1. We use the nave Bayes classifier on a dataset of 10000 instances, with 500 positive...

 

Transcribed Image Text:

1. We use the nave Bayes classifier on a dataset of 10000 instances, with 500 positive and 9500 negative instances. For a certain test instance A, nave Bayes predicts positive class with probability 30%, and negative class with probability 70%. We resample the dataset and obtain 2000 positive and 2000 negative instances. Sampling is with replacement: that is, the sample may contain multiple copies of some of the original instances. We use nave Bayes again on the resampled dataset. a) For the same test instance A, what will be the approximate probabilities of predicting positive and negative, respectively? b) Show that for Nave Bayes, resampling has (approximately) the same effect as chang- ing the classification threshold. 1. We use the nave Bayes classifier on a dataset of 10000 instances, with 500 positive and 9500 negative instances. For a certain test instance A, nave Bayes predicts positive class with probability 30%, and negative class with probability 70%. We resample the dataset and obtain 2000 positive and 2000 negative instances. Sampling is with replacement: that is, the sample may contain multiple copies of some of the original instances. We use nave Bayes again on the resampled dataset. a) For the same test instance A, what will be the approximate probabilities of predicting positive and negative, respectively? b) Show that for Nave Bayes, resampling has (approximately) the same effect as chang- ing the classification threshold.