Adaptive Tactical Pricing in Multi‐Agent Supply Chain Markets Using Economic Regimes

In today's complex and dynamic supply chain markets, information systems are essential for effective supply chain management. Complex decision making processes on strategic, tactical, and operational levels require substantial timely support in order to contribute to organizations' agility. Consequently, there is a need for sophisticated dynamic product pricing mechanisms that can adapt quickly to changing market conditions and competitors' strategies. We propose a two‐layered machine learning approach to compute tactical pricing decisions in real time. The first layer estimates prevailing economic conditions—economic regimes—identifying and predicting current and future market conditions. In the second layer, we train a neural network for each regime to estimate price distributions in real time using available information. The neural networks compute offer acceptance probabilities from a tactical perspective to meet desired sales quotas. We validate our approach in the trading agent competition for supply chain management. When competing against the world's leading agents, the performance of our system significantly improves compared to using only economic regimes to predict prices. Profits increase significantly even though the prices and sales volume do not change significantly. Instead, tactical pricing results in a more efficient sales strategy by reducing both finished goods and components inventory costs.     

Reputation for Quality

We propose a model of firm reputation in which a firm can invest or disinvest in product quality and the firm's reputation is defined as the market's belief about this quality. We analyze the relationship between a firm's reputation and its investment incentives, and derive implications for reputational dynamics. Reputational incentives depend on the specification of market learning. When consumers learn about quality through perfect good news signals, incentives decrease in reputation and there is a unique work–shirk equilibrium with ergodic dynamics. When learning is through perfect bad news signals, incentives increase in reputation and there is a continuum of shirk–work equilibria with path‐dependent dynamics. For a class of imperfect Poisson learning processes and low investment costs, we show that there exists a work–shirk equilibrium with ergodic dynamics. For a subclass of these learning processes, any equilibrium must feature working at all low and intermediate levels of reputation and shirking at the top.     

A comparison of Japanese failure models: Hybrid neural networks,logit models,and discriminant analysis

This article looks at the ability of a relatively new technique, hybrid artificial neural networks (ANNs), to predict Japanese banking and firm failures. These models are compared with traditional statistical techniques and conventional ANN models. The results suggest that hybrid neural networks outperform all other models in predicting failure for one year prior to the event. This suggests that for researchers, policymakers, and others interested in early warning systems, the hybrid network may be a useful tool for predicting banking and firm failures.     

Application of the Back Propagation Neural Network Algorithm with Monotonicity Constraints for Two-Group Classification Problems*

Neural network techniques are widely used in solving pattern recognition or classification problems. However, when statistical data are used in supervised training of a neural network employing the back-propagation least mean square algorithm, the behavior of the classification boundary during training is often unpredictable. This research suggests the application of monotonicity constraints to the back propagation learning algorithm. When the training sample set is preprocessed by a linear classification function, neural network performance and efficiency can be improved in classification applications where the feature vector is related monotonically to the pattern vector. Since most classification problems in business possess monotonic properties, this technique is useful in those problems where any assumptions about the properties of the data are inappropriate.     

The Application of Neural Networks and a Qualitative Response Model to the Auditor's Going Concern Uncertainty Decision*

An auditor gives a going concern uncertainty opinion when the client company is at risk of failure or exhibits other signs of distress that threaten its ability to continue as a going concern. The decision to issue a going concern opinion is an unstructured task that requires the use of the auditor's judgment. In cases where judgment is required, the auditor may benefit from the use of statistical analysis or other forms of decision models to support the final decision. This study uses the generalized reduced gradient (GRG2) optimizer for neural network learning, a backpropagation neural network, and a logit model to predict which firms would receive audit reports reflecting a going concern uncertainty modification. The GRG2 optimizer has previously been used as a more efficient optimizer for solving business problems. The neural network model formulated using GRG2 has the highest prediction accuracy of 95 percent. It performs best when tested with a small number of variables on a group of data sets, each containing 70 observations. While the logit procedure fails to converge when using our eight variable model, the GRG2 based neural network analysis provides consistent results using either eight or four variable models. The GRG2 based neural network is proposed as a robust alternative model for auditors to support their assessment of going concern uncertainty affecting the client company.     

An empirical study of design and testing of hybrid evolutionary–neural approach for classification

We propose a hybrid evolutionary–neural approach for binary classification that incorporates a special training data over-fitting minimizing selection procedure for improving the prediction accuracy on holdout sample. Our approach integrates parallel global search capability of genetic algorithms (GAs) and local gradient-descent search of the back-propagation algorithm. Using a set of simulated and real life data sets, we illustrate that the proposed hybrid approach fares well, both in training and holdout samples, when compared to the traditional back-propagation artificial neural network (ANN) and a genetic algorithm-based artificial neural network (GA-ANN).     

