M.S. Theses

Permanent URI for this collection

https://digitalarchive.library.bogazici.edu.tr/handle/123456789/11649

Browse

Now showing 1 - 20 of 386

The maximum acyclic matching problem
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Çelebi, Cemre.; Aşıcı, Tınaz Ekim.
The aim of this thesis is to develop exact and heuristic methods to solve the Maximum Acyclic Matching problem, which deals with obtaining maximum matching such that the subgraph induced by saturated vertices is acyclic. The maximum matching problem tries to find the most extensive possible matching set. it is a well-studied problem that can be solved with combinatorial algorithms. However, for the maximum acyclic matching problem, we are searching for not only a maximum size matching but also we require that the subgraph induced by saturated vertices does not contain any cycles. For this purpose, an additional acyclicity constraint is needed. Even though some exact and approximate algorithms are established for particular graph classes, there is no such algorithm applicable for general graphs to find a maximum acyclic matching. In this study, four algorithms are suggested. Randomly generated graphs with different density levels and sizes are used to analyze their performance. Two of the algorithms are exact algorithms, which are extensive and cutting plane formulations. Based on experimental results, the cutting plane approach performs better since it works also with larger graphs. The other two algorithms are heuristics, which are modification and construction approaches. It is observed that the construction approach performs better than the modification approach in terms of both time efficiency and quality. The construction approach yields feasible and close-to-optimal results in a shorter period. When the results of the best exact and heuristics are compared, the cutting plane algorithm performs better based on optimality. However, it is not applicable for large graphs compared to the construction algorithm. Additionally, it is seen that the effect of acyclicity constraint is increasing while graph size and density are getting larger.
Dynamics of long-term essential hypertension and pharmacotherapy options
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Hasgül, Zeynep.; Yücel, Gönenç.
Hypertension is a preventable and treatable disease however; blood pressure control remains insufficient in many countries. An incomplete understanding of the highly complex mechanism of blood pressure regulation is one of the reasons for poor hypertension management. Physiological differentiation in individuals causes optimal antihypertensive treatment strategies to vary for patients. Consequently, instead of one fits all treatments, personalized treatments can offer improved results on therapeutic goals. Due to the sheer number of feedback loops and interconnected mechanisms, we developed and validated a system dynamics model for understanding and interpreting relationships within hypertension treatment. We aim to demonstrate the use of simulation modeling to study inter-patient physiological variations and their effects on hypertension treatment outcomes. The model includes the central nervous system and renal control mechanisms, vascular elasticity, and kidney function, capturing long-term dynamics of blood pressure. Three pathogenesis mechanisms of primary hypertension (the over-active nervous system, the over-active renin-angiotensin-aldosterone system, and low sodium excretion) are examined to represent different patient profiles. Later, treatment outcomes are analyzed on the identified patient profiles by comparing the decline in kidney and vascular functions under various treatment options. This study presented the potential benefit of using dynamic simulation models to compare treatment options for patients with physiological differences. Even though obtaining granular empirical data is challenging, they can be used for calibrating simulation models such as these to get closer to individualized treatments in clinical settings. As a future study, cost-effectiveness and value of information aspects can be incorporated to assess the benefit of personalized treatments.
Money laundering detection in cryptocurrency networks
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Erdem, Elif Emine.; Ekim, Tınaz.
