Search

Small and Medium-sized Enterprises (SMEs) are private sector enterprises whose capital, workforce and assets are below the thresholds determined according to national regulations. SMEs play an important role in the economy of most countries in the world, especially in developing countries. SMEs, which make up approximately 90% of enterprises worldwide, provide more than 50% of employment. Estimating the contribution of SMEs to the economy at the country level is very important in terms of planning and investment. In this study, a deep learning-based model was developed to predict the contribution of SMEs to the economy. The developed LSTM-based deep learning model was compared with RF, SVM, CNN, GRU, MLP and RNN. Experimental results showed that the developed model had a better prediction performance than other models compared with 2.169 MSE, 1.473 RMSE, 1.175 MAE, and 0.959 R2 values. [ABSTRACT FROM AUTHOR]

Alzheimer's disease, a type of dementia, is quite common in the world. The disease has different stages and there is still no cure. With current machine learning methods, different stages of the disease can be detected. Especially with deep learning-based methods, disease detection can be made sensitively. In this study, two different deep learning architectures have developed by using special block structures of ResNet and Inception architectures. These architectures have produced effective results in the detection of Alzheimer's. With the design of special block structures, the ability of different architectures to work together has been revealed. In the experimental results, it is seen that the proposed architectures produce effective results. [ABSTRACT FROM AUTHOR]

Microbes are tiny single-celled creatures that live both inside and outside the human body and are found almost everywhere. There are microbes that can be harmful as well as harmless. Therefore, microbes play a major role in maintaining the ecological balance. However, distinguishing these microbes from each other is often a difficult task. The need for a specialist who understands the various characteristics of microbes and can interpret taxonomy nuances and the need for a laboratory makes this process both costly and time consuming. Therefore, the importance of computer aided systems in this field has increased and artificial intelligence technologies have begun to be used in the classification of microbes today. With the application of artificial intelligence technologies to this field, high level accuracy scores are obtained, and biodiversity assessments can be made automatically. In this study, microbes were classified by using deep learning models, which is one of the artificial intelligence methods. The study consisted of four stages. In the first stage, the data were obtained. In the second stage, these data were preprocessed and normalized. In the third stage, LSTM and RNN models are designed, and data were classified. In the last stage, the performances of the classifiers were determined by accuracy, precision, recall, F1-score, and AUC score. At the end of the study, an accuracy score of 92,53% was obtained with the RNN classifier and 99,85% with the LSTM classifier. [ABSTRACT FROM AUTHOR]

Electric motors have an important place in industry due to their ability to automate and facilitate various processes. Faults that occur in electric motors may affect the operation of the device or system and cause great financial losses. It is therefore critical to detect faults at an early stage. The use of computer-aided software in the detection of faults comes to the fore due to its cost and time saving potential. In this study, a deep learning-based model is proposed to detect engine bearing failure types. With this model, which uses one-dimensional convolutional neural network (1D-CNN) architecture, fault type detection is provided by using only vibration data. The proposed architecture is an efficient model that uses vibration signals to diagnose engine faults quickly and reliably. Within the scope of the study, detailed performance evaluations of the training and testing stages were provided by using different speed scenarios. With this model, which has a high generalization ability, fault detection has been made with high accuracy rates in different scenarios. [ABSTRACT FROM AUTHOR]

Reversible quantum circuits are constructed using different types and numbers of quantum gates. Optimizing the number of gates to be used while creating quantum circuits reduces the cost and complexity. It is important to know the number of inputs, outputs, and gates for obtaining and optimizing state tables in reversible quantum circuits. In addition, these parameters are also used to detect faults that may occur in quantum circuits. There is a lack of determination of the input, output, and gate numbers for quantum circuits in the literature. In addition, a limited number of standard quantum circuits are used to test the applications made in the literature. It has been determined that there are very few datasets that can be used in this context. Elimination of these deficiencies in the literature constitutes the aim of our study, and the proposed method constitutes the originality of our study. In this study, the number of gates and inputs were determined by using Yolo (You Only Look Once) based methods. A large dataset consisting of CNOT, Feynman, and Toffoli gates was created using the "MATLAB" and "RCViewer+" programs. In this study, a total of 5000 quantum circuits with 1-8 gate numbers and 3-7 input numbers were created. Gates and inputs are labeled on the obtained datasets. YoloV4, YoloV7, and YoloV7x methods were applied to the tagged datasets with a training and testing ratio of 80:20. 87.1%, 89.7% and 89.3% mAP were calculated for the YoloV4, YoloV7 and YoloV7x methods, respectively. The proposed method was run for 2800 iterations and the best result was obtained with the YoloV7 algorithm. [ABSTRACT FROM AUTHOR]

The use of computer-aided devices and systems in the field of health has become quite widespread. These devices and systems contributed to faster and earlier diagnosis of diseases. Especially imaging devices such as Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) play a major role in diseases where early diagnosis is important. In neurological diseases, the use of MRI and CT images as input images in deep learning models is becoming increasingly common. In this study, MRI images from the "Alzheimer's Parkinson's 3 Class Data Set" dataset obtained from the Kaggle website were used to diagnose Alzheimer's and Parkinson's diseases. There are three classes in this data set: 2561 Alzheimer's, 906 Parkinson's and 3010 Control (Normal). In this study, when Alzheimer's Parkinson's and Normal classes were trained with ResNet-18, VGG-16 and ConvNext architecture, accuracy rates of 96.2%, 95.4% and 98.9% were obtained respectively. In addition, Alzheimer's and Parkinson's diseases were tested with binary classifiers on the normal class. Among the models trained for Alzheimer's - Normal and Parkinson's - Normal classes, ResNet-18 architecture achieved 82.0% and 96.1%, VGG- 16 architecture achieved 95.4% and 89.4%, and ConvNext architecture achieved 99.4% and 99.5% success rates, respectively. [ABSTRACT FROM AUTHOR]