Non-Linear Hybrid Shrinkage of Weights for Forecast Selection and Combination

2022 | International Conference on Business Informatics | Citations: 0

Authors: Schulz, Felix

Abstract: We introduce Non-Linear Hybrid Shrinkage (NLHS) as a holistic model for forecast ... Expand

Abstract: We introduce Non-Linear Hybrid Shrinkage (NLHS) as a holistic model for forecast combination, shrinkage and selection. NLHS first determines the selection of forecasters based on information criteria such as forward feature selection and stores the selection status of forecasters in a selection vector. Depending on the selection status, the estimated optimal weights of the forecasters are either shrunk to zero or equal weights by the least absolute shrinkage and selection operator (LASSO). Among benchmark algorithms such as simple average, optimal weights, or linear and LASSO-based shrinkage methods, NLHS is superior for a larger number of forecasters, as shown in simulation-based experiments. Collapse

Methods: lasso regression simulation optimization model computational algorithm experiment

Theories: statistical learning theory

A Contingency Theory of Representational Complexity in Organizations

2020 | Organization Science | Citations: 6

Authors: Csaszar, Felipe A.; Ostler, James

Abstract: A long-standing question in the organizations literature is whether firms are be ... Expand

Abstract: A long-standing question in the organizations literature is whether firms are better off by using simple or complex representations of their task environment. We address this question by developing a formal model of how firm performance depends on the process by which firms learn and use representations. Building on ideas from cognitive science, our model conceptualizes this process in terms of how firms construct a representation of the environment and then use that representation when making decisions. Our model identifies the optimal level of representational complexity as a function of (a) the environment’s complexity and uncertainty and (b) the firm’s experience and knowledge about the environment’s deep structure. We use this model to delineate the conditions under which firms should use simple versus complex representations; in doing so, we provide a coherent framework that integrates previous conflicting results on which type of representation leaves firms better off. Among other results, we show that the optimal representational complexity generally depends more on the firm’s knowledge about the environment than it does on the environment’s actual complexity. We also show that the relative advantage of heuristics vis-à-vis more complex representations critically depends on an unstated assumption of “informedness”: that managers can know what are the most relevant variables to pay attention to. We show that when this assumption does not hold, complex representations are usually better than simpler ones. Collapse

Topics: organizational productivity decision making organization structure mobile application

Methods: simulation robustness check theory development artificial neural network mathematical model

Theories: contingency theory organizational theory cognitive complexity theory organizational behavior theory statistical learning theory

Don't be so emotional: How social media communication affects potential applicants' engagement

2020 | European Conference On Information Systems | Citations: 0

Authors: Moser, Kilian; Tumasjan, Andranik; Cable, Dan

Abstract: ... Expand

Abstract: Collapse

Topics: Twitter social media online labor market marketing information system viral marketing

Methods: regression analysis method support vector machine statistical hypothesis test machine learning qualitative content analysis

Theories: contract theory statistical learning theory

Text Analytics to Support Sense-Making in Social Media: A Language-Action Perspective

2018 | Management Information Systems Quarterly | Citations: 0

Authors: Abbasi, Ahmed; Zhou, Yilu; Deng, Shasha; Zhang, Pengzhu

Abstract: Social media and online communities provide organizations with new opportunities ... Expand

Abstract: Social media and online communities provide organizations with new opportunities to support their businessrelated functions. Despite their various benefits, social media technologies present two important challenges for sense-making. First, online discourse is plagued by incoherent, intertwined conversations that are often difficult to comprehend. Moreover, organizations are increasingly interested in understanding social media participants’ actions and intentions; however, existing text analytics tools mostly focus on the semantic dimension of language. The language-action perspective (LAP) emphasizes pragmatics; not what people say but, rather, what they do with language. Adopting the design science paradigm, we propose a LAP-based text analytics framework to support sense-making in online discourse. The proposed framework is specifically intended to address the two aforementioned challenges associated with sense-making in online discourse: the need for greater coherence and better understanding of actions. We rigorously evaluate a system that is developed based on the framework in a series of experiments using a test bed encompassing social media data from multiple channels and industries. The results demonstrate the utility of each individual component of the system, and its underlying framework, in comparison with existing benchmark methods. Furthermore, the results of a user experiment involving hundreds of practitioners, and a four-month field experiment in a large organization, underscore the enhanced sense-making capabilities afforded by text analytics grounded in LAP principles. The results have important implications for online sense-making and social media analytics. Collapse

