Modern Centaurs: How Humans and AI Systems Interact in Sales Forecasting

2023 | European Conference On Information Systems | Citations: 0

Authors: Beese, Jannis; Fahse, Tobias

Abstract: Recent achievements of artificial intelligence (AI) have caused organizations t ... Expand

Abstract: Recent achievements of artificial intelligence (AI) have caused organizations to increasingly bring AI capabilities into their core business processes. Such AI-supported business processes often result in human-AI hybrid systems, which consist of an AI system, which performs most of the execution, and humans, who monitor this execution and occasionally provide additional inputs and overrides. Using sales data from Walmart, we conduct an online study to investigate if human supervision can improve upon state-of-the-art AI forecasts. Furthermore, we analyze the perceptions and behavioral intentions of the human participants over time. We find that human interventions consistently lead to less accurate forecasts and that participants initially underestimate the AI system's accuracy and overestimate their own potential to improve upon AI forecasts. However, perceptions quickly shift over the course of the study, causing the participants to perceive the AI system increasingly favorably, which also leads to behavioral changes and better hybrid system performance. Collapse

Topics: artificial intelligence IT productivity information system use task productivity perceived usefulness

Methods: regression analysis method survey LightGBM conceptual modelling experiment

Theories: technology acceptance model

Algorithm-Human-Algorithm: A New Classification Approach to Integrating Judgemental Adjustments

2023 | International Conference on Information Systems | Citations: 0

Authors: Chen, Christopher; Jain, Nitish; Karamshetty, Varun

Abstract: Modern-day firms face the predicament of blending the comparative advantages of ... Expand

Abstract: Modern-day firms face the predicament of blending the comparative advantages of their two core resources: machines and humans. When forecasting demand (e.g. for a product), extant literature documents that always permitting (or prohibiting) human revision of a machine forecast is beneficial if the humans' private information role is larger (or smaller) than that of machine-accessible public information. We propose and test a complementary framework that shifts the focus to the regulation of each human revision; and in doing so, adjusts for human vulnerability to systematic biases. We collaborate with a European retailer to compile a large dataset (∼ 1.1 mn transactions) on machine-led demand forecasts. Humans revise nearly 38% of these forecasts, but revisions do not always yield an improved final forecast. Compared to the always permit or prohibit strategies, the ideal regulation of each of these revisions could reduce the absolute deviation of the final forecast from realized sales in close to 50% of instances. Any regulation to obtain the best of the machine and human-revised forecasts requires an ability to predict revision quality. In addition to private information indicators, we show that indicators of systematic bias in human judgement (e.g., indicators of cognitive load) can improve the prediction of revision quality. In an out-of-sample analysis, our revision-level regulation approach picks the best of available forecasts in 14% more instances, compared to an always-permit or prohibit strategy at the product-store level. Our paper provides practitioners with a novel approach to combining the machine-human output to improve demand forecasting performance. Collapse

Topics: privacy logistics management supply chain management analytical information system

Methods: computational algorithm machine learning descriptive statistic XGBoost gradient boosting

Interpretable cost-sensitive regression through one-step boosting

2023 | Decision Support Systems | Citations: 0

Authors: Decorte, Thomas; Raymaekers, Jakob; Verdonck, Tim

Abstract: In most practical prediction problems, such as regression and classification, th ... Expand

Abstract: In most practical prediction problems, such as regression and classification, the different types of prediction errors are not equally costly in the decision-making process. Although there exist numerous real-world costsensitive regression problems, ranging from loan charge-off forecasting to house price predictions, the literature on cost-sensitive learning mainly focuses on classification and only a few solutions are proposed for regression problems. These regressions are typically characterized by an asymmetric cost structure, where over- and underpredictions of a similar magnitude face vastly different costs. In this paper, we present a one-step boosting method (OSB) for cost-sensitive regression. The proposed methodology leverages a secondary learner to incorporate cost-sensitivity into an already trained cost-insensitive regression model. The secondary learner is defined as a linear function of certain variables deemed interesting for cost-sensitivity. These variables do not necessarily need to be the same as in the already trained model. An efficient optimization algorithm is achieved through iteratively reweighted least squares using the asymmetric cost function. The obtained results become interpretable through bootstrapping, enabling decision makers to distinguish important variables for cost-sensitivity as well as facilitating statistical inference. Applying different cost functions and various initial cost-insensitive learning methods on several public datasets consistently yields a significant reduction in the average misprediction cost, illustrating the excellent performance of our approach. Collapse

Topics: Python decision making ease of use operating system Microsoft Windows

Methods: computational algorithm regression analysis method LightGBM boosting method linear regression analysis

Punctuality Predictions in Public Transportation: Quantifying the Effect of External Factors

