Decision Support Systems最新文献

Multimorbidity, the presence of two or more chronic conditions within an individual, represents one of the most intricate challenges for global health systems. Traditional single-disease management often fails to address the multifaceted nature of multimorbidity. Network model emerges as a growing field for elucidating the interconnections among multimorbidity. However, the field lacks a standardized method to compute and visually represent of these networks. Given the challenges, this study proposes a three-stage methodology to decipher multimorbidity. First, we integrate the Failure Modes and Effects Analysis (FMEA) method with the multimorbidity encapsulation framework to develop the Multimorbidity Risk Network (MRN). Second, we use complex network techniques to identify high-risk patterns within MRN communities. Finally, we apply machine learning techniques to correlate these communities with the biological attributes of patients that have been marginalized in most studies. Our approach advocates a paradigm shift from the conventional focus on single diseases to a holistic, patient-centric approach, providing decision-makers with integrated information technology artifacts for deciphering the multimorbidity.

In customer service, emotional expressions by chatbots are considered a promising direction to improve customer experience. However, there is a lack of comprehensive understanding of how and when chatbots' emotional expressions improve customer attitudes. Although chatbots' emotional expressions of care and concern may feel inauthentic to customers in the inferential path, which can negatively affects customer attitudes, we propose that the positive effect of the affective reactions path can result in a positive effect on customer attitude based on the dual-path view of Emotions as Social Information (EASI). The relative strengths of the two EASI paths can be moderated, and we explored the moderating effects of rational thinking styles (information processing in EASI) and beliefs in computer emotion (perceived appropriateness in EASI). According to EASI, situation can affect the meaning of emotions, so we conducted experiments in two situations. With chatbot identity disclosure, we found that the chatbot's emotional expressions reduce customers' perceived authenticity (reflecting the inferential path in EASI) but ultimately improve customer attitudes. Belief in computer emotions and rational thinking style moderated the negative relationship between emotional expressions and authenticity. With chatbot identity non-disclosure, the chatbot's emotional expressions still improve customer attitudes but with no effect on authenticity. Because there is high likelihood of chatbot identities being discovered by customers, this finding of the moderating effect of perceived humanness on authenticity is highly relevant. Our findings make important contributions to research on computer emotion and service authenticity.

As a part of natural language processing (NLP), the intent of topic modeling is to identify topics in textual corpora with limited human input. Current topic modeling techniques, like Latent Dirichlet Allocation (LDA), are limited in the pre-processing steps and currently require human judgement, increasing analysis time and opportunities for error. The purpose of this research is to allay some of those limitations by introducing new approaches to improve coherence without adding computational complexity and provide an objective method for determining the number of topics within a corpus. First, we identify a requirement for a more robust stop words list and introduce a new dimensionality-reduction heuristic that exploits the number of words within a document to infer importance to word choice. Second, we develop an eigenvalue technique to determine the number of topics within a corpus. Third, we combine all of these techniques into the Zimm Approach, which produces higher quality results than LDA in determining the number of topics within a corpus. The Zimm Approach, when tested against various subsets of the 20newsgroup dataset, produced the correct number of topics in 7 of 9 subsets vs. 0 of 9 using highest coherence value produced by LDA.

We consider how to make dynamic pricing decision for Chinese Online (COL) at T time-points, an online publisher that allow authors to sell their ongoing book projects. Instead of paying for a book, readers pay for each chapter (pay-per-chapter mode) of the ongoing book project. This mode allows readers to pay for as many chapters as they want without taking the risk that the releasing of new chapters might be delayed or stopped. Despite of the dynamics of chapter-by-chapter released of COL products, the fixed pricing strategy (FPS) does not make fully use of the reading data generated by releasing chapters of the ongoing book. We propose a learning-based dynamic pricing strategy (LDPS) that exploits the newly information to maximize cumulative revenue for the publisher. The LDPS captures the ever changing features of readers. It employs the Thompson sampling method to balance the exploration of investigating different prices sufficiently with the exploitation of settling on the optimal price. Taking COL as a case study and implementing our strategy in the context of the aforementioned real-life data set, we show that LDPS outperform several classical strategies such as Greedy, Prior-Free TS and Prior-Given TS, and average revenue of LDPS is increased by 0.5 % average per time-point compared to the publisher's historical decisions. We also provide some management implications for the COL publisher by analyzing the pricing range of different genres of books and the choice of the exploration threshold parameter.

We propose a novel curvature-based indicator constructed on log-price time series that captures an interplay between trend, acceleration, and volatility found relevant to quantify risks and improve trading strategies. We apply it to diagnose explosive bubble-like behaviors in cryptocurrency price time series and provide early warning signals of impending market shifts or increased volatility. This improves significantly on standard statistical tests such as the Generalized Supremum Augmented Dickey–Fuller (GSADF) and the Backward SADF tests. Furthermore, the incorporation of our curvature-based indicator as a feature into the Light Gradient Boosting Machine enhances its predictive capabilities, as measured by classification accuracy and trading performance.

The Common Vulnerability Scoring System (CVSS) is widely used in the cybersecurity industry to assess the severity of vulnerabilities. However, manual assessments and human error can lead to delays and inconsistencies. This study employs situational awareness theory to develop an automated decision support system, integrating perception, comprehension, and projection components to enhance effectiveness. Specifically, an interpretable principal component analysis (iPCA) combined with machine learning is utilized to forecast CVSS scores using text descriptions from the Common Vulnerabilities and Exposures (CVE) database. Different forecasting approaches, including traditional machine learning models, Long-Short Term Memory Neural Networks, and Transformer architectures (ChatGPT) are compared to determine the best performance. The results show that iPCA combined with support vector regression achieves a high performance (R² = 98%) in predicting CVSS scores using CVE text descriptions. The results indicate that the variability, length, and details in the vulnerability description contribute to the performance of the transformer model. These findings are consistent across vulnerability descriptions from six companies between 2017 and 2019. The study's outcomes have the potential to enhance organizations' security posture, improving situational awareness and enabling better managerial decision-making in cybersecurity.

In today's digital economy, virtually everything from products and services to political debates and cultural phenomena can spark WOM on social media. Analyzing online WOM poses at least three challenges. First, online WOM typically consists of unstructured data that can transform into myriad variables, necessitating effective dimension reduction. Second, online WOM is often continuous and dynamic, with the potential for rapid, time-varying changes. Third, significant events may trigger symmetric or asymmetric responses across various entities, resulting in “bursty” and intense WOM from multiple sources. To address these challenges, we introduce a new computationally efficient method—multi-view sequential canonical covariance analysis. This method is designed to solve the myriad online WOM conversational dimensions, detect online WOM dynamic trends, and examine the shared online WOM across different entities. This approach not only enhances the capability to swiftly interpret and respond to online WOM data but also shows potential to significantly improve decision-making processes across various contexts. We illustrate the method's benefits through two empirical examples, demonstrating its potential to provide profound insights into online WOM dynamics and its extensive applicability in both academic research and practical scenarios.