Research Projects

Ship data distribution systems perform the collection, processing and transfer of all the data required by ships for safe navigation.

Within the scope of the project, 4 different applications will be developed by considering both hardware and software components together. In this project, Test System, Main System and Indicator System as software components; Protocol Converter embedded system will be developed as a hardware component.

In this project, it is aimed to develop a machine learning platform that can automatically learn from data sets with different data types and sizes.

A machine learning model will be created by uploading the data set owned by people who do not have sufficient knowledge in the field of machine learning to the automatic learning platform that will develop within the scope of this project.

With the platform, which will have a web interface that can be easily used by people who are experts in their field but who do not have sufficient knowledge in the field of machine learning and data science, the most suitable machine learning model for the data set will be recommended to users and training and test results for different models will be obtained and presented to users.

Mapping systems are widely used today. However, many satellite-based positioning systems that are successful in indoor areas such as shopping malls, airports, hospitals and factories and in open areas such as GPS are not sufficient due to lack of visibility or low base station reception quality. For this reason, many methods and technologies are being developed in closed areas to achieve the best performance with low cost. Within the scope of this project, suitable technologies for indoor location will be examined, a sample factory layout will be created, and an application will be developed to collect the parts whose locations are determined in the shortest way and to deliver them to the target point with the least loss of time.

Semantic search is an important field of study that promises to produce more accurate answers to the user's queries by taking advantage of the availability of explicit semantics of information. For example, when searching on "electrical equipment" using traditional word-based search technologies, it will usually return results that only include the terms "electrical equipment". However, the user's interest and electrically operated irons, ovens, brooms, etc. Since the terms "electrical devices" are not included in the devices, they will be overlooked.

In the context of Semantic search where the meaning of its content is made clear, the meaning of the keyword can be revealed. Also, the underlying semantic relationships of metadata can be used to search for relevant information and bring more accurate results. In the proposed project, a semantically search engine will be implemented for Turkish texts.

In this project, a framework is developed for collecting data from heterogeneous data sources on the Internet, measuring its consistency, automatically calculating and indexing the accuracy of the collected data, and making rapid and effective querying on indexed data. In this project, associations will be made between data obtained from many different data sources such as social media contents, data sources in tabular format, Web contents. Supervised learning methods will be used to determine the types of data sources.

With the framework to be developed, preprocessing processes are carried out for texts in English and Turkish languages, morphological analysis of the data collected on the Internet will be made and each informational term group will be reduced to triples. Also, the level of semantic relationship between the data will be determined. In the project, new algorithms will be developed in order to reduce the data in triple form, to make the accuracy through Internet resources, and to make morphological analysis of the texts.

Within the scope of this project, a dynamic window structure to be created for flow data will be used to predict forward-looking behavior and realization time according to the window frequency and the accuracy of previous predictions. In the system to be developed, a new model based on online learning and updating decision function for each new observation will be created. By determining the next occurrence time of an event in the flow data, with the model to be developed, timely and useful suggestions will be created for users. By constantly updating models to be created according to user behavior analysis with flow data, changes in user interests will be detected and useful suggestions based on predictions will be presented.

With the rapid increase in the size of information on the Internet, it has become very difficult to provide users with recommendations consisting only of quality products, information or content recommendations. For this reason, recommendation systems have begun to be developed to provide users with content that is relevant to them on very large data. The recommendation systems developed aim to increase both the satisfaction level of the users and the commercial gain by making suggestions that may be of interest to the users.

In this project, Web mining processes, recommendation systems and user behavior will be analyzed and a recommendation engine based on user behaviors and neighborhood relationships will be developed using Web mining methods on a website to be developed for application purposes. A new recommendation engine will be developed based on the products they choose, the duration of their stay on the product's detail pages, and the type of interaction with other users, without receiving any direct relevance feedback from the users. The results of the recommendation method to be developed will be compared with the collaborative filtering method widely used in today's recommendation systems. A website will be created for movie suggestions and sales for experimental studies. By taking the interactions of undergraduate and graduate students on the system, experimental results will be obtained and compared with the results of collaborative filtering method.