Bad results that I get.

[DigOrDog]

OB: I train the model using your method that you describe in the paper(BEE: A Tool for Structuring and Analyzing Bug Reports),but the result is bad, for example, the recall and precision of the eb-model is about 70% and 55% respectively. EB: I want to get the results described in your paper. S2R: First of all, I use the dict.txt to encode the sentences.Then I divide the eb-sentences into 10 folds and separate the sentences according to positive and negative examples for each fold.Futhermore,I use the upsampling.java to generate extral positive sentences.Finally I merge the positive sentences, smote_positive sentences and negative sentences. The following is my method of training （1）I use svm_learn.exe to train the model on Windows system(E:/SVM/svm_light_windows64/svm_learn.exe -z c -c 0.2 E:/SVM/svm_light_windows64/sentence/dataEB/train/_all_train_0.dat E:/SVM/svm_light_windows64/sentence/EB/02/00_model) （2）I use svm_classify.exe to evaluate the model on Windows system(E:/SVM/svm_light_windows64/svm_classify.exe E:/SVM/svm_light_windows64/sentence/dataEB/_valid_0.dat E:/SVM/svm_light_windows64/sentence/EB/02/00_model E:/SVM/svm_light_windows64/sentence/EB/02/00_valid_result )