diff --git a/Chapter_2_Background/MachineLearningForFraudDetection.md b/Chapter_2_Background/MachineLearningForFraudDetection.md
index c011bc3..ebbb8ae 100644
--- a/Chapter_2_Background/MachineLearningForFraudDetection.md
+++ b/Chapter_2_Background/MachineLearningForFraudDetection.md
@@ -86,7 +86,7 @@ y,\hat{y} &= &
 $$
 
 ```{note}
-The zero/one loss function is a standard loss function for binary classification problems. It is however not well suited for credit card fraud detection problems, due to the high-class imbalance (much more genuine than fraudulent transactions). Estimating the performance of a fraud detection system is a non-trivial issue, which will be covered in depth in [Chapter 4[(Performance_Metrics).
+The zero/one loss function is a standard loss function for binary classification problems. It is however not well suited for credit card fraud detection problems, due to the high-class imbalance (much more genuine than fraudulent transactions). Estimating the performance of a fraud detection system is a non-trivial issue, which will be covered in depth in [Chapter 4](Performance_Metrics).
 ```
 
 To obtain a fair estimate of a prediction model performance, an important methodological practice, known as *validation*, is to evaluate the performance of a prediction model on data that were not used for training. This is achieved by splitting the dataset, before training, into a *training set* and a *validation set*.  The training set is used for the training of the prediction model (that is, to find the parameters $\theta$ that minimize the loss on the training set). Once the parameters $\theta$ have been fixed, the loss is estimated with the validation set, which gives a better estimate of the performance that the prediction model is expected to have on future (and unseen) transactions.