• Korean Management Science Review
• /
• v.26 no.1
• /
• pp.183-196
• /
• 2009

Customer retention has been a pressing issue for companies to get and maintain the loyal customers in the competing environment. Lots of researchers make effort to seek the characteristics of the churning customers and the loyal customers using the data mining techniques such as decision tree. However, such existing researches don't consider relationships among customers. Social network analysis has been used to search relationships among social entities such as genetics network, traffic network, organization network and so on. In this study, a customer network is proposed to investigate the differences of network characteristics of churning customers and loyal customers. The customer networks are constructed by analyzing the real purchase data collected from a Korean cosmetic provider. We investigated whether the churning customers and the loyal customers have different degree centralities and densities of the customer networks. In addition, we compared products purchased by the churning customers and those by the loyal customers. Our data analysis results indicate that degree centrality and density of the churning customer network are higher than those of the loyal customer network, and the various products are purchased by churning customers rather than by the loyal customers. We expect that the suggested social network analysis is used to as a complementary analysis methodology with existing statistical analysis and data mining analysis.

• Proceedings of the KIEE Conference
• /
• 1997.07a
• /
• pp.171-173
• /
• 1997

This paper presents analysis of the radial force density in brushless DC motor of which distribution is not uniform in the axial direction. The analysis considering 3D shape of teeth and overhang is not only important but essential to calculate the radial force density that acts on the teeth of stator, because it is frequent source of vibration and changes at the end of teeth. For the analysis, a new 3D equivalent magnetic circuit network method taking into account movement of the rotor without remesh is proposed. The radial force density is calculated by Maxwell stress tensor and analyzed by discrete Fourier transform.

• Transactions on Control, Automation and Systems Engineering
• /
• v.2 no.4
• /
• pp.268-273
• /
• 2000

Plasma models are crucial to equipment design and process optimization. A radial basis function network(RBFN) in con-junction with statistical experimental design has been used to model a process plasma. A 2$^4$ full factorial experiment was employed to characterized a hemispherical inductively coupled plasma(HICP) in characterizing HICP, the factors that were varied in the design include source power, pressure, position of shuck holder, and Cl$_2$ flow rate. Using a Langmuir probe, plasma attributes were collected, which include typical electron density, electron temperature. and plasma potential as well as their spatial uniformity. Root mean-squared prediction errors of RBEN are 0.409(10(sup)12/㎤), 0.277(eV), and 0.699(V), for electron density, electron temperature, and Plasma potential, respectively. For spatial uniformity data, they are 2.623(10(sup)12/㎤), 5.704(eV) and 3.481(V), for electron density, electron temperature, and plasma potential, respectively. Comparisons with generalized regression neural network(GRNN) revealed an improved prediction accuracy of RBFN as well as a comparable performance between GRNN and statistical response surface model. Both RBEN and GRNN, however, experienced difficulties in generalizing training data with smaller standard deviation.

• Proceedings of the Korean Radioactive Waste Society Conference
• /
• 2009.06a
• /
• pp.186-186
• /
• 2009

The characterization strategy of fracture networks are classified into a deterministic or statistical characterization according to the type of required information. A deterministic characterization is most efficient for a sparsely fractured system, while the statistics are sufficient for densely fractured rock. In this study, the ensemble mean and variability of the effective connectivity is systematically analyzed with various density values for different network structures of a power law size distribution. The results of high resolution Monte Carlo analyses show that statistical characteristics can be a necessary information to determine the transport properties of a fracture system when fracture density is greater than a percolation threshold. When the percolation probability (II) approaches unity with increasing fracture density, the effective connectivity of the network can be safely estimated using statistics only (sufficient condition). It is inferred from conditional simulations that deterministic information for main pathways can reduce the uncertainty in estimation of system properties when the network becomes denser. Overall results imply that most pathways need to be identified when II < 0.5 statistics are sufficient when II $\rightarrow$ 1 and statistics are necessary and the identification of main pathways can significantly reduce the uncertainty in estimation of transport properties when 0.5$\ll$1. It is suggested that the proper estimation of the percolation probability of a fracture network is a prerequisite for an appropriate conceptualization and further characterization.

• Journal of information and communication convergence engineering
• /
• v.16 no.2
• /
• pp.84-92
• /
• 2018

For large wireless sensor networks running on battery power, the time synchronization of all sensor nodes is becoming a crucial task for waking up sensor nodes with exact timing and controlling transmission and reception timing. However, as network size increases, this synchronization process tends to require long processing time consume significant power. Furthermore, a naïve synchronization scheduler may leave some nodes unsynchronized. This paper proposes a power-efficient scheduling algorithm for time synchronization utilizing the notion of density, which is defined by the number of neighboring nodes within wireless range. The proposed scheduling algorithm elects a sequence of minimal reference nodes that can complete the synchronization with the smallest possible number of hops and lowest possible power consumption. Additionally, it ensures coverage of all sensor nodes utilizing a two-pass synchronization scheduling process. We implemented the proposed synchronization algorithm in a network simulator. Extensive simulation results demonstrate that the proposed algorithm can reduce the power consumption required for the periodic synchronization process by up to 40% for large sensor networks compared to a simplistic multi-hop synchronization method.

