Makale Koleksiyonu

Bu koleksiyon için kalıcı URI

https://hdl.handle.net/20.500.12511/4420

Listeleniyor 1 - 20 / 28

Analysis of city demographics in Turkiye using data mining techniques
(2023) Doğuç, Özge; Şahinbaş, Kevser; Silahtaroğlu, Gökhan
In this study, a data mining model has been developed and used to analyze how cities and regions in Turkey can be grouped, aiming to find similarities and differences between them. For this purpose, data is obtained from Turkish Statistical Institution (TUIK) and fuzzy c-means clustering algorithm was used to find categorizations. The data set contains 142 variables from 8 categories such as education, health, happiness, and development levels. The results showed that in all categories, the biggest 3 cities in Turkey, İstanbul, Ankara, and İzmir are different from the rest of the country. Also, cities located in the western and eastern regions of Turkey are mostly grouped among themselves, showing the clear distinction between those two regions. Finally, small cities with big neighbors are grouped with other big cities, showing the direct impact of big cities on their neighbors. This also implies that small cities with no big neighbors are often isolated, as their residents don’t have access to the services provided in the big cities.
Identifying the most critical side effects of antidepressant drugs: a new model proposal with quantum spherical fuzzy M-SWARA and DEMATEL techniques
(2024) Silahtaroğlu, Gökhan; Dinçer, Hasan; Yüksel, Serhat; Keskin, Abdurrahman; Yılmaztürk, Nevin; Kılıç, Alperen
Identifying and managing the most critical side effects encourages patients to take medications regularly and adhere to the course of treatment. Therefore, priority should be given to the more important ones, among these side effects. However, the number of studies that make a priority examination is limited. There is a need for a new study that determines which of these effects are more priority to increase the quality of the treatment. Accordingly, this study aims to define the most important side effects of antidepressant drugs with a novel model. Quantum Spherical fuzzy M-SWARA technique is considered to compute the importance weights of the items. The main contribution of this study is that the most critical side effects can be understood for antidepressant drugs by establishing a novel decision-making model. The findings demonstrate that psychological side effects are defined as the most critical side effects of antidepressant drugs. Furthermore, physical side effects also play a key role in this condition. Side effects in antidepressant treatment have a great impact on the effectiveness of treatment and patient compliance.
On the networks of large embeddings
(2024) Aslan, Tuğba; Khaled, Mohamed; Székely, Gergely
We define a special network that exhibits the large embeddings in any class of similar algebras. With the aid of this network, we introduce a notion of distance that conceivably counts the minimum number of dissimilarities, in a sense, between two given algebras in the class in hand; with the possibility that this distance may take the value ?. We display a number of inspirational examples from different areas of algebra, e.g., group theory and monounary algebras, to show that this research direction can be quite remarkable.
Discovering the chemical factors behind regional royal jelly differences via machine learning
(Bursa Uludag University, 2023) Özkök, Aslı; Keskin, Merve; Tanuğur Samancı, Aslı Elif; Yorulmaz Önder, Elif; Silahtaroğlu, Gökhan
This study aims to discover the characteristic chemical factors for determining the region of royal jelly using machine learning. 84 samples from 13 different regions of Turkey were used for the study, and the chemical parameters of moisture, pH, acidity, and 10-hydroxy-2-decanoic acid (10-HDA) were investigated. ANOVA test was conducted to determine whether there are differences between royal jelly from 13 locations concerning the four chemical values. In addition to the statistical tests, a machine learning model was used to find out what makes royal jelly different from each other. The descriptive statistics of the chemical analysis results of royal jelly showed the following values: moisture 63.05%±2.99, pH 3.67±0.08, acidity 45.32±3.55, and 10-HDA 2.40±0.24. Surprisingly, the machine learning model suggests that 10-HDA may be the most prominent parameter for determining the region of royal jelly. This information will help us identify royal jelly’s authenticity more easily.
