Theses supervised by Doç. Dr. Ilgın Gökaşar
11 theses · Boğaziçi University
A new public transportation pricing strategy: The evaluation of the rent-based fare
Şehirlerin büyümesi yolcuların seyahat mesafelerini artırmakta ve buna bağlı olarak çeşitli maliyetleri beraberinde getirmektedir. Bu çalışma, insanları seyahat mesafelerini kısaltacak bölgelerde ev kiralamaya teşvik ederek, yeni bir toplu ulaşım ücretlendirme sistemi geliştirmeyi amaçlamaktadır. Kiraya dayalı fiyatlandırma sistemi, Doğrusal Olmayan Optimizasyon modeli kullanılarak oluşturulmuştur. Yeni tip fiyatlandırma sistemini tanıtmak için İzmir Banliyö Sistemi seçilmiştir. Çalışma kapsamında sabit ücretlendirme, mesafe bazlı ücretlendirme, bölgesel ve kiraya dayalı ücretlendirme olarak tanımlanan dört alternatif toplu taşıma fiyatlandırma sistemi toplu taşıma uzmanları tarafından değerlendirilmektedir. Bu alternatifleri önceliklendirmek için maliyet, ulaşım, sosyal ve politik olmak üzere dört boyut belirlenmiş ve bu boyutlar altında 13 kriter yer almaktadır. Alternatifler, kriterler ve alt boyutlar kapsamlı bir literatür taraması ile tanıtılmaktadır. CRITIC (CRiteria Importance Through Intercriteria Correlation) yöntemi ve Analitik Hiyerarşi Süreci (AHP) ile kriterler ve boyutların önceliği değerlendirilmektedir. Araştırmacılar ve uygulayıcılara basit ve esnek bir karar verme aracı sağlamak için yeni T2NN (type-2 neutrosophic numbers) tabanlı MABAC (Multi-Attributive Border Approximation area Comparison) yöntemi tanıtılmaktadır. Sonuçlar, kira bazlı ücret fiyatlandırmasının en avantajlı alternatif olduğunu göstermektedir. Entegre CRITIC ve MABAC tabanlı tip-2 nötrosofik modelin yüksek güvenilirliği ve sağlamlığı karşılaştırmalı analizler ve duyarlılık analizleri ile gösterilmektedir.
Evaluation of administrative division-level road safety indices
Inadequate regional road safety studies have been conducted in developing countries such as Iran, Egypt, and Türkiye. Also, despite the existence of various regional road safety indices (RSIs), the associations between these rates rarely have been studied. Besides, there are limited studies regarding crash severity indices in the literature. Despite high road fatalities in developing countries, little attention has been given to road safety performance in these countries. Additionally, the differences between developed and developing countries regarding road safety performance rarely have been discussed. Thus, it was aimed to evaluate the regional RSIs in Iran, Türkiye, the UK, Egypt, and the USA, using correlation and regression analysis. Also, the distribution patterns of administrative divisions of these countries were assessed. Data on regional road safety and socioeconomic rates of these countries were collected. The associations between the variables were evaluated using correlation and regression analysis. Using Moran's I, local Moran indices, and Jenks natural breaks method, administrative division's spatial distributions were evaluated. Hot spot analysis was used to identify road safety deficient regions. Significant correlations between the variables were detected. Vast local clusters in terms of RSIs were detected in the countries. The distribution patterns of subdivisions regarding RSIs were cluster-like. Variable groups influencing road safety performance in regions were identified. Generally, crashes were severe in underdeveloped and remote regions. Increasing income and education levels make it possible to reduce crash severity indices in these countries. Higher exposure rates mean higher fatalities in regions. There is a nonlinear and significant association between motorization rates and TR indices of regions, and fatality risk decreases as the motorization rate increases. There is a considerable gap between developed and developing countries regarding regional RSIs. Findings suggest using the fatality per number of motor vehicles index instead of the fatality per population rates in regional road safety studies. Using distinct exposure measures in calculating RSIs leads to the inverse local cluster maps.
Real time traffic management using connected vehicles: A case study on D100 highway
Traffic congestion is one of the most crucial issue that have a negative impact on the economy and the environment of a city. The negative impacts of congestion become stronger with the increasing population and vehicle usage. Here, the metropolitan areas are the most problematic and sensitive areas by considering that they contain most of the population within itself. Even traffic congestion in a local region may be felt throughout the network and; therefore, it might deteriorate the overall traffic conditions of the road network of a city which makes metropolitans quite sensitive. The scale of this impact gets especially grander in overpopulated cities; thus, it becomes even a more critical issue. In this thesis, a real time traffic management method using connected vehicles is tested in a simulation environment with synthetic and D 100 Highway real traffic data. To test this method, a 5.4 km length network with 3 lanes was built in the SUMO (Simulation of Urban Mobility) environment. An incident was generated in each run to observe its impact on the traffic network. By changing the incident and the parameters of the connected vehicles, 189 different scenarios were tested. The results of these 189 scenarios show that the introduction of connected vehicles causes an increase up to 20.35% in terms of mean speed, a decrease up to 32.54% in terms of mean density and an increase up to 2.45% in terms of mean flow for the local system before the incident. Based on the 189 scenarios, the connected vehicles provide the maximum mean speed increase and mean density decrease when they control traffic flow 1750 meters behind the incident location.
