Learning Prediction Models with False Positives and False Negatives
In this article, you will study false positives and false negatives in Machine Learning.
Learning Prediction Models
These two concepts, i.e., false positives and false negatives, fall under the category of Learning Prediction Models in Machine Learning. So, before starting directly with the false positives and false positives, let us understand the meaning of prediction models.
Prediction Models are the trained models that help in estimating the value of some variable or attribute. In other words, some input is given to the model and on the basis of that, the machine will predict the value of the new variable. Therefore, it is a kind of Supervised Learning.
The concept is easy to understand and we will look into it through an example.
We have trained a model to differentiate between two uniforms: Doctor Uniform and Police Uniform. The function of the trained model is to identify the new uniform that comes into the factory’s warehouse.
So, there are these terms called predicted value and actual value. The predicted value is what the model has identified the variable’s category while actual value is what is the actual identity of the variable.
Coming back to our uniform example, here, the uniform is the variable and the values are doctor uniform and police uniform.
a) Doctor Uniform = Positive
b) Police Uniform = Negative
(Don’t confuse between variable and Positive-Negative. It is like we have just given names of the variables just to make it simpler.)
Suppose, the model predicts that the new uniform that has arrived in the factory’s warehouse is a Doctor Uniform. This means the model has predicted that the uniform is Positive.
Similarly, if the model has predicted that the new uniform is a Police Uniform, that means, the uniform is Negative.
Identifying True Positive:
Now, the uniform that the model predicted as Doctor Uniform has the actual attribute of Doctor Uniform; in other words, the model has predicted that one of the uniforms is Positive and it was actually Positive. This makes it a True Positive.
Identifying True Negative:
Similarly, the model has predicted that the uniform is Police Uniform and it was actually a Police Uniform. Or, we can say that the model has predicted that the attribute value of this uniform is Negative and the actual value was also Negative. This makes it a True Negative.
Identifying False Positive:
But, suppose the model has predicted the attribute value of the variable as Positive or Doctor Uniform, but it actual it was a Negative or Police Uniform. In this case, it is a False Positive.
Identifying False Negative:
In the same way, if the model has predicted the attribute value of the variable as Negative or Police Uniform and the actual attribute value of the variable is Positive or Doctor Uniform. This makes it a False Negative.
So, you see false positive and false negative is just a wrong prediction done by a trained model or supervised model. Positive and Negatives are the attribute values of the variable.