Identifying algorithm in program code based on structural features using CNN classification model

Abstract In software, an algorithm is a well-organized sequence of actions that provides the optimal way to complete task. Algorithmic thinking also essential break-down problem and conceptualize solutions in some steps. The proper selection pivotal improve computational performance software productivity as well programming learning. That is, determining suitable from given code widely relevant engineering education. However, both humans machines find it difficult identify algorithms without any meta-information. This study aims propose program classification model uses convolutional neural network (CNN) classify codes based on algorithm. First, are transformed into structural features (SFs). Second, SFs one-hot binary matrix using several procedures. Third, different structures hyperparameters CNN fine-tuned best for To do so, 61,614 real-world types collected online judge system used train, validate, evaluate model. Finally, experimental results show proposed can with high percentage accuracy. average precision, recall, F-measure scores 95.65%, 95.85%, 95.70%, respectively, indicating outperforms other baseline models.