M.A. Theses
Permanent URI for this collection
Browse
Recent Submissions
Item Realized volatility forecasting using hybrid neural networks : an application for The Istanbul Stock Exchange(Thesis (M.A.) - Bogazici University. Institute for Graduate Studies in Social Sciences, 2023., 2023) Gültekin, Mehmet Ekin.; Badur, Bertan Yılmaz.Volatility forecasting in the financial markets is important in the areas of risk management and asset pricing, among others. In this study, BIST 100’s 1-day, 5-day, and 10-day-ahead return volatilities are examined. Two types of hybrid models are utilized to improve individual GARCH-family models’ predictions. For the first hybrid model, a group of GARCH- family models is constructed to produce volatility estimates which were then fed into neural network to increase the predictive power. The second hybrid model received GARCH- family models’ specifications instead of volatility estimates as inputs for ANN to conduct the learning process. Hybrid neural networks were also fed a set of exogenous, endogenous, and dummy variables. One of the main conclusions is that both hybrid models increased the forecasting precision of individual GARCH-family models while the second hybrid model provided better volatility forecasts for all error measures used in this study. Equal forecast accuracy test also showed that the hybrid models’ out-of-sample predictions were significantly better than GARCH-family methods. All model performances deteriorated as forecast horizon was extended, although the steepest decline happened for hybrid models rather than the GARCH-family. Lastly, as the complexity of the neural network architecture was increased, the loss measures for the out-of- sample forecasts improved except on the last case where the network overfitted using the highest number of neurons per hidden layer among the searched hyperparameter grid.Item Using social media big data with machine learning to improve customer satisfaction(Thesis (M.A.) - Bogazici University. Institute for Graduate Studies in Social Sciences, 2023., 2023) Demir, Hilal.; Taşkın, Nazım.With the increasing importance of customer relationship management and improving customer support in today's competitive business landscape, there is a growing need to leverage machine learning techniques for gaining insights, forecasts, and better decision-making. Sentiment analysis, in particular, has emerged as a powerful tool for improving customer support services. In this study, we explore the use of three gradient boosting algorithms, XGBoost, CatBoost, and LightGBM, for sentiment classification on Twitter data. We employ ensemble classifications to analyze the sentiment of the data and observe improvements in performance. Our results are compared to other two algorithms that are popularly used in the context of sentiment analysis and show that the ensemble classification of the three algorithms yields the highest accuracy and F1 score. By addressing the gap in understanding how different machine learning algorithms can be used to enhance customer support processes, this research aims to contribute to the improvement of customer satisfaction and loyalty. Specifically, the study aims to improve the accuracy of sentiment classification, thereby enabling businesses to better meet customer expectations for fast and efficient customer support.Item Churn prediction in online payment sector using survival analysis(Thesis (M.A.) - Bogazici University. Institute for Graduate Studies in Social Sciences, 2023., 2023) Özalpay, Gözde.; Taşkın, Nazım.Online payment systems are rapidly becoming a vital component of fintech world. With the entry of numerous firms into the sector, competition has become intense, and customers can easily switch providers without prior notice. As a result, churn has emerged as a critical topic for organizations seeking to remain competitive. The study focuses on the key factors that influence churn in an online payment company, specifically, a well-known payment service provider in Turkiye. One of the other objectives of the study is to find out which churn period and which input variables combination is more suitable with company’s business model and customer’s behavior. To arrive at this conclusion, we have developed a survival model that can answer both questions. We analyzed the customers who had been using the services and identified them as churn and non-churn based on three different methods. We examined various variables that could influence churn, including demographic factors, payment history, and usage patterns, to build three different models with different target variables. The results of these models were compared to determine which variables were most significant in predicting churn and which type of churn period is more suitable to answer the question in hand. All the statistical models exhibited similar performance indicators and variable importance rankings. The findings indicate that commission rate change, refund rate, payment count, volume, merchant type, merchant source name, and merchant sector were significant predictors of customer churn, while settlement period and working area of the merchant were not significantly associated with churn. Moreover, the second model, which defined the churn event as occurring within a one- month