Department of Computer Engineering

Permanent URI for this collection

http://hdl.handle.net/20.500.11905/1211

Browse

Recent Submissions

Now showing 1 - 20 of 41
  • No Thumbnail Available
    Item
    EVALUATING THE QUALITY ASPECTS OF SQL AND NOSQL DATABASES
    (2022-06-22) ABBAS, Abdulbaset; TOPALLI, Damla
    With the recent advances in technology and the growth of the data to be processed, choosing the suitable databases for your software has become a highly effective element that reflects directly on the overall quality and outcomes of any work. The most popular DBMS are either relational database management systems like SQL or non-relational database management systems like NoSQL; choosing between them at the early stages is essential. In this thesis, the key features of SQL and NoSQL databases have been discussed, considering the eight main software quality attributes affecting the database quality: Availability, Efficiency, Consistency, Durability, Maintainability, Reliability, Scalability, and Recovery Time. This study aims to understand the most effective quality attributes for SQL and NoSQL database implementation. Additionally, the popular DBMS tools for SQL and NoSQL implementation are discussed, considering four tools: MySQL, PostgreSQL, MongoDB, and Redis. To better understand the current implementation preferences, a questionnaire has been conducted with IT professionals: developers, database experts, testers, and managers to understand their preferences on SQL and NoSQL databases from the quality perspective. According to the questionnaire results, the highest important quality attributes for SQL are durability, consistency, and availability whereas, scalability, durability, and efficiency were the most critical quality attributes for NoSQL.
  • No Thumbnail Available
    Item
    AN INVESTIGATION OF THE IMPACT OF DIFFERENT DATA CLEANING TECHNIQUES ON METRIC RESULT QUALITY IN MACHINE LEARNING
    (2022-06-14) ABBAS, Israa Mustafa; TOKER, Sacip
    Enormous growth of data due to e-commerce platforms and online applications has posed a big challenge for data analysis and processing. It is now a frequent practice for e-commerce web sites to enable their customers to write reviews of products that they have purchased. Such reviews provide valuable sources of information on these products. A product review has important data source for sentimental analysis is used in all online product firms. This huge volume of data influence leads to a great challenge. These datasets, however, contain different data’s issues. Typically, different data mining technique used in before deploying data in many cases. Spatially, in supervised machine learning models trained on historical and labelled data to predict unseen data, data that a model has never learned before. In this thesis, we focused on design of experiment study in machine learning too [1]. We applied Ronald Fisher theories [2] regularly to find cause- effect relationship .For carry out this design of experimental study, we chose supervised machine learning classification algorithms with sentimental analysis, it is an approach to natural language processing (NLP).This is a popular way for organizations to determine and categorize opinions about a product, service .It involves the use of data mining, machine learning and artificial intelligence to mine text for sentiment and subjective information [3].This study established with Multinominal Naïve Bays ,Random Forest and Logistic Regression to analysis impact of five experimental groups (duplicate data ,punctuation mark ,stop words, limmatezr, TF-IDF transform ) and compare with one control group (no data cleaning applied). To determine the impact experimental group on three models’ efficiency and classification ratio and explain the interesting observations. A simulation done on 353 projects chosen randomly from Amazon product review dataset from twenty-four different categories . Thus, Dataset was collected from Amazon.com by McAuley and Leskovec [4][5]. After collecting metric dataset, SPSS software used for analyzing. A repeated-measure ANOVA was performed to examine this research question and the descriptive statistics of metric used. Analysis result shows there are different impact for data cleansing on machine learning models performance . data cleaning in same cases impacted positively on Random Forest and negatively in Multinominal Naive Bays and Logistic Regression. In other cases, had no impact at all. In overall, experimental result showed Random Forest classifier more sensitive on data cleaning than Multinominal Naïve Bayes classifier and Logistic Regression classifier ,both algorithms get high classification score in un-cleaned data set. Moreover, the experiment results showed data issues behavior differ in machine learning model. We cannot consider data quality issues as irrelevant data in all machine learning algorithm. Analysis result will be explained in detail on result and discussion chapter 4 and 5.
