JournalTOCs

Veri Bilimi
Number of Followers: 0

This is an Open Access Journal

Open Access journal
ISSN (Online) 2667-582X
Published by DergiPark

[185 journals]

- Authors: Merve ÖZKAN; Caner ÖZCAN
  PubDate: Sun, 25 Dec 2022 00:00:00 +030
Classification of Some Biochemical Properties with J48 Classification Tree
Algorithms in Hyperspectral Data
- Authors: Sinan DEMİR; Levent BAŞAYİĞİT
  Abstract: Hyperspectral sensing methods have been used in agriculture for many years to determine physiological events against abiotic and biotic stress factors of plants. Studies on the determination of wavelengths associated with biochemical properties using hyperspectral data of plants in the visible and near-infrared region under sustainable agricultural production models are limited. In this research, Isparta Oil rose with the geographical indication product from organic farming medicine and aromatic plants from sustainable production models were used considering that the effect of cultural practices from abiotic and biotic stress factors is minimal. Hyperspectral measurements and feature selection of chlorophyll a (mg g-1), chlorophyll b (mg g-1), chlorophyll a+b (mg g-1), total flavonoid content (mg catechin g-1), total phenolic content (mg GAE g-1), and total antioxidant capacity (mg TEAC g-1) biochemical properties are made using the J48 classification tree algorithm in organically grown Isparta Oil Rose leaves. In the classification algorithm, different hyperspectral data (bands) are used as independent variables for each biochemical feature, while the amounts of each biochemical feature of the dependent variable are lower and higher than the mean, and the binary response variable (binary) is taken into the model. In the selection of independent variables, the correlation-based CfsSubsetEval algorithm, which does not cause multicollinearity, was used. The areas under the classification accuracies, sensitivity, specificity, and receiver operating characteristic (ROC) curves, which are the classification performances of chlorophyll a, chlorophyll b, chlorophyll a+b, total flavonoid content, total phenolic substance content, and total antioxidant capacity, are given as “72.1%, 73.3%, 76.25%, 69.6%, 71.7%, 67.5%”, “0.353, 0.440, 0.553, 0.714, 0.771, 0.657”, “0.782, 0.811, 0.802, 0.653, 0.621, 0.682” and “0.558, 0.643, 0.631, 0.638, 0.723, 0.625” were determined respectively. As a result of the study, it was concluded that the chlorophyll a+b (mg g-1) content was determined with the J48 classification tree algorithm with the highest accuracy in the classification of the biochemical contents of made the organic farming Isparta Oil rose leaves with visible and near-infrared hyperspectral data.
  PubDate: Sun, 25 Dec 2022 00:00:00 +030
Fusion-Based Hybrid Fog Removal
- Authors: Bahadır ARABALI; Kemal FİDANBOYLU
  Abstract: Aerosols in the environment we live in reduce the quality of the images we take. For different purposes, it is necessary to clean the blurs in the obtained images. There are many algorithms in the literature to achieve this. In the last 10 years, many new approaches have been developed in the problem of defogging image processing. Among these, the most successful and most used algorithm is the Dark Channel Prior algorithm. The Dark Channel Prior algorithm is based on very low pixel density values in different color channels. These low density values create a patch for the fog in the image with the algorithm and can remove or reduce the effect of foggy scenes. In this study, by using the "exposure fusion" algorithm, better exposed images were obtained by fusing different exposures according to values such as contrast and saturation. In this paper, it will be shown with different examples that more successful fog removal results are obtained by using the Dark Channel Prior algorithm together with the exposure fusion algorithm.
  PubDate: Sun, 25 Dec 2022 00:00:00 +030
Diseases Detection from Human Retina's Optical Coherence Tomography
Images using Deep Learning Models
- Authors: Batuhan METİN; Bahadir KARASULU
  Abstract: One of the most important problems that has affected the lives of individuals is the progression of the age of individuals and the inevitable of life is the visual disorders caused by retinal diseases. To avoid these diseases, diagnosis in early periods is essential to minimize the risk of life being adverse and subsequent loss of vision. Machine learning and deep learning methods that have become more prevalent with evolving technological methods have made it more common to study the method of Optical Coherence Tomography (OCT) imaging. In this study, experiments were carried out on the publicly available OCT dataset. By considering their high classification performance, the Convolutional Neural Network (CNN)-based ResNet50 and MobileNetV2 models are used in our study. Deep learning-based structures were tested in the classification of various retinal diseases in the experiments carried out in our study. In experiments, accuracy measurements of classification performance were performed by giving input on models with different parameters created. By considering both models, the result of the tests was performed that the macro average accuracy values as resulting in a performance of approximately 81% to 94%. According to these test results, when the ResNet50 and MobileNetV2 models used in the experiments are considered together that the average F1 score is obtained as 0.75 for CNV, 0.86 for Drusen, 0.90 for DME and 0.96 for normal retinas. With the comparison of the results obtained in this study and other studies in the literature, it is showed that this study has high accuracy performance. The study also includes discussions and scientific findings on the results.
  PubDate: Sun, 25 Dec 2022 00:00:00 +030
- Authors: Burçin ÖNER; Yüksel ÖNER
  PubDate: Sun, 25 Dec 2022 00:00:00 +030
The Role and Essence of Artificial Intelligence in Cybersecurity: A Survey
- Authors: Maad M. MIJWIL; Emre SADIKOĞLU, Emine CENGİZ, Hasibe CANDAN