Topics: sensemaking social media social network web forum productivity

Methods: experiment longitudinal research coherence analysis machine learning field experiment

Theories: speech act theory language action perspective statistical learning theory

Intelligent drowsy eye detection using image mining

2015 | Information Systems Frontiers | Citations: 5

Authors: Emam, Ahmed

Abstract: Cell phones, MP4 players, and GPS all distract drivers, forcing them to divide t ... Expand

Abstract: Cell phones, MP4 players, and GPS all distract drivers, forcing them to divide their attention to what is happening outside and what is happening inside the vehicle. Preventing drivers from falling asleep at the wheel and drifting off on the road is an increasingly hard task but a solution can be achieved with today’s technology: by capturing driver’s eye gaze and thus the level of their exhaustion, the potential for safer roads is real. Today, image mining systems can automatically extract important information from an image dataset which can then be used to classify images and image relationship that been applied in different areas. The basic challenge in image mining research area is to determine how image level pixel representation in an original image can be used and constructed to then identify meaningful information or form relationships. In this research Intelligent Drowsy Eye Detection, using an Image Mining (IDEDIM) system is proposed. The proposed architecture would use different feature extraction techniques and three data mining classification techniques to aid with accurate information collection. Two thousand left and right eye images were used to test the developed system by Discrete Wavelet Transform (DWT), Statistical features, and Local binary pattern (LBP) feature extraction techniques. The extracted features were used as input for Decisions Tree C5.0, K Nearest Neighbor (KNN), and the Support Vector Machine (SVM) Classifier. After several experimental sets, the C5.0 and KNN classifier were performing better than SVM classifier. Based on the results, we recommend the inclusion of LBP and DWT in conjunction with C5.0 with KNN as a classification technique. To validate the achieved results, Receiver Operating Characteristic (ROC) curve is used to compare among the proposed classifiers. The proposed system can be integrated with the existing subsystem into real time Drowsy Detection System to achieve excellent accuracy and performance. Collapse

Topics: database system Microsoft SQL server database relational database Visual Basic

Methods: decision tree classification support vector machine machine learning computational algorithm k-nearest neighbor

Theories: statistical learning theory

A comparative analysis of machine learning systems for measuring the impact of knowledge management practices

2013 | Decision Support Systems | Citations: 12

Authors: Delen, Dursun; Zaim, Halil; Kuzey, Cemil; Zaim, Selim

Abstract: Knowledge management (KM) has recently emerged as a discrete area in the study o ... Expand

Abstract: Knowledge management (KM) has recently emerged as a discrete area in the study of organizations and frequently cited as an antecedent of organizational performance. This study aims at investigating the impact of KM practices on organizational performance of small and medium-sized enterprises (SME) in service industry. Four popular machine learning techniques (i.e., neural networks, support vector machines, decision trees and logistic regression) along with statistical factor analysis (EFA and CFA) are used to developed predictive and explanatory models. The data for this study is obtained from 277 SMEs operating in the service industry within the greater metropolitan area of Istanbul in Turkey. The analyses indicated that there is a strong and positive relationship between the implementation level of KM practices and organizational performance related to KM. The paper summarizes the finding of the study and provides managerial implications to improve the organizational performance of SMEs through effective implementation of KM practices. Collapse

Topics: knowledge management financial performance organizational productivity knowledge sharing knowledge utilization

Methods: machine learning support vector machine decision tree classification survey data transformation

Theories: statistical learning theory

Effective Sentiment Analysis of Corporate Financial Reports

2013 | International Conference on Information Systems | Citations: 2

Authors: Ren, Jimmy; Ge, Huizhong; Wu, Xiaoyu; Wang, Guan; Wang, Wei; Liao, Stephen

Abstract: Sentiment analysis is widely adopted in studying various important topics in bus ... Expand