This study aims to develop scalable methods to detect suspicious wallets using historical transaction data in cryptocurrency networks such as Ethereum and Bitcoin. Different transaction networks are generated for each wallet data set using the illicit wallets dispersed around the internet. Egonet-dependent and independent features are used with a range of machine learning techniques, including logistic regression (LR), random forest (RF), and XGBoost (XGB), to predict illicit wallets. Firstly, we analyze performance of models to detect suspicious wallets in the two datasets that include suspicious bitcoin mixer services wallets such as Bitcoinfog and Helix. The area under the ROC curve value (AUC) is over 99% for XGB models. We observe that models perform better on Helix wallets than BitcoinFog wallets in terms of precision, recall, f1 score, and AUC. Secondly, we notice that egonet dependent features do not significantly improve the models’ performances. Hence, best- performing models have only egonet independent features. Thirdly, on Bitcoin datasets that do not use any mixer services, we obtain over 99% AUC. Although the performance of the models is similar in these three datasets, dominant features in terms of feature importance measure are different between the datasets including wallets using mixer services (Helix, Bitcoinfog) and the other (Bitcoin). Lastly, utilizing the same feature set as we do on Bitcoin, Bitcoinfog and Helix datasets, we train the same machine learning models on the Ethereum dataset and obtain 96% AUC. We repeated the tests with varying degrees of class imbalance to simulate real-life situations. We observe a decline in AUC up to 0.10 together with the increasing severity of the class imbalance.
Air cargo revenue management spot allocation problem
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Bütün, Yahya Umut.; Taşkın, Zeki Caner.
The focus of Air Cargo Revenue Management (ACRM) is to best estimate cargo capacity, forecast future demand, and take accept or reject decisions on the bookings accordingly. ACRM is a different problem than passenger revenue management due to uncertainty of cargo capacity, business, operations, and cargo booking behavior. These factors add additional complexity to a problem and make traditional revenue management approaches inadequate. Certain additional models need to be developed to solve the ACRM problem. The purposes of this thesis are to discuss the processes of air cargo revenue management and develop a spot allocation model. In the thesis, we develop a spot allocation optimization model. In necessary booking control conditions, this model is solved repetitively to decide on allocating expected demand to cargo capacity. A simulation study is performed after the optimization model to compare the results of our optimization model with a commonly used heuristic, First-Come First-Served, under defined scenarios and other test problem settings. Finally, we conclude that our model performed better than 26 out of 27 scenarios according to t-test statistics with a 95% confidence level.
A comprehensive dynamic model of cyclic neuropenia
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Işık, Yusuf.; Yaşarcan, Hakan.
Neutropenia is a hematological disorder that is defined as having a low level of neutrophils in the bloodstream. Low levels of absolute neutrophil counts leave the body defenseless against and vulnerable to infections. Cyclic neutropenia is a type of neutropenia that is described as the oscillations observed in the level of blood neutrophils. The disorder is mostly treated with a cytokine named recombinant granulocyte colony– stimulating factor, rG–CSF, which is administered via injection. A delicate injection schedule is called for because the treatment procedure is costly. However, treatment experiments on an actual patient require frequent sampling from bone marrow and blood, which simply cannot be allowed as it can be detrimental to the health of the patient. Therefore, modeling is a must to carry out treatment experiments. Accordingly, the main motivation in this thesis is to construct a comprehensive dynamic model of cyclic neutropenia. As human physiology is rich in dynamic complexities, system dynamics is selected as the primary methodology. We first construct a model that represents the regulatory structures of neutrophil production for a healthy person. After validating the model, the neutrophil dynamics of a cyclic neutropenia patient is obtained by simply changing the parameter values, but without changing the model structure. Neutrophil production deficiency is the most mentioned cause of cyclic neutropenia in the literature, which is also confirmed in our study. According to our simulation results, the clearance of the apoptotic neutrophils of CN patients takes longer than normal and apoptotic neutrophils can suppress both the production and effects of G– CSF. As a result of experiments with pathogens, we claim that the oscillatory behavior is a characteristic of the neutrophil–GCSF–pathogen system even for a healthy person. This may shed some light on the periodic symptoms observed in patients with diseases caused by an overactive immune system. We experiment with rG–CSF injections too.
Analysis of inventory policies under estimation of demand parameters
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Bozkaya, İrem Cansu.; Güllü, Refik.
