| 1 |
| Title |
|
Computational Eigenvector Methods in AHP for Consistency Index Calculation |
| Speaker |
|
Mustafa Sönmez |
| Advisors |
|
Süha Tuna / Başar Öztayşi |
| Date |
|
Dec 12, 2022 |
| Time |
|
13:30 (GMT +3) |
| Location |
|
Informatics Institute Room 205
|
Abstract
In this study, the consistency index calculation approaches in the Analytical Hierarchy Process method, which is one of the multi-criteria decision making methods, are emphasized. The consistency index is of critical importance in the widely used AHP method. For this reason, practitioners prefer to use computational methods as well as developing practical methods. In this study, these methods are also included in the application. |
| • |
| 2 |
| Title |
|
Optimizing Airline Flight Scheduling in an Ambiguous Market Conditions |
| Speaker |
|
Şevin Madsar |
| Date |
|
Dec 12, 2022 |
| Time |
|
14:00 (GMT +3) |
| Location |
|
Informatics Institute Room 205
|
Abstract
Airline companies use different varities of softwares to schedule their aircrafts’ flights in the most efficient way to maximize their profits. This can be achieved using their assets (aircrafts) as much as possible, and it can only be achieved by maximize their time on flight and accordingly their revenue. By considering these aims, scheduling softwares makes long term and short term flight schedules but generally these softwares cannot guess any unexpected conditions such as; aircraft failure, unexpected maintenance because of an accident, delay chains due to weather conditions, or even change of countries flight regulation rules etc. Therefore, this paper will present a system to work as a supporting system of these flight scheduling softwares to help the schedulers under ambiguous conditions and to improve decision making system to obtain better use of aircrafts and to have more efficient flight scheduling for airline companies. |
| • |
| 3 |
| Title |
|
Modeling the Pattern of Whether the Comments Made on the Tripadvisor Site Will Reach a High Level of Helping Vote |
| Speaker |
|
Seda Soykan |
| Advisor |
|
Sefer Baday |
| Date |
|
Dec 13, 2022 |
| Time |
|
13:30 (GMT +3) |
| Location |
|
Informatics Institute Room 205
|
Abstract
Since the comments made to the hotels are generally right after the hotel experience, it can be said that these comments reflect the reality to a great extent. According to the information announced by Google Analytics in the 3rd quarter of 2015, 340 million users log into TripAdvisor monthly, the site has 103 million members, 350 million user views on the site, and an average of 230 comments are made every minute. Considering the increasing number of these criteria, TripAdvisor is a very important resource for hotel reviews. If the comments made on TripAdvisor are found useful by other users, they receive a Helpful Vote. Comments with more support votes are considered more by the readers. For this reason, the number of support votes received by the comment is very important. Therefore, determining the parameters affecting the help game is an important problem. In this study, by focusing on this problem, the factors affecting the number of support votes will be examined to the degree of impact. In this context, the features of the user making the comment and the features of the comment will be detailed and their effects on the help vote will be analyzed and analyzed with machine learning models. |
| • |
| 4 |
| Title |
|
Optimal Flight Suggestion with L-GBM |
| Speaker |
|
Esma Ergün |
| Advisor |
|
Süha Tuna |
| Date |
|
Dec 13, 2022 |
| Time |
|
14:00 (GMT +3) |
| Location |
|
Informatics Institute Room 205
|
Abstract
Most of the passengers worry for the possible delays or cancellations for their flights. Therefore, I introduced a method to choose the flight least possibly delayed or cancelled via Light Gradient Boosting Tree(L-GBM). Data used in this study, was taken from United States Department of Transportation ( Air Traffic Data) and National Weather Service (Weather Data). I used some other machine learning algorithms beside L-GBM Algorithm and compared the results. |
| • |
| 5 |
| Title |
|
Sparse Coding via High Dimensional Model Representation for Hyperspectral Images |
| Speaker |
|
Kamila Muminova |
| Advisor |
|
Süha Tuna |
| Date |
|
Dec 20, 2022 |
| Time |
|
13:30 (GMT +3) |
| Location |
|
Informatics Institute Room 205
|
Abstract
Due to the high association between spectral features and noise present in spectral bands, which can considerably reduce classification performance, the classification problem was primarily tackled along with the dimension reduction when analyzing hyperspectral pictures. In this article, we will specifically address the issue of dimensionality reduction and offer a brand-new feature selection approach that is based on a technique known as sparse coding via High Dimensional Model Representation for Hyperspectral Images. Evaluating the proposed technique to traditional feature selection algorithms in terms of classification accuracy, stability of selected features, and computation time, it was tested on a few toy examples and hyperspectral datasets. The findings demonstrate that the suggested method offers high classification accuracy and reliability functions with acceptable computation times. |
| • |
| 6 |
| Title |
|
Adaptive Video Playback using Deep Learning for E-sports |
| Speaker |
|
Burak Nayır |
| Advisor |
|
Tankut Akgül |
| Date |
|
Dec 20, 2022 |
| Time |
|
14:00 (GMT +3) |
| Location |
|
Informatics Institute Room 205
|
Abstract
Simultaneously maintaining low latency and avoid stalls for live streaming has always been a challenge. In order to manage the challenge, the current systems must make a tradeoff between maintaining low latency or avoiding stalls in real time. Generally, Adaptive Playback method which changes playback speeds according to situation, works well. However Adaptive Playback method does not consider context of live-streams. For E-sport, viewers can be more sensitive to playback speed at certain screens. Using deep learning, the paper aim is to, improve a Adaptive Playback method which depends context of streams for E-sports. |