Fall

Financial Analysis Using Machine Learning Methods

Machine Learning methods have long been applied to financial data. We explore the correlation (or the lack thereof) of financial data, such as the BIST 100 index or individual stock prices, to features extracted from financial news and past data. We intend to use the advancements in text classification and Natural Language Processing capabilities in general, and machine learning models to extract the features of Turkish news documents and stock price series using statistical methods and to analyze the correlation of the extracted features with the financial data.

Chrome Extension for Spam Detection

Nowadays, spam contents in the internet could be really annoying and harmful while surfing the internet. There are some approaches to find a solution for that problem. However, those does not focus on natural language processing approach and does not give much credit for the machine learning techniques.

Constructing Risk Map of Cyber Entities Belonging to Boğaziçi University in terms of Security

The object of my dissertation is identifying the ways to exploit vulnerabilities to defeat the security features of system components used in Boğaziçi University. The goal of penetration testing will be determining whether and how a malicious user can gain unauthorized access to assets that affect the fundamental security of the systems.

Auctioning Privacy for Multi-Agent Autonomous Vehicles

Information is one of the most valuable asset in twenty-first century. While the importance of information increases, the need to preserve personal privacy is becoming essential. In the context of Internet of Things, the sensory data about our life and our surrounding is sending to the internet, which is vulnerable to violate personal privacy. In order to preserve the privacy in the context of Internet of Things, specifically in the autonomous smart traffic, we implemented a communication protocol that auctions privacy while managing intersections for an autonomous multi-agent traffic.

Graph Database for a Social Network of Musicians Throughout Universities

In this project it is aimed to design a graph database and queries for a social network of musicians and bands from different universities. The database is designed so that the recommendation queries run quickly and the system is not overloaded with endless node scans and cartesian products of sets for joining tables in contrast to relational databases. The properties of graph database are exploited to achieve this performance improvement.

IMAGE PARODY DETECTION

The project team will create a small database of famous paintings and their parodies. (Go to Google Images and query “arnolfini parody” to see what I mean).

Then, the team will develop image similarity measures to automatically detect parodies. Deep neural network and transfer learning approaches will be used.

See the following paper for more information:

Information Visualization for Choir Conductors

The project team will create a software to help choir conductors. The conductor typically assigns a piece as homework, and choir members record their voices, singing that piece. The software will take as input a folder of recorded songs (same song, recorded by different people), maybe also the midi file of the song (optionally), and analyse each file (using the OpenSmile library), and visualize each on a single page in a way to show the conductor the quality of these recordings.

AI Bots for Slither.IO

The project will involve writing AI bots for the slither.io game.

Reinforcement learning and deep reinforcement learning approaches can be implemented eventually, but the first attempts will be based on heuristics.

The bot will learn in two modes:

-Self play: many instances of the bot will play the game with different parameters, and the parameters will be adjusted based on the success of the bot.

-Learning by demonstration: a human will play the bot, and the game will be stored to learn from it.

Something theoretical

You have taken and passed CmpE350. If you didn’t like it, don’t read the rest.

There is a lot more to theoretical computer science than what you’ve seen in CmpE350. In this project, the student focuses on one subtopic, studies material that explains it, learns it, and describes it in detail to her/his peers.

Warning: In past years, this has turned out to be more difficult than the average CmpE492 project.

Automatic Logo Detection

Automated localization of logos and handwritten signatures in scanned document is a promising facility for many business activities. We describe here a discriminative framework for detecting logos and signatures from a document of any type. The method is based on the classification of segmented image regions using a set of features. The segmentation is done for both known logo and signature regions, and non-logo & non-signature regions, and feature sets are extracted for both positive (containing logo or signature) and negative segments.

Search form

Main Menu