Diabetes Mellitus Prediction Based on K Nearest Neighbor Machine Learning Algorithm

Ali Vasfi Ağlarcı; Feridun Karakurt

Makale detayları

Diabetes Mellitus Prediction Based on K Nearest Neighbor Machine Learning Algorithm

Yazarlar:

Ali Vasfi Ağlarcı | Feridun Karakurt

Yayıncı:

Zonguldak Bulent Ecevit University

Ana Başlıklar:

Endokrinoloji

Dergi:

Türkiye Diyabet ve Obezite Dergisi Vol.8 Issue 3, pp.265-276

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001337405

pISSN:

2587-0335

Anahtar kelimeler:

Özet:

 Aim: The aim of our study is to predict diabetes mellitus, which is increasing worldwide and has become an important public health problem, with machine learning method. Material and Methods: In the study, data records containing diabetes mellitus health indicators with a sample size of 253,680 obtained from the kaggle database were used. K nearest neighbor method was used to predict the diabetes mellitus status of the patients with a machine learning approach. All operations were performed with the R program. Results: Approximately 15.8 of the individuals were diagnosed with preDM or diabetes mellitus, 42.9 had high blood pressure and 42.4 had high cholesterol. 44.3 were smokers and 5.6 were heavy alcohol consumers. The rate of those who have had heart disease crisis is 9.4, and the rate of those who reported having difficulty in walking is 16.8. The rate of those with no physical activity was 24.4. The mean BMI of those without diabetes mellitus was 27.746.26, while the mean BMI of those with diabetes mellitus was 31.947.36. In the application with the k nearest neighbor method, the best prediction of diabetes mellitus was obtained when the training and test data were separated as 90.0-10.0 and the k neighborhood value was 3 three. Using the relevant markers, people with diabetes mellitus disease were correctly predicted with 97.2 accuracy and 88.9 kappa success value. Conclusion: It is reported in the literature that machine learning methods have been widely used in many fields in recent years and have yielded successful results. In this study, it has been demonstrated that the prediction of diabetes mellitus with machine learning approach is realized with a high success rate. Since diabetes mellitus is known to progress silently and in increasing numbers, early diagnosis is of vital importance. Due to the advantages of K nearest neighbor method such as easy applicability and high classification performance, it is recommended to be used by healthcare providers for early diagnosis and treatment of diabetes mellitus.

 Aim: The aim of our study is to predict diabetes mellitus, which is increasing worldwide and has become an important public health problem, with machine learning method. Material and Methods: In the study, data records containing diabetes mellitus health indicators with a sample size of 253,680 obtained from the kaggle database were used. K nearest neighbor method was used to predict the diabetes mellitus status of the patients with a machine learning approach. All operations were performed with the R program. Results: Approximately 15.8 of the individuals were diagnosed with preDM or diabetes mellitus, 42.9 had high blood pressure and 42.4 had high cholesterol. 44.3 were smokers and 5.6 were heavy alcohol consumers. The rate of those who have had heart disease crisis is 9.4, and the rate of those who reported having difficulty in walking is 16.8. The rate of those with no physical activity was 24.4. The mean BMI of those without diabetes mellitus was 27.746.26, while the mean BMI of those with diabetes mellitus was 31.947.36. In the application with the k nearest neighbor method, the best prediction of diabetes mellitus was obtained when the training and test data were separated as 90.0-10.0 and the k neighborhood value was 3 three. Using the relevant markers, people with diabetes mellitus disease were correctly predicted with 97.2 accuracy and 88.9 kappa success value. Conclusion: It is reported in the literature that machine learning methods have been widely used in many fields in recent years and have yielded successful results. In this study, it has been demonstrated that the prediction of diabetes mellitus with machine learning approach is realized with a high success rate. Since diabetes mellitus is known to progress silently and in increasing numbers, early diagnosis is of vital importance. Due to the advantages of K nearest neighbor method such as easy applicability and high classification performance, it is recommended to be used by healthcare providers for early diagnosis and treatment of diabetes mellitus.

 Aim: The aim of our study is to predict diabetes mellitus, which is increasing worldwide and has become an important public health problem, with machine learning method. Material and Methods: In the study, data records containing diabetes mellitus health indicators with a sample size of 253,680 obtained from the kaggle database were used. K nearest neighbor method was used to predict the diabetes mellitus status of the patients with a machine learning approach. All operations were performed with the R program. Results: Approximately 15.8 of the individuals were diagnosed with preDM or diabetes mellitus, 42.9 had high blood pressure and 42.4 had high cholesterol. 44.3 were smokers and 5.6 were heavy alcohol consumers. The rate of those who have had heart disease crisis is 9.4, and the rate of those who reported having difficulty in walking is 16.8. The rate of those with no physical activity was 24.4. The mean BMI of those without diabetes mellitus was 27.746.26, while the mean BMI of those with diabetes mellitus was 31.947.36. In the application with the k nearest neighbor method, the best prediction of diabetes mellitus was obtained when the training and test data were separated as 90.0-10.0 and the k neighborhood value was 3 three. Using the relevant markers, people with diabetes mellitus disease were correctly predicted with 97.2 accuracy and 88.9 kappa success value. Conclusion: It is reported in the literature that machine learning methods have been widely used in many fields in recent years and have yielded successful results. In this study, it has been demonstrated that the prediction of diabetes mellitus with machine learning approach is realized with a high success rate. Since diabetes mellitus is known to progress silently and in increasing numbers, early diagnosis is of vital importance. Due to the advantages of K nearest neighbor method such as easy applicability and high classification performance, it is recommended to be used by healthcare providers for early diagnosis and treatment of diabetes mellitus.