The aim of this thesis is to analyze the estimation performance for different characteristics of data in different supply chain systems. In this study, we analyze two different systems. The first section of this thesis covers the parameter estimation problem of a single retailer who observes price-dependent uncertain demand in additive form. To estimate the parameters of the demand, we use linear regression. Additionally, we jointly optimize the price and order quantity. We present a numerical analysis where parameter estimations are calculated for different experiment sets. We analyze the estimations for the number of observations, coefficient of variation, price, and profit margin. We show that as the number of observations increases and the coefficient of variation decreases, the accuracy of our estimations increases. The real optimal price of the data sets does not indicate a pattern of improvement and the profit margin of the product is observed to have no significant effect on the estimation performance. In the second section, we introduce a two-echelon system observing identical and non-identical exponentially distributed demand rates. The warehouse makes the allocation decision of its limited inventory based on the past demand data of retailers. The capacity constraint of the warehouse creates a conditional expected profit function. To find the expected profit, we propose a simulation method where the inventory allocation is repeated multiple times. To evaluate the performance of our method, we introduce experiment sets. Results show that for the systems with identical and non-identical demand rates, the percent loss in expected profit is smaller when the system utilization and the profit margin are larger. For the latter system, we analyze the effect of the ratio between the demand rates of two retailers. The results show no significant effect of this ratio on the percent loss in expected profit.
Multi-level production planning for multi-location chemical company
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Gökçe, Emre.; Ünal, Ali Tamer.
We investigate the decision-making process at a chemical company which is a major soda manufacturer having three production facilities in different countries. The company is one of the top ten largest soda producers in the world and serves as a major supplier to the global market with sales to over 50 countries. Complex sales and operation structures of company require synchronization of optimized decisions. Our study focuses on optimizing decisions from various levels which are strategic, tactical, operational and synchronizing them within and among organizations. In order to reach this purpose, we designed an integrated solution platform based on various optimization models. We fully implemented and deployed the solution platform, which has been used since 2020.
Joint replenishment and pricing of a single product under exchange rate uncertainty
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Koyucan, Gizem Güneş.; Bilgiç, Taner, 1965- .
Purchasing cost uncertainty in the future subject to exchange rate fluctuation is modeled as a markov chain transition matrix, and it is combined with supply chain profit maximization problem. This problem is modeled as a multi-period stochastic inventory control problem on USD/TRY dataset. Replenishment problem is considered under the myopic and dynamic inventory policies. Excess demand is lost, and salvage cost is zero. The procedures to compute order up to inventory levels of both inventory policies are determined. It is verified that (1-P) value, which is an indicator of myopic solution effectiveness, shows the closeness of the dynamic and myopic inventory policies. Average profit is computed with a simulation which includes multi-period purchasing and selling steps. Demand is taken as a random variable with gamma distribution, since it can take only positive values. Price dependent demand is also evaluated. Moreover, the effect of variance in demand on both of order up to inventory level and average profit is analyzed. It is seen that the variance in demand increases the volatility in order up to inventory levels with respect to purchasing costs, and average profits. Optimal inventory level and pricing are also assessed together after replenishment problem is evaluated. Two different pricing systems are used, namely best constant pricing and best pricing. Best constant pricing represents that price is announced before purchasing cost is determined and it cannot be changed period by period. Best pricing represents that price can be updated with respect to purchasing cost in each period. A procedure to find out optimal order up to inventory level and best price combination is formed to handle exchange rate uncertainty. It is observed that purchasing cost volatility promotes the best pricing to maximize profit.
Integer programming formulations and cutting plane algorithms for the maximum selective tree problem
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Onar, Ömer Burak.; Ekim, Tınaz.; Taşkın, Zeki Caner.