  Abstract: Today, artificial intelligence technologies are entering into many areas, including the area of cybersecurity, which is a branch of technology. The role of cybersecurity is to defend systems, networks and data from hacking and threats, preventing unauthorised access, tampering with, or deleting data, or blackmailing users. This article strives to determine the significance and role of artificial intelligence in achieving cybersecurity in the digital environment and seeking to preserve user data and information from the penetration and destruction of systems by highlighting the essential dimensions and concepts of artificial intelligence techniques. In this article, the authors decided to confirm the role of artificial intelligence in conducting cybersecurity for users of social networking sites. This article concluded that it is necessary to think about artificial intelligence techniques in employing them to protect the privacy and data of individuals and users across various digital platforms and in all areas.
  PubDate: Sun, 25 Dec 2022 00:00:00 +030
Common Future of Blockchain and Quantum Computing in Information
Technologies: Quantum Blockchain
- Authors: Sevdanur GENÇ
  Abstract: Bitcoin, Ethereum and other cryptocurrencies are developed on blockchain technology. On this technology, data is encrypted and divided into parts, and this encryption is used in financial transactions. Quantum computing is offered as an alternative for encryption algorithms to work faster. However, there is more interest in quantum computers and quantum programming when it comes to the existence of quantum computing. Quantum computers analyze algorithms more powerfully and faster than today's classical computers. The problem is that if quantum computing is used in the cryptocurrency market in the future, all transactions on the blockchain will both be able to resume quickly and be vulnerable to security threats and fraud.In this study, first of all, information about blockchain and quantum computing is given. The future impact of quantum computers on crypto money and blockchain technologies is discussed and the concept of quantum blockchain is mentioned. At the same time, the possible positive and negative effects of quantum computing on the blockchain were mentioned. Finally, the software industry's attempts on quantum computing, which is seen as a threat to the field of cryptography, are mentioned. It is thought that this study will contribute to the Turkish literature.
  PubDate: Sun, 25 Dec 2022 00:00:00 +030
Investigating Word Association Mining Techniques
- Authors: Duygu BAĞCI DAŞ; Sevdanur GENÇ
  Abstract: This study presents the investigation of the effect of conditional entropy, mutual information (MI) values, log-likelihood ratio (LLR), and simple co-occurrences on extracting strong syntagmatic relationships. Experiments are conducted by using the Yelp Academic Dataset, which includes extracted 10.000 restaurant reviews. The mutual information values of word pairs are considered to extract the top syntagmatically related words from the corpus. For this purpose, Spyder 3.3.6 and Python Natural Language Toolkit (NLTK) Library are used. The mutual information values are then compared with simple co-occurrences count. The analysis results indicated that the three Word collocation techniques give similar results and therefore, all of those can be employed for Word collocations effectively.
  PubDate: Sun, 25 Dec 2022 00:00:00 +030
- Authors: Mert ERSEN; Ali Hakan BÜYÜKLÜ, Semra ERPOLAT TAŞABAT
  PubDate: Sun, 25 Dec 2022 00:00:00 +030
Analysis of PDB Files and Calculating Similarity Between Two Proteins
- Authors: Emre YILDIRIM; Volkan ALTUNTAŞ
  Abstract: Examining biological structures is a critical process for many broad applications, such as protein engineering, drug design and production. The PDB (Protein Data Bank) file format is the most convenient and popular extension/format in use for the identification, identification of biological molecules such as proteins. In this format, which has a typeface structure, lines indicate a particular molecule or atom and contain technical details such as the name of the atom, the name of its residue, the chain to which it is attached, and its coordinates. Various applications and intermediaries have been developed to extract various details from these files and make sense of them. These developed tools make the study and design of molecular formations difficult. Examples of these technical applications are packages such as ProDy and BioJava. However, these packages may not be easy to integrate into new applications and are often integrated into large codebases for the purpose of examining their simulation on a molecular basis. They also lack techniques for updating/changing and manipulating coordinates. The Python package pdb2sql is used, which can pull files with the PDB extension into the database and present them by making them easier to parse and manipulate. In terms of different solutions, manually writing SQL code is one of the most common solutions used to query a database. However, SQL commands and queries are challenging and confusing for those outside the field and hinder the technical use of the SQL language. Right here, pdb2sql provides end users with SQL queries as an intermediary to use simple Python techniques. However, with the technique used in the article, the positive and effective sides of SQL queries are utilized and it will also be a solution to SQL query difficulties. In addition, several types of reliable classes have been created and used in pdb2sql to rotate biological and molecular structures around their axis of origin, interfere with interface and technical information, and examine structure affinity between two proteins. The result is the Python package pdb2sql; It is a PDB extension file manipulation tool that is lightweight and versatile, easy to modify, design and integrate.
  PubDate: Sun, 25 Dec 2022 00:00:00 +030