Prediction of hotel booking cancellations: Integration of machine learning and probability model based on interpretable feature interaction

2023 | Decision Support Systems | Citations: 0

Authors: Chen, Shuixia; Ngai, Eric W.T.; Ku, Yaoyao; Xu, Zeshui; Gou, Xunjie; Zhang, Chenxi

Abstract: Reliable hotel cancellation prediction can help establish appropriate operationa ... Expand

Abstract: Reliable hotel cancellation prediction can help establish appropriate operational strategies for hotel manage­ ment. In this sector, personal name records (PNR) data may be the most representative information source for prediction tasks. Despite the popularity of PNR, its inherent lack of availability has been commonly disregarded in the literature. Existing studies have directly input PNR into high-dimensional machine learning (ML) models to achieve cancellation predictions. Another type of model generates cancellation prediction based on the proba­ bility modeling of samples. In this study, we propose an interpretable feature interaction method to enrich the existing PNR information. Thereafter, we empirically assess the prediction performance of the two model classes. This study specifically determines whether or not the two methods can cross-fertilize each other to improve cancellation prediction. To do so, we propose a model integrating Bayesian networks (BNs) and Lasso regression for this prediction task. This study utilizes BNs for the probability model consistent with our correlated variables and dichotomous prediction setting. Moreover, we use a linear ML model (i.e., Lasso regression), given its ad­ vantages in reducing ineffective predictors and transparency for ranking feature importance. Empirical results show that the proposed integration model has better prediction performance, and the obtained BN estimators and interactive features are the most important predictors. This study contributes to the booking cancellation liter­ ature by proposing an interpretable feature interaction and a prediction method integrating two types of effective models. The obtained accurate and interpretable cancellation prediction further contributes to offering practical implications to hoteliers in managerial decision-making. Collapse

Topics: missing data decision making competitive advantage price management marketing management

Methods: lasso regression experiment computational algorithm robustness check parametric test

Economic Impacts of Platform-Endorsed Quality Certification: Evidence from Airbnb

2023 | Management Information Systems Quarterly | Citations: 0

Authors: Dewan, Sanjeev; Kim, Jooho; Nian, Tingting

Abstract: We contribute to the emerging literature on quality certification by digital pl ... Expand

Abstract: We contribute to the emerging literature on quality certification by digital platforms by studying the launch of the Airbnb Plus service, wherein the platform inspects properties and provides a badge that presumably signals the quality of the property and the reliability of the host. Our identification strategy relies on the fact that the Airbnb Plus service was launched in different cities at different times, and listings within the cities received the certification at different times. Using a staggered difference-in-differences estimation strategy in conjunction with suitable matching methods, we found that the Airbnb Plus certification increased the weekly booking rate of Plus listings by about 6.8% on average (direct effect). We also found some evidence that non-Plus listings saw a temporary decline in booking rate when one or more nearby properties received a Plus certification (externality effect). The net impact of the Airbnb Plus service on the platform itself was an annual increase in revenue of about $37,500 for the average 2-kilometer zone in a U.S. city that included one or more Plus listings, as compared to matched zones without any Plus listings (local platform effect). We performed additional analyses, including a randomized experiment, to demonstrate the robustness of our findings. Overall, our results suggest that platform-endorsed quality certification has significant economic impacts-not just on the listings that receive the certification but on other listings on the platform as well as on the platform itself. Collapse

Topics: Airbnb website sharing platform peer-to-peer model Google+

Methods: experimental group experiment propensity score method product review robustness check

Detection of Stock Manipulation Influencer Content using Supervised Learning

2023 | International Conference on Information Systems | Citations: 0

Authors: Haase, Frederic; Rath, Oliver; Lauten, Julia; Schoder, Detlef

Abstract: In recent years, social media influencers have emerged as key players in stock ... Expand

Abstract: In recent years, social media influencers have emerged as key players in stock manipulation schemes. Despite their growing impact, methods to detect such activities remain scarcely explored. In this study, we examine the social media content of stock manipulation influencers (SMIs) implicated in a $100 million fraud case by the U.S. Securities and Exchange Commission (SEC) in 2022. Leveraging natural language processing (NLP) techniques, we first investigate the linguistic characteristics present in the social media content published by SMIs. Next, we develop and evaluate supervised learning models to detect manipulative content. Our results have significant implications for investors, regulators, and the broader financial community. They reveal the unique linguistic characteristics of SMI content and demonstrate the potential of machinelearning and deep-learning-based techniques in advancing fraud detection systems. Collapse

Topics: Twitter social media fraud detection database system social media analytics

Methods: machine learning RoBERTa DistilBERT computational algorithm natural language processing

Depicting Risk Profile over Time: A Novel Multiperiod Loan Default Prediction Approach

2023 | Management Information Systems Quarterly | Citations: 0

Authors: Wang, Zhao; Jiang, Cuiqing; Zhao, Huimin

Abstract: With the rapid development of fintech, the need for dynamic credit risk evaluat ... Expand

