Bilgisayar Mühendisliği Bölümü Koleksiyonu

Bu koleksiyon için kalıcı URI

https://hdl.handle.net/20.500.12428/2626

Listeleniyor 1 - 20 / 27

Exploring the perspectives of university students on post-COVID-19 rental housing demands: a case study of Çanakkale, Türkiye
(Springer, 2025) Uşaklı, Metin; Yücebaş, Sait Can; Genç, Levent
This study focused on utilizing Machine Learning (ML) to examine the housing preferences of students. Employing a survey method, the study utilized decision tree, a widely favored ML approach, to present findings. The analysis focused on the post-COVID-19 rental housing preferences of students and their impact on rental prices. Furthermore, the research identified the number of rooms as a crucial factor for male students, particularly for first-year students, with gender becoming significant for second-, third-, and fourth-year students in Barbaros neighborhood. A noteworthy post-COVID-19 trend was the observation that students, in general, preferred communal living arrangements, sharing rental costs. Additionally, the study found that under different circumstances, male students were more inclined to lease housing in & Ccedil;anakkale province compared to their female counterparts.
Discrimination accuracy of haploid and diploid maize seeds using NIR spectroscopy coupled with different machine learning algorithms and data pretreatment methods
(Taylor & Francis Inc, 2025) Kahrıman, Fatih; Polat, Adem; Tiryaki, Ali Murat; Eskizeybek, Volkan; Fidan, Sertuğ; Songur, Umut
Spectral data collected at the single seed level allows determination of the biochemical content of the seed sample, as well as to identify the seed class. NIR (Near Infrared) spectroscopy provides a more precise method for differentiating haploid and diploid seeds in maize than traditional visual examination. In this study, classification models that can be used in the separation of haploid and diploid maize seeds were developed using spectra collected between 900-1700 nm from a single maize seed. In the study, 427 diploid and 311 haploid samples obtained by crossing 10 donor materials and 3 inducer lines and separated by eye according to the Navajo marker were used. Spectral measurements were conducted over the wavelength range of 900 to 1700 nm for each sample. The robust PCA (Principal Component Analysis) method was used to detect spectral outliers. Spectral data were treated with none, FD (First Derivative), SD (Second Derivative), SNV (Standard Normal Variate), and their binary combinations. Logistic Regression, Support Vector Machine with a linear kernel (SVM-C), Random Forest, and XGBoost methods were employed as machine learning techniques. The performance of the developed machine learning models was assessed using metrics such as Sensitivity, Specificity, Recall, F1-Score, and Accuracy. The Boosting method demonstrated the best performance with 94.9% accuracy, 95.1% sensitivity, 94% specificity, and an F1 Score of 96%, particularly when using raw reflectance data. These results obtained from raw data show that high accuracy can be achieved in classification models without requiring additional preprocessing steps. D2 preprocessing was found to be unsuitable for intact seed spectra, whereas SNV and D1 applications improved the classification success of other modeling techniques. The study revealed that the Boosting-Raw combination is a powerful and feasible method for classifying haploid and diploid samples.
ANALYZING THE IMPACT OF THE 2023 GENERAL ELECTIONS ON LAND PRICES USING MACHINE LEARNING: A CASE STUDY IN ÇANAKKALE, TURKEY
(Konya Teknik Univ, 2025) Doğan, Simge; Genç, Levent; Yücebaş, Sait Can; Yalpır, Şükran
This study analyses the impact of the general elections to be held on 14 May 2023 on the real estate market in Turkey. The aim of the study is to develop a model to predict land unit prices (₺/m²) by analysing land prices, exchange rates and gold values observed before (February-March-April) and after (May-June-July) elections for Ayvacık, Bayramiç, Biga, Çan, Eceabat, Ezine, Gelibolu, Lapseki, Merkez and Yenice districts of Çanakkale province. Daily fluctuations in foreign exchange and gold values, which are the main economic parameters in the study, were recorded during the election period. The findings of this research, which predicts price movements in the property market using machine learning methods such as regression trees, reveal that unit prices of land generally tend to increase with increases in exchange rates, but in some districts where gold prices increase, the unit price shows a reverse trend. This is attributed to the fact that investors prefer gold as a safer asset in times of economic uncertainty. The results obtained can help investors and buyers to predict future trends in property prices, as well as contribute to the development of economic policies by experts to stabilise fluctuations in investment instruments.
