Personalized Recommendation System for College Innovation and Entrepreneurship Curriculum Resources in Universities using Non-Negative Factor Normalization Based Collaborative Filtering

In the current scenario, with the vast number of online resources, students’ facing difficulty in locating the courses where they are interested and causing a lack of interest. The existing collaborative filtering based recommendation systems have scalability and sparsity issues. In this paper, Non-negative factor normalization based Collaborative Filtering (NMF-CF) is proposed to improve the scalability and sparsity. At first, the input is taken from the MOOC dataset and then behaviors of students are extracted with the help of nodes. After that, these nodes are further processed to create a knowledge map, which is a behavior map in Amazon Neptune database. Next, the nodes from the knowledge map are fetched to create a behavior path in order to understand in which innovation and entrepreneurial curriculum courses the students are interested. Here, path vectorization is introduced to convert previously used text data into digital data with Kera’s Tokenizer. Then, the path similarity is calculated with the help of Euclidean distance to give efficient recommendation inputs to the proposed NMF-CF recommendation algorithm. Finally, the proposed NMF-CF recommends the college innovation and entrepreneurship curriculum resources based on student interest. From the results, the proposed NMF-CF recommendation system provided better results than existing Collaborative Filtering Recommendation (CFR) in terms of accuracy, precision and recall as 75.68 %, 69.12% and 79.32% respectively.