2023 | International Conference on Business Informatics | Citations: 0

Authors: Meyer-Hollatz, Tim; Schwarz, Nina; Werner, Tim

Abstract: Increasing availability of large-scale datasets for automatic vehicle location ... Expand

Abstract: Increasing availability of large-scale datasets for automatic vehicle location (AVL) in public transportation (PT) encouraged researchers to investigate data-driven punctuality prediction models (PPMs). PPMs promise to accelerate the mobility transition through more accurate prediction delays, increased customer service levels, and more efficient and forward-looking planning by mobility providers. While several PPMs show promising results for buses and longdistance trains, a comprehensive study on external factors' effect on tram services is missing. Therefore, we implement four machine learning (ML) models to predict departure delays and elaborate on the performance increase by adding realworld weather and holiday data for three consecutive years. For our best model (XGBoost) the average MAE performance increased by 17.33 % compared to the average model performance when only trained on AVL data enriched by timetable characteristics. The results provide strong evidence that adding informationbearing features improve the forecast quality of PPMs. Collapse

Topics: decision making explainable artificial intelligence artificial intelligence

Methods: literature study XGBoost machine learning Cross-Industry Standard Process for Data Mining LightGBM

Learning Personalized Privacy Preference From Public Data

2023 | International Conference on Information Systems | Citations: 0

Authors: Wang, Wen; Li, Beibei

Abstract: Understanding consumers' privacy preferences is crucial for firms and policymak ... Expand

Abstract: Understanding consumers' privacy preferences is crucial for firms and policymakers to establish trust and encourage innovation and competition. With the widespread use of digital technologies, individuals generate and share vast amounts of data about them selves in the public domain. Even without knowing a person's private information, the psychosocial traits revealed in public data can provide valuable insights into their pri vacy preferences. In this study, we aim to predict personalized privacy preferences us ing social media posts. Our prediction model shows that psychosocial traits such as per sonality, lifestyles, riskpreference, economic thinking, emotions, etc extracted from posts provide significantly greater predictive power than demographic characteristics. Fur thermore, we demonstrate the practical value and impact of our model for business and society through a simulation analysis. Our tool can help platforms and policymakers es timate the impact of privacy policies and prevent potential harms such as discrimination. Collapse

Topics: privacy social media Twitter shapley additive explanation information privacy concern

Methods: BERT machine learning experiment computational algorithm topic model

Theories: game theory

Assuring quality and waiting time in real-time spatial crowdsourcing

2023 | Decision Support Systems | Citations: 0

Authors: Wu, Zhibin; Peng, Lijie; Xiang, Chuankai

Abstract: With the rapid development of mobile devices, spatial crowdsourcing has become a ... Expand

Abstract: With the rapid development of mobile devices, spatial crowdsourcing has become an important way to collect data. Task assignment is an important aspect of spatial crowdsourcing. How to improve the quality of the results and decrease the travel distance has been extensively studied in recent years. Existing studies often assume that moving speed is constant or real-time road network information is known. In this paper, the travel time is predicted based on historical data. A framework for time-prediction-based task assignment approach in spatial crowdsourcing (TP-TASC) is proposed. Firstly, a prediction model based on the light gradient boosting machine (LightGBM) is used to predict the travel time of workers with the consideration of the spatial features, the temporal features, and the climate features. Secondly, a heuristic algorithm is proposed to assign the spatial crowdsourcing tasks to appropriate workers. When a task is assigned to a worker, the payment of the worker is also determined automatically. Finally, the simulation experiments based on a real-world taxi-hailing dataset show that the proposed method can not only effectively minimize the task requesters’ waiting time, but also maximize the results’ quality. Collapse

Topics: crowdsourcing decision support system Python decision making

Methods: LightGBM experiment computational algorithm simulation descriptive statistic

Efficient fraud detection using deep boosting decision trees

2023 | Decision Support Systems | Citations: 0

Authors: Xu, Biao; Wang, Yao; Liao, Xiuwu; Wang, Kaidong

Abstract: Fraud detection is to identify, monitor, and prevent potentially fraudulent acti ... Expand

Abstract: Fraud detection is to identify, monitor, and prevent potentially fraudulent activities from complex data. The recent development and success in AI, especially machine learning, provides a new data-driven way to deal with fraud. From a methodological point of view, machine learning based fraud detection can be divided into two categories, i.e., conventional methods (decision tree, tree boosting methods, etc.) and deep learning, both of which have significant limitations in terms of the lack of representation learning ability for the former and interpretability for the latter. Furthermore, due to the rarity of detected fraud cases, the associated data is usually imbalanced, which seriously degrades the performance of classification algorithms. Collapse

