Chih-Ping Wei

Nearest-neighbor-based approach to time-series classification

Many interesting applications involve predictions based on a time-series sequence or a set of time-series sequences, which are referred to as time-series classification problems. Prior classification analysis research predominately focuses on constructing a classification model from training instances that involve non-time-series attributes. Direct application of traditional classification analysis techniques to time-series classification problems requires the transformation of time-series attributes into non-time-series ones by applying some statistical operations (e.g., average, sum, variance). However, such statistical-transformation-based approach often results in information loss and, in turn, imperils classification effectiveness. In this study, we propose a time-series classification technique based on the k-nearest-neighbor (kNN) classification approach. Using churn prediction of the mobile telecommunications industry as an evaluation application, our empirical evaluation results show that the proposed kNN-based time-series classification (kNN-TSC) technique achieves better performance (measured by miss and false alarm rates) than the statistical-transformation-based approach does. Highlights? We highlight the importance of time-series classification problems. ? We develop the kNN-TSC technique for time-series classification problems. ? We evaluate kNN-TSC using churn prediction as the evaluation application. ? kNN-TSC generally outperforms the benchmark technique in miss and false alarm rates. ? kNN-TSC with the stratified average method is capable of handling imbalanced data sets.

Predicting the length of hospital stay of burn patients: Comparisons of prediction accuracy among different clinical stages

A burn injury is a disastrous trauma and can have wide-ranging impacts on burn patients, their families, and society. Burn patients generally experience long hospital stays, and the accurate prediction of the length of those stays has strong implications for healthcare resource management and service delivery. In addition to prediction accuracy, the timing of length of hospital stay (LOS) predictions is also relevant, because LOS predictions during earlier clinical stages (e.g., admission) can provide an important component for service and resource planning as well as patient and family counseling, whereas LOS predictions at later clinical stages (e.g., post-treatment) can support resource utilization reviews and cost controls. This study evaluates the effectiveness of LOS predictions for burn patients during three different clinical stages: admission, acute, and post-treatment. In addition, we compare the prediction effectiveness of two artificial intelligence (AI)-based prediction techniques (i.e., model-tree-based regression and support vector machine regression), using linear regression analysis as our performance benchmark. On the basis of 1080 burn cases collected in Taiwan, the empirical evaluation suggests that the accuracy of LOS predictions at the acute stage does not improve compared with those during the admission stage, but LOS predictions at the post-treatment stage are significantly more accurate. Moreover, the AI-based prediction techniques, especially support vector machine regression, appear more effective than the regression technique for LOS predictions for burn patients across stages.

Green Purchasing by MNC Subsidiaries: The Role of Local Tailoring in the Presence of Institutional Duality†

To examine the essential determinants of green purchasing by multinational corporations� (MNC) subsidiaries, this study takes institutional theory as a foundation and focuses on the institutional duality associated with localization and globalization. Specifically, we develop a model to explain subsidiaries� green purchasing and empirically test the model with data from 141 purchasing managers and senior purchasing staff members from subsidiaries in 39 countries. Our results suggest that pressures from headquarters and the local environment do not affect subsidiaries� green purchasing directly; rather, they exert indirect influences through local tailoring. This study contributes to extant literature by revealing the significance of local tailoring in an MNC context. In addition, our findings offer several implications for practice by providing a roadmap for disseminating green purchasing across the subsidiaries of an MNC, as well as highlighting the importance of both clear communication about the benefits of green purchasing and internal audits.

Exploiting Technological Indicators for Effective Technology Merger and Acquisition (M&A) Predictions†

Mergers and acquisitions (M&A) play increasingly important roles for contemporary business, especially in high-tech industries that conduct M&As to pursue complementarity from other companies and thereby preserve or extend their competitive advantages. The appropriate selection (prediction) of M&A targets for a given bidder company constitutes a critical first step for an effective technology M&A activity. Yet existing studies only employ financial and managerial indicators when constructing M&A prediction models, and select candidate target companies without considering the profile of the bidder company or its technological compatibility with candidate target companies. Such limitations greatly restrict the applicability of existing studies to supporting technology M&A predictions. To address these limitations, we propose a technology M&A prediction technique that encompasses technological indicators as independent variables and accounts for the technological profiles of both bidder and candidate target companies. Forty-three technological indicators are derived from patent documents and an ensemble learning method is developed for our proposed technology M&A prediction technique. Our evaluation results, on the basis of the M&A cases between January 1997 and May 2008 that involve companies in Japan and Taiwan, confirm the viability and applicability of the proposed technology M&A prediction technique. In addition, our evaluation also suggests that the incorporation of the technological profiles and compatibility of both bidder and candidate target companies as predictors significantly improves the effectiveness of relevant predictions.

