Pradan, Benedictus Jullian (2022) RANCANG BANGUN APLIKASI UNTUK MENDETEKSI PARASIT MALARIA PADA CITRA APUSAN DARAH TIPIS DENGAN MENERAPKAN TRANSFER LEARNING PADA CONVOLUTIONAL NEURAL NETWORK. Tugas Akhir/Skripsi thesis, Universitas Ma Chung.
|
Text
13 311810007_BENEDICTUS JULLIAN PRADANA.pdf Download (3MB) | Preview |
Abstract
Malaria is an infectious disease caused by Plasmodium parasite. In 2019
there are 229 million Malaria cases with 400.900 deaths. In 2020 the number of
cases increased to 241 million of cases with 627.000 deaths. Malaria diagnosis is
done by sampling the patient blood cell and smear it with giemsa stain then observe
the sample under microscope. This process is done manually by medical personnel,
if the diagnosis isn’t done correctly it can lead to misdiagnosis.
Deep learning approach can be used to help Malaria diagnosis by applying
classification on blood smear images. This research use Transfer Learning method
on Convolutional Neural Network architecture to fasten the process of model
development and to obtain model with high accuracy. The model that is used for
Transfer Learning is EfficientNetB0.
The model that has been developed is embedded on python-based web
application, containerize using docker container, and deploy it on Google App
Engine so it can be used by medical personnel to diagnose patient with Malaria.
The model training result in 0,9664 training accuracy, 0,0937 training loss, 0,9734
validation accuracy, and 0,0816 validation loss. The evaluation using test data result
in 96,8% accuracy and 0,968 F1-score.
Keywords: Convolutional Neural Network, Deep Learning, Docker Container,
Google App Engine, Malaria, Transfer Learning, Web Application
| Item Type: | Thesis (Tugas Akhir/Skripsi) |
|---|---|
| Subjects: | T Technology > T Technology (General) |
| Divisions: | Fakultas Teknologi dan Desain > S1 Teknik Informatika |
| Depositing User: | umclibs1 |
| Date Deposited: | 20 May 2026 03:48 |
| Last Modified: | 20 May 2026 03:48 |
| URI: | http://repository.machung.ac.id/id/eprint/1184 |
Actions (login required)
 |
View Item |