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

Box office sales and social media: A cross-platform comparison of predictive ability and mechanisms

2021 | Decision Support Systems | Citations: 0

Authors: Bogaert, Matthias; Ballings, Michel; Van den Poel, Dirk; Oztekin, Asil

Abstract: This paper aims to determine the power of social media data (Facebook and Twitte ... Expand

Abstract: This paper aims to determine the power of social media data (Facebook and Twitter) in predicting box office sales, which platforms, data types and variables are the most important and why. To do so, we compare several models based on movie data, Facebook data, and Twitter data. We benchmark these model comparisons using various prediction algorithms. Next, we apply information-fusion sensitivity analysis to evaluate which variables are driving the predictive performance. Our analysis shows that social media data significantly increases the predictive power of traditional box office prediction models. Facebook data clearly outperform Twitter data and including user-generated content next to marketer-generated always improves predictive power. Our sensitivity analysis reveals that volume and valence based combination variables pertaining to Facebook comments are the most important variables. Furthermore, we provide an in-depth analysis of the potential mechanisms driving differential predictive ability of Facebook and Twitter. Our findings suggest that Twitter has less of an impact on box office sales than Facebook because Twitter users have less source credibility than Facebook users. Our results are important for practitioners, marketers and academics who want to employ social media data for box office sales predictions. Collapse

Topics: Twitter Facebook social media application programming interface user-generated content

Methods: computational algorithm sensitivity analysis decision tree classification sentiment analysis random forest classification

Theories: signaling theory

Spline-rule ensemble classifiers with structured sparsity regularization for interpretable customer churn modeling

2021 | Decision Support Systems | Citations: 0

Authors: De Bock, Koen W.; De Caigny, Arno

Abstract: An important business domain that relies heavily on advanced statistical- and ma ... Expand

Abstract: An important business domain that relies heavily on advanced statistical- and machine learning algorithms to support operational decision-making is customer retention management. Customer churn prediction is a crucial tool to support customer retention. It allows an early identification of customers who are at risk to abandon the company and provides the ability to gain insights into why customers are at risk. Hence, customer churn pre­ diction models should complement predictive performance with model insights. Inspired by their ability to reconcile strong predictive performance and interpretability, this study introduces rule ensembles and their extension, spline-rule ensembles, as a promising family of classification algorithms to the customer churn pre­ diction domain. Spline-rule ensembles combine the flexibility of a tree-based ensemble classifier with the simplicity of regression analysis. They do, however, neglect the relatedness between potentially conflicting model components which can introduce unnecessary complexity in the models and compromises model inter­ pretability. To tackle this issue, a novel algorithmic extension, spline-rule ensembles with sparse group lasso regularization (SRE-SGL) is proposed to enhance interpretability through structured regularization. Experiments on fourteen real-world customer churn data sets in different industries (i) demonstrate the superior predictive performance of spline-rule ensembles with sparse group lasso over a set well yet powerful benchmark methods in terms of AUC and top decile lift; (ii) show that spline-rule ensembles with sparse group lasso regularization significantly outperform conventional rule ensembles whilst performing at least as well as conventional splinerule ensembles; and (iii) illustrate the interpretable nature of a spline-rule ensemble model and the advantage of structured regularization in SRE-SGL by means of a case study on customer churn prediction for a telecommu­ nications company. Collapse

Topics: partial dependence plot logistics management marketing intelligence decision making missing data

Methods: computational algorithm machine learning case study experiment logistic regression

PRESCRIPTIVE PROCESS ANALYTICS WITH DEEP LEARNING AND EXPLAINABLE ARTIFICIAL INTELLIGENCE

2020 | European Conference On Information Systems | Citations: 0

Authors: Mehdiyev, Nijat; Fettke, Peter

Abstract: The proliferation of enterprise information systems allows to capture the digit ... Expand

Abstract: The proliferation of enterprise information systems allows to capture the digital footprints generated over various user interaction phases. Such transaction data describe the details of the user interactions and the underlying processes on a fine granular level. Building capabilities to analyse the transactional process data is a key success differentiation factor that enables to grasp the user behavior more effectively. In this study, we aim to propose a prescriptive process analytics approach by combining the approaches from the machine learning, process mining and explainable artificial intelligence (XAI) research domains. After examining predictability of the business processes by employing an advanced deep learning approach, this study applies for the first time both in the business process prediction and customer journey analytics research domains an XAI technique, Partial Dependence Plots (PDP), to generate causal explanations. The real-life process data delivered by various information systems of a Dutch autonomous administrative authority were used to investigate the appropriateness of the proposed prescriptive analytics approach. The applied deep learning approach achieves a very good performance with an Area Under ROC Curve of 0.933. The generated explanations with PDP give insights to identify a set of alternative courses-of-actions to prevent the undesired outcomes. Collapse