  • No Thumbnail Available
    Item
    A METHODOLOGICAL APPROACH FOR SERIOUS GAME SOFTWARE DEVELOPMENT: AN APPLICATION FOR LANGUAGE DISORDERS
    (2012-01-25) ÇAĞATAY, Mehmet; EGE, Pınar; ÇAĞILTAY, Nergiz
    The computer software has been actively used in education area in different ways today. However, for several reasons educational institutions are failing to integrate this software to current educational environments. Educational institutions have been criticized for using technologies similar to the ones used hundred years ago. We believe that, one of the reasons for this failure, integration of educational software technologies into current educational environments, is the complexity of these systems. Hence developing efficient software by addressing the real life problems is a complex process. There are various software development methodologies especially for complex software, with regular and planned development processes. So far, these software development methodologies are appropriate for almost all software, though, in terms of unique needs and developmental process of educational software, they may be inadequate. In other words, development of the educational software process requires some other considerations, such as the domain experts that are not considered during the development process of commercial software projects. In this thesis, a new educational software development methodology by involving the domain experts and their interactions with the end users is recommended. Additionally, this software development methodology is used in a serious game development process that supports the therapy process of children with impaired speech and language. Primarily in this study, the contribution of the serious game on the current therapy sessions is evaluated which is developed by using the proposed educational software development methodology. It is aimed to better address the problems of current therapy sessions by developing the software according to the new methodological approach. In other words, this study is a case study to show how the proposed methodology is applied on the development process of a serious game as well as its impact on current therapy sessions.
  • No Thumbnail Available
    Item
    A DATABASE DESIGN METHODOLOGY FOR COMPLEX SYSTEMS
    (2013-07-14) TOPALLI, Damla; ÇAĞILTAY, Nergiz
    The quality of the software is directly related to addressing the users' needs and their level of satisfaction. To reflect user requirements to the software processes, correct design of the database model provides a critical stage during software development. Database design is a fundamental tool for modeling all the requirements related to users' data. The possible faulty conditions in database design have adverse effects on all of the software development processes. The possible faulty conditions can also cause continuous changes in the software and the desired functionality of the targeted system which may result in user dissatisfaction. In this context, reflecting the user requirements accurately in the database model and understanding of the database model correctly by every stakeholder involved in the software development process is the factor that directly affects the success of the software systems. In this study, a two-stage conceptual data modeling approach is proposed to reduce the level of complexity, to improve the understandability of database models and to improve the quality of the software. This study first describes the proposed two-stage conceptual data modeling. Then the proposed method’s impact on software engineers’ comprehension is also investigated and the results are examined. Results of this study show that, the proposed two-stage conceptual modeling approach improves the understanding level of software engineers and eliminates possible defects in this stage.
  • No Thumbnail Available
    Item
    ANALYSIS OF FILTERING AND QUANTIZATION PREPROCESSING STEPS IN IMAGE SEGMENTATION
    (2013-08-14) ÇALAMAN, Seda; KOYUNCU, Murat
    There is a series of processes to extract semantic information from an image and one of them is the image segmentation. Image segmentation splits the image into smaller parts (segments) such that each segment has similar features such as similar colors or textures. In this thesis, the effects of preprocessing methods on image segmentation process are analyzed from different perspectives. Firstly, Peer Group Filtering, which is one of the preprocessing methods used before image segmentation, is applied on the images and its effect on image segmentation is analyzed. Peer Group Filtering algorithm is used to eliminate noises and to smooth color changes on images. Secondly, Lloyd’s quantization algorithm, which is another preprocessing method used before image segmentation, is applied and its contribution on image segmentation is investigated. Lloyd’s quantization algorithm reduces the number of colors in images. Finally, two different segmentation algorithms (fast scanning algorithm and JSEG algorithm) are compared using preprocessed images. Natural and synthetic images have been experimentally tested in this study. The results obviously indicate that after Peer Group Filtering preprocessing, segmentation achievement increases while run time of the segmentation decreases. On the other hand, the experiments related with the quantization show that, selected quantization level is very important to get benefit from Lloyd’s quantization algorithm. If correct quantization level is selected, then quantization helps segmentation process.