This thesis considers the Maximum Selective Tree Problem (MSelTP) as a gen eralization of the Maximum Induced Tree problem. Given an undirected graph with a partition of its vertex set into clusters, MSelTP aims to choose the maximum number of vertices such that at most one vertex per cluster is selected and the graph induced by the selected vertices is a tree. To the best of our knowledge, MSelTP has not been studied before although several related optimization problems have been investigated in the literature. We propose two mixed integer programming formulations for MSelTP; one based on connectivity constraints, the other based on cycle elimination constraints. In addition, we develop two exact cutting plane procedures to solve the problem to op timality. On graphs with up to 25 clusters, up to 250 vertices, and varying densities, we conduct computational experiments to compare the results of two solution procedures with solving a compact integer programming formulation of MSelTP. Our experiments indicate that the algorithm CPAXnY outperforms the other procedures overall except for graphs with low density and large cluster size, and that the algorithm CPAX yields better results in terms of the average time of instances optimally solved and the overall average time.
Designing an efficient storage location assignment approach for through-flow warehouses
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Görür, Fırat.; Yücel, Gönenç.; Aras, Necati.
Order placement and retrieval operations in a warehouse with separated incom ing and outgoing storage handling areas are analyzed. An approach, which consists of a model and a heuristic algorithm, is devised in order to determine how stock keeping units could be assigned to the storage locations so that the total distance traveled is optimized. All necessary data for the model are generated. Because of the complexity of the problem, a new model with a reformulated objective function, which is derived from number of occurrences in the same SKUs entries and retrievals, is defined to be minimized instead of the SLAP itself. In the objective function; movements between storage handling area of the incoming SKUs and storage locations during the place ment operations, movements between the storage handling area of outgoing SKUs and storage locations during the retrieval operations, and movements between the storage locations during the picking operations are included as three components. After model ing, the problem is solved by a heuristic algorithm. Computational results are obtained by 42 different scenarios. The results reveal that good improvements are obtained by the approach, especially when there is no capacity constraint for stock handlers.
Option pricing under stochastic interest rate
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Işık, Mert.; Güllü, Refik.
This study proposes a newly generated model to price options under a stochastic interest rate environment. The developed model introduces a numerical procedure with the arbitrage-free condition by working on a binomial tree. The model handles price changes in stocks and interest rates. Principally, it is assumed that the movements in the stock prices are defined by the Cox, Ross, Rubinstein (CRR) model. The CRR model proposes a numerical method to price options by assuming the interest rate is constant throughout the option's life. Moreover, the thesis claims that interest rates vary based on the Black, Derman, Toy (BDT) model. The BDT model defines the evolution of interest rates in the future. It presents a numerical procedure by using the binomial tree. Crucially, the interest rate is log-normally distributed in the BDT model; hence, the short rate cannot take a negative value. Also, the BDT model assumes that it has a mean-reverting property which means that the interest rate shows a tendency to converge to the average of interest rates in the long term. Additionally, this study utilizes the CRR and BDT model in order to derive a new option valuation framework. Also, the proposed model gives a numerical solution rather than an analytical formula due to the BDT model's structure. This thesis focuses on pricing the European options that can expire only on the maturity date. Furthermore, a group of options with different strike prices and different maturities is valued according to the developed model and the CRR model to observe interest rate impacts under two parameters: strike price and time-to-maturity. Finally, the estimated prices by both models are compared with actual-market prices to determine the accuracies of the models. Then, it is detected that the effect of the stochastic interest rate behavior on which maturities is significant.
Deep learning approaches for multi-site wind power forecasting in the West Aegean region of Turkey
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Sarıkaya, Mert.; Baydoğan, Mustafa Gökçe.
