Assessing and Enhancing Adversarial Robustness of Predictive Analytics: An Empirically Tested Design Framework

2022 | Journal of Management Information Systems | Citations: 4

Authors: Li, Weifeng; Chai, Yidong

Abstract: As predictive analytics increasingly applies supervised machine learning (SML) m ... Expand

Abstract: As predictive analytics increasingly applies supervised machine learning (SML) models to inform mission-critical decision-making, adversaries become incentivized to exploit the vulnerabilities of these SML models and mislead predictive analytics into erroneous decisions. Due to the limited understanding and awareness of such adversarial attacks, the predictive analytics knowledge and deployment need a principled technique for adversarial robustness assessment and enhancement. In this research, we leverage the technology threat avoidance theory as the kernel theory and propose a research framework for assessing and enhancing the adversarial robustness of predictive analytics applications. We instantiate the proposed framework by developing a robust text classification system, the ARText system. The proposed system is rigorously evaluated in comparison with benchmark methods on two tasks extensively enabled by SML: spam review detection and spam email detection, which then confirmed the utility and effectiveness of our ARText system. Results from numerous experiments revealed that our proposed framework could significantly enhance the adversarial robustness of predictive analytics applications. Collapse

Topics: email spam electronic mail artificial intelligence decision making risk management

Methods: machine learning ensemble learning experiment computational algorithm meta design

Theories: technology threat avoidance theory

A multivariate approach for multi-step demand forecasting in assembly industries: Empirical evidence from an automotive supply chain

2021 | Decision Support Systems | Citations: 0

Authors: Gonçalves, João N.C.; Cortez, Paulo; Carvalho, M. Sameiro; Frazão, Nuno M.

Abstract: Demand forecasting works as a basis for operating, business and production plann ... Expand

Abstract: Demand forecasting works as a basis for operating, business and production planning decisions in many supply chain contexts. Yet, how to accurately predict the manufacturer’s demand for components in the presence of endcustomer demand uncertainty remains poorly understood. Assigning the proper order quantities of components to suppliers thus becomes a nontrivial task, with a significant impact on planning, capacity and inventory-related costs. This paper introduces a multivariate approach to predict manufacturer’s demand for components throughout multiple forecast horizons using different leading indicators of demand shifts. We compare the autoregressive integrated moving average model with exogenous inputs (ARIMAX) with Machine Learning (ML) models. Using a real case study, we empirically evaluate the forecasting and supply chain performance of the multivariate regression models over the component’s life-cycle. The experiments show that the proposed multivariate approach provides superior forecasting and inventory performance compared with traditional univariate benchmarks. Moreover, it reveals applicable throughout the component’s life-cycle, not just to a single stage. Particularly, we found that demand signals at the beginning of the life-cycle are predicted better by the ARIMAX model, but it is outperformed by ML-based models in later life-cycle stages. Collapse

Topics: supply chain management logistics management procurement management enterprise resource planning decision support

Methods: longitudinal research time series analysis regression analysis method autoregressive integrated moving average case study

Machine Learning Based Differentiation Between High and Low Self-Discipline Using EEG Data

2021 | European Conference On Information Systems | Citations: 0

Authors: Sauter, Daniel; Butz, Lars; Buettner, Ricardo

Abstract: Since the detection of high and low self-discipline based on self-reported meas ... Expand

Abstract: Since the detection of high and low self-discipline based on self-reported measures is problematic due to validity problems, we have proposed an objective method based on physiological EEG data. Using a fine grained EEG spectrum in combination with machine learning, we were able to build a stable, reliable classifier, which reaches an accuracy of almost 76 percent. Moreover, we show that there are systematic differences in the theta and beta frequencies. Predicting low self-discipline and high selfdiscipline is an important topic in learning (especially e-learning) research as well as in medicine. Collapse

Topics: personality

Methods: electroencephalography computational algorithm survey machine learning oculometry

Leveraging Multisource Heterogeneous Data for Financial Risk Prediction: A Novel Hybrid-Strategy-Based Self-Adaptive Method

2021 | Management Information Systems Quarterly | Citations: 0

Authors: Wang, Gang; Chen, Gang; Zhao, Huimin; Zhang, Feng; Yang, Shanli; Lu, Tian

Abstract: Emerging phenomena of ubiquitous multisource data offer promising avenues for ma ... Expand