A Deep Learning Approach based on Ensemble Classification Pipeline and Interpretable Logical Rules for Bilingual Fake Speech Recognition
(Gazi University, 2025) Boztepe, Emre Beray; Karasulu, Bahadır
The essential steps of our study are to quantify and classify the differences between real and fake speech signals. In this scope, the main aim is to use the salient feature learning ability of deep learning in our study. With the use of ensemble classification pipeline, the interpretable logical rules were used for generalized reasoning with the class activation maps to discriminate the different speech classes as correctly. Fake audio samples were generated by using Deep Convolutional Generative Adversarial Neural Network. Our experiments were conducted on three different language dataset such as Turkish, English languages and Bilingual. As a result of higher classification and recognition accuracy with the use of classification pipeline as compiled into a majority voting-based ensemble classifier, the experimental results were obtained for each individual language performance approximately as 90% for training and as 80.33% for testing stages for pipeline, and it reached as 73% for majority voting results considered together with the appropriate test cases as well. To extract semantically rich rules, an interpretable logical rules infrastructure was used to infer the correct fake speech from class activations of deep learning's generative model. Discussion and conclusion based on scientific findings are included in our study.
Enhancing Picture Book Reading Experiences: Empowering Children Through Participatory Technology Solutions in Early Years
(Springer Nature, 2024) Bus, Adriana G.; Broekhof, Kees; Coenraads, Christiaan; Mifsud, Charles L.; Sarı Uğurlu, Burcu; Uğurlu, Bora; Vaessen, Karin
The Erasmus Plus project, Stimulating Adventures for Young Learners (SAYL), aims to establish a digital library of picture books, empowering young children to engage in book reading without adult guidance. Our primary focus is on leveraging technology to address challenges that might impede the cognitive processing of stories conveyed through a combination of words and images. The challenge in creating digital books lies in effectively guiding the reader’s cognitive processing during reading, without overwhelming or underutilizing their working memory capacity. Following the principles of multimedia learning theory, we must acknowledge the limited memory capacity allocated to the two channels processing visual and verbal information. This restriction necessitates minimizing any unnecessary cognitive load, for instance due to the effort required to navigate through the book. Similarly, it requires swiftly identifying the essential information necessary for comprehending the story, even in the presence of various distractions. Furthermore, according to multimedia learning theory, we cannot expect the effortless integration of essential information. Many young children may require incentives to achieve that. This chapter explores how enhancements, such as highlighting specific details or incorporating animations or musical elements, can assist young children in overcoming these challenges. This is supported by illustrative examples from one of the books digitized as part of the SAYL project. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.
Kripto Varlık Özelinde Üniversite Öğrencilerinin Blok Zincir Teknolojisi Bilgi Düzeylerinin İncelenmesi
(2024) Demir, Ümit; Uğurlu, Bora; Ataç, Sezgin
İnternet ve bilişim teknolojileri araçlarında meydana gelen gelişimler gelişmeler birçok alan ve işkolunun yeniden düzenlenmesine hem olanak sağlamış hem de zorunlu kılmıştır. Blok zinciri kavramı ülkemizde kripto varlıklar ile bilinirliği artan bir kavram olsa bile bilgi güvenliğinin esas olduğu dağıtık yapıda birçok sektörde gelişme potansiyeli bulunmaktadır. Bu nedenle blok zincir kavramının farklı alanlarda kullanımına yönelik bilgi yeterliliğinin önemli olduğu düşünülmektedir. Bu kapsamda gerçekleştirilen bu çalışma ile 344 lisans ve önlisans öğrencinin katılımı ile üniversite öğrencilerinin kripto para özelinde blok zinciri teknolojisine yönelik bilgi düzeylerini belirlenmesi amaçlanmıştır. Elde edilen veriler ışığında öğrencilerin çoğunun blok zinciri konusunda az ya da hiç bilgi sahibi olmadığı sonucuna ulaşılmıştır. Fakat öğrenciler kripto varlıklara yatırım düzeylerinin yüksek olduğu görülmüştür. Ayrıca öğrencilerini kripto varlıklara yatırım konusunda olumlu algılara sahip oldukları ve yatırım sürecinde sosyal çevre ve medya ortamlarının da etkisinin büyük olduğu çalışma sonucunda elde edilen önemli bulgulardandır. Elde edilen veriler ışığında tüm eğitim kademelerinde finansal okuryazarlık ve lisans seviyesinde blok zincir eğitimlerinin verilmesine yönelik öneriler getirilmiştir.