Early prediction of the severe course, survival, and ICU requirements in acute pancreatitis by artificial intelligence
(Elsevier B.V., 2023) İnce, Ali Tüzün; Silahtaroğlu, Gökhan; Seven, Gülseren; Koçhan, Koray; Yıldız, Kemal; Şentürk, Hakan
Objective: To evaluate the success of artificial intelligence for early prediction of severe course, survival, and intensive care unit(ICU) requirement in patients with acute pancreatitis(AP).Methods: Retrospectively, 1334 patients were included the study. Severity is determined according to the Revised Atlanta Classification(RAC). The success of machine learning(ML) method was evaluated by 13 simple demographic, clinical, etiologic, and laboratory features obtained on ER admission. Additionally, it was evaluated whether Balthazar-computerized tomography severity index(CTSI) at 48-h contributed to success. The dataset was split into two parts, 90% for ML(of which 70% for learning and 30% for testing) and 10% for validation and 5-fold stratified sampling has been utilized. Variable Importance was used in the selection of features during training phase of machine. The Gradient Boost Algorithm trained the machine by KNIME analytics platform. SMOTE has been applied to increase the minority classes for training. The combined effects of the measured features were examined by multivariate logistic regression analysis and reciever operating curve curves of the prediction and confidence of the target variables were obtained.Results: Accuracy values for the early estimation of Atlanta severity score, ICU requirement, and survival were found as 88.20%, 98.25%, and 92.77% respectively. When Balthazar-CTSI score is used, results were found as 91.02%, 92.25%, and 98% respectively.Conclusions: The ML method we used successfully predicted the severe course, ICU requirement and survival, with promising accuracy values of over 88%. If 48-h Balthazar-CTSI is included in the calculation, the severity score and survival rates increase even more.
Detecting Turkish fake news via text mining to protect brand integrity
(Gazi University, 2022) Doğuç, Özge
Fake news has been in our lives as part of the media for years. With the recent spread of digital news platforms, it affects not only traditional media but also online media as well. Therefore, while companies seek to increase their own brand awareness, they should also protect their brands against fake news spread on social networks and traditional media. This study discusses a solution that accurately classifies the Turkish news published online as real and fake. For this purpose, a machine learning model is trained with tagged news. Initially, the headlines were analyzed within the scope of this study that are collected from Turkish online sources. As a next step, in addition to the headlines of these news, news contexts are also used in the analysis. Analysis are done with unigrams and bigrams. The results show 95% success for the headlines and 80% for the texts for correctly classifying the fake Turkish news articles. This is the first study in the literature that introduces an ML model that can accurately identify fake news in Turkish language.
An intelligent system for document-based banking processes
(Kahramanmaras Sutcu Imam University, 2022) Doğuç, Özge
Business process automation has been helping companies by eliminating mundane and repetitive tasks. Automation tools have been used in many sectors, providing high full-time employee (FTE) savings and low error rates to the companies. Banks have been utilizing the automation tools for core banking and branch operations. In addition, banks receive hundreds of notice documents from courts, municipalities and third parties; which usually contain private and time-sensitive information about their customers. Automated document processing requires special solutions such as optical character recognition (OCR), natural language processing (NLP) and taxonomy; and modern process automation tools can utilize these solutions to provide end-to-end automation. This paper discusses how the notice documents such as account closure, payment blockage in Turkey can be automated. It also shows how automation can efficiently and effectively process the documents by providing experimental results for each document type.
