ANALISIS PERBANDINGAN ARSITEKTUR REST DAN GRAPHQL UNTUK APLIKASI PENGENALAN EMOSI PADA PEMBELAJARAN DARING SINKRONIS

• Main Author: Derry Dwi Aditya Hendarto, - (Author)
• Format: Book
• Published: 2023-01-12.
• Subjects: Thesis; NonPeerReviewed
• Online Access: Link Metadata
• Notes: http://repository.upi.edu/87468/1/S_RPL_1907630_Title.pdf; http://repository.upi.edu/87468/2/S_RPL_1907630_Chapter1.pdf; http://repository.upi.edu/87468/3/S_RPL_1907630_Chapter2.pdf; http://repository.upi.edu/87468/4/S_RPL_1907630_Chapter3.pdf; http://repository.upi.edu/87468/5/S_RPL_1907630_Chapter4.pdf; http://repository.upi.edu/87468/6/S_RPL_1907630_Chapter5.pdf; http://repository.upi.edu/87468/7/S_RPL_1907630_Appendix.pdf
• Keywords: L Education (General); QA76 Computer software; T Technology (General)

Emosi memiliki peran yang penting dalam proses pembelajaran, karena dapat mencerminkan perasaan dan keterlibatan pelajar selama mengikuti pembelajaran. Saat ini telah tersedia berbagai macam Application Programming Interface (API) pengenalan emosi, sehingga membuka peluang untuk mengimplementasikan aplikasi pengenalan emosi dan aplikasi untuk visualisasinya secara real-time. Data real-time umumnya diperoleh melalui suatu API, seperti Representational State Transfer (REST) dan Graph Query Language (GraphQL). Beberapa penelitian sebelumnya belum ditemukan penggunaan GraphQL serta belum ditemukan kajian mendalam terkait arsitektur yang paling optimal untuk digunakan. Penelitian ini ditujukan untuk mengembangkan aplikasi pengenalan emosi dengan arsitektur yang paling optimal diantara REST dan GraphQL, lalu menganalisis performa aplikasi back-end sebagai data primer dan front-end sebagai data sekunder. Performa aplikasi back-end diukur menggunakan metrik QoS meliputi response time, throughput, memory utilization, dan CPU load. Analisis data yang digunakan adalah efisiensi. Uji-t dan visualisasi boxplot digunakan untuk membuktikan dan memvisualisasikan adanya perbedaan signifikan antara kedua arsitektur. Hasil pengujian pada endpoint recognition grup response time REST lebih efisien sebesar 4,42%, throughput REST lebih efisien sebesar 4,82%, memory utilization REST lebih efisien sebesar 10,83%, serta CPU load REST lebih efisien sebesar 20,51%. Adapun pada endpoint recognition individual, response time REST lebih efisien sebesar 5,92%, throughput REST lebih efisien sebesar 9,31%, memory utilization REST lebih efisien sebesar 9,89%, serta CPU load GraphQL lebih efisien sebesar 22,01%. Secara umum REST adalah pilihan arsitektur yang tepat untuk aplikasi dengan kebutuhan performa serta stabilitas tinggi, sedangkan GraphQL cocok diimplementasikan pada aplikasi dengan kebutuhan field data yang kerap berubah. ----- Emotions have an important role in the learning process, because they can reflect the feelings and involvement of learners during learning. Currently, there are various kinds of emotion recognition Application Programming Interfaces (APIs) available, thus opening up opportunities to implement emotion recognition applications and applications for their visualization in real-time. Real-time data is generally obtained through an API, such as Representational State Transfer (REST) and Graph Query Language (GraphQL). Several previous studies have not found the use of GraphQL and no in-depth studies have been found regarding the most optimal architecture to use. This research is aimed at developing emotion recognition applications with the most optimal architecture between REST and GraphQL, and then analyzing the performance of back-end applications as primary data and front-end as secondary data. Back-end application performance is measured using QoS metrics including response time, throughput, memory utilization, and CPU load. The data analysis used is efficiency. T-test and boxplot visualization are used to prove and visualize the existence of significant differences between the two architectures. The test results on group recognition endpoint showed that REST response time is more efficient by 4,42%, REST throughput is more efficient by 4,82%, memory utilization REST is more efficient by 10,83%, and CPU load REST is more efficient by 20,51%. As for individual recognition endpoint, REST response time was more efficient by 5,92%, REST throughput was more efficient by 9,31%, REST memory utilization was more efficient by 9,89%, and GraphQL CPU load was more efficient by 22,01%. In general, REST is the right choice of architecture for applications with high performance and stability needs, while GraphQL is suitable for applications with frequently changing data field needs.