Yüksek Lisans Tezleri
Permanent URI for this collectionhttps://hdl.handle.net/20.500.11779/1785
Browse
7 results
Search Results
Master Term Project A Comprasion of Ensemble Learning Methods in Retail Sales Forecasting(MEF Üniversitesi, Fen Bilimleri Enstitüsü, 2019) Süer, Serhan; Güney, EvrenForecasting has always been an essential skill which companies try to have and implement in various areas. Sales forecasting is one of the major usage areas of forecasting which is used in almost all sectors. This study refers to forecasting sales of Walmart Stores based on several features such as store id, department id, date, and store size. Walmart sales data which was used in this study contains information of stores between 2010 and 2012. At the beginning of the study, the introduction of the dataset and exploratory data analysis were made to identify dependent/independent variables and their characteristics. To apply machine learning algorithms, data preprocessing methods such as missing value treatment, outlier treatment, and feature selection was applied. Ensemble learning methods in machine learning algorithms were applied in the modeling stage. These methods were addressed in three parts such as Bootstrap Aggregation, Boosting, and Stacked Generalization and these parts consist of six different algorithms in total. The models were compared based on four regression metrics as Root Mean Square Error, Mean Absolute Error, R-Squared, and runtime. After selecting the main metric which models were evaluated, cross-validation was applied to achieve unbiased estimates. Finally, parameters of the model which have the highest score in cross-validation were tuned in the hyperparameter optimization stage and a machine learning model which can be used in forecasting sales of Walmart stores and its success score were obtained.Master Term Project Machine Learning Applications To Increase Customer Satisfaction In Finance Sector(MEF Üniversitesi, Fen Bilimleri Enstitüsü, 2019) Yiğit, Leyla; Çakar, TunaIn this project, consumers’ complaints about financial data are analyzed. After the analysis, we aim to provide a tool for financial companies such as banks, Lenders that will help them in managing communication with the consumers. Our main aim is to answer the question “How do consumers feel?” This analysis will give a complete picture of consumers’ feedback. We start the project by clustering the customers into different groups. In order to classify customers, we use classification algorithms XGBOOST and Random Forest. XGBOOST is used to predict the probability of getting a complaint. XGBOOST is also tested as an ensemble learning technique. By Using Random Forest the comparison of Bagging and Boosting is performed. This kind of model is very useful for a customer service department that wants to classify the complaints they receive from their customers. These kinds of models can also be expanded into a system that can recommend automatic solutions to future complaints as they come. The topic is motivated by the researcher’s experience in finance where she intends to increase credit sell numbers by anticipating customer feelings. The data set that we use has many measures and dimensions that facilitate to use more than 3 machine learning algorithms. The complaints database is published by the Consumer Financial Protection Bureau (https://www.consumerfinance.gov/). It provides consumers’ feedback in a string format. We also aim to analyze consumers’ complaints dataset from the perceptive of a consumer dispute.Master Term Project Mortality Prediction of Countries(MEF Üniversitesi, Fen Bilimleri Enstitüsü, 2018) Üşenmez, Elif Efser; Koç, UtkuIn this study mortality reasons of countries detailed by sex and age-group is analyzed and different forecasting models are developed by using different machine learning algorithms. The dataset is obtained from the World Health Organization(WHO) Mortality Database. In WHO database there are different datasets for countries mortality reason number. The study used the dataset that used ICD-10 for classifying mortality reasons.ICD-10 is the 10 revision of International Statistical Classification of Diseases and Related Health Problems published by the World Health Organization. In addition to main mortality reason datasets, we add different independent variables and try to find the best features to fit models without biasing and overfitting and reaching high R2 and Mean Square Errors. To find the best model for forecasting mortality reasons by age-groups and sex different machine learning algorithms are fitted and results of these algorithms are analyzed.Master Term Project Predicting Transaction Numbers İn Atm(MEF Üniversitesi, Fen Bilimleri Enstitüsü, 2018) Karasu, Ahsen Ceren; Özlük, ÖzgürATMs continue to be one of the most important channels for banks to touch their customers. They play an active role in life in terms of cash access and banking experience. The ability of a bank to predict the number of transactions that will occur from ATMs is crucial for the proper control of the budgetary source. When cash is loaded into ATMs, the average transaction made from that ATM is taken into consideration and alarm mechanisms can be activated when a decreasing trend is observed on transaction basis.Before a new ATM is set up, the banks investigate how often customers in that area use other bank ATMs and calculate the commission costs incurred from those uses. As a result, the number of transactions made from ATMs is one of the most monitored KPIs of a bank and has important place in the cash management of the bank.The aim of this study is to estimate the number of future transactions with Auto Regressive Moving Average (ARIMA) method based on the number of transactions that occurred from ATMs.Master Term Project E-Commerce Customer Shurn Prediction Based Machine Learning Algortihms(MEF Üniversitesi, Fen Bilimleri Enstitüsü, 2018) Eser, Ahmet Yetkin; Arısoy Saraçlar, EbruWith the development and popularization of a digital world, human behavior has changed so remarkably. A lot of sectors affected because of this change. One of the most affected areas is the retail sector. People have left their regular shopping habits and started shopping on e-commerce sites. Thanks to increasing of variety and volume of collected data and velocity of new machines, companies can use sophisticated algorithms efficiently on their data. In this paper, we discuss about how companies can predict potential churned customers with machine learning methods.Master Term Project Risk Parameter Calculation Using Princpal Component Analysis of Yield Curves: the Case of Borsa İstanbul Fixed Income Market(MEF Üniversitesi, Fen Bilimleri Enstitüsü, 2017) Konuk, Hayrettin; Güntay, LeventTo enable a trustworthy clearing operation, clearinghouses require conservative margins to avoid the risk of incurring a loss in case one counterparty defaults. When margin requirements for fixed income instruments are calculated, yield curves of each instrument are stressed using their first three principal components. All instruments in an account are then evaluated against each stressed yield curve and the margin requirement is calculated as the difference of the combined value of these instruments calculated with the worst of the stressed yield curves between their combined values calculated with related unstressed curves. The aim of this project is to construct a tool for applying principle component analysis (PCA) on daily zero coupon yield curve of Turkish Treasury Securities. The analysis employs a yield curve panel data set obtained consisting historical zero coupon yield curves. The data set includes interest rates of 60 different maturities varying between overnight and 15 years and 1250 daily observations between December 2010 and December 2015. The result of this analysis provides a method that could be run at the end of each clearing day to determine the major components of the yield curve such as level/height, slope and curvature that describes at least 95% of the variation in interest changes and subject to stress shocksMaster Term Project Loaction Based Recommendation System Using Flickr Geotagged Photos(MEF Üniversitesi, Fen Bilimleri Enstitüsü, 2018) Çakar, Erbil; Arslan, Şuayb Ş.Flickr is a worldwide photo sharing application. In this project, photos with location information provided by Flickr will be used. By utilizing this location information, a recommendation system that recommends users the most suitable photo shooting locations based on their historical data will be designed. During this system design phase, different methodologies will be tried and a research will be conducted on the methods that provide the most appropriate solution to the problem.