Intelligent early warning system for epidural acute hematomas
(International Balkan University, 2020) Doğuç, Özge
Epidural hematoma (EAH) is the accumulation of blood in the space between the outer membrane of the brain (dura mater) and the bone. Acute subdural and epidural hematoma appears on CT scan as a hyper-dense collection often located in brain convexity. Such bleeding can become fatal by increasing intracranial pressure and creating a mass effect. Therefore, it is very important to recognize these bleedings promptly in an emergency trauma setting. Thus, early diagnosis is essential to reduce mortality and morbidityratesin these cases. There has been a growing interest in artificial intelligence (AI) and machine learning (ML) algorithms for diagnostics in medical fields. In this study, a supervised learning method was used in which the decision tree ML algorithm is trained with the patients'statuses(EAH or Normal). This study proposes an early warning system (EWS) that scans all cranial CTs obtained at the trauma center. The EWS in this study, trained with CT scans from about 100 patients, can predict EAH with 100% accuracy usingimage recognition and supervised learning algorithms. Each MR section obtained for each patient is individually analyzedbyimage processing and EAH detection is made. For this, the decision tree method, which is a supervised learning algorithm, was trained and used to detect EAH in MR sections. The algorithm has been developed in such a way that it will immediately alert the emergency physician and consultant neurosurgeon by e-mail when it detects EAH in more than 10 sections in any patient.
Diagnosis of Covid-19 via patient breath data using artificial intelligence
(Ital Publication, 2023) Doğuç, Özge; Silahtaroğlu, Gökhan; Canbolat, Zehra Nur; Hambarde, Kailash; Yiğitbaşı, Ahmet Alperen; Gökay, Hasan; Yılmaz, Mesut
Using machine learning algorithms for the rapid diagnosis and detection of the COVID-19 pandemic and isolating the patients from crowded environments are very important to controlling the epidemic. This study aims to develop a point-of-care testing (POCT) system that can detect COVID-19 by detecting volatile organic compounds (VOCs) in a patient's exhaled breath using the Gradient Boosted Trees Learner Algorithm. 294 breath samples were collected from 142 patients at Istanbul Medipol Mega Hospital between December 2020 and March 2021. 84 cases out of 142 resulted in negatives, and 58 cases resulted in positives. All these breath samples have been converted into numeric values through five air sensors. 10% of the data have been used for the validation of the model, while 75% of the test data have been used for training an AI model to predict the coronavirus presence. 25% have been used for testing. The SMOTE oversampling method was used to increase the training set size and reduce the imbalance of negative and positive classes in training and test data. Different machine learning algorithms have also been tried to develop the e-nose model. The test results have suggested that the Gradient Boosting algorithm created the best model. The Gradient Boosting model provides 95% recall when predicting COVID-19 positive patients and 96% accuracy when predicting COVID-19 negative patients.
Performance comparison of K-means and DBSCAN methods for airline customer segmentation
(Uğur Şen, 2022) Şahinbaş, Kevser
Organizations are now fully embracing ideas such as customer success, customer loyalty, customer experience management and customer satisfaction. The application of these concepts must be based on three pillars of technology, process and people, to ensure that the organization ultimately has satisfied, loyal and successful customers. In today's competitive environment, as in all sectors, gaining great services in the aviation industry can provide a competitive advantage. With this study, it is aimed to help aviation companies to know how their services should meet the needs of customers and to obtain passenger satisfaction. Customer segmentation is widely used, which groups objects according to the similarity difference on each object and provides a high level of homogeneity in the same cluster or a high level of heterogeneity between each group. The aim of this study is to examine airline passenger satisfaction by using data mining methods including K-Means and DBSCAN clustering algorithms to reveal the service quality importance for customer satisfaction. K-Means algorithm achieved better results than DBSCAN algorithm with a Silhouette value of 0.1450671.
Predicting order cancellations for e-commerce domain: A proposed model based on retailing experience
(Mustafa Süleyman ÖZCAN, 2022) Şahinbaş, Kevser
E-Commerce technologies enable contact between businesses and their suppliers for the aim of exchanging information such as purchase orders, invoices, and payments thank to the rapid development in information technologies. E-Commerce has become a particularly important concept and has revolutionized the retail space. Understanding customer behavior patterns is key to gaining competitive advantage and achieving business goals. Predicting the probability of order cancellations has become a very urgent need as it causes loss of revenue for the retailer. When dealing with day-to-day operations such as order processing, tracking and order cancellations, finding enough time to grow the business is difficult. Cancellations are an important aspect of retail industry revenue management. In fact, little is known about the factors that cause customers to cancel or how to avoid them. The aim of this study is to propose a model that predicts the tendency to cancel an order and the parameters that affect the cancellation of the order. This solution can identify key factors that cause orders to be canceled by analyzing historical transaction data. A custom modeling application has been created that helps automate the process of tracking order cancellations in real time and predict the probability of an order being cancelled. For this purpose, machine learning techniques (ML) such as Artificial Neural Network, Support Vector Machine, Linear and Logistic Regression, XGBoost, Random Forest are applied to provide a tool for predicting order cancellations. The Random Forest algorithm achieves the best performance with 86% accuracy and 88% F1-Score compared to the other algorithm. This work will help firms manage their inventories well and strengthen their actions regarding customer behavior.