Topics: fraud detection marketing management

Methods: machine learning computational algorithm decision tree classification artificial neural network gradient boosting

Financial risk prediction in supply chain finance based on buyer transaction behavior

2023 | Decision Support Systems | Citations: 0

Authors: Yi, Zelong; Liang, Zhuomin; Xie, Tongtong; Li, Fan

Abstract: Financial risk in supply chain finance (SCF) is defined as the possibility that ... Expand

Abstract: Financial risk in supply chain finance (SCF) is defined as the possibility that suppliers fall into liquidity crisis due to delayed payment. Predicting financial risk is important for supply chain stability. In this paper, a financial risk prediction model is developed using XGBoost and then evaluated by applying buyer transaction behavior data. We further construct single and hybrid models, respectively, and compare their performance using receiver operating characteristic curve (ROC), area under the ROC curve (AUC), and F1-Score. Last, feature importance and partial dependence plots (PDPs) are employed for model interpretation. The results show that XGBoost model can effectively predict potential financial risks, and shed lights on managers’ payment practice. This paper is one of the few studies that develop new models to examine financial risks in SCF empirically. Collapse

Topics: artificial intelligence supply chain management big data decision making partial dependence plot

Methods: machine learning XGBoost LightGBM computational algorithm random forest classification

Introduction to the Special Issue on Pattern-Driven Mining, Analytics, and Prediction for Decision Making, Part 1

2022 | ACM Transactions on Management Information Systems | Citations: 0

Authors: Cwei, Lin Jerry; Sahoo, Nachiketa; Srivastava, Gautam; Ding, Weiping

Abstract: Data Mining is an analytic process to explore data in search of consistent patt ... Expand

Abstract: Data Mining is an analytic process to explore data in search of consistent patterns and/or systematic relationships between variables, and then to validate the findings by applying the detected patterns to new sets of data. More specifically, pattern-driven mining, analytics, and prediction have received a lot of attention in the last two decades since information discovered in data can be used to support decision and strategy making. The results can also be utilized in decision support or information management systems (IMS). Different types of patterns and knowledge can be mined (extracted) from various applications and domains. Many previous studies focused on designing and implementing new methodologies to handle different constraints and requirements. This special issue focuses on discovering the knowledge, rules, and information for decision support and management information systems. Innovative methodologies, principles, methods, techniques, framework, theory, and applications are thus considered to deal with the challenges for decision support and management information systems. In this special issue there were 47 submissions. For Part 1, we are publishing eight articles, with more planned for a future issue. All accepted manuscripts have made a significant scientific contribution and presented a rigorous evaluation of the Information Systems outcomes in real-world practices and analysis. The summary of the accepted papers is stated below.In DSSAE: Deep stacked sparse autoencoder analytical model for COVID-19 diagnosis by fractional Fourier entropy [1], the authors proposed a novel artificial intelligence model to diagnose COVID-19 based on chest CT images. First, the two-dimensional fractional Fourier entropy was presented to extract features. A custom deep-stacked sparse autoencoder (DSSAE) model was then created to serve as the classifier. Improved multiple-way data augmentation was proposed to resist overfitting. Results showed that the designed DSSAE model obtains a micro-averaged F1 score of 92.32% in handling a four-class problem. In addition, the designed model outperforms 10 state-ofthe-art approaches.In TRG-DAtt: The target relational graph and double attention network based sentiment analysis and prediction for supporting decision making [2], the authors designed a TRG-DAtt model for sentiment analysis based on target relational graph (TRG) and double attention network (DAtt) to analyze the emotional information for supporting decision making. A dependency tree-based TRG is firstly introduced to independently and fully mine the semantic relationships. A dependency graph attention network (DGAT) is then designed, as well as the interactive attention network (IAT) to form the DAtt, and obtained the emotional features of the target words and reviews. DGAT models the dependency of the TRG by aggregating the semantic information, and the target emotional enhancement features are obtained by the DGAT as an input to the IAT.Finally, the target emotional enhancement features are weighted by the impact factor to generate the review's emotional features.In Mining high utility itemsets with hill climbing and simulated annealing [3], the authors proposed two algorithms for high-utility itemset mining (HUIM) based on Hill Climbing (HUIM-HC) and Simulated Annealing (HUIM-SA). Both algorithms transform the input database into a bitmap for efficient utility computation and search space pruning. To improve population diversity, high-utility itemsets (HUIs) discovered by evolution are used as target values for the next population instead of keeping the current optimal values in the next population. Through experiments on real-life datasets, it was found that the proposed algorithms are faster than stateof-the-art heuristic and evolutionary HUIM algorithms, that HUIM-SA discovers similar HUIs, and that HUIM-SA evolves linearly with the number of iterations.In Cross-modality transfer learning for image-text information management [4], the authors proposed a novel heterogeneous Cross-Modality Transfer Learning (CMTL) algorithm, which requires only a tiny set of unlabeled target data and labelled source data with associate text tags. In addition, the authors introduced a latent semantic information extraction (LSIE) method to connect the information learned from the image data and the text data. The proposed method can thus effectively handle the information transfer across different modalities (text-image). From the experimental results, the authors stated the designed model has achieved up to 5% higher classification accuracy non-transfer algorithms and up to 9% higher than existing CMTL models.In AI-based vehicular network toward 6G and IoT: Deep learning approaches [5], the authors compared the image recognition performance of various models, specifically to determine the presence or absence of the target in the frame, whereas previous research had mostly focused on object recognition used to recognize fixed objects. Thus, to address the potentially dangerous driver behaviour, the authors applied deep learning (YOLOv2+CNN) approaches for image recognition to develop an AI-based detection system that can detect potentially dangerous driving behaviour. Multiple convolutional neural networks (CNN)-based techniques including VGG16, VGG19, Densenet, and Openpose were compared in terms of their ability to detect and identify problematic driving. From the results, it can be seen that although OpenPose is used for human skeleton recognition, it slightly underperforms the YOLOv2 method, which conducts posture recognition through locating objects.In Novel machine learning for big data analytics in intelligent support information management systems [6], the authors utilized machine learning for the optimized Light Gradient Boosting Machine (LightGBM) algorithm. An Android-based intelligent support information management system is designed based on LightGBM for the big data analysis and classification management of information in the intelligent support information management system. The system is designed with modules of employee registration and login, company announcement notice, attendance and attendance management, self-service, and daily tools with the company as the subject. Results demonstrate that the training time of the optimized LightGBM algorithm can stabilize at about 100s, and the test time can stabilize at 0.68s. Besides, its accuracy rate can reach 89.24%, which is at least 3.6% higher than other machine learning algorithms. In summary, the constructed intelligent support system can attain high accuracy while minimizing error, with an apparent acceleration effect.In Distance based pattern driven mining for outlier detection in high dimensional big dataset [7], the authors investigated a novel method for detecting cluster outliers in a multidimensional dataset, which is capable of identifying the clusters and outliers for datasets containing noise. The proposed method can detect the groups and outliers left by the clustering process, like instant irregular sets of clusters (C) and outliers (O), to boost the results. The results obtained after applying the algorithm to the dataset improved in terms of several parameters. For the comparative analysis, Collapse