  • No Thumbnail Available
    Item
    A COMPARISON OF IMAGE DETECTION ALGORITHMS YOLO AND FASTER R-CNN IN DIFFERENT CONDITIONS
    (2022-06-13) ABDULGHANI, ABDULGHANI MAWLOOD A.GHANI; Dalveren, Gonca Gökçe Menekşe
    In this thesis, we compare YOLOv4 with YOLOv3 and Faster R-CNN in terms of better object detection in both challenging weather conditions and darkness. Moving objects such pedestrians, cars, buses and motorcycles can be difficult to detect in rainy, foggy and snowy weather conditions or even at night. This study is aimed at evaluating the three modules to determine which perform best in such circumstances, bearing in mind that none of them was initially intended to perform in bad weather conditions or at night. This Study is done by utilizing Tesla P4 GPU, with 12GB RAM. We trained these algorithms with an Open-Image dataset, where YOLOv4 has scored the best results at 40,000 iterations, 72 mAP, and 0.63 Recall. On the other hand, YOLOv3 has scored maximum at 36000 iterations, 65.53 mAP, and 0.54 Recall. Finally, Faster R-CNN scored 36,000 iterations, 51 mAP, and 0.49 Recall. In terms of detection performance evaluation, YOLOv4 performed at 42 FPS, while YOLOv3 was at 37 FPS and Faster R-CNN at 10 FPS in video with 30 FPS. Based on the results, YOLOv4 has performed the best in comparison to YOLOv3 and Faster R-CNN.
  • No Thumbnail Available
    Item
    SITUATIONAL METHOD ENGINEERING FOR REQUIREMENT ENGINEERING PHASE
    (2009-05-30) AYDIN, Seçil; MISHRA, Deepti
    This thesis focuses on requirements engineering phase and reviews the existing requirement engineering methods and compares them according to the constraints in the software projects. It has been found that some techniques are better suited to particular project teams and circumstances. Besides, methods are normally general in nature and they can not be used directly without adapting them according to the characteristics of the project. This is the concern of situational method engineering, where the term situational method is used to refer to a method tailored to the needs of a particular development setting. A criterion methodology is established to distinguish requirement engineering methods from each other according to different characteristics of the project. A tool is implemented to store different methods according to this criterion methodology by using Situational Method Engineering. This tool is compared with other tools that exist in the literature. The tool is published and validated by collecting data from the industry. The results gathered from the industry are presented and discussed for the improvement of proposed approach.
  • No Thumbnail Available
    Item
    COMPARISON OF VARIOUS TRANSITION MECHANISMS FROM IPv4 TO IPv6
    (2015-06-25) AL-FAYYADH, Faris; KOYUNCU, Murat
    IPv4 which is the old version of Internet Protocol has a new successor named IP Next Generation (IPng) or IPv6 developed by IETF. This new version is developed especially to resolve some issues of IPv4 like scalability, performance and reliability. Although new version is ready for usage, it is obvious that it will take years to transit fully from IPv4 to IPv6. We have to use these two versions together for a long time. Therefore, we have to investigate and know transition mechanisms that we can use during transition period to achieve a transition with minimum problem. This thesis analyzes present IP transition techniques. Here an attempt has also been made to make empirical evaluation of the three most generally used transition mechanisms which are Automatic 6to4 Tunneling, Manual 6in4 Tunneling and Dual Stack. The obtained test results are also compared with the results of native IPv6 and native IPv4 environments. Empirical evaluation is based on simulations which are carried out using OPNET simulation framework. The outcomes of the thesis are important for providing an insight for choosing an appropriate transition technique, an idea about network capacity planning and migration.
