Search results

The purpose of this paper is to consolidate existing knowledge and provide a deeper understanding of the use of social media (SM) data for predictions in various areas, such as disease outbreaks, product sales, stock market volatility and elections outcome predictions.

The scientific literature was systematically reviewed to identify relevant empirical studies. These studies were analysed and synthesized in the form of a proposed conceptual framework, which was thereafter applied to further analyse this literature, hence gaining new insights into the field.

The proposed framework reveals that all relevant studies can be decomposed into a small number of steps, and different approaches can be followed in each step. The application of the framework resulted in interesting findings. For example, most studies support SM predictive power, however, more than one-third of these studies infer predictive power without employing predictive analytics. In addition, analysis suggests that there is a clear need for more advanced sentiment analysis methods as well as methods for identifying search terms for collection and filtering of raw SM data.

The proposed framework enables researchers to classify and evaluate existing studies, to design scientifically rigorous new studies and to identify the field's weaknesses, hence proposing future research directions.

Partial least squares (PLS) has been introduced as a “causal-predictive” approach to structural equation modeling (SEM), designed to overcome the apparent dichotomy between explanation and prediction. However, while researchers using PLS-SEM routinely stress the predictive nature of their analyses, model evaluation assessment relies exclusively on metrics designed to assess the path model’s explanatory power. Recent research has proposed PLSpredict, a holdout sample-based procedure that generates case-level predictions on an item or a construct level. This paper offers guidelines for applying PLSpredict and explains the key choices researchers need to make using the procedure.

The authors discuss the need for prediction-oriented model evaluations in PLS-SEM and conceptually explain and further advance the PLSpredict method. In addition, they illustrate the PLSpredict procedure’s use with a tourism marketing model and provide recommendations on how the results should be interpreted. While the focus of the paper is on the PLSpredict procedure, the overarching aim is to encourage the routine prediction-oriented assessment in PLS-SEM analyses.

The paper advances PLSpredict and offers guidance on how to use this prediction-oriented model evaluation approach. Researchers should routinely consider the assessment of the predictive power of their PLS path models. PLSpredict is a useful and straightforward approach to evaluate the out-of-sample predictive capabilities of PLS path models that researchers can apply in their studies.

Future research should seek to extend PLSpredict’s capabilities, for example, by developing more benchmarks for comparing PLS-SEM results and empirically contrasting the earliest antecedent and the direct antecedent approaches to predictive power assessment.

This paper offers clear guidelines for using PLSpredict, which researchers and practitioners should routinely apply as part of their PLS-SEM analyses.

This research substantiates the use of PLSpredict. It provides marketing researchers and practitioners with the knowledge they need to properly assess, report and interpret PLS-SEM results. Thereby, this research contributes to safeguarding the rigor of marketing studies using PLS-SEM.

The purpose of this study is to investigate the relationship between gold–platinum price ratio (GP) and stock returns in international stock markets. The study addresses three empirical questions: (1) Does GP have robust predictive power in international stock markets? (2) Does GP outperform other macroeconomic variables in international stock markets? (3) What is the relationship between GP and stock market returns during economic recessions?

The study mainly uses OLS regressions to perform empirical tests for a comprehensive set of 17 advanced international stock markets and overall world market. The monthly data is used for the period January 1978 to July 2019, 499 observations for each market.

The study finds that the first-difference of GP (ΔGP), not the initial-level of GP, has strong predictive power for stock returns, both in short- and long-time horizons. The results remain robust after controlling for a number of macroeconomic predictors. The out-of-sample test results are significant, confirming the robustness of the predictive power of ΔGP.

This study is the first to examine the ability of the ΔGP to predict stock returns, and provide novel evidence on the relationship between ΔGP and international stock markets. The study draws on behavioral finance theory, specifically the myopic loss aversion, the herd effect and the limited attention theory, to explain the predictability of stock returns in international stock markets.

The purpose of this paper is to implement a finite set model predictive control algorithm to a shunt (or parallel), multilevel (cascaded H-bridge) active power filter (APF). Specifically, the purpose is to get a controller that could compensate the mains current and, at the same time, to control the voltages of its capacitors. This strategy avoids the use of multiple PWM carriers or another type of special modulator, and requires a relatively low processing power.

This paper is focussed in the application of the predictive controller to a single-phase parallel APF composed for two H-bridges connected in series. The same methodology can be applied to a three-phase APF. In the DC buses of each H-bridge, a floating capacitor was connected, whose voltage is regulated by the predictive controller. The controller is composed by, first, a model for the charge/discharge dynamics for each floating capacitor and a model for the output current of the APF; second, a cost function; and third, an optimization algorithm that is able to control all these variables at the same time, choosing in each sample period the best combination of firing pulses.

