Geri Dön

Bayes ağları ile endüstri rekabet senaryolarının oluşturulması: Bulaşık makinesi sektörü uygulaması

Bayesian network modelling of industry competitiveness scenarios: Dishwasher industy application

  1. Tez No: 510634
  2. Yazar: BEGÜM ÜNLÜ
  3. Danışmanlar: PROF. DR. SEÇKİN POLAT
  4. Tez Türü: Yüksek Lisans
  5. Konular: Endüstri ve Endüstri Mühendisliği, Industrial and Industrial Engineering
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 2018
  8. Dil: Türkçe
  9. Üniversite: İstanbul Teknik Üniversitesi
  10. Enstitü: Fen Bilimleri Enstitüsü
  11. Ana Bilim Dalı: Endüstri Mühendisliği Ana Bilim Dalı
  12. Bilim Dalı: Mühendislik Yönetimi Bilim Dalı
  13. Sayfa Sayısı: 143

Özet

İlgili tez çalışmasının amacı değişen ve belirsizlik seviyesi hızla artmakta olan çevre dinamiklerinde firmaların genel ve görev çevresinin incelenmesi ve olası pazar girişlerinin göz önünde bulundurulması ile rekabet seviyesindeki değişimi farklı senaryolar altında modellemek ve gerçeği ortaya koyabilir bir model geliştirmektir. Amaç doğrultusunda çevre analizi ve analiz metotları gözden geçirilmiş; genel çevre ve görev çevresinin bütünsel olarak incelenmesi yardımıyla endüstriye etki edebilecek tüm faktörlerin kapsanması hedeflenmiştir. Her ne kadar çevre modelleri teorik ve sektörel bilgiye göre modellenebilse de rekabet seviyesinin ölçülebilir hale getirilmesi amacıyla da Bayes ağları metodu kullanılmıştur. Tezin amacı doğrultusunda çevre analizi, senaryo geliştirme ve Bayes ağları metotları olmak üzere üç ana odak çerçevesinde literatür araştırması ve ardından sektör uygulaması gerçekleştirilmiştir. Akademik çalışmalar incelendiğinde rekabetçilik ve rekabet seviyesinin belirlenmesi kapsamında sayısal ve yapısal metotlar içerisinde çevre analizi de birlikte düşünüldüğünde Porter'ın geliştirmiş olduğu 5F metodu tez kapsamında uygun bulunmuş ve belirli düzenlemelerle kullanılabilir hale getirilmiştir. Sayısal yöntemlerin endüstri rekabet seviyesinde pazar payları gibi değerleri dikkate aldığı; tez çalışmasındaki amaç doğrultusunda ise firmalardan ziyade genel olarak endüstrinin incelenmesinin daha faydalı olacağı düşünülmektedir. Genel çevre analizi için literatürde genel kabul görmüş PEST metodu üzerinde yoğunlaşılmıştır. Genel çevre, görev çevresine kıyasla daha dolaylı yoldan ve daha uzun vadede firmalara etkide bulunacak faktörleri barındırmaktadır. Modelde kullanılacak alt faktörlerin belirlenmesinde incelenen çalışmaların çoğunluğunda bahsi geçen ve model karmaşıklığını çözülebilir seviyelerde tutabilecek kadar seçim yapılmasına özen gösterilmiştir. Literatürde bu iki spesifik metodun birlikte kullanımı ve Bayes ağları ile modellenmesine yönelik bir çalışmaya rastlanmamıştır. Dolayısıyla alanında ilk kez çalışılacak modelin doğruluk seviyesini artırmak adına sektör, coğrafya ve zaman dilimi kısıtlanmıştır. Çalışmanın odağı Türkiye bulaşık makinesi üreticileri ve Türkiye tüketici pazarı için önümüzdeki 5-10 yıllık geleceğe dair farklı rekabetçilik artış senaryoları kurulması olarak belirlenmiştir. Devam eden aşamada teorik bilgi ile bu iki analiz metot faktörleri arasındaki ilişkiler belirlenmeye çalışılmış; kaynak ve veri yetersizliği sebebiyle eksik kalan bilgiler alanda ve sektörde belirli kriterleri sağlamakta olan uzmanlarla birebir görüşmeler yardımıyla elde edilmiştir. Modelin son haline kavuşmasını takiben ölçülebilir hale getirmek için Bayes ağları metodu doğrultusunda soru tabloları oluşturulmuştur. Önceden belirlenen zaman dilimi doğrultusunda geleceğe dair bahsi geçen bilgi gerekliliği veriler ile sağlanamamış; belirlenen uzmanların görüş ve deneyimlerine dayalı olarak birebir görüşmelerde doldurulmuştur. Literatür ve uzman görüşleri doğrultusunda belirlenen senaryolar için modifiye edilen model çıktılarının incelenmesi, değerlendirilmesi sonucunda uzmanlar ile model sonuçları teyit edilmiş ve çalışma başarıyla tamamlanmıştır. Çalışmanın klasik Bayes ağları metodu üzerinden yürütülmesi sebebiyle önerilen bazı döngüsel ilişkiler gözardı edilmiş ve yine literatürdeki belirsizlikler sebebiyle uzman görüşleri ayrı ayrı değerlendirilmiştir. Devam edecek akademik çalışmalar modelin daha geniş bir coğrafi alan ve pazar için uygulanması ya da model metodolojisinin iyileştirilmesi üzerine gerçekleştirilebilecektir.