Wind energy is one of the most economical and promising ways of producing renewable energy among today's technologies. But the uncertainties arising from the chaotic nature and the variability of weather events are major problems to the market stabilization and regular maintenance of the wind power systems. Wind farms are usually built in high wind speed potential areas. West Marmara and Aegean regions have the highest number of wind farms in Turkey. Forecasting the wind power production in a region is usually done separately for each wind farm, but forecasting in multi-site context contains more spatial information, thus enables learning from the neighbor wind plants. In this thesis, incorporation of multi- site spatial information, besides the temporal information, to deep learning models is studied. Six alternative deep learning methods, i.e. multilayer perceptrons, recurrent networks, graph neural networks, convolutional networks and their variants are implemented for this purpose. Each model is enhanced with numerical weather predictions to create more accurate long term forecasts, and model parameters are tuned with a hyperparameter optimization. Finally, these models are compared with tree- based boosting, penalized regression and persistence benchmarks over a one-year period. In order to investigate the positive effect of using multi-site approach, recurrent model is trained both separately for each plant and for all the plants at the same time in a multi-site context. Single plant based recurrent model performed better than the multi-site recurrent model, but methods using convolutional layers, significantly outperforms single recurrent, benchmarks and remaining deep learning models.
Routing in trucking service network design problem with transsipments and driver regulations
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Özgürbüz, Ekin.; Bilge, Ümit.
Most of the large transportation companies are currently trying to have the scale to operate their middle-mile or long-distance transportation plans, while using both their resources and outsourced vehicles. The aim is to deliver customer shipments from stocking origins to last-mile delivery transfer stations. Freight transportation planning is done manually in most logistics companies, often leading to poor transportation plans or plans whose quality is hard to assess. Furthermore, since transportation plans are needed almost every day, it is crucial to have a solution procedure that is as effi cient as possible. We have tried to model this environment without loss of generality while developing a methodology to solve this problem efficiently. This is an operational problem that falls into the domain of trucking service network design and operation. We offer a mathematical model and a solution methodology for this problem taking consolidation, transshipment, synchronization, and driver regulations into considera tion. Our model employs a pre-processing strategy for creating the data representing the transportation environment. Moreover, a matheuristic is presented for considering multiple periods of operation. The proposed models can be used for companies oper ating both with their own and rental vehicles. Distribution of the vehicles throughout the transportation network is crucial for the companies that operate with their own vehicles, since the aim is to plan over a horizon, rather than a single day. The assump tions and the parameters used in the models are based on data coming from already operating large companies. The benefit of using this mathematical model is to achieve a better transportation plan with a reasonable amount of time.
Dynamics and management of vector-borne viral epidemics
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Durak, Gizem.; Yaşarcan, Hakan.; Barlas, Yaman.
In this thesis, we model the dynamics of vector-borne viral epidemics. We choose dengue fever as the specific viral epidemic for the study, which is a mosquito-borne viral infection caused by the dengue virus. Dengue virus relies on the human-vector interaction to spread; thus, we need to model a host-vector system. We use system dynamics methodology and construct a dynamic model. We choose Rio de Janeiro as the area. We obtain the average parameter values from the relevant literature or use our close estimations. We observe that the number of infectious human and infected vectors converge to zero in the model, which means the virus decays and eradicates. However, the viral epidemic of dengue is persistent in the region. We try to understand the reasons behind the persistent existence of this specific viral epidemic. We identify the leverage parameters, and calibrate them to obtain persistence in the epidemic. We compare the outputs with the data for validation to show that the model can reproduce the dynamics of the real system with its own internal causal feedback structure. The model we construct does not only aim to generate valid dynamics, but it also goes beyond the existing models in the sense that it serves as an experimental platform for scenario analyses to support the understanding and management of the disease. The second aim is to demonstrate at a conceptual level that the internal structure of our model makes scenario analyses possible. In scenario experiments, we try to eliminate the virus with two parameters related to the biting rates and vector births. Biting rates and vector births can be decreased by taking precautions such as wearing clothes with more skin coverage, using repellents, and cleaning the water storages regularly. We conclude that the virus can be eliminated if humans take precautions against the interaction with vectors or against vector breeding areas.
Exact solution methods for the assignment problem with conflict constraints
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Arslan, Elif.; Altınel, İ. Kuban.