  • No Thumbnail Available
    Item
    COMPARISON OF QUALITY OF SERVICE (QoS) SUPPORTS IN IPV4 AND IPV6
    (2015-06-25) AL-FAYYADH, Hayder; KOYUNCU, Murat
    Providing guaranteed services in the current Internet has become extremely essential to fulfill the requirements of Internet users. With an increase in the number of users and demand for multimedia applications like video streaming, VoIP and video conferencing, larger bandwidth requirement increases drastically since such applications are very sensitive to delay, packet loss, and jitter. However, it is not feasible to provide enough bandwidth to satisfy such a high demand. Quality of Service (QoS) is an important network performance parameter having a significant impact on multimedia applications. IPv6 was designed to improve QoS supported by IPv4, as well as other improvements. In this thesis, we discuss quality of service and various parameters affecting it on the Internet. We compare IPv4 and IPv6 protocols in terms of their performance for multimedia applications. In this scope, various queuing algorithms (First in First Out FIFO, Priority Queuing-PQ, and Weighted Fair Queuing-WFQ), which are typical scheduling algorithms used by routers, are examined to see their effects for multimedia applications in IPv4 and IPv6 environments. In addition, integrated services (IntServ) and differentiated services (DiffServ), which are two important techniques developed to provide QoS on the Internet, are elaborated again in IPv4 and IPv6 environments. For comparisons, a simulation framework based on OPNET Modeler is used for modeling, simulating and analyzing network behaviors. The results obtained from the simulations clearly show that the use of IPv6 with different QoS techniques has better performance in terms of jitter and delay compared to IPv4 protocol.
  • No Thumbnail Available
    Item
    HUMAN IDENTIFICATION AND VERIFICATION BY HAND GEOMETRY INFORMATION
    (2022-03-10) SHAKIR, MUSTAFA KANAAN; ŞENGÜL, Gökhan
    In this thesis, a hand geometry-based human identification system is proposed. The hand is a vital component of the human body. It consists of many unique features that can be used for human identification and verification systems. This study presents an approach for recognizing the human using the features extracted from hand images. The proposed method is implemented in three steps, namely preprocessing, feature extraction and classification phases. In the preprocessing step, the hand images are resized for the feature extraction model. In the feature extraction phase, the convolutional neural network (AlexNet model) is used in order to extract hand features. The extracted features are classified using the well*known Support Vector Machines (SVM) and k-nearest-neighbourhood classifiers. The proposed method is tested on the CASIA-MS-Palmprint dataset, using a different number of training and testing images. in used for the hand geometry-based recognition system. The average accuracy, sensitivity, and specificity were 94.36, 89.96 and 90.36. We conclude that the recognition accuracy rate is reasonable when the system is trained with an adequate number of images.
  • No Thumbnail Available
    Item
    DESIGN AND IMPLEMENTATION OF IOT-BASED REFERENCE ARCHITECTURE FOR SMART HOMES
    (2022-03-10) Bello, Abdussamad; KARAKAYA, Ziya
    Interoperability, which is the ability of applications, systems, devices, and services to work together reliably and consistently, is one of the most important problems of current smart home systems. The lack of interoperability and the accompanying technological partitioning (that is, managing different devices in the same environment with different applications) creates a major barrier for smart home developers and resistance to consumer acceptance of IoT products in their home environments. The aim of this study is to create a IoT-based Smart Home Reference Architecture that will manage, monitor, and process the data coming from devices and services in the smart home environment while increasing the interoperability. This study also presents a proof-of-concept solution for the proposed reference architecture using FIWARE components. In conducting this study, we made use of constructive research methodology and the ProSA-RA process to analyze existing works, their underlying issues, and then propose solutions to them. To further bridge the gaps in the smart home ecosystem, it is suggested that smart home developers adopt the reference architecture proposed in this study while developing their architectures, applications, and devices.
