Klasifikasi Laporan Pengaduan Warga Jakarta Melalui Aplikasi Jaki Menggunakan Algoritma Machine Learning

Hanin, Nafi’ah (2021) Klasifikasi Laporan Pengaduan Warga Jakarta Melalui Aplikasi Jaki Menggunakan Algoritma Machine Learning. Technical Report. Pustakawan, Perpustakaan Institut Teknologi Telkom Purwokerto. (Unpublished)

	Text Cover.pdf Download (1MB)
	Text Abstract.pdf Download (1MB)
	Text Abstrak.pdf Download (1MB)
	Text BAB I.pdf Download (1MB)
	Text BAB II.pdf Download (1MB)
	Text BAB III.pdf Download (1MB)
	Text BAB IV.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text BAB V.pdf Download (1MB)
	Text Daftar Pustaka.pdf Download (1MB)
	Text Lampiran.pdf Restricted to Registered users only Download (1MB) | Request a copy

Abstract

JAKI is a super app for the DKI Jakarta government that offers a variety of services and can be accessed via smartphone. Residents of DKI Jakarta use one of JAKI's mainstay features, JakLapor, to report various complaints about the community or the environment in their area. The solution for updating the system by adding a Machine Learning model as an execution program in categorizing reports must be handled by officers immediately to make the process of handling the report more efficient and on target. The advancement of technology and information resulted in the creation of Artificial Intelligence (AI), which is used to assist humans in their work, such as Machine Learning. To process text data in citizen complaints reports on the JAKI application feature, JakLapor, a Supervised Learning method learning with the concept of Natural Language Processing and variations of Long Short Term Memory (LSTM) is used. The manual labelling of some of the data used as samples is a step that allows the model to learn from the existing dataset and then predict the next data. Sentiment Analysis is used as an analytical method for complaints classified as "Urgent" or "Not Urgent." Based on the model evaluation, the results are 99.9% accuracy and 0.02% loss in the model training process. Meanwhile, the model validation process yields an accuracy of 90.8% and a loss of 3.95%. The model is saved in a tflite extension file after it has been tested and proven to be accurate so that it can be implemented in the Android system. Keywords: JakLapor, Supervised Learning, Natural Language Processing, LSTM, dan Sentiment Analysis

Actions (login required)

View Item