Chieh-Yuan Tsai

A purchase-based market segmentation methodology

Abstract Market segmentation is critical for a good marketing and customer relationship management program. Traditionally, a marketer segments a market using general variables such as customer demographics and lifestyle. However, several problems have been identified and make the segmentation result unreliable. This paper develops a novel market segmentation methodology based on product specific variables such as purchased items and the associative monetary expenses from the transactional history of customers to resolve these problems. A purchase-based similarity measure, clustering algorithm, and clustering quality function are defined in this paper. A genetic algorithm approach is adopted to ensure that customers in the same cluster have the closest purchase patterns. After completing segmentation, a designated RFM model is used to analyze the relative profitability of each customer cluster. The findings from a practical marketing implementation study will also be discussed.

Developing a feature weight self-adjustment mechanism for a K-means clustering algorithm

K-means is one of the most popular and widespread partitioning clustering algorithms due to its superior scalability and efficiency. Typically, the K-means algorithm treats all features fairly and sets weights of all features equally when evaluating dissimilarity. However, a meaningful clustering phenomenon often occurs in a subspace defined by a specific subset of all features. To address this issue, this paper proposes a novel feature weight self-adjustment (FWSA) mechanism embedded into K-means in order to improve the clustering quality of K-means. In the FWSA mechanism, finding feature weights is modeled as an optimization problem to simultaneously minimize the separations within clusters and maximize the separations between clusters. With this objective, the adjustment margin of a feature weight can be derived based on the importance of the feature to the clustering quality. At each iteration in K-means, all feature weights are adaptively updated by adding their respective adjustment margins. A number of synthetic and real data are experimented on to show the benefits of the proposed FWAS mechanism. In addition, when compared to a recent similar feature weighting work, the proposed mechanism illustrates several advantages in both the theoretical and experimental results.

A Web services-based collaborative scheme for credit card fraud detection

A Web services-based collaborative scheme for credit card fraud detection is proposed. With the proposed scheme, participant banks can share the knowledge about fraud patterns in a heterogeneous and distributed environment and further enhance their fraud detection capability and reduce financial loss.

A hybrid ANP model in fuzzy environments for strategic alliance partner selection in the airline industry

Strategic airline alliances are an increasingly common strategy for enhancing airline competitiveness and satisfying customer needs, especially in an era characterized by blurring industry boundaries, fast-changing technologies, and global integration. Airlines have been very active in utilizing this form of strategic development. However, the selection of a suitable partner for a strategic alliance is not an easy decision, involving a host of complex considerations by different departments. Furthermore the decision-makers may hold diverse opinions and preferences arising due to incomplete information and knowledge or inherent conflict between various departments. In this study fuzzy preference programming and the analytic network process (ANP) are combined to form a model for the selection of partners for strategic alliances. The effects of uncertainty and disagreement between decision-makers as well as the interdependency and feedback that arise from the use of different criteria and alternatives are also addressed. This generic model can be easily extended to fulfill the specific needs of a variety of companies.

A case-based reasoning system for PCB defect prediction

The manufacturing process for a new Printed Circuit Board (PCB) design is often instable and might generate a number of defects during the complicated production process. Defects reduce the yield rate and increase the production costs. Although skilled engineers can predict the possible defect items for a new PCB product, this approach requires strong engineering experience and is time consuming. To conquer this problem, this research applies case-based reasoning (CBR) methodology to develop a defect prediction system for new PCB products. In the CBR system, each case is represented using the design specifications, defect items and corresponding costs. A vantage-based case indexing mechanism is developed to accelerate the case retrieval efficiency. In addition, a reasoning algorithm that considers the defect cost is proposed to infer the defect items that are interesting to PCB manufacturers. The system performance is analyzed to show the efficiency and accuracy of the proposed system. A practical implementation using a case-base provided by a PCB manufacturer is demonstrated.

A dynamic Web service based data mining process system

This research introduces a dynamic data mining process (DDMP) system based on service-oriented architecture (SOA). Each activity in data mining process is viewed as a Web service operated on Internet. The Web services provide functions of data preprocessing, data mining algorithms, and visualization analysis. Depending on users requirement, the Web services are dynamically linked using Business Process Execution Language for Web service (BPEL4WS) to construct a desired data mining process. Finally, the result model described by Predictive Model Markup Language (PMML) is returned for further analysis. A practical credit card data set provided by a commercial bank is implemented using the proposed system. It is found that the DDMP system can provide dynamical and satisfied analysis result to the enterprise.

A k-Anonymity Clustering Method for Effective Data Privacy Preservation

Data privacy preservation has drawn considerable interests in data mining research recently. The k-anonymity model is a simple and practical approach for data privacy preservation. This paper proposes a novel clustering method for conducting the k-anonymity model effectively. In the proposed clustering method, feature weights are automatically adjusted so that the information distortion can be reduced. A set of experiments show that the proposed method keeps the benefit of scalability and computational efficiency when comparing to other popular clustering algorithms.

A case-based reasoning system for PCB principal process parameter identification

The Printed Circuit Board (PCB) manufacturing process usually consists of lengthy production activities. Each activity is controlled by a number of process parameters. Although numerous process parameters must be determined before fabrication, only a number of parameters called principal process parameters because they affect the quality of a PCB product. As long as the principal process parameters are identified efficiently and controlled well, the manufacturing lead-time can be shortened and the quality of the new PCB product can be assured. This research proposes a Case-Based Reasoning (CBR) system to infer the principal process parameters for a new PCB product. Each case in the case-base stores design specifications, process parameters, and the corresponding production quality specifications. A Significant Nearest Neighbor (SNN) search is developed to retrieve similar cases from a case-base. A Mutual Correlation Parameter Selection (MCPS) method and a correlation-based parameter setting method are developed to identify the principal parameters and infer their reasonable value range. A set of experiments and a practical implementation case are demonstrated to show the efficiency and accuracy of the proposed system.

A change detection method for sequential patterns

Recent trends in customer-oriented markets drive many researchers to develop sequential pattern mining algorithms to explore consumer behaviors. However, most of these studies concentrated on how to improve accuracy and efficiency of their methods, and seldom discussed how to detect sequential pattern changes between two time-periods. To help business managers understand the changing behaviors of their customers, a three-phase sequential pattern change detection framework is proposed in this paper. In phase I, two sequential pattern sets are generated respectively from two time-period databases. In phase II, the dissimilarities between all pairs of sequential patterns are evaluated using the proposed sequential pattern matching algorithm. Based on a set of judgment criteria, a sequential pattern is clarified as one of the following three change types: an emerging sequential pattern, an unexpected sequence change, or an added/perished sequential pattern. In phase III, significant change patterns are returned to managers if the degree of change for a pattern is large enough. A practical transaction database is demonstrated to show how the proposed framework helps managers to analyze their customers and make better marketing strategies.

A two-stage fuzzy approach to feature-based design retrieval

In the conceptual design stage, designers usually begin the design process by referencing previous design cases similar to the one they are currently working on. Reviewing similar design cases can inspire a new idea and also shorten the design lead-time. To enhance this need, a two-stage fuzzy approach to feature-based design retrieval is developed in this research. First, a Fuzzy ART network is developed to search for designs based on geometric similarity. By adjusting a Fuzzy ART network parameter, designers can control the similarity of retrieved reference designs. Second, a fuzzy evaluation procedure is developed for finding reference designs with similar technological specifications. This procedure is capable of dealing with qualitative and quantitative technological attribute data. Implementation examples show the robustness of the proposed system in that it can flexibly retrieve reference designs based on geometric features and technological attributes even when incomplete query is provided.