valid-method-ids-cnn

Example:

edu.lu.uni.data.preparing/*.java

First, preprocess data for deep learning.

Package: edu.lu.uni.data.preparing

1. append 0 to all vectors, make all vectors hold the same size.

JavaFile: ZeroAppender.java

Input: inputData/integer-vectors/

Output: outputData/WithoutNormalization/

2. Normalize the values in the data vectors of results of 1.

JavaFile: DataForNormalization.java

Input: outputData/WithoutNormalization/

Output: outputData/Normalization/

3. Standardize the values in the data vectors of results of 1.

JavaFile: DataForStandardization.java

Input: inputData/integer-vectors/

Output: outputData/Standardization/

Second, extract(encode) features of data vectors by deep learning (CNN).

Package: edu.lu.uni.deeplearning.extractor

###1. extract(encode) features of method body. Input: outputData/Normalization/method-body/, outputData/WithoutNormalization/method-body/, outputData/Standardization/method-body/

Output: outputData/CNN/

2. extract(encode) features of method name.

Input: outputData/Normalization/method-name/, outputData/WithoutNormalization/method-name/, outputData/Standardization/method-name/

Output: outputData/CNN/

The results of CNN will be used in clustering.