period, outperformed the other models. Additionally, the results suggest that the risk of customer churn increases after 25 months of doing business with the company.Item Blockchain-based decision support system for measuring environmental and social sustainability in the supply chain(Thesis (M.A.) - Bogazici University. Institute for Graduate Studies in the Social Sciences, 2022., 2022.) Okay, Can Nilüfer.; Sencer, Aslı.Globalization has caused supply chains to become more complex and has created problems of misinformation, lack of transparency, traceability, and control. There is an inappropriate use of natural resources and exploitation of people and the environment at lower-tier levels of opaque value chains. Companies are pressured to be held accountable for these malpractices at the lower-tier levels of their suppliers. Consumers and governments demand that organizations reveal the environmental and social impacts of their supply chain activities. It is expected from companies to transform their business models to prioritize transparency and the environmental and social sustainability of their operations. Blockchain technology offers the essential properties that can make this transformation possible, it enables transparency, traceability, security, and real-time information sharing across supply chain participants. It has the potential to overcome the challenges in sustainable supply chain management. In this thesis, a quantitative sustainability measurement model for the environmental and social sustainability of supply chains is presented. Subsequently, a blockchain-based decision support system is developed to track and trace products throughout the supply chain and assess the environmental and social sustainability of the products and supply chain actors. Companies can use this system to gain more information about their supply chain activities and measure their sustainability performance.Item Challenges in service catalog management and recommendations for higher success(Thesis (M.A.) - Bogazici University. Institute for Graduate Studies in Social Sciences, 2022., 2022.) Bal, Efsun.; Sencer, Aslı.The service catalog is a priceless resource, providing a single point of access to all of the services that the company provides to external and internal clients through IT and other departments. It delivers several benefits when it works effectively and is structured holistically, including clarity in pricing the services given, cost reduction, operational efficiency, successful service level management, and increased customer satisfaction. Furthermore, because it is linked to so many other processes, unsuccessful applications here have a detrimental impact on many other operations. Despite all its criticality and usefulness, many companies fail to build the service catalog structure successfully. This research aims to identify the challenges companies face while trying to implement and manage service catalogs and, based on their criticality, share some recommendations for better adoption. The stages of service catalog process implementation are grouped under four categories for service identification, service catalog implementation, maintenance, and adoption. Based on the data collected from 98 respondents, the critical challenges in providing higher success in IT service catalog management are identified and recommendations are given for higher success. Accordingly, keeping service catalog up-to-date, identifying the services and service relations, and creating ownership and adoption have been identified as the top three most important challenges for successful service catalog management. Companies that have strategic plan, assign service catalog manager and implement best practices have higher success in SCM.Item Cyber insurance adoption in SMEs as a risk management tool in digitalization(Thesis (M.A.) - Bogazici University. Institute for Graduate Studies in Social Sciences, 2022., 2022.) Özkeleş Yıldırım, Aslı.; Metin, Bilgin.; Taşkın, Nazım.Small – medium sized enterprises (SMEs) create the backbone of the Turkish economy. Digitalization is a key advancement for SMEs in order to create efficiency and open up new opportunities for innovation. However, digitalization makes SMEs vulnerable to cyber threats by opening an outlet to other systems. The lack of awareness of cyber protection and the increasing advancements in cyberattacks puts SMEs at risk of data breaches which in turn causes damage to the company. Cyber insurance is considered a risk management tool for the coverage of costs in the event of an unexpected cyber incident. Even though the coverages are beneficial for the insured, the cyber insurance market is far from reaching its full potential. The study aims to find the factors of cyber insurance adoption for SMEs and the effects of cyber insurance on digitalization through cyber readiness, organizational security performance and information and communication technologies (ICT) adoption. The model created for the study was based on technology-organization-environment (TOE) context extended with individual context and the post adoption effects of cyber insurance. A quantitative survey aimed towards SMEs was conducted to test the model. Methods for increasing adoption of cyber insurance among SMEs were suggested based on the model outcomes.