Abstract: Emerging phenomena of ubiquitous multisource data offer promising avenues for making breakthroughs in financial risk prediction. While most existing methods for financial risk prediction are based on a single information source, which may not adequately capture various complex factors that jointly influence financial risks, we propose a hybrid-strategy-based self-adaptive method to effectively leverage heterogeneous soft information drawn from a variety of sources. The method uses a proposed new feature- sparsity learning method to adaptively integrate multisource heterogeneous soft features with hard features and a proposed improved evidential reasoning rule to adaptively aggregate base classifier predictions, thereby alleviating both the declarative bias and the procedural bias of the learning process. Evaluation in two cases at the individual level (concerning borrowers at a P2P lending platform) and the company level (concerning listed companies in the Chinese stock market) showed that, compared with relying solely on hard features, effectively incorporating multisource heterogeneous soft features using our proposed method enabled earlier prediction of financial risks with desirable performance. Collapse

Topics: peer-to-peer lending crowdsourcing social media electronic commerce communication service infrastructure

Methods: machine learning ensemble learning support vector machine sensitivity analysis XGBoost

High-performance diagnosis of sleep disorders: a novel, accurate and fast machine learning approach using electroencephalographic data

2020 | HICSS | Citations: 0

Authors: Buettner, Ricardo; Grimmeisen, Annika; Gotschlich, Anne

Abstract: While diagnosing sleep disorders by physicians using electroencephalographic da ... Expand

Abstract: While diagnosing sleep disorders by physicians using electroencephalographic data is protracted and inaccurate, we report promising results from a novel, fast and reliable machine learning approach. Our approach only needs an electroencephalographic recording snippet of 10 minutes instead of eight hours to correctly classify the disorder with an accuracy of over 90 percent. The Rapid Eye Movement sleep behavior disorder can lead to secondary diseases like Parkinson or Dementia. Therefore, it is important to classify the disorder fast and with a high level of accuracy -which is now possible with our approach. Collapse

Topics: database system health information system personality mindfulness

Methods: electroencephalography machine learning random forest classification computational algorithm oculometry

Financial crisis prediction model using ant colony optimization

2020 | International Journal of Information Management | Citations: 0

Authors: J, Uthayakumar; Metawa, Noura; Shankar, K.; Lakshmanaprabu, S.K.

Abstract: Financial decisions are often based on classification models which are used to a ... Expand

Abstract: Financial decisions are often based on classification models which are used to assign a set of observations into predefined groups. Different data classification models were developed to foresee the financial crisis of an organization using their historical data. One important step towards the development of accurate financial crisis prediction (FCP) model involves the selection of appropriate variables (features) which are relevant for the problems at hand. This is termed as feature selection problem which helps to improve the classification performance. This paper proposes an Ant Colony Optimization (ACO) based financial crisis prediction (FCP) model which incorporates two phases: ACO based feature selection (ACO-FS) algorithm and ACO based data classification (ACO-DC) algorithm. The proposed ACO-FCP model is validated using a set of five benchmark dataset includes both qualitative and quantitative. For feature selection design, the developed ACO-FS method is compared with three existing feature selection algorithms namely genetic algorithm (GA), Particle Swarm Optimization (PSO) algorithm and Grey Wolf Optimization (GWO) algorithm. In addition, a comparison of classification results is also made between ACO-DC and state of art methods. Experimental analysis shows that the ACO-FCP ensemble model is superior and more robust than its counterparts. In consequence, this study strongly recommends that the proposed ACO-FCP model is highly competitive than traditional and other artificial intelligence techniques. Collapse

Topics: artificial intelligence information management database system

Methods: machine learning computational algorithm support vector machine AdaBoost bagging method

Impact of Heterogeneous Prior Contribution on Reciprocity in Online Sellers’ Community

2019 | International Conference on Information Systems | Citations: 0

Authors: Yi, Chen

Abstract: This study explores how different types of resources a seller contributes in the ... Expand

Abstract: This study explores how different types of resources a seller contributes in the online community will trigger others’ reciprocity, reflected by the responses the seller’s threads receive. Drawing on social exchange theory and using machine learning techniques, we identify two important types of resources transferred in an online sellers’ community: informational resource and instrumental resource. Our findings reveal that a seller’s provision of informational resource is positively associated with the responses the seller’s threads receive, while a seller’s provision of instrumental resource is negatively associated with the responses the seller’s threads receive. Moreover, the effect is moderated by the types of resources sought by a certain thread. Specifically, both effects are strengthened for threads that seek informational resource and undermined for threads that seek instrumental resource. The study contributes to the understanding of online reciprocity by uncovering the differing impact of different contribution and its boundary condition. Collapse

Topics: online community information exchange electronic commerce

Methods: computational algorithm qualitative coding classification method forums posting robustness check

