SPARRING: Sistem Rekomendasi Peneliti Terintegrasi Google Scholar via SerpAPI dan Latent Dirichlet Allocation pada Konteks Perguruan Tinggi

Abstract

The researcher partner recommendation system plays a crucial role in fostering academic collaboration in universities, where a challenge for new users is finding suitable research partners. In addressing the limitations of Naïve Bayes classifiers, this article introduces an innovative approach in the form of a non-linear sigmoid activation function. We highlight the urgency of this solution, detail its implementation steps, and describe its substantial contribution to research partner recommendations. This article not only identifies existing obstacles but also proposes revolutionary solutions to enhance the effectiveness of consultation systems in academic environments. A gap in this research is the manual input method for data retrieval, creating weaknesses, susceptibility to human errors, and reduced efficiency in collecting journal data. We propose SPARRING, a researcher recommendation system connected to Google Scholar, in the context of higher education. This approach uses a dataset of faculty members from the Faculty of Information Technology and Business at a private university in Indonesia. The results from Google Scholar extraction, with topic keywords determined by Latent Dirichlet Allocation, are then classified using the Naïve Bayes algorithm. Additionally, we integrate web scraping tools, particularly SerpAPI, to access data from Google Scholar. Through the integration of SerpAPI, the proposed web-based system is capable of providing more accurate recommendations, especially for new users with limited collaboration experience. By incorporating SerpAPI, the proposed web-based system can offer more accurate recommendations, particularly for new users without extensive collaboration experience.

Keywords

• google scholar
• naive bayes
• latent dirichlet allocation
• recommendation system
• higher education