From the course: Natural Language Processing with ML.NET by Microsoft Press

Unlock this course with a free trial

Join today to access over 24,700 courses taught by industry experts.

Training and evaluating a classification model

Training and evaluating a classification model - ML.NET Tutorial

From the course: Natural Language Processing with ML.NET by Microsoft Press

Training and evaluating a classification model

- We have previously mentioned that evaluation is a step in the process that comes after training to have an estimate of the performance of our model on unseen and new data. When talking about classification, one of the most common and intuitive evaluation matrix is accuracy. Accuracy measures among all the predictions, how many were correct. However, you need to be careful about using accuracy as a measurement of how well a model works. For example, if you have skew data. Suppose that only 3% of the population is diabetic, you could create a model that always predicts zero and it would be 97% accurate, but it would not help correctly predict cases of diabetes, right? For this reason, there are also other matrix that can help evaluate our model. One common approach for classification is building a confusion matrix. It shows cases where both the predicted and actual values were, one, known as true positives, in our example, true spam. And they are at the top left of the matrix. And…

Contents