Doğal gaz tüketiminin modellenmesi: Türkiye için MARS yöntemiyle bir analiz
(Gaziantep Üniversitesi Sosyal Bilimler Enstitüsü, 2022) Aydın, Rıdvan; Yüksel, Serhat; Silahtaroğlu, Gökhan; Dinçer, Hasan
Bu çalışmada Türkiye’deki doğal gaz talebinin tahmin edilmesine yönelik model ortaya konması amaçlanmaktadır. Doğal gaz tüketimi bağımlı değişken olarak ele alınmış, buna bağlı olarak makroekonomik veriler, iklim koşulları, enerji ve fiyat verileri ile toplumsal ve kültürel veriler bağımsız değişken olarak kullanılmaktadır. Aylık verilerin değerlendirildiği bu çalışmada değişkenlere ait 2015 yılı ocak ayı ile 2021 yılı haziran ayı arasındaki 78 gözlem kapsama dâhil edilmiştir. Madelin analiz sürecinde MARS yönteminden faydalanılmıştır. Modelde 3 temel fonksiyon ile sıcaklık oranı ve yenilenebilir enerji oranı olarak 2 değişken yer almaktadır. Sıcaklık azalışları makul seviyelerde olduğu sürece yenilenebilir enerjideki artış oranının doğal gaz tüketiminin düşmesinde çok önemli etkiye sahip olduğu görülmektedir. %99 oranında ithalata bağımlı olan doğal gazın makroekonomik dengeler içerisindeki payını azaltmak için, sıcaklık değişikliklerimi kontrol edemediğimiz bir durum olduğundan, yenilenebilir enerji yatırımlarına hız kesmeden devam edilmesi gerektiği analiz sonucunda ortaya çıkmaktadır.
Data mining applications in banking sector while preserving customer privacy
(Ital Publication, 2022) Doğuç, Özge
In real-life data mining applications, organizations cooperate by using each other’s data on the same data mining task for more accurate results, although they may have different security and privacy concerns. Privacy-preserving data mining (PPDM) practices involve rules and techniques that allow parties to collaborate on data mining applications while keeping their data private. The objective of this paper is to present a number of PPDM protocols and show how PPDM can be used in data mining applications in the banking sector. For this purpose, the paper discusses homomorphic cryptosystems and secure multiparty computing. Supported by experimental analysis, the paper demonstrates that data mining tasks such as clustering and Bayesian networks (association rules) that are commonly used in the banking sector can be efficiently and securely performed. This is the first study that combines PPDM protocols with applications for banking data mining.