Gömülü Sistem Cihazları ile Kuantum Ağların Dağıtık Simülasyonu
(2024) Ceylan, Osman Semi; Yılmaz, İhsan
Günümüz kuantum ağ yapısı mimari bakımdan klasik ağ yapıları referans alarak kurulmuştur. Fakat kuantum ağı oluşturan aygıtların parçacık kontrol mekanizmaları nedeniyle klasik aygıtlara göre daha maliyetlidir. Deneysel bir kuantum ağın gerçekleştirme maliyetini arttıran bu sebepten dolayı bir kuantum ağ kurmadan önce olurluk benzetimleri yapılmaktadır. Fakat kuantum dolaşıklığın mevcut benzetim yöntemlerindeki üstel artan veri depolama kapasitesi gerektirdiğinden dolayı güçlükler yaşanmaktadır. Bu çalışmada bu probleme bir çözüm olarak kuantum ağların deneysel olarak birden fazla klasik cihaz yardımıyla dağıtık benzetim kullanarak daha az maliyetli olarak nasıl gerçekleştirilebileceği önerilmektedir. Bu amaç için geliştirilen eklenti yazılımı ile farklı ağ senaryolarında elde edilen sonuçlar dağıtık benzetim modelinin mevcut modellere göre daha etkin olduğunu göstermektedir.
QDNS: Quantum Dynamic Network Simulator Based on Event Driving
(Institute of Electrical and Electronics Engineers Inc., 2021) Ceylan, Osman Semi; Yılmaz, İhsan
After the no-cloning theory was presented in the quantum physics field, researchers offered more secure but theoretical protocols than classic ones shaped around this theorem. Without much time passed, we then observed that experimental studies have been made particularly for the national security concerns. In line with these events, with this study, we are presenting the QDNS, an event driven quantum network simulation framework for enthusiasts of the filed to simulate their custom protocols in quantum network topology. With the event triggered way, we tried to make a more understandable and user-friendly environment yet powerful enough to take into account of complex nature of the quantum world.
High-dimensional Grover multi-target search algorithm on Cirq (vol 137, 244, 2022)
(Springer Heidelberg, 2022) Acar, Erdi; Gündüz, Sabri; Akpınar, Güven; Yılmaz, İhsan
[Anstract Not Available]
Strange quark matter solutions for Marder's universe in f (R, T) gravity with Λ
(Springer, 2016) Aygün, Sezgin; Aktaş, Can; Yılmaz, İhsan
In this paper, we investigate homogeneous cylindrically symmetric Marder's universe in the presence of strange quark matter (SQM) source in f (R, T) gravity with cosmological constant Lambda. For this aim we have used the anisotropy feature (sigma(x)(x)/theta) of Marder type universe and equation of state (EoS) strange quark matter to obtain solutions in two classes f (R, T) gravity (Harko et al. in Phys. Rev. D 84:024020, 2011). Finally, some physical and kinematical properties are discussed.
Determination of the relationship between housing characteristics and housing prices before and after the Kahramanmaraş earthquake using machine learning: A case study of Adana, Turkiye
(Yildiz Technical Univ, Fac Architecture, 2024) Doğan, Simge; Genç, Levent; Yücebaş, Sait Can; Uşaklı, Metin; Dumlu, Cengizhan
Earthquakes have a significant impact on the real estate sector. Damage caused by earthquakes leads to an imbalance in the supply and demand for housing, thus temporarily causing stagnation in the real estate sector. Two earthquakes occurred in Pazarcık and Elbistan districts of Kahramanmaraş on February 6, 2023 at 04.17 am with a magnitude of 7.7 and 13.24 am with a magnitude of 7.6. A machine learning based model was created to analyze the change in house prices and the variables affecting the price during the earthquake, which is called “the Disaster of the Century”. After the earthquake, the prices of houses for sale in the central districts of Adana province (Seyhan, Yüreğir, Sarıçam and Çukurova) where there was the least damage were collected from the relevant website with a web scraper. These data were classified as categorical and numerical datasets, and the necessary pre-processing stage for machine learning algorithms was performed. The characteristics that change and are effective in housing preferences before the earthquake (February 2022) and after the earthquake (February 2023) were determined by the decision tree method, which is one of the machine learning algorithms. In this context, it is aimed to determine the housing variables that are effective in before and after-earthquake pricing in the central districts of Adana province. In the study, while 'Building Age and Number of Room’ are effective in determining the price in 2022; 'Housing shape and Facade' feature comes to the fore in 2023. The housing characteristics that affect the price change in two years. The change in housing preference criteria after the earthquake shows that the lifestyle in cities has also changed. According to this change, it requires the development of new approaches in urban design and planning approaches and is expected to be a reference for future studies.