 Amaç: Çalışmamızın amacı dünya çapında giderek artan ve önemli bir halk sağlığı sorunu hâline gelen diabetes mellitus hastalığının makine öğrenme yöntemi ile tahmin edilmesidir. Gereç ve Yöntemler: Çalışmada diabetes mellitus sağlık göstergelerini içeren ve kaggle veri tabanından elde edilen 253.680 örnek hacmine sahip veri kayıtları kullanılmıştır. K en yakın komşu yöntemi ile hastaların diabetes mellitus durumları makine öğrenme yaklaşımıyla tahmin edilmeye çalışılmıştır. Tüm işlemler R programı ile gerçekleştirilmiştir. Bulgular: Kişilerin yaklaşık 15,8i preDM ya da diabetes mellitus tanılıdır, 42,9unde yüksek tansiyon, 42,4ünde yüksek kolesterol bulunmaktadır. Sigara içenlerin oranı 44,3, ağır alkol tüketenlerin oranı ise 5,6dır. Kalp hastalığıkrizi geçirenleri oranı ise 9,4, yürüyüşte zorluk çektiğini bildirenlerin oranı ise 16,8dir. Fiziksel aktivitesi bulunmayanların oranı 24,4tür. Diabetes mellitus tanısı olmayanların BMI ortalaması 27,746,26 iken diyabet hastası olanların BMI ortalaması 31,947,36 olarak bulunmuştur. K en yakın komşu yöntemi ile yapılan uygulamada diabetes mellitus tahmini en iyi eğitim ve test verisinin 90,0-10,0 olarak ayrıldığı ve K komşuluk değerinin 3 üç alındığı durumda elde edilmiştir. İlgili belirteçler kullanılarak 97,2 doğruluk ve 88,9 kappa başarı değeri ile diabetes mellitus hastalığına sahip kişiler doğru tahmin edilebilmiştir. Sonuç: Makine öğrenme yöntemlerinin son yıllarda birçok alanda kullanımının yaygınlaştığı ve başarılı sonuçlar verdiği literatürde bildirilmektedir. Bu araştırmada da makine öğrenme yaklaşımıyla diabetes mellitus tahmininin yüksek başarı oranı ile gerçekleştirildiği uygulamalı olarak gösterilmiştir. Diabetes mellitus hastalığının sessiz ve artan sayıda ilerlediği bilindiğinden erken tanı hayati öneme sahiptir. K en yakın komşu yönteminin kolay uygulanabilirliği ve yüksek sınıflama performansı gibi avantajlarından dolayı diabetes mellitus hastalığının erken tanı ve tedavisi için sağlık hizmeti sağlayıcıları tarafından kullanılması önerilmektedir.

Makale detayları

Diabetes Mellitus Prediction Based on K Nearest Neighbor Machine Learning Algorithm

Yazarlar:

Ali Vasfi Ağlarcı | Feridun Karakurt

Yayıncı:

Zonguldak Bulent Ecevit University

Ana Başlıklar:

Endokrinoloji

Dergi:

Türkiye Diyabet ve Obezite Dergisi Vol.8 Issue 3, pp.265-276

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001337405

pISSN:

2587-0335

Diabetes mellitus

Health and disease

Machine learning

Prediction

Healthcare applications

Intelligent systemDiabetes mellitus

Sağlık ve hastalık

Makine öğrenmesi

Tahmin

Sağlık uygulamaları

Akıllı sistem

Sayı

Makale detayları

Yazarlar: Ali Vasfi Ağlarcı | Feridun Karakurt

Yayıncı: Zonguldak Bulent Ecevit University

Ana Başlıklar: Endokrinoloji

Dergi: Türkiye Diyabet ve Obezite Dergisi Vol.8 Issue 3, pp.265-276

Belge Tipi: Makale

Yayın Tarihi: 12/2024

Belge Dili: Türkçe

Document ID (Org No.): 2001337405

pISSN: 2587-0335

Sayı

Yazarlar:

Ali Vasfi Ağlarcı | Feridun Karakurt

Yayıncı:

Zonguldak Bulent Ecevit University

Ana Başlıklar:

Endokrinoloji

Dergi:

Türkiye Diyabet ve Obezite Dergisi Vol.8 Issue 3, pp.265-276

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001337405

pISSN:

2587-0335