Abstract: Sentiment analysis is widely adopted in studying various important topics in business intelligence. Though many studies reported interesting results by using machine learning, the lack of theoretic analysis and the shortage of practical guidance are hurdles of theory development. Besides, due to the difficulty in labelling data, the effectiveness of sentiment analysis with only labelled data needs to be questioned. In this paper, we drew on statistical learning theory to perform extensive theoretic analysis in sentiment analysis by using real corporate financial reports. We investigated when and why machine learning methods provide preferred performance under the guidance of the theory. We also provided practical suggestions in applying machine learning methods for both researchers and practitioners. In addition, we utilized the cheap and ubiquitous unlabelled data to further improve the sentiment analysis performance. This has the potential to largely reduce the manual data labelling work and to scale up the experiments. Collapse

Methods: experiment naïve bayes computational algorithm natural language processing machine learning method

Theories: statistical learning theory

A note comparing support vector machines and ordered choice models’ predictions of international banks’ ratings

2011 | Decision Support Systems | Citations: 0

Authors: Bellotti, Tony; Matousek, Roman; Stewart, Chris

Abstract: We find that support vector machines can produce notably better predictions of i ... Expand

Abstract: We find that support vector machines can produce notably better predictions of international bank ratings than the standard method currently used for this purpose, ordered choice models. This appears due to the support vector machine's ability to estimate a large number of country dummies unrestrictedly, which was not possible with the ordered choice models due to the low sample size. Collapse

Topics: database system

Methods: support vector machine longitudinal research time series analysis cross sectional research experiment

Theories: statistical learning theory

Detection of financial statement fraud and feature selection using data mining techniques

2011 | Decision Support Systems | Citations: 24

Authors: Ravisankar, P.; Ravi, V.; Raghava Rao, G.; Bose, I.

Abstract: Recently, high profile cases of financial statement fraud have been dominating t ... Expand

Abstract: Recently, high profile cases of financial statement fraud have been dominating the news. This paper uses data mining techniques such as Multilayer Feed Forward Neural Network (MLFF), Support Vector Machines (SVM), Genetic Programming (GP), Group Method of Data Handling (GMDH), Logistic Regression (LR), and Probabilistic Neural Network (PNN) to identify companies that resort to financial statement fraud. Each of these techniques is tested on a dataset involving 202 Chinese companies and compared with and without feature selection. PNN outperformed all the techniques without feature selection, and GP and PNN outperformed others with feature selection and with marginally equal accuracies. Collapse

Topics: fraud detection

Methods: support vector machine machine learning experiment decision tree classification discriminant analysis

Theories: statistical learning theory

Detecting Fake Websites: The Contribution of Statistical Learning Theory

2010 | Management Information Systems Quarterly | Citations: 145

Authors: Abbasi, Ahmed; Zhang, Zhu; Zimbra, David; Chen, Hsinchun; Nunamaker, Jr., Jay F.

Abstract: Fake websites have become increasingly pervasive, generating billions of dollars ... Expand

Abstract: Fake websites have become increasingly pervasive, generating billions of dollars in fraudulent revenue at the expense of unsuspecting Internet users. The design and appearance of these websites makes it difficult for users to manually identify them as fake. Automated detection systems have emerged as a mechanism for combating fake websites, however most are fairly simplistic in terms of their fraud cues and detection methods employed. Consequently, existing systems are susceptible to the myriad of obfuscation tactics used by fraudsters, resulting in highly ineffective fake website detection performance. In light of these deficiencies, we propose the development of a new class of fake website detection systems that are based on statistical learning theory (SLT). Using a design science approach, a prototype system was developed to demonstrate the potential utility of this class of systems. We conducted a series of experiments, comparing the proposed system against several existing fake website detection systems on a test bed encompassing 900 websites. The results indicate that systems grounded in SLT can more accurately detect various categories of fake websites by utilizing richer sets of fraud cues in combination with problem-specific knowledge. Given the hefty cost exacted by fake websites, the results have important implications for e-commerce and online security. Collapse

Topics: phishing website website web page source code internet fraud

Methods: support vector machine experiment machine learning design artifact Student's t-test

Theories: statistical learning theory