Topics: artificial intelligence internet of things mobile system big data information management

Methods: computational algorithm machine learning LightGBM convolutional neural network data transformation

Novel Machine Learning for Big Data Analytics in Intelligent Support Information Management Systems

2022 | ACM Transactions on Management Information Systems | Citations: 0

Authors: Lv, Zhihan; Lou, Ranran; Feng, Hailin; Chen, Dongliang; Lv, Haibin

Abstract: Two-dimensional1 arrays of bi-component structures made of cobalt and permalloy ... Expand

Abstract: Two-dimensional1 arrays of bi-component structures made of cobalt and permalloy elliptical dots with thickness of 25 nm, length 1 mm and width of 225 nm, have been prepared by a self-aligned shadow deposition technique. Brillouin light scattering has been exploited to study the frequency dependence of thermally excited magnetic eigenmodes on the intensity of the external magnetic field, applied along the easy axis of the elements. Scientific information technology has been developed rapidly. Here, the purposes are to make people's lives more convenient and ensure information management and classification. The machine learning algorithm is improved to obtain the optimized Light Gradient Boosting Machine (LightGBM) algorithm. Then, an Android-based intelligent support information management system is designed based on LightGBM for the big data analysis and classification management of information in the intelligent support information management system. The system is designed with modules of employee registration and login, company announcement notice, attendance and attendance management, self-service, and daily tools with the company as the subject. Furthermore, the performance of the constructed information management system is analyzed through simulations. Results demonstrate that the training time of the optimized LightGBM algorithm can stabilize at about 100s, and the test time can stabilize at 0.68s. Besides, its accuracy rate can reach 89.24%, which is at least 3.6% higher than other machine learning algorithms. Moreover, the acceleration efficiency analysis of each algorithm suggests that the optimized LightGBM algorithm is suitable for processing large amounts of data; its acceleration effect is more apparent, and its acceleration ratio is higher than other algorithms. Hence, the constructed intelligent support information management system can reach a high accuracy while ensuring the error, with apparent acceleration effect. Therefore, this model can provide an experimental reference for information classification and management in various fields. Collapse

Topics: information management internet technology top management support artificial intelligence internet of things

Methods: computational algorithm machine learning LightGBM cluster analysis decision tree classification