• The Journal of the Acoustical Society of Korea
• /
• v.17 no.4E
• /
• pp.3-10
• /
• 1998

This paper concerns the dynamical behavior, in probabilistic sense, of a feedforward neural network performing auto association for novelty. Networks of retinotopic topology having a one-to-one correspondence between and output units can be readily trained using back-propagation algorithm, to perform autoassociative mappings. A novelty filter is obtained by subtracting the network output from the input vector. Then the presentation of a "familiar" pattern tends to evoke a null response ; but any anomalous component is enhanced. Such a behavior exhibits a promising feature for enhancement of weak signals in additive noise. As an analysis of the novelty filtering, this paper shows that the probability density function of the weigh converges to Gaussian when the input time series is statistically characterized by nonsymmetrical probability density functions. After output units are locally linearized, the recursive relation for updating the weight of the neural network is converted into a first-order random differential equation. Based on this equation it is shown that the probability density function of the weight satisfies the Fokker-Planck equation. By solving the Fokker-Planck equation, it is found that the weight is Gaussian distributed with time dependent mean and variance.

• Textile Coloration and Finishing
• /
• v.19 no.6
• /
• pp.7-11
• /
• 2007

A prediction model for yarn density profile was developed using the neural network methodology. The neural network model developed traces mass densities of a yarn within a section and predicts the mass profiles of the next yarn segment yet to be measured. The model does not require an assumption on the existence of a relationship between the past and future data sets. Four high-draft yarns made under different processing conditions were employed in order to test the performance of the model developed. It was shown that the model could predict the yarn density profiles without a significant error.

• Economic and Environmental Geology
• /
• v.50 no.1
• /
• pp.61-71
• /
• 2017

The effects of joint density and size distribution on the hydrogeologic characteristics of jointed rock masses are addressed through numerical experiments based on the 2-D DFN (discrete fracture network) fluid flow analysis. Using two joint sets, a total of 51 2-D joint network system were generated with various joint density and size distribution. Twelve fluid flow directions were chosen every $30^{\circ}$ starting at $0^{\circ}$, and total of 612 $20m{\times}20m$ DFN blocks were prepared to calculate the directional block conductivity. Also, the theoretical block conductivity, principal conductivity tensor and average block conductivity for each generated joint network system were determined. The directional block conductivity and chance for the equivalent continuum behavior of the 2-D DFN system were found to increase with the increase of joint density or size distribution. However, the anisotropy of block hydraulic conductivity increases with the increase of density discrepancy between the joint sets, and the chance for the equivalent continuum behavior were found to decrease. The smaller the intersection angle of the two joint sets, the more the equivalent continuum behavior were affected by the change of joint density and size distribution. Even though the intersection angle is small enough that it is difficult to have equivalent continuum behavior, the chance for anisotropic equivalent continuum behavior increases as joint density or size distribution increases.

• Journal of Distribution Research
• /
• v.16 no.2
• /
• pp.95-118
• /
• 2011

The purpose of this study is as follows. First, we investigate empirically the effects of social network properties such as social network density and centrality of a franchisee on its information sharing with various subjects such as the franchisor and other franchisees in the franchise system. Second, we examine exploratively if tie strength between a franchisee and its franchisor plays a moderating role on the relationship between social network properties and information sharing. The study model was established as shown in

• Journal of Korea Technology Innovation Society
• /
• v.4 no.2
• /
• pp.157-171
• /
• 2001

This paper discusses the economics of local loop architecture focusing on existing technologies, ADSL, HFC, and new one, PLC, and suggests a new modeling approach of access network system and the numerical equations. To modelize access network system and drive the numerical equations, we consider the double star and the tree & branch architecture and made block diagram of each access system. In addition, we introduce the density of subscriber as a variable and the equation of seeking the optimal number of cell in a service area. The economics of local loop architecture is analyzed in two ways, i.e. with and without consideration of the cost of cable and infrastructure. From the numerical analysis, we find that in case of not including the cost of cable and infrastructure, there is no much difference in the cost per one subscriber, while, in case of including it, there is remarkable difference among technologies. Therefore we conclude that the economics of local loop architecture is depend on the density of subscriber and existing network infrastructures.