A new submarket approach using distances to transit lines for the prediction of real estate prices
Prediction of a price of a real estate has been one of the trend topics in recent years. There are many studies conducted on both prediction of a value of a real estate or the affecting parameters. In this study, the prediction of a real estate price using submarket near transit lines is studied on two neighbor counties in Istanbul, Beylikduzu and Esenyurt. So, the data should be analyzed into parts to investigate altered dynamics of different districts, which is also called submartket analysis. After the whole data of real estates in these counties are collected and analyzed, the data are divided into three parts (Esenyurt, Beylikduzu and transition zone) and analyzed in order to investigate altered dynamics of different districts. A total of 3487 real estate data with one dependent variable and 13 independent variables collected from aforementioned districts are analyzed with two machine learning (Multiple Linear Regression (MLR) and Spatial Auto Regression (SAR) and one deep learning tool (MultiLayer Perceptron (MLP)). According to the results of the both whole data and submarket analysis, Spatial Auto Regressive model is superior to the others in a metric of R-squared. Moreover, with the submarket analysis, prediction power of all the algorithms (MLR, SAR, and MLP) are significantly increased. Significant independent variables of each model differ from each other so that it can be concluded that submarket analysis in real estate prediction is improving the prediction models and showing different dynamics of each specific district of a county.
Personel servislerinin toplu taşıma ile entegrasyonu: Boğaziçi Üniversitesi vaka çalışması
To provide comfortable and reliable means of transportation, most of the employers offer personnel services to their employees. Although it is a commonly used concept, for both employers and employees, there are downsides. From employers' point of view, financial load of the concept is high for both private sector and state institutions. Extra budget spent on such systems leads to missed opportunities for other investments or researches. On the other hand, some employees have complaints about conditions of the concept. Purpose of this study is to provide a systematic approach for the optimization of personnel services and furthermore, integrating the concept with means of public transportation. In accordance with this purpose, two methods are represented in this study. The first method mostly focuses on optimization of service routes, latter, by building top of it, aims to integrate personnel services with public transportation. The study is conducted with data of current personnel system of Bogazici University, which is one of the most prestigious universities in Turkey. A Geographic Information Systems (GIS) software was used to enable the purpose of the study. Study has shown that, compared to conventional approach to the concept, offered methods have better results. Besides providing improvements to a single case, in the long term, by spreading the methods of the study, it is believed that improved traffic conditions may be provided.
AHP-based risk identification risk assessment and risk allocation approach for the micromobility sector
In today's world, where the micromobility sector, which is one of the most important sub-headings of the sharing economy model, in which people pay for the products and services they need for a short time, without owning them, is rapidly becoming widespread, many people and institutions are directly and indirectly affected by this situation. To minimize the negative effects and take the necessary precautions, the current situation should be revealed and the risks that may arise should be determined. After examining the sharing economy and risk allocation concepts, first of all, the risks arising from the micromobility sector were determined in this study. Afterward, the AHP method, which is one of the multi-criteria decision-making methods, was explained and the surveys prepared were evaluated by the experts, the determined risks were distributed among the alternatives, and the opportunity to guide and share responsibility for the applications to be made after that was provided. Based on this study, it was easier to determine the main topics for future research.
Evaluation of the environmental effects of connected autonomous vehicles in traffic incident scenarios on uninterrupted facilities
Traffic incidents can occur due to both human errors and the inadequacy of road networks. These incidents can cause not only material damage but also loss of life. In case of an incident, it causes the vehicles in the traffic network to stay on the road longer and consume more fuel. The increase in fuel consumption increases the emission of CO2 (Carbon Dioxide), which is an effect of climate change. To reduce the negative effects of incidents on the environment, incident detection, and real-time traffic management methods are important. In this thesis, an uninterrupted road network was created utilizing SUMO traffic simulation to evaluate the environmental effects of incidents. This road network was evaluated over different scenarios with the integration of incident detection algorithms which are California and Standard Normal Deviation and real-time traffic management algorithms which are VSL and LCS. Environmental results were obtained by analyzing these different scenarios. Two types of vehicles were used: human-driven and connected autonomous vehicles. 11 different percentages of autonomous vehicles in increments of 10 from 0 to 100 were based on the research. It was seen that the increase in the use of connected autonomous vehicles in countries such as Turkey, which provide their electricity needs from nonrenewable energy sources, harms the environment. In the countries that provide their energy sources mostly from non-renewable sources, the scenario with the least CO2 emissions in the CAL-LCS and CAL-VSL scenarios was achieved in conditions with 40% connected autonomous vehicle traffic. Finally, a relationship of up to 80% was found between CO2 and speeds two by using KNN and Decision Tree Regressor models.
