Yüksek Lisans Tezleri
Permanent URI for this collectionhttps://hdl.handle.net/20.500.11779/1785
Browse
7 results
Search Results
Master Term Project Convolutional Neural Network for Facial Emotion Recognition With Geometrical Features of Face(MEF Üniversitesi Fen Bilimleri Enstitüsü, 2021) Arslan, İlker; Tuna ÇakarOne of the recent challenging machine learning problems is to make predictions on image datasets. The aim of the project is to construct a convolutional neural network to guess emotions for a face of a human given in an image file considering the face. After the geometrical features are extracted using pretrained models, we construct five models which are convolutional networks fed with handcrafted geometrical features extracted. The last model uses the outputs of other four models to predict more accurately.Master Term Project QPICAR Deep Learning(MEF Üniversitesi Fen Bilimleri Enstitüsü, 2021) Beğde, Özge; Tuna ÇakarThe aim of the project is to train a smart tool kit named "Sunfounder Raspberry Pi Robot Car" to move without hitting the walls in a closed area. The goal is to maximize the driving time without crashing by reducing the number of hits. Ultrasonic sensor data collected from the vehicle are processed with reinforcement learning and deep reinforcement learning algorithms and the results are compared. In this study, Python programminglanguage is used. In this study, firstly, the Q-Learning method, which is a reinforcement learning algorithm based on Markov decision processes, is used. The method basically relies on a memory table, Q-Table, in which the Q-values of the agent moving from one state to another are kept. This table is updated according to the results of the Bellman equation in every action of the agent, and as a result of this iterative process, it is optimized to provide that the agent moves to maximize its rewards. Deep Q-Learning (DQN) is used as a deep reinforcement learning algorithm. This algorithm was developed by the DeepMind Technologies team in 2013. Basically, it is based on the use of the Bellman equation, which is an element of the Q-Learning method, incombination with neural networks. This method is often used for training agents in complex and multidimensional environments such as video games. Due to the different type of the data used on the algorithm, minor changes were made to adapt it to the study. RElu and Softplus are used as activation functions. The results of the training process show that the DQN algorithm has an important advantage in terms of training the agent in a short time. At this point, the results are in accordance with other academic studies demonstrating the success of the DQN algorithm for complex environments.For future work, by differentiating the equipment that collects data on the vehicle, different data types such as image, temperature value, oxygen value can be collected and processed. At the same time, with changes to the reward setup in the algorithm, the agent can be trained to move to a specific target or to take actions to avoid a specific target.Master Term Project The Automatic Identification of Butterfly Species Using Deep Learning Methodologies(MEF Üniversitesi Fen Bilimleri Enstitüsü, 2020) Tek Kara, Seda Emel; Tuna ÇakarAutomatic identification of butterflies, especially at an expert level, is needed for important topics such as species conservation studies, minimizing the insect damage on plants in agriculture, and biodiversity conservation. An efficient and performing model which can define species even in small datasets may reduce the need for experts on the subject or reduce the time spent for identification. By the model proposed in this study, automatic taxonomic classification of butterflies was studied. Convolutional Neural Network (CNN) applications were applied on 7148 photographs of six butterfly species used in the study. 80 percent of the data set was reserved for training and 20 percent for testing, and the model was run with the relevant parameters. At the end of the study, an accuracy degree of 92.73% was obtained.Master Term Project Turkish Private Pension Fund Size Forecasting as an Application of Data Analytics(MEF Üniversitesi Fen Bilimleri Enstitüsü, 2020) Kara, Serdar Ufuk; Tuna ÇakarIn this study univariate and multivariate models are used to forecast the net changes in total pension fund size of a private pension company in Turkey, using the daily data between November 2003 and November 2020. Univariate models include the naïve, autoregressive (AR), moving average (MA), and autoregressive moving average (ARMA) models. Multivariate models include vector autoregression (VAR) and multiple linear regression models. Our findings suggest that multivariate model predictions outperform univariate model predictions. Univariate