Abstract: With the rapid development of fintech, the need for dynamic credit risk evaluation is becoming increasingly important. While previous studies on credit scoring have mostly focused on single-period loan default prediction, we call for a new avenue-multiperiod default prediction (MPDP)-to depict risk profiles over time. To address the challenges raised by MPDP, such as monotonic default probability prediction and complex relationship accommodation, we propose a novel approach, hybrid and collective scoring (HACS). We design a hybrid modeling strategy to predict whether and when a borrower will default separately through a default discrimination model and a default time estimation model, respectively, and synthesize them through a probabilistic framework. To accommodate various possible patterns of default time and measure the distribution of default probability over successive time intervals, we propose a joint default modeling method to train the default time estimation model. Empirical evaluations at the model (time-to-default prediction performance and discrimination performance) and mechanism (identifiability and discriminability) levels, as well as impact analyses at the application (granting performance and profitability performance) level, show that HACS outperforms the benchmarked survival analysis and multilabel learning methods on all fronts. It can more accurately predict time-to-default and provide financial institutions and investors better decisionsupport in granting loans and selecting loan portfolios. Collapse

Topics: online lending risk management decision making analytical information system

Methods: survival analysis case study machine learning design science design artifact

Theories: portfolio theory

Tune Down the Misinformation, Please: Generating Corrective Messages for COVID-19 Misinformation

2022 | HICSS | Citations: 0

Authors: Meyer, Dylan; Tao, Jie; Kris, Alison

Abstract: The ongoing COVID-19 pandemic drastically changed our lives in multiple aspects, ... Expand

Abstract: The ongoing COVID-19 pandemic drastically changed our lives in multiple aspects, one of which is the reliance on social media during quarantine, both for social interaction and information-seeking purposes. However, the wide dissemination of misinformation on social media has impacted public health negatively. Previous studies on COVID-19 misinformation mainly focused on exploration of impacts and explanation of motivations, with few exceptions. In this study, we propose an analytical pipeline that generates corrective messages toward COVID-19 misinformation in a semiautomatic fashion, and then evaluate it against a large amount of data. Both the automated and manual evaluation results suggest the efficiency of the proposed pipeline, which can be used in combination with human intelligence by individuals and public health organizations in fighting COVID-19 misinformation. Collapse

Topics: misinformation Twitter social media covid 19 fake news

Methods: machine learning BERT transformer model experiment term frequency–inverse document frequency

Peak cubes in service operations: Bringing multidimensionality into decision support systems

2021 | Decision Support Systems | Citations: 0

Authors: Barfar, Arash; Padmanabhan, Balaji; Hevner, Alan

Abstract: Companies like Ritz Carlton, Disney and Verizon are among many who have invested ... Expand

Abstract: Companies like Ritz Carlton, Disney and Verizon are among many who have invested in analytics to improve their customers’ service experiences with the firms. Extensive data are collected on all aspects of how customers interact or experience the products or services. Research has shown the importance of the “peak-end” rule in service design; that is, providing a customer with good “peak” service levels and “ending” the service experience with high quality can enhance customer satisfaction and build loyalty. However, previous studies have examined this phenomenon only in contexts with unidimensional service levels. We introduce peak cubes, which enable service designers and scholars to pinpoint prominent service levels in multidimensional service experience pro­ files—thereby extending current research on behavioral economics and service design to more general settings. Results indicate the potential of multidimensional peak-end models to better predict customer satisfaction in various service scenarios. Using Shapley values in coalitional game theory, the resulting models can also inform service designers about the quality dimensions that are critical from the perspective of multidimensional peakend heuristic and customer satisfaction. Our research contributions and proposed methodology will enhance decision support systems with multidimensional capabilities and have applications to fields as diverse as service operations and healthcare. Collapse

Topics: service quality shapley additive explanation logistics management customer satisfaction database system

Methods: simulation computational algorithm time series analysis longitudinal research LightGBM

Theories: behavioral economics theory of bounded rationality game theory cooperative game theory

Just DM Me (Politely): Direct Messaging, Politeness, and Hiring Outcomes in Online Labor Markets

2021 | Information Systems Research | Citations: 1

Authors: Hong, Yili; Peng, Jing; Burtch, Gordon; Huang, Ni

Abstract: This study examines the role of text-based direct messaging systems in online la ... Expand

