1. Given a classification problem and a dataset, where each record has several attributes and a class label, a learning

[answerhappygod]

1 Given A Classification Problem And A Dataset Where Each Record Has Several Attributes And A Class Label A Learning 1 (196.88 KiB) Viewed 52 times

1. Given a classification problem and a dataset, where each record has several attributes and a class label, a learning algorithm can be applied to the data in order to determine a classification model. The model is then used to classify previously unseen data (data without a class label) to predict the class label. (a) Consider a classification model which is applied to a set of records, of which 100 records belong to class A (the positive class) and 900 records to class B. The model correctly predicts the class of 20 records in A and incorrectly predicts the class of 100 records in class B. Compute the confusion matrix. (5 marks) (b) Write down the definitions of accuracy and error rate. (2 marks) Compute the accuracy and error rate for the example in part (a). (2 marks) (c) Write down the definitions of precision, recall and Fl-measure. (3 marks) Compute the precision, recall and F1-measure for the example in part (a). (3 marks) a (d) Discuss the limitations of accuracy as a performance metric for evaluating a classification model under class imbalance. How can these limitations be overcome with a cost function? (5 marks)