Design-Pattern-detection

This application identifies design pattern instances from a given dataset. Currently, the code is compatible with Python 2.7

Note: The files below require the matplotlib package for plotting and readable output

Data Source

• Java file Corpus

Dataset Generation

The dataset used for the classification is generated in a series of steps. First, make_ngram.py is used to construct a Word2Vec model using a set of files (verbose files) containing class names and ngrams associated with the class. make_glove.py performs a similar task, but instead constructs a GloVe model given the verbose files. The word embedding model is used to construct a dataset by running make_class_features.py followed by

detector.py

Outputs .verbose files in the output directory taking .java files from input directory

Usage:

python detector.py --input ./input --output ./output --tasks all

make_class_features.py

Outputs vector representation Java classes from the ngram model in dataset.csv along with labels merged from output-refined.csv

Usage:

python make_class_features.py [VERBOSE_ROOT]

Arguments:

• VERBOSE_ROOT: root of the directory containing the verbose files.

NOTE: This script uses the Word2Vec implementation in the Gensim package. And may need to install sudo apt-get install -y liblzma-dev to run it for linux machines.

Classification

Classification is performed in the python files listed below. The file calls the dataset.csv file, which is used to train a machine learning model and generate predictions.

Each script outputs

• confusion matrix csv and plot
• Per class Precision, Recall barplots
• report containing the precision, recall, and f-score values for the prediction.

classifier.py

Uses a various to classify the design patterns contained in the test files.

Usage:

python classifier.py [CLASSIFIERNAME]

NOTE: Assumes the dataset is called 'dataset.csv'

Arguments:

• CLASSIFIERNAME - Which Classification algorithm you want to use.

CLASSIFIERNAME supported are

• RF: RandomForest
• SVM: Support Vector Machine
• ADABOOST: Adaboost
• ADABOOST_LOGISTIC: Adaboost
• LOGISTIC: Logistic Regression
• GBTREE: GradientBoosted Tree
• RIDGE: Ridge classifier
• VOTER: Voting classifier composed of RF, SVM, ADABOOST

Miscallenous Scripts [currently not used]

print_ngram_features.py

Prints the vector representations of the ngrams in readable format.

Usage:
print_ngram_features.py [MODEL_JAVA_FILE] [MODEL_GLOVE_FILE] [OUTPUT_JAVA_FILE] [OUTPUT_GLOVE_FILE]

Arguments:

MODEL_JAVA_FILE: file name of the ngram vector model

MODEL_GLOVE_FILE: file name of the GloVe vector model

OUTPUT_JAVA_FILE: text file to write the vector representations of the ngrams

OUTPUT_GLOVE_FILE: text file to write the vector representations of GloVe