How it works...

In this recipe, we built a classifier based on SVM. Once the classifier was obtained, we used a set of points to measure the distance of those points from the boundary and then measured the confidence levels for each of those points. When estimating a parameter, the simple identification of a single value is often not sufficient. It is therefore advisable to accompany the estimate of a parameter with a plausible range of values ​​for that parameter, which is defined as the confidence interval. It is therefore associated with a cumulative probability value that indirectly, in terms of probability, characterizes its amplitude with respect to the maximum values ​​assumed by the random variable that measures the probability that the random event described by that variable in question falls into this interval and is equal to this area graphically, subtended by the probability distribution curve of the random variable in that specific interval.