COMPARATIVE STUDY OF NAÏVE BAYES AND ARTIFICIAL NEURAL NETWORK TO DETECT PATIENTS’ DISEASE TYPES BY USING STRUCTURAL AND UNSTRUCTURAL DATA

THOHARI, IBROHIM IMAM (2021) COMPARATIVE STUDY OF NAÏVE BAYES AND ARTIFICIAL NEURAL NETWORK TO DETECT PATIENTS’ DISEASE TYPES BY USING STRUCTURAL AND UNSTRUCTURAL DATA. S1 thesis, Universitas Mercu Buana Jakarta.

Preview	Text (HAL COVER) 01 Cover.pdf Download (1MB) | Preview
	Text (BAB I) 02 Bab 1.pdf Restricted to Registered users only Download (392kB)
	Text (BAB II) 03 Bab 2.pdf Restricted to Registered users only Download (411kB)
	Text (BAB III) 04 Bab 3.pdf Restricted to Registered users only Download (482kB)
	Text (BAB IV) 05 Bab 4.pdf Restricted to Registered users only Download (371kB)
	Text (BAB V) 06 Bab 5.pdf Restricted to Registered users only Download (502kB)
	Text (BAB VI) 07 Bab 6.pdf Restricted to Registered users only Download (604kB)
	Text (DAFTAR PUSTAKA) 08 Daftar Pustaka.pdf Restricted to Registered users only Download (306kB)
	Text (LAMPIRAN) 09 Lampiran.pdf Restricted to Registered users only Download (425kB)

Abstract

The main concept of the hospital is the provision of health services to the community. To ensure that healthy services can be maximized, the hospital uses technology to record all activity data in the hospital. However, currently the data is only stored in the database and used as history without any further use. Many experiences show that by optimizing the data usage it will greatly assist doctors in making decisions to minimize medical errors. For example, examination data that among others of medical abstract, blood pressure, temperature, etc. can be used for the classification of the kind of disease. This paper presents the result study of the using a comparative study of naïve bayes and artificial neural network to examine the data in classifying the kind of disease based on structural and unstructural examination data. Natural language processing is used to represent unstructured text medical abstracts in a vector form using Word2Vec word embedding. That way, medical abstract and other examination data can be processed using the Naïve Bayes algorithm and the Artificial Neural Network. By using these two algorithms, the results of the classification of the kind of disease. The performed experiments show that ANN model gives the better performance with the best accuracy average of 91.46% compared to Naive Bayes which is 68.33%. In addition, the involvement of unstructured data from the dataset in the word2vec training process improves the performance ANN even though it is not significant with an accuracy average of 90.27% compared without the involvement of unstructured data which is 89.82%. Key words: Natural Language Processing, Word2Vec, Medical abstract, Naïve bayes, Artificial Neural Network. Konsep utama dari rumah sakit adalah penyediaan layanan kesehatan kepada masyarakat. Untuk memastikan pelayanan kesehatan bisa maksimal, rumah sakit memanfaatkan teknologi untuk merekam semua data kegiatan yang ada di rumah sakit. Namun, saat ini data tersebut hanya disimpan di database dan digunakan sebagai history tanpa digunakan lebih lanjut. Banyak pengalaman yang menunjukkan bahwa dengan mengoptimalkan penggunaan data akan sangat membantu dokter dalam mengambil keputusan untuk meminimalisir kesalahan medis. Misalnya data pemeriksaan yang antara lain anamnesis (abstrak medis), tekanan darah, suhu, dll dapat digunakan untuk klasifikasi jenis penyakit. Makalah ini memaparkan hasil studi penggunaan comparative study antara Algoritma Naïve Bayes, dan Artificial Neural Network untuk mengkaji data dalam pengklasifikasian jenis penyakit berdasarkan data pemeriksaan dokter yang terstruktur dan tidak terstruktur. Natural Language Processing digunakan untuk merepresentasikan unstructured text medical abstract ke dalam bentuk vector menggunakan Word2vec word embedding. Dengan begitu, medical abstract dan data pemeriksaan lainnya dapat diolah dengan menggunakan algoritma Naïve Bayes dan Artificial Neural Network. Dengan menggunakan kedua algoritma tersebut maka didapatkan hasil klasifikasi jenis penyakit. Eksperimen yang dilakukan menunjukkan bahwa model ANN memberikan kinerja yang lebih baik dengan rata-rata akurasi terbaik sebesar 91,46% dibandingkan dengan Naive Bayes yaitu 68,33%. Selain itu, keterlibatan unstructured text dari dataset dalam proses pelatihan word2vec dapat meningkatkan kinerja ANN meskipun tidak signifikan yaitu rata-rata akurasi 90.27% dibandingkan tanpa melibatkan unstructured text data yaitu 89.82%. Kata kunci: Natural Language Processing, Word2vec, Medical abstract, Naïve Bayes, Artificial Neural Network

Actions (login required)

View Item