Geri Dön

Predicting dementia: A machine learning approach

Başlık çevirisi mevcut değil.

PDF İndir

  1. Tez No: 771865
  2. Yazar: GÖZDE ORHAN
  3. Danışmanlar: Belirtilmemiş.
  4. Tez Türü: Yüksek Lisans
  5. Konular: Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Bilim ve Teknoloji, Computer Engineering and Computer Science and Control, Science and Technology
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 2019
  8. Dil: İngilizce
  9. Üniversite: Goldsmiths, University of London
  10. Enstitü: Yurtdışı Enstitü
  11. Ana Bilim Dalı: Belirtilmemiş.
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: 54

Özet

Dementia is a generalised term associated with the deterioration of an individual's cognitive functioning in a broad range from an occasional memory loss to a severe impairment inhibiting day-to-day activities. The main challenge regarding to this illness is the fact that there is no standardised diagnosis test which gives a rise to undiagnosed or misdiagnosed cases. Since there is no available cure yet, the diagnosis of dementia prior to the irreversible stage is crucial. This study aims to accomplish diagnosis prediction within a machine learning framework by exploring various number of feature selection algorithms and demonstrating different classifier capabilities. The datasets utilised in this study were the subsets of the adnimerge data, which was a collection of a longitudinal study. The subset datasets only included baseline variables as well as a manipulated diagnosis target column. This study is designed to exploit four different datasets which all have binary output variables but not three classes as in the original dataset. Regarding to a feature selection method: Recursive Feature Elimination, Boruta and ReliefF methods are implemented to all datasets in order to be able to compare performances of feature selection methods as well as comparing the classifiers. Four different classifiers are exploited during the training, tuning and testing phases while a 10-fold cross validation is being carried out: Random Forests, Support Vector Machines, eXtreme Gradient Boosting and Neural Networks. The datasets utilised are consisted of: - dfCN: 44 input variables, 1 output variable and 1851 observations - dfMCI: 44 input variables, 1 output variable and 1851 observations - dfD: 44 input variables, 1 output variable and 1851 observations - dfD2: 44 input variables, 1 output variable and 965 observations A total number of 48 learning algorithms are implemented. By analysing the results, xGBoost achieved to be the most robust algorithm regardless of datasets (93%-95% accuracy) and feature selection algorithms utilised. In addition, SVM established remarkable performances (~96% accuracy) on dfD2 but wasn't as successful in other datasets. Boruta feature elimination technique was the most common technique appeared in the highest-ranking models which is a useful result, reinforcing the other studies stating Boruta is a successful technique especially in a medical framework. By demonstrating the success achieved by machine learning algorithms, this study can light the way for a more extensive and exhaustive machine learning study in dementia prediction. Through exploring different classifiers or even improving the ones existing via hyperparameter tuning, a continuation of this project will provide a major benefit on an individual basis as well as a societal basis.

Özet (Çeviri)

Özet çevirisi mevcut değil.

Benzer Tezler

  1. Tez No

    856463

    Diagnosis of alzheimer's disease with deep learning: A hybrid 3D CNN and RNN approach

    Derin öğrenme ile alzheımer hastalığının tanısı: Hibrit 3B CNN ve RNN yaklaşımı

    IDRIS CHENTOUF

    Yüksek Lisans

    İngilizce

    İngilizce

    2024

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolBahçeşehir Üniversitesi

    Yapay Zeka Ana Bilim Dalı

    YRD. DOÇ. DR. ZAFER İŞCAN

  2. Tez No

    830502

    Application of classical and machine learning models on longitudinal data with binary response

    İkı̇lı̇ yanıtlı boylamsal verı̇lerde klası̇k ve makı̇ne öğrenme modellerı̇nı̇n uygulanması

    RÜMEYSA RANA ARSLAN

    Yüksek Lisans

    İngilizce

    İngilizce

    2023

    BiyoistatistikOrta Doğu Teknik Üniversitesi

    İstatistik Ana Bilim Dalı

    PROF. DR. ÖZLEM İLK DAĞ

  3. Tez No

    890137

    A smart wristband connected mobile application for the management and machine learning based assesment of dementia symptoms

    Demans semptom yönetimi için makine öğrenmesi tabanlı akıllı bileklik bağlantılı mobil uygulama

    BARIŞ CEYHAN

    Doktora

    İngilizce

    İngilizce

    2023

    Bilim ve TeknolojiMuğla Sıtkı Koçman Üniversitesi

    Biyoenformatik Ana Bilim Dalı

    DR. ÖĞR. ÜYESİ TUĞBA SÜZEK

    PROF. DR. VEDAT SEMAİ BEK

  4. Tez No

    827740

    Ayırıcı tanının güç olduğu bir durumda makine öğrenme yöntemleri ile demans ve hafif kognitif bozukluk ayırt ediciliğinin değerlendirilmesi

    Evaluation of dementia and mild cognitive disorders with machine learning methods in a situation where the differential diagnosis is difficult

    CELAL ŞENER

    Yüksek Lisans

    Türkçe

    Türkçe

    2023

    Bilim ve TeknolojiDokuz Eylül Üniversitesi

    Medikal İnformatik Ana Bilim Dalı

    PROF. DR. PEMBE KESKİNOĞLU

  5. Tez No

    612634

    Makine öğrenmesi yöntemleri ile demans tahmini

    Prediction of dementia by machine learning methods

    TUĞBA TUNA

    Yüksek Lisans

    Türkçe

    Türkçe

    2019

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrolİstanbul Teknik Üniversitesi

    Hesaplamalı Bilimler ve Mühendislik Ana Bilim Dalı

    PROF. DR. FETHİYE AYLİN SUNGUR

    PROF. DR. MUSTAFA ERSEL KAMAŞAK

Geri Dön