- Hands-On Ensemble Learning with R
- Prabhanjan Narayanachar Tattar
- 147字
- 2021-07-23 19:10:52
Na?ve Bayes classifier
The na?ve Bayes classifier is a simplistic implementation based on the Bayes formula. It is based on simple empirical and conditional probabilities, as evidenced in the actual data. Beyond the simplest assumption of observation independence, we don't have any restrictions in using this model.
Na?ve Bayes for hypothyroid classification
A na?ve Bayes classifier is fit using the naiveBayes function from the e1071 R package. The prediction and accuracy assessment is carried out using two functions, predict and sum:
> NB_fit <- naiveBayes(HT2_Formula,data=HT2_Train) > NB_predict <- predict(NB_fit,newdata=HT2_TestX) Warning message: In data.matrix(newdata) : NAs introduced by coercion > NB_Accuracy <- sum(NB_predict==HT2_TestY)/nte > NB_Accuracy [1] 0.9732704403
The accuracy of the na?ve Bayes classifier is 97.33%, which is the same as the logistic regression model and less than the one provided by the neural network. We remark here that it is only a coincidence that the accuracy of this method and logistic regression is the same.