Abstract: This study examines the role of text-based direct messaging systems in online labor markets, which provide a communication channel between workers and employers, adding a personal touch to the exchange of online labor. We propose the effect of workers’ use of the direct messaging system on employers’ hiring decisions and conceptualize the information role of direct messaging. To empirically evaluate the information role of the direct messaging system, we leverage data on the direct messaging activities between workers and employers across more than 470,000 job applications on a leading online labor market. We report evidence that direct messaging with a prospective employer increases a worker’s probability of being hired by 8.9%. However, the degree to which workers benefit from direct messaging is heterogeneous, and the effect amplifies for workers approaching employers from a position of disadvantage (lacking tenure or fit with the job) and attenuates as more workers attempt to message the same prospective employer. The effects also depend on message content. In particular, we find that the benefits of direct messaging for workers depend a great deal on the politeness of the workers, and this “politeness effect” depends on several contextual factors. The beneficial effects are amplified for lower-status workers (i.e., workers lacking tenure and job fit) and workers who share a common language with the employer. At the same time, the beneficial effects weaken in the presence of typographical errors. These findings provide important insights into when and what to message to achieve favorable hiring outcomes in online employment settings. Collapse

Topics: online labor market information overload platform worker database system PHP

Methods: data imputation instrumental variables estimation machine learning simulation XGBoost

Theories: lemon market theory

Automated Feature Selection for Anomaly Detection in Network Traffic Data

2021 | ACM Transactions on Management Information Systems | Citations: 0

Authors: Nakashima, Makiya; Sim, Alex; Kim, Youngsoo; Kim, Jonghyun; Kim, Jinoh

Abstract: Variable selection (also known as feature selection) is essential to optimize th ... Expand

Abstract: Variable selection (also known as feature selection) is essential to optimize the learning complexity by prioritizing features, particularly for a massive, high-dimensional dataset like network traffic data. In reality, however, it is not an easy task to effectively perform the feature selection despite the availability of the existing selection techniques. From our initial experiments, we observed that the existing selection techniques produce different sets of features even under the same condition (e.g., a static size for the resulted set). In addition, individual selection techniques perform inconsistently, sometimes showing better performance but sometimes worse than others, thereby simply relying on one of them would be risky for building models using the selected features. More critically, it is demanding to automate the selection process, since it requires laborious efforts with intensive analysis by a group of experts otherwise. In this article, we explore challenges in the automated feature selection with the application of network anomaly detection. We first present our ensemble approach that benefits from the existing feature selection techniques by incorporating them, and one of the proposed ensemble techniques based on greedy search works highly consistently showing comparable results to the existing techniques. We also address the problem of when to stop to finalize the feature elimination process and present a set of methods designed to determine the number of features for the reduced feature set. Our experimental results conducted with two recent network datasets show that the identified feature sets by the presented ensemble and stopping methods consistently yield comparable performance with a smaller number of features to conventional selection techniques. Collapse

Topics: information system use IT security IT security defense denial of service attack Python

Methods: experiment anomaly detection computational algorithm machine learning ensemble learning

Theories: set theory

Unsupervised Deep Learning for Fake Content Detection in Social Media

2021 | HICSS | Citations: 1

Authors: Tao, Jie; Fang, Xing; Zhou, Lina

Abstract: Fake content is ever increasing in the online environment, driven by various mot ... Expand

Abstract: Fake content is ever increasing in the online environment, driven by various motivations such as gaining commercial and political advantages. The interactive and collaborative nature of social media further fuels the growth of fake content by exerting fast and widespread influence. Despite growing and interdisciplinary efforts in detecting fake content in social media, some common research challenges remain to be addressed such as humans’ cognitive bias and scarcity of labeled data for training supervised machine learning models. This study aims to tackle both challenges by developing unsupervised deep learning models for the detection of fake content in social media. In view that traditional linguistic features fail to capture context information, our proposed method learns feature representations from the context in social media content. The empirical evaluation results with fake comments from YouTube demonstrate that our proposed methods not only outperform baseline models with traditional unsupervised machine learning techniques, but also achieve comparable performance to the state-of-the-art supervised models. The proposed analytical pipeline provides an end-toend solution to detecting fake social media contents, which largely reduce the human labor required in collaborative data science teams (i.e., particularly the data labeling). The findings of this study can be used to facilitate collaboration in data science by reducing humans’ cognitive bias and improve the collaboration efficiency. Collapse

Topics: social media YouTube fake review

Methods: unsupervised learning supervised learning data transformation cluster analysis machine learning

Predicting product adoption intentions: An integrated behavioral model-inspired multiview learning approach

2021 | Information & Management | Citations: 0

Authors: Zhang, Zhu; Wei, Xuan; Zheng, Xiaolong; Zeng, Daniel Dajun

Abstract: Mining product adoption intentions from social media could provide insights for ... Expand

Abstract: Mining product adoption intentions from social media could provide insights for many business practices, such as social media marketing. Existing methods mainly focus on text information but overlook other types of data. In light of the Integrated Behavioral Model (IBM), in this study, we argue that it is valuable to consider users’ social connections in addition to postings for identifying product adoption intentions. Based on this rationale, we propose a novel multiview deep learning framework to identify product adoption intentions. Extensive experi­ ments show our proposed approach is effective, and demonstrate the benefit of incorporating social network information for intention identification. Collapse

Topics: Apple iPhone Twitter social media social network mobile phone

Methods: deep learning machine learning experiment support vector machine binary classification