The Assignment Problem with Conflict Constraints (APC) deals with finding a maximum weight assignment in the presence of incompatibilities. The incompatibilities are constituted by the conflicting edge pairs such that both edges in a conflicting pair cannot be in a feasible solution. Unlike the assignment problem, APC is a N P hard problem; therefore, it brings about the importance of finding efficient solution procedures for APC. Yet, there are only a few studies on APC that put forward exact solution procedures. To the best of our knowledge, this is the first study which proposes Branch & Cut algorithms for the solution of APC. Within the computational analysis, we first evaluated the performance of Branch & Bound with different branching rules. After wards, we assessed the performance of additional algorithms with the best performing branching rule. Finally, we checked the contribution of the added cuts to the overall problem with the selected branching rule and the default branching rule of the mixed integer linear programming commercial solver. The computational results showed that the Branch & Bound algorithm with the best performing branching rule, the Branch & Cut algorithm with clique cuts and the Branch & Cut algorithm with combination of clique and cycle cuts performed better compared to the state-of-art commercial solver.
Dynamic analysis of public health insurance programmes
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Güz, Şanser.; Yücel, Gönenç; Karanfil, Özge
Effectiveness and continuity of public health insurance systems are facing signif icant threats from structural and demographic challenges such as principle-agent prob lems and population aging. In this thesis, we construct a multi-sector broad- boundary dynamic simulation model: i) to explore the underlying causal structure surrounding public health insurance systems, and ii) to analyze the operational causality that leads to the development of structural problems for single-payer health insurance models that is financed by employee contributions. The model is simulated under different parameter configurations to analyze the system behavior under different scenarios. We show that the austerity periods carry the danger of overlooking the critical essential demand which would leave collateral damage on population health, lasting long after the first reduction in coverages. Several policy options such as public revenue injections and labor immigration policies are tested to preserve the programme benefits in the long run. We find that greater magnitude interventions generally perform better even with low frequency when the early action is taken. We also show that the effectiveness of some policies are dependent on the scenario conditions. Although highly related, the population health and public programme benefits are found not to share the same exact dynamics and eventual fate. Consequently, the stabilization of programme vari ables does not necessarily mean a better population health outcome for all cases. As a further research avenue to this study, a multi-payer sector can be added to the model in order to conduct a comparative simulation analysis on the health systems performance of single-payer, multi-payer and hybrid insurance models.
Analysing the efficacy of feature elimination and adaptive sampling in meta-model based exploration of agent-based simulation models
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Yıldız, Ecemnaz.; Yücel, Gönenç.
In this thesis, an advanced procedure is constructed to investigate agent-based simulation models. A meta-modeling approach, that utilizes adaptive sampling and feature elimination methods is used in the proposed procedure. The procedure aims to build a machine learning model that replicates the input-output relationships of the original agent-based simulation model and accurately predicts the output of inter est. Thanks to feature importance measurements, the proposed procedure also enables researchers to analyse the relationships between the agent-based simulation model pa rameters and the output of interest. The Random Forest algorithm is used for building the meta-model. The adaptive sampling method is utilized to create a high-quality data set to train the meta-model. The feature elimination process is applied to enable meta-model to prevent the curse of dimensionality and keep the focus on important fea tures regarding the output of interest. The proposed procedure is applied to a complex agent-based meta-model to evaluate its performance. A recent agent-based simulation model, that is analyzing socio-dynamic systems, is selected for application considering its probabilistic nature and wide range of parameters. Moreover, previously proposed meta-modeling approaches in the literature are reviewed and performance comparisons are assessed with the proposed procedure. Both the accuracy of output predictions and the validation of feature elimination decisions are analysed in detail. The conducted experiments and analysis showed that the proposed advanced procedure estimates the output of the original simulation model in an accurate and efficient way, and it out performed the previously proposed meta-modeling approach in terms of accuracy.