Özet (Çeviri)

The purpose of the related thesis is to develop a model that can determine the change in competition level under different scenarios. The model built up with general and task environment methods in the literature. Our main goal is to see the impact of changes by examining the general and task environment of firms and increasing the level of uncertainty in the rapidly changing environment dynamics. Environmental analysis and analysis methods are examined; it is aimed to cover all the factors that can affect the industry with the help of holistic examination of the general environment and the task environment. Although environmental models can be modeled according to theoretical and sectoral knowledge, Bayesian networks are used to make the level of competition measurable. In the direction of the thesis, literature survey and sector application were carried out in three main focuses: environmental analysis, scenario development and Bayesian network methods. The PEST analysis methodology has been choosen for analyzing and modeling the general environment. Although main focus and factors of PEST methodology covers up politic, economic, social and technological changes, we search for sub-factors in these four main groups. Different academic papers has benn asnalyzed and built up an chart to decide the sub-factors. The general environment contains factors that are more indirect than the task environment and affect the firm in the longer run. In the majority of the studies investigated in determining the sub-factors to be used in the model, care has been taken to select minimum level of sub-levels of the sub-factors to hold the model complexity at resolvable levels. When academic studies are examined, the 5F method developed by Porter, in which the level of competitiveness and competition is determined and environmental analysis is considered together, has been deemed appropriate within the scope of the thesis and made available in certain arrangements. Classical point of view of the 5F analysis is to determine new level of the competition level within the industry in case of new entrants to the industry. Thesis's objective is to determine the change in the competition level in case of new entrants to the industry. That is why we mainly took the vital elements of the classical 5F methodology, but the final goal, which is the change in the competition level, has been changed. On the other hand, the sub-factors of the industry analysis took from classical 5F literature as detailed in the related sections in this paperwork. There has been no paper on the use of these two specific methods together and their modeling with Bayesian networks. Therefore, sector, geography and timeframe are restricted in order to increase the accuracy level of the model to be worked on for the first time in its field. The focus of activities for Turkey dishwasher manufacturers and consumer market in Turkey has been identified as the establishment of different scenarios for the future competitiveness increase next 5-10 years. In the ongoing phase, the theoretical knowledge and the relations between these two analysis method factors were tried to be determined; missing information due to inadequacy of resources and data were obtained with the help of one-by-one interviews with experts. Five experts have certain amount of dishwasher industry experience and enough time and knowledge for this study. There has been three phased face-to-face interviews. Within first interviews, experts analyzed the model that has been prepared from the informations gathered from literature. Suggestions of the experts reviewed based on predetermined criterias. All the experts suggested a gate that will connect the increase or decrease of the competition level to the current competition level to make the model more accurate and dynamic. These suggestions were made model cyclical. Because the study was conducted through the method of classical Bayesian networks, we ignore this relationship that has been suggested. In order to make it measurable following the finalization of the model, questionnaires were prepared in the direction of Bayesian networks method. The information requirement of the future in the direction of the predetermined time slot has not been provided with the data requirement. Second interviews contain the gathering informations of probability distributions of the factors to build up an Bayesian network. As there has been conflicts within the literature to compound the experts' knowledge, we split up the factors of model due to the expertise area of the experts. We choose a control factor group to cross check the knowledge differencies of the experts. There was no major distinctness on the knowledge level, so we proceed to the next level of the research. Literature research has been made to find an effective software to build the Bayesian network to analyze the outcomes. Although NETICA software has widespred usage in the area, the GENIE software has been choosen as it has free academic usage option. The final model and probability distributions have defined on the software and sensitivity analysis had conducted to evaluate the model. Sensitivity tables interpreted for the final node and all the nodes that has importance for the model. Afterwards the model has been proved to work correctly, scenario-building process began. As the prior examinations has been made within literature research on scenario building and techniques, we choose the appropriate sceanario building process due to aim and scope of the research. The optimum number of scenario is also hot topic within the literature. Although there are conflicts about the exact number, three to six is adequate. Five different scenario has been developed due to this statement and decrease the complexity of the interpretation process between different scenarios. Two main factor groups that has changed through different scenarios are the enterance levels and PEST factors. PEST factors took three different statement as same with current state, wors case, best case. The enterance levels are also has three states as same with current state, all the enterance will be made and there will be no enterance to the marketplace. All five scenarios has been reviewed both separately and together to understand the results deeply. First scenario has same future with current state has only enterance by competitors that has production facilities abroad and thinking to open a new facility in Turkey and same PEST factors. The current state of Turkey can be sum up as social factors that cover up the technology enthusiasm, increase of the population and the marriage level has positive scala. The political factors as political stability and tax policies are all positive. Economic factors as inflation ratio and exchange rates are all negative but the economic statement of the company has stabile state. Due to these differencies within the sub-factors' different scalas, different factor groups are affected on different directions on the model. For example the increase threat for the cost is affected negatively, but the potential demand affected positively. As a result of these diversities, the increase level of the competition still high. Second scenario is about the same PEST factors, but all possible enterances will be made within five to ten years. As these possible enterances has affected the potential demand reallocation, the buyer's power, the supplier's power and also the cost threats due to these reallocations in the demands, they are very important. The result is also bad, as the increase level on the competition has the highest level. Third scenario has the same PEST factors but there were no new enterance to the marketplace. It's increase level on the competition level is bigger than first scenario bu smaller than the second scenario as expected. Fourth scenario is about the optimistic PEST factors and current market entrances. The optimistic PEST factors can be explained as all the political factors for example the political stability is high, economical factors for example the exchange rates are low, social factors for example marriage level is high, technological factors as change ratio is low. We can compare the fourth scenario with the first due to the PEST factor changes. As expected the increase of the competition level is relatively low for the fourth scenario. Last scenario is about the worst PEST factors and the current market enterances. As the PEST factors effecs different factor groups in different directions, fifth scenario seems to effect positively the increase of the sector competition level. The last interviews are about to analyze the first outcome and discuss further details of scenarios and suggestions. Examination of the model outputs modified for the scenarios determined in line with the literature and expert opinions, model results with the experts have been confirmed, and the study has been completed successfully. After the research and the empirical study completed, the results has been reviewed and the new topics for further researches determined. As mentioned earlier, the research made in a restricted geographical area with restricted producer levels. Further researches can be made on wider areas to see the changes of the factors and also the probability distribution differencies for different scopes. Secondly, the study has been conducted due to the classic Bayesian networks and it restrain the cyclical relations between factors. New studies can be conducted with dynamic Bayesian networks or different methods in the literature to see the effect and develop the model further. Thirdly, the dishwasher sector has it's own dynamics. The new studies can be made for different industries to see the differencies on the models. Last but not the least, the number of experts may be increased and also merging the views of the expers may become another study field.

