision and recall.
Taking the above example,
Precision × Recall
F1 Score = 2 ×
Precision + Recall
0.84 × 0.66
= 2 ×
0.84 + 0.66
= 0.73
A good F1 score means that both the false positives and false negatives are low, so the AI model is identifying the real
threats properly and is not giving any false alarms. F1 score ranges between 0 and 1. If the F1 score is 1 (100%), the
model is considered perfect, and if the score is 0, that means that the model has failed completely.
False Positive or False Negative in Medical Science
A false positive, in medical testing, is actually an error. A medical test may wrongly report the presence of a disease (as
the test result is positive), when in reality the patient does not suffer from that disease.
A false negative is also an error. In this case, the test result improperly shows the absence of a disease, when in reality the
patient is suffering from that disease. These are the two kinds of errors given by a binary classification model.
While many medical tests conducted nowadays are accurate and reliable, however, there are still, few cases of false
positives or false negatives. Their implications on the patient or his family are quite severe.
Classification & Clustering 305