Theories: social exchange theory

Rebuilding sample distributions for small dataset learning

2018 | Decision Support Systems | Citations: 0

Authors: Li, Der-Chiang; Lin, Wu-Kuo; Chen, Chien-Chih; Chen, Hung-Yu; Lin, Liang-Sian

Abstract: Over the past few decades, a few learning algorithms have been proposed to extra ... Expand

Abstract: Over the past few decades, a few learning algorithms have been proposed to extract knowledge from data. The majority of these algorithms have been developed with the assumption that training sets can denote populations. When the training sets contain only a few properties of their populations, the algorithms may extract minimal and/or biased knowledge for decision makers. This study develops a systematic procedure based on fuzzy theories to create new training sets by rebuilding the possible sample distributions, where the procedure contains new functions that estimate domains and a sample generating method. In this study, two real cases of a leading company in the thin film transistor liquid crystal display (TFT-LCD) industry are examined. Two learning algorithms—a back-propagation neural network and support vector regression—are employed for modeling, and two sample generation approaches—bootstrap aggregating (bagging) and the synthetic minority over-sampling technique (SMOTE)—are employed to compare the accuracy of the models. The results indicate that the proposed method outperforms bagging and the SMOTE with the greatest amount of statistical support. Collapse

Topics: IT manager

Methods: experiment computational algorithm sensitivity analysis paired t-test experimental group

Identification of Human Factors in Aviation Incidents Using a Data Stream Approach

2017 | HICSS | Citations: 0

Authors: Shi, Donghui; Zurada, Jozef; Guan, Jian

Abstract: This paper investigates the use of data streaming analytics to better predict th ... Expand

Abstract: This paper investigates the use of data streaming analytics to better predict the presence of human factors in aviation incidents with new incident reports. As new incidents data become available, the fresh information can help not only evaluate but also improve existing models. First, we use four algorithms in batch learning to establish a baseline for comparison purposes. These are NaiveBayes (NB), Cost Sensitive Classifier (CSC), Hoeffdingtree (VFDT), and OzabagADWIN (OBA). The traditional measure of the classification accuracy rate is used to test their performance. The results show that among the four, NB and CSC are the best classification algorithms. Then we test the classifiers in a data stream setting. The two performance measure methods Holdout and Interleaved Test-Then-Train or Prequential are used in this setting. The Kappa statistic charts of Prequential measure with a sliding window show that NB exhibits the best performance, and is better than the other algorithms. The two different measure methods, batch learning with 10-fold cross validation and data stream with Prequential measure, get one consistent result. CSC is a suitable for unbalanced data in batch learning, but it is not best in Kappa statistic for data stream. Valid incremental algorithms need to be developed for the data stream with unbalanced labels. Collapse

Topics: ergonomics usability open source online learning system

Methods: computational algorithm classification method natural language processing machine learning simulation

Cost-sensitive and ensemble-based prediction model for outsourced software project risk prediction

2015 | Decision Support Systems | Citations: 0

Authors: Hu, Yong; Feng, Bin; Mo, Xizhu; Zhang, Xiangzhou; Ngai, E. W. T.; Fan, Ming; Liu, Mei

Abstract: Nowadays software is mainly developed through outsourcing and it has become one ... Expand

Abstract: Nowadays software is mainly developed through outsourcing and it has become one of the most important business practice strategies for the software industry. However, outsourcing projects are often affiliated with high failure rate. Therefore to ensure success in outsourcing projects, past research has aimed to develop intelligent risk prediction models to evaluate the success rate and cost-effectiveness of software projects. In this study, we first summarized related work over the past 20years and observed that all existing prediction models assume equal misclassification costs, neglecting actual situations in the management of software projects. In fact, overlooking project failure is far more serious than the misclassification of a success-prone project as a failure. Moreover, ensemble learning, a technique well-recognized to improve prediction performance in other fields, has not yet been comprehensively studied in software project risk prediction. This study aims to close the research gaps by exploring cost-sensitive analysis and classifier ensemble methods. Comparative analysis with T-test on 60 different risk prediction models using 327 outsourced software project samples suggests that the ideal model is a homogeneous ensemble model of decision trees (DT) based on bagging. Interestingly, DT underperformed Support Vector Machine (SVM) in accuracy (i.e., assuming equal misclassification cost), but outperformed in cost-sensitive analysis under the proposed framework. In conclusion, this study proposes the first cost-sensitive and ensemble-based hybrid modeling framework (COSENS) for software project risk prediction. In addition, it establishes a new rigorous evaluation standard for assessing software risk prediction models by considering misclassification costs. Collapse

Topics: IT project IS management

Methods: Student's t-test parametric test experiment support vector machine ensemble learning