The controller can track the voltage references, compensate the current harmonics and compensate reactive power with an algorithm that evaluates only the three nearest voltage levels to the last voltage level applied in the inverter. This strategy decreases the number of calculations required by the predictive algorithm. This controller can be applied to the general case of a single-phase multilevel APF of N-levels and extend it to the three-phase case without major problems.

The implemented controller, when the authors consider a constant sample time, gives a mains current with a Total Harmonic Distortion (THD-I) slightly greater in comparison with the base algorithm (that evaluates all the voltage levels). However, when the authors consider the processing times under the same processor, the implemented algorithm requires less time to get the optimal values, can get lower sampling times and then a best performance in terms of THD-I. To implement the controller in a three-phase APF, a faster Digital Signal Processor would be required.

The implemented solution uses a model for the charge/discharge of the capacitors and for the filter current that enable to operate the cascaded multilevel inverter with asymmetrical voltages while compensates the mains currents, with a predictive algorithm that requires a relatively low amount of calculations.

Generalized structured component analysis (GSCA) and partial least squares path modeling (PLSPM) are two key component-based approaches to structural equation modeling that facilitate the analysis of theoretically established models in terms of both explanation and prediction. This study aims to offer a comparative evaluation of GSCA and PLSPM in a predictive modeling framework.

A simulation study compares the predictive performance of GSCA and PLSPM under various simulation conditions and different prediction types of correctly specified and misspecified models.

The results suggest that GSCA with reflective composite indicators (GSCA_R) is the most versatile approach. For observed prediction, which uses the component scores to generate prediction for the indicators, GSCA_R performs slightly better than PLSPM with mode A. For operative prediction, which considers all parameter estimates to generate predictions, both methods perform equally well. GSCA with formative composite indicators and PLSPM with mode B generally lag behind the other methods.

Future research may further assess the methods’ prediction precision, considering more experimental factors with a wider range of levels, including more extreme ones.

When prediction is the primary study aim, researchers should generally revert to GSCA_R, considering its performance for observed and operative prediction together.

This research is the first to compare the relative efficacy of GSCA and PLSPM in terms of predictive power.

Artificial intelligence (AI) offers many benefits to improve predictive marketing practice. It raises ethical concerns regarding customer prioritization, market share concentration and consumer manipulation. This paper explores these ethical concerns from a contemporary perspective, drawing on the experiences and perspectives of AI and predictive marketing professionals. This study aims to contribute to the field by providing a modern perspective on the ethical concerns of AI usage in predictive marketing, drawing on the experiences and perspectives of professionals in the area.

The study conducted semistructured interviews for 6 weeks with 14 participants experienced in AI-enabled systems for marketing, using purposive and snowball sampling techniques. Thematic analysis was used to explore themes emerging from the data.

Results reveal that using AI in marketing could lead to unintended consequences, such as perpetuating existing biases, violating customer privacy, limiting competition and manipulating consumer behavior.

The authors identify seven unique themes and benchmark them with Ashok’s model to provide a structured lens for interpreting the results. The framework presented by this research is unique and can be used to support ethical research spanning social, technological and economic aspects within the predictive marketing domain.

La Inteligencia Artificial (IA) ofrece muchos beneficios para mejorar la práctica del marketing predictivo. Sin embargo, plantea preocupaciones éticas relacionadas con la priorización de clientes, la concentración de cuota de mercado y la manipulación del consumidor. Este artículo explora estas preocupaciones éticas desde una perspectiva contemporánea, basándose en las experiencias y perspectivas de profesionales en IA y marketing predictivo. El estudio tiene como objetivo contribuir a la literatura de este ámbito al proporcionar una perspectiva moderna sobre las preocupaciones éticas del uso de la IA en el marketing predictivo, basándose en las experiencias y perspectivas de profesionales en el área.

Para realizar el estudio se realizaron entrevistas semiestructuradas durante seis semanas con 14 participantes con experiencia en sistemas habilitados para IA en marketing, utilizando técnicas de muestreo intencional y de bola de nieve. Se utilizó un análisis temático para explorar los temas que surgieron de los datos.

Los resultados revelan que el uso de la IA en marketing podría tener consecuencias no deseadas, como perpetuar sesgos existentes, violar la privacidad del cliente, limitar la competencia y manipular el comportamiento del consumidor.

El estudio identifica siete temas y los comparan con el modelo de Ashok para proporcionar una perspectiva estructurada para interpretar los resultados. El marco presentado por esta investigación es único y puede utilizarse para respaldar investigaciones éticas que abarquen aspectos sociales, tecnológicos y económicos dentro del ámbito del marketing predictivo.