New fully automated software for assessment of brachial artery flow-mediated dilation with advantages of continuous measurement
(Turkish Soc Cardiology, 2012) Ercan, Ertugrul; Kırılmaz, Bahadır; Kahraman, İsmail; Bayram, Vildan; Doğan, Hüseyin
Objective: Flow-mediated dilation (FMD) is used to evaluate endothelial functions. Computer-assisted analysis utilizing edge detection permits continuous measurements along the vessel wall. We have developed a new fully automated software program to allow accurate and reproducible measurement. Methods: FMD has been measured and analyzed in 18 coronary artery disease (CAD) patients and 17 controls both by manually and by the software developed (computer supported) methods. The agreement between methods was assessed by Bland-Altman analysis. Results: The mean age, body mass index and cardiovascular risk factors were higher in CAD group. Automated FMD% measurement for the control subjects was 18.3+/-8.5 and 6.8+/-6.5 for the CAD group (p=0.0001). The intraobserver and interobserver correlation for automated measurement was high (r=0.974, r=0.981, r=0.937, r=0.918, respectively). Manual FMD% at 60th second was correlated with automated FMD % (r=0.471, p=0.004). Conclusions: The new fully automated software (c) can be used to precise measurement of FMD with low intra-and interobserver variability than manual assessment. (Anadolu Kardiyol Derg 2012; 12: 553-9)
Higher dimensional FRW universe solutions with quark and strange quark matter in creation field cosmology
(Elsevier Science Bv, 2016) Aygün, Sezgin; Aktaş, Can; Yılmaz, İhsan; Şahin, Mustafa
In this study, firstly we have studied the behavior of quark and strange quark matter for a (n + 2) -dimensional Friedmann-Robertson-Walker (FRW) universe which is homogeneous and isotropic in creation field (C-field) cosmology. Using the deceleration parameter two different exact solutions of the modified Einstein equations in C-field cosmology are obtained. In addition, we obtain exact solutions of the quark and strange quark matter for a (n + 2)-dimensional homogeneous and isotropic static Einstein universe (SEU) and a maximally symmetric de Sitter vacuum universe in four dimensions. Also, using. C = 0 in C-field theory, we get the SEU and de Sitter vacuum universes in Riemann geometry. Finally, some physical and kinematical quantities are discussed. (C) 2016 The Physical Society of the Republic of China (Taiwan). Published by Elsevier B.V. All rights reserved.
Teleparallel energy-momentum distribution of various black hole and wormhole metrics
(World Scientific Publ Co Pte Ltd, 2018) Aygün, Sezgin; Baysal, Hüsnü; Aktaş, Can; Yılmaz, İhsan; Sahoo, P. K.; Tarhan, İsmail
Using the Einstein, Bergmann-Thomson (BT) and Landau-Lifshitz (LL) energy and momentum formulations in teleparallel gravity (TG), we obtain the total energy and momentum distributions for phantom black hole metric. We get different energy distributions similar to the earlier study and the momentum distributions vanish for phantom black hole metric in TG. These momentum solutions agree with the study of Sahoo et al. in general relativity. However, using Einstein, Bergmann-Thomson and Landau-Lifshitz energy and momentum complexes, we investigate regular black hole, asymptotically flat wormhole, anti-de Sitter wormhole and Ellis wormhole solutions in TG. We obtain i) same BT and LL energy density solutions for regular black hole metric, (ii) same and zero energy distribution for asymptotically flat wormhole, (iii) proportion with Einstein and BT energy density solutions for anti-de Sitter wormhole, (iv) same and negative Einstein and BT energy density solutions for Ellis wormhole in TG.
Improving effectiveness of different deep learning-based models for detecting COVID-19 from computed tomography (CT) images
(Springer Science and Business Media Deutschland GmbH, 2021) Acar, Erdi; Şahin, Engin; Yılmaz, İhsan
COVID-19 has caused a pandemic crisis that threatens the world in many areas, especially in public health. For the diagnosis of COVID-19, computed tomography has a prognostic role in the early diagnosis of COVID-19 as it provides both rapid and accurate results. This is crucial to assist clinicians in making decisions for rapid isolation and appropriate patient treatment. Therefore, many researchers have shown that the accuracy of COVID-19 patient detection from chest CT images using various deep learning systems is extremely optimistic. Deep learning networks such as convolutional neural networks (CNNs) require substantial training data. One of the biggest problems for researchers is accessing a significant amount of training data. In this work, we combine methods such as segmentation, data augmentation and generative adversarial network (GAN) to increase the effectiveness of deep learning models. We propose a method that generates synthetic chest CT images using the GAN method from a limited number of CT images. We test the performance of experiments (with and without GAN) on internal and external dataset. When the CNN is trained on real images and synthetic images, a slight increase in accuracy and other results are observed in the internal dataset, but between 3 % and 9 % in the external dataset. It is promising according to the performance results that the proposed method will accelerate the detection of COVID-19 and lead to more robust systems.
