Tez

Permanent URI for this collectionhttp://acikerisim.bau.edu.tr:4000/handle/123456789/160

Browse

Filters

20112

Settings

Subject: search.filters.subject.Churn management

Search Results

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Churn management by using fuzzy c-means
    (Bahçeşehir Üniversitesi Fen Bilimleri Enstitüsü, 2011-09) Arifoğlu, Evren; Karahoca, Adem
    Nowadays, Global Service of Mobile Communication (GSM) market is a huge sector in nations' economies. Voice quality is an important factor for a customer to choose a GSM operator and hence GSM companies increases their voice quality via 3G technologies. Also there are other factors which affect a consumer to prefer a particular GSM operator. Due to several reasons, customers change their current GSM operators. It is very important for GSM operators to predict if a subscriber will cancel the service and switch to another GSM operator. Therefore, companies that provide GSM services have to monitor the behavior of each subscriber and predict one step ahead. In this study, using fuzzy c-means algorithm, we aim to predict whether a subscriber will change her current GSM operator or not. We also compare fuzzy c-means algorithm with Decision Tree, Nai"ve Bayes and Support Vector Machine and Probabilistic Neural Network. At the end of this study we expect that fuzzy c-means will give best result.
  • No Thumbnail Available
    Item
    Customer segmentation for churn management by using ant colony
    (Bahçeşehir Üniversitesi Fen Bilimleri Enstitüsü, 2011-09) Güllüoğlu, Batuhan; Karahoca, Adem
    Data mining is interested in clustering, by similarities of data. Some of clustering techniques are evolutionary and optimization techniques. Characteristic selection is used for novel hybrid modeling. Customer priorities are very important for companies. Moreover, customer priorities must be determined, and campaigns must be ordered according to these priorities. Customer segmentation was done with Ant Colony algorithm. Shortest path approach is used in Ant Colony algorithm. Moreover, clustering is done by the euclidean distance formula in Ant Colony algorithm. Customer segmentation attributes are mostly related with the satisfaction factors, but some of them were eliminated by using ranker. These results are mostly related with the customer's income, tenure, equip, callcard and reside. These attributes are the most important satisfaction factors not to lose customers as expected. There are many reasons in changing GSM operator for subscribers and it is very important for companies to predict if subscriber will change GSM operator or not. For this reason companies that gives GSM services have to monitor subscribers behavior and predict one step forward. In this study changing subscribers’ GSM operator will be predicted by using data mining techniques.