Topics: business process management explainable artificial intelligence privacy process mining open source

Methods: deep learning machine learning binary classification artificial neural network design artifact

The added value of Facebook friends data in event attendance prediction

2016 | Decision Support Systems | Citations: 8

Authors: Bogaert, Matthias; Ballings, Michel; Van den Poel, Dirk

Abstract: This paper seeks to assess the added value of a Facebook user's friends data in ... Expand

Abstract: This paper seeks to assess the added value of a Facebook user's friends data in event attendance prediction over and above user data. For this purpose we gathered data of users that have liked an anonymous European soccer team on Facebook. In addition we obtained data from all their friends. In order to assess the added value of friends data we have built two models for five different algorithms (Logistic Regression, Random Forest, Adaboost, Neural Networks and Naive Bayes). The baseline model contained only user data and the augmented model contained both user and friends data. We employed five times two-fold cross-validation and the Wilcoxon signed rank test to validate our findings. The results suggest that the inclusion of friends data in our predictive model increases the area under the receiver operating characteristic curve (AUC). Out of five algorithms, the increase is significant for three algorithms, marginally significant for one algorithm, and not significant for one algorithm. The increase in AUC ranged from 0.21%-points to 0.82%-points. The analyses show that a top predictor is the number of friends that are attending the focal event. To the best of our knowledge this is the first study that evaluates the added value of friends network data over and above user data in event attendance prediction on Facebook. These findings clearly indicate that including network data in event prediction models is a viable strategy for improving model performance. Collapse

Topics: Facebook partial dependence plot mobile application social media database system

Methods: computational algorithm machine learning random forest classification artificial neural network logistic regression

The added value of auxiliary data in sentiment analysis of Facebook posts

2016 | Decision Support Systems | Citations: 12

Authors: Meire, Matthijs; Ballings, Michel; Van den Poel, Dirk

Abstract: The purpose of this study is to (1) assess the added value of information availa ... Expand

Abstract: The purpose of this study is to (1) assess the added value of information available before (i.e., leading) and after (i.e., lagging) the focal post's creation time in sentiment analysis of Facebook posts, (2) determine which predictors are most important, and (3) investigate the relationship between top predictors and sentiment. We build a sentiment prediction model, including leading information, lagging information, and traditional post variables. We benchmark Random Forest and Support Vector Machines using five times twofold cross-validation. The results indicate that both leading and lagging information increase the model's predictive performance. The most important predictors include the number of uppercase letters, the number of likes and the number of negative comments. A higher number of uppercase letters and likes increases the likelihood of a positive post, while a higher number of comments increases the likelihood of a negative post. The main contribution of this study is that it is the first to assess the added value of leading and lagging information in the context of sentiment analysis. Collapse

Topics: Facebook personality partial dependence plot social media Twitter

Methods: sentiment analysis random forest classification machine learning computational algorithm support vector machine

Adaptive Advertisement Recommender Systems for Digital Signage

2015 | Americas Conference on Information Systems | Citations: 0

Authors: Addicam, Sanjay; Balkan, Sule; Baydogan, Mustafa

Abstract: With the incorporation of new technologies, digital signages can adopt their con ... Expand

Abstract: With the incorporation of new technologies, digital signages can adopt their content in real time to the audience demographic and temporal features. This research proposes an adaptive advertisement recommender system for digital signage. Our objective is to create a quantitative method for targeted advertising. After analyzing digital signage advertisement viewing data collected over the course of two months, our results show that learning-to-rank approach using Stochastic Gradient-Boosted Trees (SGBT) yields the best adaptive advertisement recommender system. Our system can identify the best sequence of advertisements to attract the most viewing. More importantly, we can use the same method for different business objectives like attracting the longest time of viewing or targeting a certain age groups or genders. Collapse

Topics: recommender system partial dependence plot advertising management accounting information system use

Methods: experiment post-hoc analysis ANOVA parametric test regression analysis method