A novel approach to distinguish complicated and non-complicated acute cholecystitis: Decision tree method
(Lippincott Williams and Wilkins, 2023) Gojayev, Afig; Karakaya, Emre; Erkent, Murathan; Yücebaş, Sait Can; Aydın, Hüseyin Onur; Kavasoğlu, Lara; Aydoğan, Cem; Yıldırım, Sedat
It is difficult to differentiate between non-complicated acute cholecystitis (NCAC) and complicated acute cholecystitis (CAC) preoperatively, which are two separate pathologies with different management. The aim of this study was to create an algorithm that distinguishes between CAC and NCAC using the decision tree method, which includes simple examinations. In this retrospective study, the patients were divided into 2 groups: CAC (149 patients) and NCAC (885 patients). Parameters such as patient demographic data, American Society of Anesthesiologists (ASA) score, Tokyo grade, comorbidity findings, white blood cell (WBC) count, neutrophil/lymphocyte ratio, C-reactive protein (CRP) level, albumin level, CRP/albumin ratio (CAR), and gallbladder wall thickness (GBWT) were evaluated. In this algorithm, the CRP value became a very important parameter in the distinction between NCAC and CAC. Age was an important predictive factor in patients with CRP levels >57 mg/L, and the critical value for age was 42. After the age factor, the important parameters in the decision tree were WBC and GBWT. In patients with a CRP value of ≤57 mg/L, GBWT is decisive and the critical value is 4.85 mm. Age, neutrophil/lymphocyte ratio, and WBC count were among the other important factors after GBWT. Sex, ASA score, Tokyo grade, comorbidity, CAR, and albumin value did not have an effect on the distinction between NCAC and CAC. In statistical analysis, significant differences were found groups in terms of gender (34.8% vs 51.7% male), ASA score (P < .001), Tokyo grade (P < .001), comorbidity (P < .001), albumin (4 vs 3.4 g/dL), and CAR (2.4 vs 38.4). By means of this algorithm, which includes low-cost examinations, NCAC and CAC distinction can be made easily and quickly within limited possibilities. Preoperative prediction of pathologies that are difficult to manage, such as CAC, can minimize patient morbidity and mortality.
An Ontology based product recommendation system for next generation e-retail
(Taylor and Francis Ltd., 2023) Tiryaki, Ali Murat; Yücebaş, Sait Can
The number of e-commerce resources has increased considerably. Thus, it has become important for sellers to be able to quickly recommend products to potential buyers. Some product recommendation systems developed for this purpose. However, due to the lack of semantics, the systems’ success in recommending accurate products according to user preferences is low. In this study carried out within the scope of a state-funded R&D project, an ontology-based personalized product recommendation system named E-Prod was developed. E-Prod tracks various e-commerce systems in real time and transfers the product information to the ontology model. E-Prod uses a novel recommendation approach that combines machine learning and semantic matching to provide personalized recommendations. The system learns user’s preferences based on semantic relationships between products by monitoring their behaviors. In this way, accurate recommendations are made by semantic matching between products and user preferences. E-Prod has been tested with over 250 registered users and compared to traditional collaborative recommendations in terms of accuracy, precision, and recall. As a result, E-Prod outperformed traditional methods by 92.79% accuracy, 92.93% precision, and 90.58% recall. Within the scope of this study, E-Prod covers the clothing, shoes, and bag retail sectors. However, it provides a generic infrastructure for new generation e-commerce systems. Its reusable modules can be adapted to any domain.