Dynamic analysis of false information spread over social media : 5G-covid 19 conspiracy theory
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) İrsoy, Orkun.; Yücel, Gönenç.; Barlas, Yaman.
The spread of false information via online social networks is a critical societal issue with various potential harms. Although there are huge efforts both in research and application to mitigate this problem, it persists with an increasing magnitude of results ranging from political manipulation to violent attacks. In our research, we built a causal simulation model to combine the existing accumulated knowledge in the liter ature and provide a formal model to evaluate the governing dynamics for the specific case of the viral spread of the 5G-COVID-19 conspiracy theory. The model makes use of both qualitative and quantitative data and successfully generates the observed dynamics for the 5g narrative. Results from the base run suggest that the dominance of believers in the active discussion on social media is overrepresented relative to the total population. Moreover, common mitigation strategies proposed in the literature such as limiting the interaction with believers of the misinformation often seem to pro duce worse outcomes for specific cases which indicates policy resistance. In addition, scenario analysis suggests that the involvement of neutral people in sharing misinforma tion or superspreader actors might be enough to induce the system to pass the tipping point and generate an infodemic. The current analysis presents several trade-offs while discussing the underlying reasons through posterior analysis. In further research, we plan to expand our analysis by the inclusion of other user profiles, experiment with other mitigation strategies, and discuss the potential similarities and differences of our case with other types of false information dynamics.
Multiple queues with simultaneous arrivals
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Emre, Toygar.; Güllü, Refik.
Queuing theory problems have been the topic of deep research owing to the fact that so many difficulties are in existence and their significance in real life cases can not be ignored. Those problems can be observed in numerous sectors such as telecommu nications, airlines, logistics, hospitals, computing, production and inventory. Besides, speed is the key word in today’s world because population is almost at the peak, thus demands or requests must be met as much as possible. However, our world has limited sources that is why there has to be some delays and queues. Additionally, game theory is one of the most important topics and it comes into prominence due to increasing competition in the world. There are lots of organizations which dwell in aforemen tioned sectors and they need to compete with each other to maximize their benefits. Just as in queuing theory, application of game theory spans the huge part of real life problems involving so much burden. So, there are abundance of works which dive into the distinct branches of game theory. In this study, both queueing theory and game theory are taken into consideration. We include the concept of game analysis, server rate optimization, multiple queues, loss systems and simultaneous arrivals at the same time whereas the studies in literature just focus on some of them. In our first case, we apply a game theoretic approach to two loss queuing systems under specific assump tions. With the deployment of server rate optimization we reach Nash equilibrium points. We also provide some analytical derivations and validate them using simula tions. In our second case, we deal with one loss system with an uncapacitated queue involving quasi birth death process. We find the steady state probabilities employing two different computation techniques and calculate the expected profit for each queue in the system.
Oblique random forest algorithm using lasso regression for wind power forecasting
(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2022., 2022) Tabak, Burak.; Baydoğan, Mustafa Gökçe.
With the increasing trend towards the use of renewable energy sources, wind power has been the subject of many researches. Wind power has stochastic nature due to uncertainties in atmospheric conditions, especially in wind speed, which makes it hard to forecast accurately. To solve the problem, statistical methods using Numerical Weather Prediction (NWP) models as inputs are proposed in the literature. Random Forest is a statistical model frequently used in wind power forecasting with proven success. Random Forest ensembles decision trees that partition the feature space over a single variable at each node. However, partitions based on a single vari able may fail to provide a proper distinction. Thus, oblique decision tree algorithms evaluating the partitions over linear combinations of variables are proposed in the lit erature, especially on classification problems. There are a limited number of studies in the literature on oblique decision tree-based methods applied in time series regression problems. This thesis proposes a novel strategy to be applied in regional wind power fore casting tasks that ensembles oblique decision trees. The proposed method is compared with its univariate counterparts in three wind power forecasting tasks. Computational results show that the proposed method performs better on all tasks.

Browse

Recent Submissions