Macroscopic modeling of the motorway traffic in Sweden: A calibration study
Traffic congestion is an important factor in today's urban life which has a direct influence on people's quality of life. This phenomenon usually happens on peak times when people are commuting to work. As people are stuck in the traffic congestion, not only it will take more time to reach their destination, but also this wasted time can be of their work's time meaning a loss of money. In addition to that, congestion may cause pollution. Since more vehicles are stuck in traffic and cannot advance properly, higher amount of gas is ejected to the air. Thus, tackling congestion problem is critical. There are several factors that lead to traffic congestion including the times when roads are full of vehicles, or when there are constructions, maintenance or incidents on roads. These factors cause a disruption in traffic flow. It is important to note that resolving congestion problem cannot only be done by increasing capacity and by constructing new roads and motorways; It also requires developing a transportation management technology. Although the solution of building new roads may seem interesting at first sight, it is not practical as it may cause even more problems. Therefore, utilizing the existing roads and enhancing the traffic management technology would be a better solution. Having a calibrated traffic flow model that reproduces the behavior of drivers on the real road can be of benefit. A calibrated and simulated microscopic traffic flow model is used in this thesis to calibrate a second-order macroscopic traffic flow model to represent the traffic flow on the motorway in Sweden by validation. For this purpose, different scenarios which represent free flow, intermediate flow and congestion have been studied. This model can be used to better manage the traffic and implement control strategies on motorways.
Çok nitelikli fayda teorisi ile köprü sistemlerini önceliklendirme
Bridge maintenance planning is a complex problem because of the complexity of the objectives. Handling this issue optimally is a challenge. There is a need of bridge planning process to organize and control the bridge inventory for a better decision-making process and the overall public utility. Among the other alternatives, multi-attribute utility theory (MAUT) is applied as a viable multi-objective decision-making method to prioritize 20 bridge networks in Western Turkey. The aim of this study is applying MAUT to a bridge maintenance planning problem in Turkey and providing improvements to the existing Turkish bridge management system (BMS) decision-making method in the literature. Existing attributes are examined and another criterion that intrigues the safety for the work-zone is added. The prioritization methodology is replaced with the MAUT approach to explain the uncertainty situation by taking risk preferences of the decision makers. Additionally, additive utility independence assumption of MAUT is inquired to for a more optimal solution. The results provide that how calculating risk preferences, interrogating more decision makers, unique point of views of the various expertise, and adding a new objective criterion might change the priority order for the same bridge dataset.
Insights into the effects of Covid-19 on taxi use near M2metro line in Istanbul
COVID-19 has become one of the most significant events in this century and the effects are felt in daily life all around the world economically and socially. Therefore, investigating the effects of COVID-19 must be prioritized to minimize future damages. The goal of this thesis is to reveal different travel reactions to COVID-19 based on spatial and socioeconomic characteristics such as public transport connectivity, education level, female percentage, etc. Towards this goal, taxi GPS data is used and the M2 subway line in Istanbul is selected as the case study area since the line covers many residential and commercial centers. The prepared COVID-19 timeline is divided into five phases based on critical events such as the announcement of governments or unexpected peaks in daily case numbers. The analyses are conducted for the average trip counts in four time periods of a day, namely total, off-peak, morning, and evening. K-means clustering is used to observe the relationship between stations and the data is analyzed by ordinary least squares (OLS), spatial auto regression (SAR), and geographically weighted regression (GWR) models based on daily average trip counts and characteristics of stations. The best results are obtained by the GWR model. According to the results, the population size is one of the most significant parameters, that explains the change in trip counts. The morning peak shows a unique characteristic that can be explained by the socioeconomic index, which is a weighted average of many parameters including education and income level. In general, the decrease in taxi trips is higher for the areas with a higher socioeconomic index. Other significant variables are the number of shopping malls, the existence of another public transportation option, and the population density.
A biosequence based dynamic ride-matching algorithm that takes into account social factors
Increasing traffic congestion and advancements in technology have fostered the growth of alternative transportation modes such as dynamic ride-sharing. Smartphone technologies enable dynamic ride-sharing, which aims to establish ride matches between people with similar routes and schedules at short notice. Many automated matching methods are designed to improve system performance, such as minimizing process time, minimizing total system cost or maximizing total distance savings; however, the results may not provide the maximum benefits for the participants. In this dissertation, an attempt is made to develop an algorithm to optimize matches when considering participants' gender, age, employment status and social tendencies. A biosequence algorithm, namely the Needleman-Wunsch algorithm, is used to quantify the similarity of participants' itineraries. A stated preference survey was conducted among 604 students and members of staff at Turkish-German University in 2018. An extensive simulation study was then performed by utilizing the survey data to compare the performance of the proposed algorithm with that of traditional bipartite and optimization algorithms. The simulation results indicate that when compared to the traditional bipartite and optimization algorithms, the proposed algorithm significantly increases performance in terms of computation times and the potential success rate of the matches. A sensitivity analysis for the proposed algorithm is also performed.