Database Reference
In-Depth Information
Evaluating the performance of regression
models
We saw in Chapter 5 , Building a Classification Model with Spark , that evaluation methods
for classification models typically focus on measurements related to predicted class mem-
berships relative to the actual class memberships. These are binary outcomes (either the
predicted class is correct or incorrect), and it is less important whether the model just
barely predicted correctly or not; what we care most about is the number of correct and in-
correct predictions.
When dealing with regression models, it is very unlikely that our model will precisely pre-
dict the target variable, because the target variable can take on any real value. However, we
would naturally like to understand how far away our predicted values are from the true val-
ues, so will we utilize a metric that takes into account the overall deviation.
Some of the standard evaluation metrics used to measure the performance of regression
models include the Mean Squared Error ( MSE ) and Root Mean Squared Error
( RMSE ), the Mean Absolute Error ( MAE ), the R-squared coefficient, and many others.
Search WWH ::




Custom Search