model predictions can be improved with walk forward approach. Increased lag size can help improve AR, MA, ARMA and VAR model predictions. Naïve model produces the weakest predictions.Master Term Project Calculation of the Capacity of a Retail Clothing Store(MEF Üniversitesi Fen Bilimleri Enstitüsü, 2021) Türkoğlu, Murat; Tuna ÇakarThe purpose of this study is to define and improve the capacity calculation for clothing store companies. It is important to know how many products need to be sent to the relevant store in order to sell more products, whether more or less products will be sent to the stores, how much the product can sell and how much capacity the stores will have for the relevant products during the season. Planning and producing more products than necessary may also cause insufficient capacity to consume the stocks of that product in the relevant season. For these reasons, a detailed capacity management system is needed. The capacity of certain product groups in the stores in certain seasons can be determined by calculating the capacities of the relevant units in the stores and the relations of these units with the product groups. The relevant system will produce output for both planning and allocation units. At the same time, converting the capacity of products and store display units into a common unit (LCM) will be one of the factors that facilitate our work in capacity calculation. A short version of the LC Waikiki capacity system platform is used to obtain the data. ASP.Net Core Web API, ADO.NET, T-SQL, C # programming were used as program tools. Azure Microsoft SQL Server was used as a database server. Azure App Services has been used to keep the business codes.Master Term Project Customer Clustring With Machine Learning(MEF Üniversitesi Fen Bilimleri Enstitüsü, 2021) Kara, Ömer Faruk; Tuna ÇakarWhen analyzing a company that sells in very different product ranges, you are likely to encounter different types of customers. Grouping customers correctly can set standard actions while serving them. Standardization of marketing processes leads to speed and they are easy to improve. While making this classification, the KMeans algorithm was used in Machine Learning. Inertia and Silhouette Points values were used to find the most suitable cluster number. Principal Components Analysis (PCA) was used to show customers with multidimensional features in 2 dimensions.Master Term Project Retail Data Predictive Analysis Using Machine Learning Models(MEF Üniversitesi Fen Bilimleri Enstitüsü, 2020) Güner, Müjde; Tuna ÇakarMachine Learning (ML) is a popular field which deals with training the system with data (experience), performing some task (regression or classification) and evaluating the system with the desired performance metrics. ML automatically extracts useful and meaningful insights from the data. ML models for sales prediction applies computational intelligence in many real world applications such as stock market, production, economics, weather, retail, census analysis and so on. Sales prediction can be viewed as a regression problem and various algorithms can be applied. In this project, real life data analysis has been done to predict the sales for four categories of products like Cold Cereal, Bag Snacks, Oral Hygiene Products, and Frozen Pizza. Exploratory Data Analysis (EDA) has been applied to the dataset to make exact predictions even during an unpredictable environment. The different phases of EDA used in this project are Data Preprocessing and Analysis, Feature Selection and Feature Extraction, Model Building and Regression Analysis, Clustering, Time Series Analysis and Model Evaluation using the Performance Metrics. For outlier detection, InterQuartile Range (IQR) method is used. For Filter Based Feature Selection, Univariate Feature Analysis using SelectK-Best and SelectPercentile, Decision Tree Regressor method has been used. For Wrapper Based Feature Selection, Sequential Feature Selector method has been deployed. For Regression Analysis, various algorithms such as Linear Regression, XGBoost Regression and Support Vector Regression (SVR) are analyzed. K-Means Clustering Algorithm has been used on the dataset to generate 4 different clusters. In Time Series Analysis, the week end date and average weekly basket attributes are analyzed, and the sequential data has been rendered for a given time period of occurrence. In model evaluation phase, the Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), R2 and Adjusted R2 accuracy has been calculated and validated. The project has been implemented in an open source software called Anaconda which includes Jupyter Notebook platform for scientific computations. Python programming language with different packages such as Numpy, Pandas, Scikit learn has been used.