Benzer Tezler

  1. Bayes ağları ile uluslararası rekabetçilik ölçümü

    Measuring international competitiveness with Bayesian network

    RABİA YILMAZ

    Yüksek Lisans

    Türkçe

    Türkçe

    2024

    Endüstri ve Endüstri Mühendisliğiİstanbul Teknik Üniversitesi

    Endüstri Mühendisliği Ana Bilim Dalı

    PROF. DR. SEÇKİN POLAT

  2. Değer akış haritalarında belirlenen darboğazların çözümü için Bayes ağları ile senaryo üretimi: Çamaşır makinesi fabrikasında bir uygulama

    Bayesian network approach for solving bottlenecks exposed by value stream maps: Case study in washing machine manufacturing

    ÖZCAN AKÇAOĞLU

    Yüksek Lisans

    Türkçe

    Türkçe

    2012

    Endüstri ve Endüstri Mühendisliğiİstanbul Teknik Üniversitesi

    Endüstri Mühendisliği Ana Bilim Dalı

    DOÇ. DR. GÜLGÜN KAYAKUTLU

  3. Tedarik zinciri yönetiminde risk bileşenlerinin etki analizi için bütünleşik bir model önerisi

    An integrated model proposal for impact analysis of risk components in supply chain management

    HASİBE BERFU ÖZTÜRK

    Doktora

    Türkçe

    Türkçe

    2024

    Endüstri ve Endüstri Mühendisliğiİstanbul Üniversitesi-Cerrahpaşa

    Endüstri Mühendisliği Ana Bilim Dalı

    PROF. DR. ŞULE ÖNSEL EKİCİ

  4. Veri madenciliği ve yapay sinir ağları yöntemleri ile hazır giyim sektöründe talep tahmini

    Demand forecasting in the apparel industry with data mining methods and artificial neural networks

    ESRA ÇETİN

    Yüksek Lisans

    Türkçe

    Türkçe

    2023

    Endüstri ve Endüstri MühendisliğiKocaeli Üniversitesi

    Endüstri Mühendisliği Ana Bilim Dalı

    DR. ÖĞR. ÜYESİ ATAKAN ALKAN

  5. Destek vektör makineleri yardımıyla tüketici kredilerinin sınıflandırılması

    Classifying consumer loans by means of support vector machines

    KAYAHAN KAYA

    Yüksek Lisans

    Türkçe

    Türkçe

    2016

    Endüstri ve Endüstri Mühendisliğiİstanbul Teknik Üniversitesi

    İşletme Mühendisliği Ana Bilim Dalı

    PROF. DR. FERHAN ÇEBİ