A Systematic Clone Detection Performance Analysis on BigCloneBench Dataset

BigCloneBench dataset is a large repository that is widely used to study code clones. The usability of the dataset extends to the making and deployment of code clone detectors to code clone detection validation too. The dataset comes in many forms and sizes. We worked with a portion of the dataset containing 50000 entries. We first generated tokens to extract features from our entries. Simple models like neural network and decision tree performed extremely well but Naive Bayesian model and Random Forest classifier failed to perform in the testing phase, leading to poor model efficacy. The results leave us to further study the dataset with different setups.