FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2282008049> ?p ?o ?g. }

• W2282008049 abstract "Mobile phones have entered the daily lives of people around the globe. Various mobile data services are provided to customers these days and they are becoming more and more popular. In Hong Kong the penetration rate of mobile phones is more than 100% which means on an average people own more than one mobile phone. On the other hand there are six mobile telecommunication operators in Hong Kong. So the competition to acquire and retain customers among mobile service providers is fierce. The key to survival in this competitive industry lies in knowing the customers better. Different people have different preferences in using mobile telecommunication services and mobile phones. According to IDC group’s study of usage patterns of mobile data services across the Asia Pacific region, SMS is the most popular mobile service used [1]. About 65% of customers send SMS everyday. Only 35% of customers do not use SMS that frequently. Treating all customers without differentiation may lead to the situation that some customers have to choose services they do not want and this may lead to loss of customers. One of the approaches used to understand customers is customer clustering. Clustering classifies customers into different groups in order to see similarities and dissimilarities between customers. Mobile service providers can develop different mobile services for different clusters in order to match the services to the customers’ preferences. In mobile marketing, the usage information of mobile telecommunication services is the best data that can be used to reflect customers’ behaviors and preferences. However, relatively little attention has been paid in academic research on applying clustering techniques using mobile telecommunication usage data for customer profiling. In this research, we perform customer clustering using mobile telecommunication usage data to identify interesting facts about customers which may be of use to mobile marketers. Data mining techniques are employed in this research and the data mining process model of CRoss Industry Standard Process, CRISP-DM in short, is followed [2]. This process model describes a standard data mining project lifecycle [3] [4]. The CRISP-DM process has 6 phases: business understanding, data understanding, data preparation, modeling, evaluation, and deployment. Unsupervised clustering techniques are used because we are exploring the data set and do not have any target. We use K-means clustering and Kohonen Vector Quantization networks (KVQ). Kohonen Vector Quantization networks are an unsupervised clustering technique closely related to K-means cluster analysis and Self-Organizing Maps (SOM) [5]. K-means begins clustering by selecting K (specified number of clusters) seeds (centers of clusters) according to the distribution of the data set. And these seeds selected by K-means tend to be approximately uniformly distributed. In other words K-means assumes a uniform distribution of the data set. In contrast, KVQ selects code book vectors, randomly and the distribution probability density function is approximated by a set of optimally placed discrete parameter vectors. Code book vectors which are closest to each cluster are found and moved closer to the clusters by a certain portion. The portion is specified by the learning ratio. KVQ and SOM have similar learning algorithms but SOM considers both distances in input space and distances in the map while KVQ only considers the former. We collected mobile telecommunication data from a major local mobile telecommunication operator. The data set contains information on customers’ demographics, handset features, usage records of mobile services, registered services, and amount of revenue contributed by customers. There are about 52000 records and each record is characterized by 200 attributes. We divide the attributes into four groups: usage attributes, revenue attributes, service attributes, and handset attributes. By clustering customers based on these four groups of attributes we create four profiles of customers: usage profile, revenue profile, service profile, and handset profile. The analysis of the clustering results contains two steps. First the results obtained using the two clustering techniques are compared. The criteria used for comparison are cluster numbers, average distances, average CV indexes (ratio of standard deviation and mean of distances). Distance in the context of clustering refers to Euclidean distances between data points and the seeds of clusters. Lower average distance or lower CV index means higher cluster concentration and hence higher cluster quality. The second step involves examination of the clustering results obtained by KVQ to discover relationships between different profiles of customers. Several interesting facts about customers are discovered that may be of help to mobile marketers for carrying out targeted marketing campaigns. Our results show that KVQ approach is able to identify clusters with lower average distances than K-means approach. On the other hand clusters identified by the K-means approach have lower CV indexes. We chose to further analyze the PROFILING CUSTOMERS OF MOBILE TELECOMMUNICATION SERVICES 389 results of KVQ because the clusters it identified had smaller average distances while the CV indexes of them are acceptable compared to K-means. We identified six groups using usage data. Each of these groups showed a different usage pattern. Some customers tended to use more mobile IDD calls. Some tended to use more SMS. Others tended to use more of other services. For marketers this might indicate that there are opportunities to recommend services to customers according to the profile of customers. For researchers it might be interesting to compare these customers to find out the reason for the discrepancy in usage patterns for a certain type of mobile telecommunication service. The four profiles of customers are not closely related. By comparing service profile and revenue profile we found that most people register available services but do not use them often because most services charge customers money only when they use them. This might indicate that either the customers think the usefulness of the mobile services provided to them is low or they have not realized the usefulness of these services. Thus, researchers and marketers of mobile telecommunications could think of studying the customers’ preferences and develop more personalized mobile communication services for them. By comparing usage profile and revenue profile we found that customers’ usage of mobile communication services and revenues they generated are unbalanced. Some customers who had low usage of services contributed higher revenues. This might be good for the operators at present but might be dangerous in future. When these customers would realize the unbalance between the money they paid and the service they used, they might either switch to plans with low revenue or even switch to other companies. Marketers should take suitable actions to either provide these customers more valuable services or advise them to use more suitable plans with lower fees. For researchers this result implies that when studying customer churn type problems the factors related to unbalance between usage and revenue should be taken into careful consideration. We also noticed that customers that had high usage contributed less money and this indicated the customers’ potential to generate more profit. Packages with higher level of services might be promoted to these customers. Finally we compared handset profile and revenue profile and found that if researchers wanted to include the factor of handset in models studying customers’ mobile telecommunication behavior patterns it might not be necessary to develop many levels of value for that factor." @default.
