Master Degree / Yüksek Lisans Tezleri

Permanent URI for this collectionhttps://hdl.handle.net/11147/3008

Browse

Filters

2002 - 20044

Settings

Author: search.filters.author.Püskülcü, HalisSubject: search.filters.subject.Data mining

Search Results

Now showing 1 - 4 of 4
  • Master Thesis
    Comparison of Different Algorithms for Exploting the Hidden Trends in Data Sources
    (Izmir Institute of Technology, 2003) Özsevim, Emrah; Püskülcü, Halis
    The growth of large-scale transactional databases, time-series databases and other kinds of databases has been giving rise to the development of several efficient algorithms that cope with the computationally expensive task of association rule mining.In this study, different algorithms, Apriori, FP-tree and CHARM, for exploiting the hidden trends such as frequent itemsets, frequent patterns, closed frequent itemsets respectively, were discussed and their performances were evaluated. The perfomances of the algorithms were measured at different support levels, and the algorithms were tested on different data sets (on both synthetic and real data sets). The algorihms were compared according to their, data preparation performances, mining performance, run time performances and knowledge extraction capabilities.The Apriori algorithm is the most prevalent algorithm of association rule mining which makes multiple passes over the database aiming at finding the set of frequent itemsets for each level. The FP-Tree algorithm is a scalable algorithm which finds the crucial information as regards the complete set of prefix paths, conditional pattern bases and frequent patterns by using a compact FP-Tree based mining method. The CHARM is a novel algorithm which brings remarkable improvements over existing association rule mining algorithms by proving the fact that mining the set of closed frequent itemsets is adequate instead of mining the set of all frequent itemsets.Related to our experimental results, we conclude that the Apriori algorithm demonstrates a good performance on sparse data sets. The Fp-tree algorithm extracts less association in comparison to Apriori, however it is completelty a feasable solution that facilitates mining dense data sets at low support levels. On the other hand, the CHARM algorithm is an appropriate algorithm for mining closed frequent itemsets (a substantial portion of frequent itemsets) on both sparse and dense data sets even at low levels of support.
  • Master Thesis
    Categorization of Web Sites in Turkey With Svm
    (Izmir Institute of Technology, 2004) Şimşek, Kadir; Püskülcü, Halis
    In this study of topic .Categorization of Web Sites in Turkey with SVM. after a brief introduction to what the World Wide Web is and a more detailed description of text categorization and web site categorization concepts, categorization of web sites including all prerequisites for classification task takes part. As an information resource the web has an undeniable importance in human life. However the huge structure of the web and its uncontrolled growth led to new information retrieval research areas to be risen in last years. Web mining, the general name of these studies, investigates activities and structures on the web to automatically discover and gather meaningful information from the web documents. It consists of three subfields: .Web Structure Mining., .Web Content Mining. and .Web Usage Mining.. In this project, web content mining concept was applied on the web sites in Turkey during the categorization process. Support Vector Machine, a supervised learning method based on statistics and principle of structural risk minimization is used as the machine learning technique for web site categorization. This thesis is intended to draw a conclusion about web site distributions with respect to thematic categorization based on text. The popular web directory Yahoo.s 12 top level categories were used in this project. Beside of the main purpose, we gathered several statistical descriptive informations about web sites and contents used in html pages. Metatag usage percentages, html design structures and plug-in usage are some of these information. The processes taken through solution, start with employing a web downloader which downloads web page contents and other information such as frame content from each web site. Next, manipulating, parsing and simplifying the downloaded documents takes place. At this point, preperations for categorization task are completed. Then, by applying Support Vector Machine (SVM) package SVMLight developed by Thorsten Joachims, web sites are classified under given categories. The classification results obtained in the last section show that there are some over-lapping categories exist and accuracy and precision values are between 60-80. In addition to categorization results, we saw that almost 17 of web sites utilize html frames and 9367 web sites include metakeywords.
  • Master Thesis
    Finding and Evaluating Patterns in Web Repository Using Database Technology and Data Mining Algorithms
    (Izmir Institute of Technology, 2002) Özakar, Belgin; Püskülcü, Halis
    Web mining is a very hot research topic, which combines two of the active research areas: Data Mining and World Wide Web. The Web mining research relates to several research communities such as Database, Statistics, Artificial Intelligence and Visualization. Although there exists some confusion about the Web mining, the most recognized approach is to categorize Web mining into three areas: Web content mining, Web structure mining, and Web usage mining. Web content mining focuses on the discovery/retrieval of the useful information from the Web contents/data/documents, while the Web structure mining emphasizes to the discovery of how to model the underlying link structures of the Web. Sometimes the distinction between these two categories is not very clear. Web usage mining is relatively independent, but not isolated category, in which the following studies continue; General Web Usage Mining, Site Modification, Systems Improvement and Personalization. General Web Usage Mining systems aim to discover general trends and patterns from the log files by adapting data mining techniques. The objective of the Site Modification systems is to improve the design of a web site by suggesting modifications in its content and structure. The research on System Improvement focuses on using the web usage mining for improving the web traffic. Finally, personalization systems aim to understand individual trends used for personalizing the web sites. The study subject to this thesis, IYTE Web Usage Mining (WUM) System was an example of system development in the field of General Web Usage Mining with a database approach where the flexible query capability of SQL (Structured Query Language) was explored. The data mining and database techniques were applied on the access/error/user logs of the web server of Izmir Institute of Technology. The main objective was to create a site improvement tool for the web administrator by reporting the distribution of the hits received by the web server according to the time stamp, users, service and URL types and at the same time revealing the nature of the errors generated by the web server. All data cleaning and transaction identification processes were handled by the software routines coded in Java. Clean transactions were imported into IYTE Web Usage Mining (IYTE WUM) relational database. Flexible features of SQL were utilized for application of algorithm Apriori to discover most frequent pair of URL s visited, in addition to extraction of general knowledge from data.
  • Master Thesis
    Improvements in K-Means Algorithm To Execute on Large Amounts of Data
    (Izmir Institute of Technology, 2004) Sülün, Erhan; Püskülcü, Halis
    By the help of large storage capacities of current computer systems, datasets of companies has expanded dramatically in recent years. Rapid growth of current companies. databases has raised the need of faster data mining algorithms as time is very critical for those companies.Large amounts of datasets have historical data about the transactions of companies which hold valuable hidden patterns which can provide competitive advantage to them. As time is also very important for these companies, they need to mine these huge databases and make accurate decisions in short durations in order to gain marketing advantage. Therefore, classical data mining algorithms need to be revised such that they discover hidden patterns and relationships in databases in shorter durations.In this project, K-means data mining algorithm has been proposed to be improved in performance in order to cluster large datasets in shorter time. Algorithm is decided to be improved by using parallelization. Parallelization of the algorithm has been considered to be a suitable solution as the popular way of increasing computation power is to connect computers and execute algorithms simultaneously on network of computers. This popularity also increases the availability of parallel computation clusters day by day. Parallel version of the K-means algorithm has been designed and implemented by using C language. For the parallelisation, MPI (Message Passing Interface) library hasbeen used. Serial algorithm has also been implemented by using C language for the purpose of comparison. And then, algorithms have been run for several times under same conditions and results have been discussed. Summarized results of these executions by using tables and graphics has showed that parallelization of the K-means algorithm has provied a performance gain almost proportional by the count of computers used for parallel execution.