Optimized Decision Tree and Black Box Learners for Revealing Genetic Causes of Bladder Cancer
(Tech Science Press, 2023) Yücebaş, Sait Can
The number of studies in the literature that diagnose cancer with machine learning using genome data is quite limited. These studies focus on the prediction performance, and the extraction of genomic factors that cause disease is often overlooked. However, finding underlying genetic causes is very important in terms of early diagnosis, development of diagnostic kits, preventive medicine, etc. The motivation of our study was to diagnose bladder cancer (BCa) based on genetic data and to reveal underlying genetic factors by using machine-learning models. In addition, conducting hyperparameter optimization to get the best performance from different models, which is overlooked in most studies, was another objective of the study.Within the framework of these motivations, C4.5, random forest (RF), artificial neural networks (ANN), and deep learning (DL) were used. In this way, the diagnostic performance of decision tree (DT)-based models and black box models on BCa was also compared. The most successful model, DL, yielded an area under the curve (AUC) of 0.985 and a mean square error (MSE) of 0.069. For each model, hyper-parameters were optimized by an evolutionary algorithm. On average, hyper-parameter optimization increased MSE, root mean square error (RMSE), LogLoss, and AUC by 30%, 17.5%, 13%, and 6.75%, respectively. The features causing BCa were extracted. For this purpose, entropy and Gini coefficients were used for DT-based methods, and the Gedeon variable importance was used for black box methods. The single nucleotide polymorphisms (SNPs) rs197412, rs2275928, rs12479919, rs798766 and rs2275928, whose BCa relations were proven in the literature, were found to be closely related to BCa. In addition, rs1994624 and rs2241766 susceptibility loci were proposed to be examined in future studies.
Predictors of rapidly progressive glomerulonephritis in acute poststreptococcal glomerulonephritis
(Springer Science and Business Media Deutschland GmbH, 2023) Karakaya, Deniz; Güngör, Tülin; Kargin Çakıcı, Evrim; Yazılıtaş, Fatma; Çelikkaya, Evra; Yücebaş, Sait Can; Bülbül, Mehmet
Background Acute post-streptococcal glomerulonephritis (APSGN) is an immune-mediated inflammatory respsonse in the kidneys caused by nephritogenic strains of group A (3-hemolytic streptococcus (GAS). The present study aimed to present a large patient cohort of APSGN patients to determine the factors that can be used for predicting the prognosis and progression to rapidly progressive glomerulonephritis (RPGN).Methods The study included 153 children with APSGN that were seen between January 2010 and January 2022. Inclusion criteria were age 1-18 years and follow-up of >= 1 years. Patients with a diagnosis that could not be clearly proven clinically or via biopsy and with prior clinical or histological evidence of underlying kidney disease or chronic kidney disease (CKD) were excluded from the study.Results Mean age was 7.36 +/- 2.92 years, and 30.7% of the group was female. Among the 153 patients, 19 (12.4%) progressed to RPGN. The complement factor 3 and albumin levels were significantly low in the patients who had RPGN (P = 0.019). Inflammatory parameters, such as C-reactive protein (CRP), platelet-to-lymphocyte ratio, CRP/albumin ratio, and the erythrocyte sedimentation rate level at presentation were significantly higher in the patients with RPGN (P < 0.05). Additionally, there was a significant correlation between nephrotic range proteinuria and the course of RPGN (P = 0.024).Conclusions We suggest the possibility that RPGN can be predicted in APSGN with clinical and laboratory findings.
İnsanın günlük aktivitelerinin yeni bir veri kümesi: Derin öğrenme tekniklerini kullanarak sınıflandırma performansı için kıyaslama sonuçları
(Gazi Üniversitesi, 2021) Metin, İbrahim Ali; Karasulu, Bahadır
Studies to classify human activities can contribute to the development of new systems that will facilitate daily life by evaluating the interaction of individuals with their environment. In this study, a novel data set is presented to be used in classifying the activities that individuals perform during the day. First of all, various deep architectural models presented in the study were tested with publicly available datasets well-known in the literature. Afterwards, various classification experiments were carried out by using our novel dataset, which was created with the sensor data collected with the smartphone located onto the belly region of ten volunteer individuals consisting of five males and five females aged between 25 and 55 years. Data of each activity at two different positions were taken, and also, 15 seconds raw data including 4 dynamic and 3 static activities were acquired. With 20 Hz sampling frequency for each activity position, 20 readings are made per signal window in 1 second. Thanks to the software tool developed for the study, various human activities were succesfully classified in experiments by allowing different network parameters and layer selection for the deep learning architectures including recurrent neural network models and convolutional neural network model. The novel dataset contains raw data, as well as, it involves some alternative subsets created with the use of Butterworth filter. As a result of experiments, the classification performance at accuracy rate of 97% to 99% for various activities of individuals was obtained on various datasets. The suitability of using the novel data set in studies on classification and prediction of human activities has been proven.

Güncel Gönderiler