PERANCANGAN SISTEM PEMANTAUAN PENGEMBANGAN ILMU DAN TEKNOLOGI BERDASARKAN INDIKATOR CAPAIAN PENELITIAN DENGAN TEXT MINING

• Main Author: Safira Shofa, Mediana
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/87597
• Institution: Institut Teknologi Bandung

The Directorate for the Application of Multidisciplinary Science and Technology at the Bandung Institute of Technology (DPITM ITB) plays a crucial role in coordinating the performance evaluation of 27 Centers and 7 Research Centers within ITB. This evaluation aims to ensure that each Center and Research Center not only functions optimally but also delivers maximum impact in achieving ITB's strategic goals. However, in practice, DPITM ITB faces several challenges that hinder the evaluation process. Through root cause identification using the 5 Whys diagram, it was concluded that DPITM ITB's difficulties stem from three main root causes: data known only to relevant parties, the lack of an integrated ITB information system, and the unclear definition of targets for each Center and Research Center. This study aims to address these root causes by developing a Business Intelligence (BI)-based solution focusing on three key aspects: the formulation of more comprehensive and relevant KPIs, the design of a user interface in the form of a dashboard, and the development of data mining models to process unstructured data for certain specific KPIs. The research methodology adopts the Business Intelligence Roadmap to manage the stages of BI solution development, combined with Eva Hariyanti's methodology emphasizing user-centric dashboard design and the CRISP-DM methodology for implementing data mining processes. The outcome of this study is a BI dashboard prototype, with each section presenting five main categories of standard KPIs: Publications, Research Projects, Community Services, Intellectual Property, and Human Resources. Each dashboard page is equipped with interactive visualizations, and for specific KPIs, data processing is supported by data mining models such as LaBSE, KeyBERT, MiniLM, and regex matching.