人工智能（AI）为改进预测营销实践带来了诸多益处。然而, 这也引发了与客户优先级、市场份额集中和消费者操纵等伦理问题相关的观点。本文从当代角度深入探讨了这些伦理观点, 充分借鉴了人工智能和预测营销领域专业人士的经验和观点。旨在通过现代视角提供关于在预测营销中应用人工智能时所涉及的伦理观点, 为该领域做出有益贡献。

本研究采用了目的性和雪球抽样技术, 与14位在人工智能营销系统领域具有丰富经验的参与者进行为期六周的半结构化访谈。研究采用主题分析方法, 旨在深入挖掘数据中显现的主要主题。

研究结果表明, 在营销领域使用人工智能可能引发一系列意外后果, 包括但不限于加强现有偏见、侵犯客户隐私、限制竞争以及操纵消费者行为。

本研究通过明确定义七个独特的主题, 并采用阿肖克模型进行基准比较, 为读者提供了一个结构化的视角, 以解释研究结果。所提出的框架具有独特之处, 可有效支持在跨足社会、技术和经济领域的预测营销中展开的伦理研究。

Since stock return and volatility matters to investors, this study proposes to incorporate the textual sentiment of annual reports in stock price crash risk prediction.

Specific sentences gathered from management discussions and their subsequent analyses are tokenized and transformed into numeric vectors using textual mining techniques, and then the Naïve Bayes method is applied to score the sentiment, which is used as an input variable for crash risk prediction. The results are compared between a collection of predictive models, including linear regression (LR) and machine learning techniques.

The experimental results find that those predictive models that incorporate textual sentiment significantly outperform the baseline models with only accounting and market variables included. These conclusions hold when crash risk is proxied by either the negative skewness of the return distribution or down-to-up volatility (DUVOL).

It should be noted that the authors' study focuses on examining the predictive power of textual sentiment in crash risk prediction, while other dimensions of textual features such as readability and thematic contents are not considered. More analysis is needed to explore the predictive power of textual features from various dimensions, with the most recent sample data included in future studies.

The authors' study provides implications for the information value of textual data in financial analysis and risk management. It suggests that the soft information contained within annual reports may prove informative in crash risk prediction, and the incorporation of textual sentiment provides an incremental improvement in overall predictive performance.

This study aims to evaluate the usage of partial least squares structural equation modeling (PLS-SEM) in journals related to logistics and supply chain management (LSCM).

Based on a structured literature review approach, the authors reviewed 401 articles in the field of LSCM applying PLS-SEM published in 15 major journals between 2014 and 2022. The analysis focused on reasons for using PLS-SEM, measurement model and structural model evaluation criteria, advanced analysis techniques and reporting practices.

LSCM researchers sometimes did not clarify the reasons for using PLS-SEM, such as sample size, complex models and non-normal distributions. Additionally, most articles exhibit limited use of measurement models and structural model evaluation techniques, leading to inappropriate use of assessment criteria. Furthermore, progress in the practical implementation of advanced analysis techniques is slow, and there is a need for improved transparency in reporting analysis algorithms.

This study contributes to the field of LSCM by providing clear criteria and steps for using PLS-SEM, enriching the understanding and advancement of research methodologies in this field.

To identify predictors of corporate financial distress, using the discriminant and logit models, in an emerging market over a period of economic turbulence and to reveal the comparative predictive and classification accuracies of the models in this different environmental setting.

The research relies on a sample of 27 failed and 27 non‐failed manufacturing firms listed in the Istanbul Stock Exchange over the 1996‐2003 period, which includes a period of high economic growth (1996‐1999) followed by an economic crisis period (2000‐2002). The two well‐known methods, discriminant analysis and logit, are compared on the basis of a better overall fit and a higher percentage of correct classification under changing economic conditions. Furthermore, this research attempts to reveal the changes, if any, in the bankruptcy predictors, from those found in the earlier studies that rested on the data from the developed markets.

The logistic regression model is found to have higher classification power and predictive accuracy, over the four years prior to bankruptcy, than the discriminant model. In this research, the discriminant and logit models identify the same number of significant predictors out of the total variables analyzed, and six of these are common in both. EBITDA/total assets is the most important predictor of financial distress in both models. The logit model identifies operating profit margin and the proportion of trade credit within total claims ratios as the second and third most important predictors, respectively.

This paper reveals the accuracy with which the discriminant and logit models work in an emerging market over a period when firms face high uncertainty and turbulence. This study may be extended to other emerging markets to eliminate the limitation of the small sample size in this study and to further validate the use of these models in the developing countries. This can serve to make the methods important decision tools for managers and investors in these volatile markets.