Sürdürülebilir ekonomik kalkınma bağlamında havayolu sektörüne yönelik etkin yatırım stratejilerinin belirlenmesi
(Gaziantep Üniversitesi, 2022) Mızrak, Filiz; Yüksel, Serhat; Silahtaroğlu, Gökhan; Dinçer, Hasan
Bu çalışmanın amacı sürdürülebilir ekonomik kalkınma bağlamında havayolu sektörüne yönelik müşteri memnuniyeti bazlı yatırım stratejilerinin belirlenmesidir. Bu amaca yönelik olarak, havacılık sektöründe müşterilerin taleplerini etkileyen faktörlerin tespit edilmesi amaçlanmıştır. Bu bağlamda, 2019 yılında en iyi 75 havayolu şirketi arasına giren şirketler inceleme kapsamına alınmıştır. Bahsi geçen şirketlere yönelik 2011-2020 yılları arasında Skytrax web sitesinden yapılan yorumlar analiz edilmiştir. İlgili müşteri yorumları metin madenciliği yöntemiyle KNIME platformu üzerinde analiz edilmiştir. Bu kapsamda, en fazla geçen tek, ikili ve üçlü kelime grupları belirlenmiştir. Elde edilen analiz sonuçlarına göre, müşterilerin tüm havayolu şirketlerinden genel beklentileri sunulan hizmetin kaliteli olmasıdır. Koltuk rahatlığı, kabin ekibinin kibar ve yardımsever olması ve yemeklerin lezzetli olması gibi hususlar tüm ülkeler için ön plana çıkmıştır. Belirtilen hususlara ek olarak, müşteriler havayolu şirketlerinden genel olarak meydana gelen rötarlara ilişkin şikayetlerde bulunmaktadır. Dolayısıyla, bu problemi yaşayan havayolu şirketlerinin gerekli tedbirleri almadığı durumda diğerlerine kıyasla önemli ölçüde rekabet avantajı kaybedecekleri ortadadır. Öte yandan, Afrika, Kuzey Amerika ve Güney Amerika’lı müşterilerin uçaklardaki temizliğe daha fazla önem gösterdiği sonucuna ulaşılmıştır. Son olarak, Güney Amerika’lı müşterilerin fiyatların yüksek olmasına yönelik bazı rahatsızlıkları bulunduğu belirlenmiştir. Bu bilgiler göz önünde bulundurularak, özellikle bu kıtaya hizmet veren havayolu şirketlerinin düşük fiyat odaklı yatırım stratejisini belirlemeleri yerinde olacaktır. Belirtilen bu öneriler ülkelerin sürdürülebilir kalkınma hedeflerine ulaşabilmelerine yardımcı olacaktır.
Differentiating gastrointestinal stromal tumors from leiomyomas using a neural network trained on endoscopic ultrasonography images
(Karger, 2022) Seven, Gülseren; Silahtaroğlu, Gökhan; Seven, Özden Özlük; Şentürk, Hakan
Background: Endoscopic ultrasonography (EUS) is crucial to diagnose and evaluate gastrointestinal mesenchymal tumors (GIMTs). However, EUS-guided biopsy does not always differentiate gastrointestinal stromal tumors (GISTs) from leiomyomas. We evaluated the ability of a convolutional neural network (CNN) to differentiate GISTs from leiomyomas using EUS images. The conventional EUS features of GISTs were also compared with leiomyomas. Patients and Methods: Patients who underwent EUS for evaluation of upper GIMTs between 2010 and 2020 were retrospectively reviewed, and 145 patients (73 women and 72 men; mean age 54.8 ± 13.5 years) with GISTs (n = 109) or leiomyomas (n = 36), confirmed by immunohistochemistry, were included. A total of 978 images collected from 100 patients were used to train and test the CNN system, and 384 images from 45 patients were used for validation. EUS images were also evaluated by an EUS expert for comparison with the CNN system. Results: The sensitivity, specificity, and accuracy of the CNN system for diagnosis of GIST were 92.0%, 64.3%, and 86.98% for the validation dataset, respectively. In contrast, the sensitivity, specificity, and accuracy of the EUS expert interpretations were 60.5%, 74.3%, and 63.0%, respectively. Concerning EUS features, only higher echogenicity was an independent and significant factor for differentiating GISTs from leiomyomas (p < 0.05). Conclusions: The CNN system could diagnose GIMTs with higher accuracy than an EUS expert and could be helpful in differentiating GISTs from leiomyomas. A higher echogenicity may also aid in differentiation.