• W2282008049 created "2016-06-24" @default.
• W2282008049 creator A5045996720 @default.
• W2282008049 creator A5049351897 @default.
• W2282008049 date "2006-01-01" @default.
• W2282008049 modified "2023-10-02" @default.
• W2282008049 title "Profiling Customers of Mobile Telecommunications Services" @default.
• W2282008049 cites W1679913846 @default.
• W2282008049 cites W2050605423 @default.
• W2282008049 hasPublicationYear "2006" @default.
• W2282008049 type Work @default.
• W2282008049 sameAs 2282008049 @default.
• W2282008049 citedByCount "0" @default.
• W2282008049 crossrefType "journal-article" @default.
• W2282008049 hasAuthorship W2282008049A5045996720 @default.
• W2282008049 hasAuthorship W2282008049A5049351897 @default.
• W2282008049 hasConcept C111919701 @default.
• W2282008049 hasConcept C144133560 @default.
• W2282008049 hasConcept C187191949 @default.
• W2282008049 hasConcept C2781307350 @default.
• W2282008049 hasConcept C41008148 @default.
• W2282008049 hasConcept C76155785 @default.
• W2282008049 hasConcept C95491727 @default.
• W2282008049 hasConceptScore W2282008049C111919701 @default.
• W2282008049 hasConceptScore W2282008049C144133560 @default.
• W2282008049 hasConceptScore W2282008049C187191949 @default.
• W2282008049 hasConceptScore W2282008049C2781307350 @default.
• W2282008049 hasConceptScore W2282008049C41008148 @default.
• W2282008049 hasConceptScore W2282008049C76155785 @default.
• W2282008049 hasConceptScore W2282008049C95491727 @default.
• W2282008049 hasLocation W22820080491 @default.
• W2282008049 hasOpenAccess W2282008049 @default.
• W2282008049 hasPrimaryLocation W22820080491 @default.
• W2282008049 hasRelatedWork W1022845921 @default.
• W2282008049 hasRelatedWork W1496346451 @default.
• W2282008049 hasRelatedWork W1736042454 @default.
• W2282008049 hasRelatedWork W1789954112 @default.
• W2282008049 hasRelatedWork W182232572 @default.
• W2282008049 hasRelatedWork W2031072151 @default.
• W2282008049 hasRelatedWork W2050590272 @default.
• W2282008049 hasRelatedWork W2085221034 @default.
• W2282008049 hasRelatedWork W2134711594 @default.
• W2282008049 hasRelatedWork W2152553657 @default.
• W2282008049 hasRelatedWork W2166200805 @default.
• W2282008049 hasRelatedWork W2167435361 @default.
• W2282008049 hasRelatedWork W2167752114 @default.
• W2282008049 hasRelatedWork W2377782253 @default.
• W2282008049 hasRelatedWork W2392324524 @default.
• W2282008049 hasRelatedWork W2541146452 @default.
• W2282008049 hasRelatedWork W744965803 @default.
• W2282008049 hasRelatedWork W1786870252 @default.
• W2282008049 hasRelatedWork W2136075006 @default.
• W2282008049 hasRelatedWork W3118436349 @default.
• W2282008049 isParatext "false" @default.
• W2282008049 isRetracted "false" @default.
• W2282008049 magId "2282008049" @default.
• W2282008049 workType "article" @default.