Exploiting temporal characteristics of features for effectively discovering event episodes from news corpora

An organization performing environmental scanning generally monitors or tracks various events concerning its external environment. One of the major resources for environmental scanning is online news documents, which are readily accessible on news websites or infomediaries. However, the proliferation of the World Wide Web, which increases information sources and improves information circulation, has vastly expanded the amount of information to be scanned. Thus, it is essential to develop an effective event episode discovery mechanism to organize news documents pertaining to an event of interest. In this study, we propose two new metrics, Term Frequency × Inverse Document FrequencyTempo (TF×IDFTempo) and TF×Enhanced‐IDFTempo, and develop a temporal‐based event episode discovery (TEED) technique that uses the proposed metrics for feature selection and document representation. Using a traditional TF×IDF‐based hierarchical agglomerative clustering technique as a performance benchmark, our empirical evaluation reveals that the proposed TEED technique outperforms its benchmark, as measured by cluster recall and cluster precision. In addition, the use of TF×Enhanced‐IDFTempo significantly improves the effectiveness of event episode discovery when compared with the use of TF×IDFTempo.

Exploiting poly-lingual documents for improving text categorization effectiveness

With the globalization of business environments and rapid emergence and proliferation of the Internet, organizations or individuals often generate, acquire, and then archive documents written in different languages (i.e., poly-lingual documents). Prevalent document management practice is to use categories to organize this ever-increasing volume of poly-lingual documents for subsequent searches and accesses. Poly-lingual text categorization (PLTC) refers to the automatic learning of text categorization models from a set of preclassified training documents written in different languages and the subsequent assignment of unclassified poly-lingual documents to predefined categories on the basis of the induced text categorization models. Although PLTC can be approached as multiple, independent monolingual text categorization problems, this naive PLTC approach employs only the training documents of the same language to construct a monolingual classifier and thus fails to exploit the opportunity offered by poly-lingual training documents. In this study, we propose a feature-reinforcement-based PLTC (FR-PLTC) technique that takes into account the training documents of all languages when constructing a monolingual classifier for a specific language. Using the independent monolingual text categorization (MnTC) approach as a performance benchmark, the empirical evaluation results show that our proposed FR-PLTC technique achieves higher classification accuracy than the benchmark technique. In addition, our empirical results suggest the superiority of the proposed FR-PLTC technique over its counterpart across a range of training sizes.

An ontology-based technique for preserving user preferences in document-category evolutions

Influxes of new documents over time necessitate reorganization of document categories that a user has created previously. As documents are available in increasing quantities and accelerating frequencies, the manual approach to reorganizing document categories becomes prohibitively tedious and ineffective, thus making a system-oriented approach appealing. Previous research (Larsen & Aone, 1999; Pantel & Lin, 2002) largely has followed the category-discovery approach, which groups documents by using a document-clustering technique to partition a document corpus. This approach does not consider existing categories a user created previously, which in effect reflect his or her document-grouping preference. A handful of studies (Wei, Hu, & Dong, 2002; Wei, Hu, & Lee, 2009) have taken a category-evolution approach to develop lexicon-based techniques for preserving user preference in document-category reorganizations, but have serious limitations. Responding to the significance of document-category reorganizations and addressing the fundamental problems of salient, lexicon-based techniques, we develop an ontology-based category evolution (ONCE), a technique that first enriches a concept hierarchy by incorporating important concept descriptors (jointly referred to as an ontology) and then employs the resulting enriched ontology to support category evolutions at a concept level rather than analyzing and comparing feature vectors at the lexicon level. We empirically evaluate our proposed technique and compare it with two benchmark techniques: CE2 (a lexicon-based category-evolution technique) and hierarchical agglomerative clustering (HAC; a conventional hierarchical document-clustering technique). Overall, our results show that the ONCE technique is more effective than are CE2 and HAC, across all the scenarios studied. Furthermore, the completeness of a concept hierarchy has important impacts on the performance of the proposed technique. Our results have some important implications for further research.