If you like what we're working on, please  star us on GitHub.
DEEPCHECKS GLOSSARY

# Sensitivity and Specificity of Machine learning

Any Machine Learning program is relying on accuracy. To make it reliable for further operations, it Is mandatory to filter data through a classification process known as Confusion Matrix. The confusion matrix contains evaluation metrics relayed to machine learning such as machine learning specificity, machine learning accuracy, and machine learning sensitivity. Performance metrics are usually shown by 4 values:

True Positive – TP

False Positive – FP

True Negative – TN

False Negative – FN

## All of them are equally important in data evaluation processing.

But first, let’s define  these  terms more precisely

### Machine Learning Specificity definition

Specificity itself can be described as the algorithm/model’s ability to predict a true negative of each category available. In literature, it is also known simply as the true negative rate. Formally it can be calculated by the equation below

Specificity = TN / TN + FP (True Negative/True Negative + False Positive)

### Machine Learning Sensitivity definition

Sensitivity in Machine Learning can be described as the metric used for evaluating a model’s ability to predict the true positives of each available category.  In literature, this term can be also recognized as a true positive rate and it can be calculated with the following equation:

Sensitivity = TP / TP+FN (True Positive/True Positive + False Negative)

Specificity and sensitivity are important metrics, but nothing less important term is Machine Learning Accuracy is actually the proportion of true results, (true positive or true negative) and commonly being used with terms of model specificity, and sensitivity in the machine learning area. Machine Learning Accuracy is formally calculated through the following equation:

Accuracy = TP + TN/TP+FP+TN+FN (True Positive + True Negative/True Positive + False Positive + True Negative + False Negative)

Please have in mind that this equation gives valuable response/result only if each class contained an equal number of samples.

### Predictivenness in the Machine Learning process

Machine Learning Accuracy, Sensitivity, and Specificity are constituent parts of predictiveness training. Predictive modeling in machine learning is dependent on the dataset size specified and model performances included in the training. The potential problem can occur when the relationship is unknown or maybe doesn’t even exist for chosen datasets and performances of the model. Problem-solving can be performed through model sensitivity testing, which results often in rough limits of datasets included and needed for effective performances on larger datasets. To simplify, model sensitivity analysis is a procedure that can help discover how the effectiveness of your datasets varies with less or more data included. It is useful for model optimization and time and cost-effectiveness.

To conclude, all the measures listed above (sensitivity, accuracy, and specificity) give us equally important pieces of information about our classification model’s real value. They are important indicators to look at all of them with equal care and dedication because, if you, for example, omit specificity but include all other measurements you could create a model with high precision and recall that simply recognizes everything as true, and it couldn’t be accepted as trustworthy. Metric sensitivity is one of the major pillars of Machine Learning processing. Spare your time and resources by following exact equations for each evaluation metric so your work can be represented and recognized in the Machine Learning community worldwide.