An early warning system using machine learning for the detection of intracranial hematomas in the emergency trauma setting
(Turkish Neurosurgical Society, 2022) Aydoseli, Aydın; Ünal, Tuğrul Cem; Kardeş, Onur; Doğuç, Özge; Dolaş, İlyas; Adıyaman, Ali Ekrem; Ortahisar, Emircan; Silahtaroğlu, Gökhan; Aras, Yavuz; Sabancı, Pulat Akın; Sencer, Serra; Sencer, Altay
AIM: To present an early warning system (EWS) that employs a supervised machine learning algorithm for the rapid detection of extra-axial hematomas (EAHs) in an emergency trauma setting. MATERIAL and METHODS: A total of 150 sets of cranial computed tomography (CT) scans were used in this study with a total of 11,025 images. Of the CTs, 75 were labeled as EAH, the remaining 75 were normal. A random forest algorithm was utilized for the detection of EAHs. The CTs were randomized into two groups: 100 samples for training of the algorithm (split evenly between EAH and normal cases), and 50 samples for testing. In the training phase, the algorithm scanned every CT slice separately for image features such as entropy, moment, and variance. If the algorithm determined an EAH on two or more images in a CT set, then the workflow produced an alert in the form of an email. RESULTS: Data from 50 patients (25 EAH and 25 controls) were used for testing the EWS. For all CTs with an EAH, an alert was produced, with a 0% false-negative rate. For 16% of the cases, the practitioner received an email from the EWS that the patient might have an EAH despite having a normal CT scan. Positive and negative predictive values were 86% and 100%, respectively. CONCLUSION: An EWS based on a machine learning algorithm is an efficient and inexpensive way of facilitating the work of emergency practitioners such as emergency physicians, neuroradiologists, and neurosurgeons.
Use of artificial intelligence in the prediction of malignant potential of gastric gastrointestinal stromal tumors
(Springer, 2022) Seven, Gülseren; Silahtaroğlu, Gökhan; Koçhan, Koray; Tüzün İnce, Ali; Arıcı, Dilek Sema; Şentürk, Hakan
Background and Aims This study aimed to investigate whether AI via a deep learning algorithm using endoscopic ultrasonography (EUS) images could predict the malignant potential of gastric gastrointestinal stromal tumors (GISTs). Methods A series of patients who underwent EUS before surgical resection for gastric GISTs were included. A total of 685 images of GISTs from 55 retrospectively included patients were used as the training data set for the AI system. Convolutional neural networks were constructed to build a deep learning model. After applying the synthetic minority oversampling technique, 70% of the generated images were used for AI training and 30% were used to test AI diagnoses. Next, validation was performed using 153 EUS images of 15 patients with GISTs. In addition, conventional EUS features of 55 patients in the training cohort were evaluated to predict the malignant potential of GISTs and mitotic index. Results The overall sensitivity, specificity, and accuracy of the AI system for predicting malignancy risk were 83%, 94%, and 82% in the training dataset, and 75%, 73%, and 66% in the validation cohort, respectively. When patients were divided into low-risk and high-risk groups, sensitivity, specificity, and accuracy increased to 99% in the training dataset and 99.7%, 99.7%, and 99.6%, respectively, in the validation cohort. No conventional EUS features were found to be associated with either malignant potential or mitotic index (P > 0.05). Conclusions AI via a deep learning algorithm using EUS images could predict the malignant potential of gastric GISTs with high accuracy. Graphic
Lemmatizer: Akıllı Türkçe kök bulma yöntemi
(International Balkan University, 2020) Doğuç, Özge; Aytaç, Ömer Berkay; Silahtaroğlu, Gökhan
