• Journal of Intelligence and Information Systems
• /
• v.8 no.2
• /
• pp.15-35
• /
• 2002

Data mining has received a lot of attention from practitioners. That is partly because it allows company to extract a set of useful knowledge about customers from database, thereby retaining current customers and magneting potential customers. This logic is especially essential in the field of credit card industry, where just 5% increase of number of customers is hewn to cause 120% increase in profit. The problem is how to retain current customers and even make them more loyal to company. However, previous studies lacked proposing extensive strategies of reducing the churning rate. In this sense, this study attempts to suggest such strategies by applying neural network, logistic regression, and C5.0 techniques to credit card data. We used a real data set of four years from 1997 to 2000, which were gathered from a credit card company. Experimental results revealed that our approach could yield robust strategies for retaining customers by reducing the churning rate.

• Journal of Korea Multimedia Society
• /
• v.20 no.1
• /
• pp.1-9
• /
• 2017

Explosive deployment of smart mobile devices such as smart phones, and tablets along with expansion of wireless internet bandwidth have enabled the deployment of mobile video streaming such that video traffic becomes the most important service in wireless networks. Recently, for more efficient video streaming services, the ISO MPEG group standardized a protocol called DASH (Dynamic Adaptive Streaming over HTTP) and the standard has been quickly adopted by many service providers such as YouTube and Netflix. Despite of the convenience of mobile streaming services, users also suffer from low QoE(Quality of Experience) due to dynamic channel fluctuations and unnecessary downloading due to high churning rates. This paper proposes a noble efficient video rate selection algorithm considering user buffer level, channel condition and churning rate. Computer simulation based performance study showed that the proposed algorithm improved the QoE significantly compared to a method that determines the video rate based on current channel conditions. Especially, the proposed method reduced the rebuffering rate, one of the most important performance factors of the QoE, to a nonnegligible level.

• Journal of Intelligence and Information Systems
• /
• v.7 no.1
• /
• pp.17-25
• /
• 2001

Tree model is the most popular classification algorithm in data mining due to easy interpretation of the result. In CART(Breiman et al., 1984) and C4.5(Quinlan, 1993) which are representative of tree algorithms, the split fur classification proceeds to attain the homogeneous terminal nodes with respect to the composition of levels in target variable. But, fur instance, in the chum prediction modeling fur CRM(Customer Relationship management), the rate of churn is generally very low although we are interested in mining the churners. Thus it is difficult to get accurate prediction modes using tree model based on the traditional split rule, such as mini or deviance. Buja and Lee(1999) introduced a new split rule, one-sided purity for classifying minor interesting group. In this paper, we compared one-sided purity with traditional split rule, deviance analyzing churning vs. non-churning data of ISP company. Also reviewing the result of tree model based on one-sided purity with some simulated data, we discussed problems and researchable topics.

• Asia Marketing Journal
• /
• v.7 no.1
• /
• pp.21-42
• /
• 2005

Customer Relationship Management (CRM) is a corporate marketing strategy maintaining and managing customers. And with CRM companies maximize the customer's value through a series of processes of new customer retention, VIP customer retention, customer value increase, potential customer activation, and customers for lifetime by collecting the customer information and taking advantage of it effectively. In particular, as the competitive environment is changing rapidly and getting more intense, maintaining the customer retention through customer churn management becomes more important in order to increase the customer value for maximizing the company's profit and to build up the relationship with customers. For example, the financial industry has managed the customer churn with the concept of customer segmentation. Recently the customer retention and churn management is becoming increasingly important in all business fields as well as financial industry since the companies expect the effect of preventing the customer churn by identifying characteristics of customers. However, despite the increasing interest and importance of the management of the customer churn, not many of studies are systematically executed by analyzing the data of customer churn. In this study we analyze the actual data of CRM activities for the customer retention, specifically the data of TV home-shopping. By doing so, we hope to identify the differences of demographic attributes and transaction specific characteristics in consumer behaviors between the churning customer and the retained customers. In addition, we try to find out the variables which can impact the churning of the customers and to predict the churn rate of individual customer through our proposed model of customer churn. In the end, based on our findings we suggest the possible marketing strategies for TV home-shopping companies.

• Industrial Engineering and Management Systems
• /
• v.13 no.4
• /
• pp.454-462
• /
• 2014

Retention of possible churning customer is one of the most important issues in customer relationship management, so companies try to predict churn customers using their large-scale high-dimensional data. This study focuses on dealing with large data sets by reducing the dimensionality. By using six different dimension reduction methods-Principal Component Analysis (PCA), factor analysis (FA), locally linear embedding (LLE), local tangent space alignment (LTSA), locally preserving projections (LPP), and deep auto-encoder-our experiments apply each dimension reduction method to the training data, build a classification model using the mapped data and then measure the performance using hit rate to compare the dimension reduction methods. In the result, PCA shows good performance despite its simplicity, and the deep auto-encoder gives the best overall performance. These results can be explained by the characteristics of the churn prediction data that is highly correlated and overlapped over the classes. We also proposed a simple out-of-sample extension method for the nonlinear dimension reduction methods, LLE and LTSA, utilizing the characteristic of the data.