  • No Thumbnail Available
    Item
    AUDITOR TECHNOLOGY AND PRIVACY CONTROL TO SECURE E-LEARNING INFORMATION ON CLOUD STORAGE
    (2017-01-07) AL-KHAFAJI, Khalid Muhammad Kareem; ERYILMAZ, MELTEM
    The aim of this study, proposal to establish and applying a safe mechanism is: Technique Relationship Protected (TRP), that use a privacy control with the Auditor on the information shared between two of the most important services offered by modern technology, namely the E-learning platforms and service of cloud storage, as a matter of fact these technologies required relevant features in the databases made up of information that need safety. This study aims to shed light on the concept of ensuring privacy to protect information shared between the E-learning system and Cloud platform by proposing a mechanism to preserve the privacy of quite distinctive that supports public scrutiny on the information shared between these important technologies. It is worth mentioning that the study has a tendency to need to pick advantage of looping the signatures to verify the authenticity of the information required to review the validity the information shared encrypted. With our mechanism, entity location on each block in the joint information is unbroken figure of the year investigators, administrative body is able to check the efficiency of the safety of the common information, without retrieving the entire file. The proposed system keep a privacy of auditing mechanism for shared information with cloud keeping-safe. With use AES (Advanced Encryption Standard) for protecting the shared information by applying an encryption mechanism on the information that is being uploaded from its owners. They have been certified by the E-learning system administrators, and the application of a decrypt the encrypted information after user's requested, they trusted users has E-learning system after sponsorship by the owners of the information they need. We have to take advantage of ring autographs to build homomorphism authenticators so that a general verifier has the capacity to audit shared information integrity without retrieving the complete data or access to content. Addition to it cannot recognize who's the signer on file, this means that the power to administrator to reveal the identity of the signer. As well as the presence of additional functions in the work of the Auditor that contribute to the consolidation of this technique relationship protected. This is for the protect and the success of that relationship between the technical services which has become an urgent necessity in the modern technical world that contains a very massive amount of data and information those shared and transmitted daily between branches of our technological world.
  • No Thumbnail Available
    Item
    IRIS RECOGNITION BY USING IMAGE PROCESSING TECHNIQUES
    (2017-01-07) ALHAMROUNI, MOHAMED; Şengül, Gökhan
    Iris recognition system has become very important, especially in the field of security, because it provides high reliability. Many researchers have suggested new methods to iris recognition system in order to increase the efficiency of the system. In this thesis, various methods have been proposed to achieve high performance in iris recognition. In the proposed system, three feature extraction approaches, Histogram of Oriented Gradient (HOG), Gray Level Co-Occurrence Matrix (GLCM) and Local Binary Pattern (LBP) are used to extract the features from iris image. On other hand, two classifiers; K Nearest Neighbors (KNN) and Support Vector Machine (SVM) are used in the classification stage. The iris image passes through several stages before extracting features stage; first, pre-processing stage which includes image resizing that unifies all images' size, second, segmentation stage which determines the iris region in eye image, finally, normalization stage which converts the iris region to suitable shape with specific dimensions. The proposed methods have been applied on two iris databases, UPOL and IITD. However, the proposed system achieved recognition rate of 100% when HOG+KNN method is used.
  • No Thumbnail Available
    Item
    EMOTION ESTIMATION FROM FACIAL IMAGES
    (2017-01-07) NAJAH, GOMA MOHAMED SALEM; Şengül, Gökhan
    Prediction of emotions from facial images is one of the popular and active researches, and it’s implemented via many methods. In this thesis, the proposed system to predict emotions from facial expressions images contains several stages, first stage of this system is the pre-processing stage which is applied by detecting the face in images, then resizing the images, and then Histogram Equalization (HE) technique is applied to normalize the effects of illumination. The second stage is extracting features from facial expressions images using Histogram of Oriented Gradient (HOG), and Local Binary Pattern (LBP) feature extraction algorithms, which generates the training dataset and the testing dataset that contains expressions of Anger, Contempt, Disgust, Embarrass, Fear, Happy, Neutral, Pride, Sad, and Surprised. Then Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) classifiers are used for the classification stage in order to predict the emotion. In addition, Confusion Matrix (CM) technique is used to evaluate the performance of these classifiers. The proposed system is tested on JAFFE, KDEF, MUG, WSEFEP, TFEID and ADFES databases. However, the proposed system achieved prediction rate of 96.13% when HOG+SVM method is used.
