Arnan Bonny

Threshold Independent Binary Classification Model Performance Measure Press enter or click to view image in full size Machine Learning models are used to either predict numbers or predict classes. The prior ones are called regression models and the latter are called classification models. Examples of classification problems are predicting whether or not a person has a disease or condition, or whether an email is spam or not. But how do we measure the performance of such models?

How Machine Learning model performance is evaluated for Binary Classification Press enter or click to view image in full size Classification models come in all shapes and sizes. Some models can perform better in some cases than in others. We either change how the model works through using different model parameters or we can use a completely separate model. To take this decision we must use a robust method to assess the performance of a model.

Ever wondered how your email service identifies spam emails and throws them into the abyss? Or how social media sites know the sentiment of a user when writing a post or a comment? Or when financial institutions have systems in place that can automatically detect fishy financial transactions? (At least the more obvious ones). Well Naive Bayes lies at the center of all such classification. Let's see how it works! What is the Naive Bayes Model?