Pembuatan Aplikasi Complaint Management System pada Universitas Kristen Petra dengan menggunakan Metode Support Vector Machine Multiclass One vs Rest

Isak Imanuel Leong (Program Studi Informatika)
Rolly Intan (Program Studi Informatika)
Leo Willyanto Santoso (Program Studi Informatika)

Jurnal Infra
Vol 8, No 2 (2020)

One of the facilities at Petra Christian University for the accommodation of aspirations from the academic community is the suggestion box. However, the use of suggestion boxes is considered less effective and objective because it has a reluctance to fill out and lacks strong evidence through photo documentation.To answer the above problem, an Android-based complaint management system was created that supports the aspiration process that is more effective and objective through the use of mobile devices that are more widely used by the academic community. Application supported by the model. Supported by Vector Support Engine (SVM) with the approach of multiclass One vs. Rest to classify the bureau or related units to overcome errors in determining the recipient of the aspirational recipient.The results of the research conducted show that preprocessing parameter such as normalization, stemming and stopwords removal affect the accuracy of the model. The best kernel type in SVM for aspiration text classification is linear with value of C = 1 which results in an accuracy of 95,441%. In addition, the results of a survey to administrator who manage the management of aspiration shows that the application created already answer the needs and problems, in this case supporting the objectivity and effectiveness of aspiration data delivery.