  • No Thumbnail Available
    Item
    AN ADAPTIVE EDUCATIONAL MODEL FOR FLIPPED CLASSROOM
    (2017-03-07) Ahmed, Aisha Abdulaali Abdulla; ERYILMAZ, MELTEM
    This study aimed to develop a flipped classroom demonstrate utilizing versatile innovations for primary school understudies and identify the individual contrasts among the third-grade basic understudies in the English Language in Libya by a versatile method in flipped learning, and customary training at three levels recalling, comprehension and applying of Bloom's Taxonomy independently. This study attempted to answer the following question: Are there any differences between traditional style of education, flipped learning , and adaptive technique in flipped learning in achievement tests according to Bloom's Taxonomy ( Remembering, Understanding, and Applying) for understudies in the third grade of essential in English ? To accomplish the targets of the review and answer its question, three tests were constructed and afterward ensure its earnestness and steadiness in proper ways, and selected the study sample and divided it randomly into three gatherings:- 1. The experimental group (1) educated by adaptive technique in flipped learning. 2. The control group (1) instructed through customary training. 3. The control group (2) (The trial bunch (2)) instructed by flipped learning. The review presumed that in the pre-test the gatherings were homogeneous; however, in midterm test and post- test, there were contrasts measurably huge for the experimental assembly (1).
  • No Thumbnail Available
    Item
    THE ROLE OF USING MOBILE SOCIAL MEDIA LEARNING IN LIBYAN HIGHER EDUCATION
    (2017-05-02) Alhadad, Salha; Ertürk, Korhan Levent
    The use of social media (e.g., Facebook, Twitter, YouTube, Google+) by smartphones and the using of mobile learning applications in education increased day by day, and this reflects the importance of information and communication technologies ICT and its active role in the development of methods and tools of education, and enhance learning among students as well as promote active learning for them, where they have become the most important technology education tools. Mobile apps and the use of social media created opportunities for interaction and collaboration among students, as well as allowed students to engage in creating content and communication by using social media, Web 2.0 tools and mobile web 2.0. In this quantitative study was employed utilizing survey method to presents a portion of the findings on students' perceptions of learning with mobile devices and the role of social media in Libyan higher education. Data were collected through two surveys. In this study was designed mobile learning application based on the results of the surveys, to introduce the concept of Mobile social media learning to the Libyan Higher Education to develop its tools and old methods. The initial experimental results are very positive and this reflects the importance of the mobile applications and social media in developing and increasing the educational performance of students.
  • No Thumbnail Available
    Item
    THE AWARENESS TO CREATE THE DIGITAL RESOURCE LIBRARY BASED ON RESPONSIVE WEBSITE DESIGN
    (2017-05-02) Elbaraasi, Asma; Ertürk, Korhan Levent
    This research study analyzed the awareness of digital resource library based on the responsive web design. The background of the present study is signifying that digital resource library has been significant impact on the user experience and enhances the knowledge and opportunity to get more information from anytime and anywhere. The rapid development and growth in the online website the importance of responsive web design provides the significant impact on the user to get access on the digital library through several kinds of IT devices such as PCs, tablet, mobile and others. A quantitative research methodology was used by researcher to find out the research questions answers. This present study research analysis projects lead several research advantages such as responsive web design, digital devices, user interface, display format, search features, storage, processing, flexibility, availability and reduce cost. While digital library system at Higher Institute of Education is moderately taking place in the academic system, development, library research and activities are most effective elements of education. The present digital library system also includes images, video, maps, audio, academic resources, documents, personal record and combination of other relevant material. The study shows that RWD significantly address the challenges of in today’s web development and effectiveness for the digital resources library in order to improve the user experience more effectively and efficiently.