Yakın zamanda Türkçe doğal dil işleme alanında çeşitli çalışmalar yapılmıştır. Bu çalışmalar, üretilen akıllı bir sistemin Türkçe soru cevaplama, yazıyı başka bir dile çevirme, yazıyı özetleme,e-postalara otomatik yanıt gönderme gibi kabiliyetlere sahip olmasını öngörmektedir. Bahsedilen kabiliyetlerin temelinde, Türkçe kelimelerin köklerinin doğru şekilde bulunması gereksinimi yatmaktadır. Literatürde çeşitli Türkçe kök bulma yöntemleri verilmiş olsa da, Türkçe kelimelerin kompleks yapılarından dolayı başarı oranları genelde düşük kalmıştır. Bu çalışmada, Türkçe’nin sondan eklemeli yapısı kullanılarak bir kök bulma sistemigeliştirilmiş (Lemmatizer) ve bu konuda daha önce yapılmış olan Zemberek ve Snowball yöntemleriyle karşılaştırması verilmiştir. Lemmatizer sistemi Python ile yazılmıştır ve Türkçe’de en sık kullanılan 130’dan fazla ekive TDK sözlüğünübaz almaktadır. Ayrıca Knime platformu kullanılarak istatistiksel analiz yapılmıştır. Bu çalışma için öncelikle Lemmatizer sistemi çok sayıda Türkçe makale ve kitapla eğitilmiş ve Lemmatizer sistemi dağarcığını sürekli geliştirmiştir. Aynı zamanda, Kalbur isimli Türkçe ek ve kök veritabanıkullanılarak alınan geri beslemeler sayesinde, doğruluk oranı sürekli artmıştır.Lemmatizer sistemisonuçları hem sayı hem doğruluk açısından daha önce yapılmış olan Zemberek ve Snowball yöntemleriyle karşılaştırılmıştır. Karşılaştırmada farklı uzunluklarda Türkçe metinler kullanılmıştır. Lemmatizer yönteminin TDK sözlüğü kullanarak öğrenebilme özelliği sayesinde, Snowball ve Zembere yöntemlerine yakın sonuçlar verdiği ve kullanılan her yeni metinle başarı oranının diğer yöntemlere göre arttığı gösterilmiştir.
A generative model based adversarial security of deep learning and linear classifier models
(Slovene Society Informatika, 2021) Sivaslıoğlu, Samed; Çatak, Ferhat Özgür; Şahinbaş, Kevser
In recent years, machine learning algorithms have been applied widely in various fields such as health, transportation, and the autonomous car. With the rapid developments of deep learning techniques, it is critical to take the security concern into account for the application of the algorithms. While machine learning offers significant advantages in terms of the application of algorithms, the issue of security is ignored. Since it has many applications in the real world, security is a vital part of the algorithms. In this paper, we have proposed a mitigation method for adversarial attacks against machine learning models with an autoencoder model that is one of the generative ones. The main idea behind adversarial attacks against machine learning models is to produce erroneous results by manipulating trained models. We have also presented the performance of autoencoder models to various attack methods from deep neural networks to traditional algorithms by using different methods such as non-targeted and targeted attacks to multi-class logistic regression, a fast gradient sign method, a targeted fast gradient sign method and a basic iterative method attack to neural networks for the MNIST dataset.
An early prediction and diagnosis of sepsis in intensive care units: An unsupervised machine learning model
(Mugla University, 2020) Canbolat, Zehra Nur; Silahtaroğlu, Gökhan
Sepsis infection, which is one of the most important causes of death in intensive care units, is seen as a severe global health crisis. If an early diagnosis of sepsis infection cannot be made, and treatment is not started rapidly, septic shock may result in multiple organ failure and death is almost inevitable. Therefore, it is vital to establish an early diagnosis and start the treatment at once. This study aims to accomplish a new model of unsupervised machine learning using lactate and Ph laboratory test values, which are considered to be important parameters to diagnose sepsis infection. The data used in the study have been obtained from MIMIC-III international clinical database. Unsupervised machine learning has been performed via the Fuzzy-C algorithm along with validity indexes like Xie Beni on patients’ data diagnosed sepsis and non-sepsis. The machine-generated ten labels at the end of the training session considering-designed validity indexes. The labelled cluster representatives have been reduced to two dimensions by Principal Component Analysis method in order to monitor the learning in a two-dimensional space. The study contributes to the literature by conducting unsupervised learning through two parameters (Lactate and Ph) and leading to multi-parameter studies. In addition, the study reports that there are five types of sepsis patterns in terms of Lactate and PH laboratory tests.

Güncel Gönderiler