pardeepshokeen/Question-Independent-Automated-Code-Analysis-and-Grading-using-Bag-of-Words-and-Machine-Learning

Our first hand experience with the recruiting process of companies helped us realize that existing systems judge coding responses solely on the basis of test cases passed and often additional manual involvement is needed to determine the quality of code. We aim to build an open source code base to automatically grade coding responses. This work will help save the tedious effort on the part of Subject Matter Experts (SMEs) by proving useful for grading responses in MOOCs (Massive Open Online Courses) as well as coding rounds in the recruitment process. We have built our work on foundations introduced in extant research in the area. We have achieved better precision (89\%) in addition to better Pearson Correlation Coefficient (0.981) as compared to existing methods by using bag of words technique to calculate the distance vector. Various machine learning models were explored to achieve optimal results. These results and observations provide promise of successful application of this technique in other areas requiring automated grading.

CGPL-2.0