  • No Thumbnail Available
    Item
    THE EFFECT OF PAIR PROGRAMMING ON THE UNDERSTANDABILITY OF FLOWCHARTS: A CASE STUDY IN A C PROGRAMMING COURSE
    (2022-03-01) Khudhur, Omar Mohammed; Topallı, Damla
    Constructing algorithms is one of the most important concepts to be learned for the first-year computer science students in learning programming. However, it is observed that most students find it challenging to construct algorithms and represent it by using a design, such as flowcharts. Based on this issue, in this thesis, it is aimed to investigate the effect of pair programming approach on the understandability of flowcharts and evaluate the progress of CS1 students in a C programming course on this topic. Pair programming is a technique which enables students to interact and communicate with their peers when working on a code or design project. Accordingly, in this thesis, an experimental study is conducted with two groups: solo and pair to better understand the effect of pair programming on defect detection performance while working on the flowcharts. Secondly, it is investigated if pair and solo groups’ performance can be classified by using classification algorithms. Finally, both groups’ progress in the course regarding the flowcharts is compared and analyzed based on a pre-test and post-test experimental design. The results of the study reveal that the pair group detects more defects correctly on the flowchart when compared to the solo group. To differentiate between solo and pair groups, after applying feature selection methods and classification algorithms, the highest accuracy obtained is approximately 70% when the decision tree (J48) and rule-based PART algorithms are applied, which is considerably low. According to the pre-test and post-test analysis results, at the beginning, there is no significant difference between those groups, but post-test results reveal that pair groups progress is higher considering the exam scores compared to the solo groups. These results may provide insight that adopting pair programming in programming courses can increase students’ motivation and success in constructing algorithms and learning programming. In future studies, it is expected that the use of this technique for software and IT companies and the creation of effective paired groups that can work in synchronization may increase the efficiency of the projects.
  • No Thumbnail Available
    Item
    A NEW METHOD FOR SOFTWARE DEFECT PREDICTION BASED ON OPTIMIZED MACHINE LEARNING TECHNIQUES
    (2022-03-01) HASSEN, SHAHO ISMAEL HASSEN; YAZICI, Ali; MISHRA, Alok
    In this thesis a novel and robust heuristic driven neuro-computing model was developed for software defect prediction. Unlike other classical machine learning models, neuro-computing, especially Levenberg Marquardt Neural Network (LM ANN), is considered to be more robust in terms of adaptive learning, which can be vital towards non-linear feature learning and hence defect data. However, similar to the other machine learning models, the likelihood of local minima and convergence could not be avoided due to exceedingly high weight estimation for 17 input features. Considering this fact, this research contributed a novel improved genetic algorithm, say heuristic model was developed to assist ANN for adaptive weight estimation and update during learning. Here, the key purpose of heuristic model was to help LM-ANN gaining superior weight estimation, update and hence learning without undergoing any local minima and convergence problem. This as a result helped the proposed neuro computing model to achieve higher accuracy than the classical neural network over targeted software fault datasets. In addition to the classifier or machine learning improvement, in this research the focus was made on feature engineering as well that helped alleviating any probability of class imbalance, over-fitting and convergence.
  • No Thumbnail Available
    Item
    INVESTIGATING THE USAGE OF MOBILE APPLICATIONS VERSUS WEB BROWSERS FOR SMARTPHONES
    (2022-02-28) OTHMAN, Safa; KOYUNCU, Murat
    The utilization of mobile applications and services by the smartphone users is ever increasing over the last two decades. A considerable amount of research work has been done in this field with different aims and objectives. The aim of the work done in this thesis is to understand the preferences of smartphone users regarding the usage of the Internet in their devices either by mobile applications or via mobile websites. Our findings will be useful for mobile application developers to understand the main aspects of the average Turkish consumers and what they expect from their mobile usage experience. The study was conducted in two phases. In the first phase, a survey was conducted with 158 participants from Atilim University, Ankara. In the second phase, interviews were conducted with 30 participants based on testing two implemented applications (a native mobile application and a mobile website). The collected data is analyzed using different statistical methods such as One Way ANOVA. Obtained results show that mobile applications are mostly preferred, especially for the most common and frequently used applications among users. Moreover, respondents find that mobile applications are better in most aspects. However, we cannot ignore the proportion of participants who make their preferences according to sites.