Learning from international business affiliates: developing resource-based learning capacity through networks and knowledge acquisition

Resource-based learning capacity (RBLC) is an organization's specific resources – both human and tangible – that can be organized to enhance learning processes. This study develops and tests a model that examines the relationship between the learning efforts of focal firms from their international business affiliates (IBAs) – organizations located outside the focal firm's domestic market with whom the focal firm has a relationship – and the focal firms' RBLC. This learning process refers to the transfer of knowledge from the IBA to the focal firm. Results indicate that while learning effectiveness positively influences the RBLC of the focal firm, learning efficiency has a negative impact on RBLC. The IBA's home country network centrality and the tie strength between the focal organization and the IBA are found to influence learning effectiveness positively. Tie strength also enhances learning efficiency. Finally, the findings indicate that the IBA's home country network centrality enhances the strength of the ties between the focal organization and its IBA.     

Comparing backpropagation with a genetic algorithm for neural network training

This article shows that the use of a genetic algorithm can provide better results for training a feedforward neural network than the traditional techniques of backpropagation. Using a chaotic time series as an illustration, we directly compare the genetic algorithm and backpropagation for effectiveness, ease-of-use, and efficiency for training neural networks.     

Robust Structural Equations for Designing and Monitoring Strategic International Facility Networks

Using predictive global sensitivity analysis, we develop a structural equations model to abstract from the details of a large‐scale mixed integer program (MIP) to capture essential design trade‐offs of global manufacturing and distribution networks. We provide a conceptual framework that describes a firm's network structure along three dimensions: market focus, plant focus, and network dispersion. Normalized dependent variables are specified that act as proxies for a company's placement into our conceptual network classification via the calculation of just a few key independent variables. We provide robust equation sets for eight cost structure clusters. Many different product types could be classified into one of these groups, which would allow managers to use the equations directly without needing to run the MIP for themselves. Our numerical tests suggest that the formulas representing the network structure drivers—economies of scale, complexity costs, transportation costs, and tariffs—may be sufficient for managers to design their strategic network structures, and perhaps more importantly, to monitor them over time to detect potential need for adjustment.     

A learning vector quantization neural network model for the classification of industrial construction projects

In several key functional areas of contemporary engineering and management science, neural networks have steadily been gaining recognition as robust and reliable tools for classification problems. This paper describes a new application of the learning vector quantization neural network: the classification of the degree of modularization appropriate for the construction of an industrial facility. This neural network uses variables related to plant location, labor issues, organizational issues, plant characteristics, project risks, and environmental issues as inputs to perform the classification. The neural network training and performance evaluation is also discussed.     

Costs of partner search and selection in strategic alliances

This study examines the costs associated with alliance partner search and selection as well as their antecedents. Based on transaction cost economics and the network perspective on inter-organizational relationships, the findings drawing on survey-based data from a sample of 83 firms in the German telecommunications industry reveal that partner search and selection costs are closely connected but differentially affected by task- and company-related factors. When firms must make alliance-specific investments, search and selection costs increase. A firm’s number of current alliances decreases search and selection costs, whereas neither alliance scope nor firm performance significantly affect them. Additional analyses show that alliance-specific investments especially increase search costs but do not affect selection costs, whereas a firm’s performance decreases search costs but does not reduce selection costs.     

Neural networks in production planning and control

The purpose of this paper is to present a selection of the most significant ANN (artificial neural network) applications used to solve PPC (production planning and control) problems. Therefore, the paper itself is a reasoned classification rather than a simple survey. The PPC domain is divided into various areas, and for each area one or more examples of ANNs are discussed, together with the benefits and drawbacks of their application to that particular area. Finally, two case studies developed by the authors are reported and discussed to analyse the main implementation problems to be considered when using ANNs.     

A machine-learning approach for a sintering process using a neural network

This paper presents a machine-learning approach using a multi-layered neural network (NN) with application to a sintering process in an iron- and steel-making plant. Our method induces 'operational rules' that determine operational conditions to obtain products that meet a given quality specification. In our application, an operational condition decides the appropriate ranges of chemical composition and heat input to obtain sinter with desirable properties. Our approach consists of two stages. First, backpropagation (BP) training is performed to obtain a NN which decides whether a given condition is appropriate or not. Secondly, from the trained NN, we extract rules which explain what operational conditions are appropriate. In spite of the effective learning capability, a major drawback of a NN is 'unreadability' of the learned knowledge, or the lack of an explanatory capability, which is crucial in the second stage. We developed a rule extraction algorithm which contributes to overcoming this 'unreadability'. The extracted rules are found to agree well with the knowledge in material science.     

The Effect of Misclassification Costs on Neural Network Classifiers

The potential of neural networks for classification problems has been established by numerous successful applications reported in the literature. One of the major assumptions used in almost all studies is the equal cost consequence of misclassification. With this assumption, minimizing the total number of misclassification errors is the sole objective in developing a neural network classifier. Often this is done simply to ease model development and the selection of classification decision points. However, it is not appropriate for many real situations such as quality assurance, direct marketing, bankruptcy prediction, and medical diagnosis where misclassification costs have unequal consequences for different categories. In this paper, we investigate the issue of unequal misclassification costs in neural network classifiers. Through an application in thyroid disease diagnosis, we find that different cost considerations have significant effects on the classification performance and that appropriate use of cost information can aid in optimal decision making. A cross-validation technique is employed to alleviate the problem of bias in the training set and to examine the robustness of neural network classifiers with regard to sampling variations and cost differences.     

基于ARTⅡ网络的煤炭资源资产分类研究

本文建立了用于煤炭资源资产分类的ARTⅡ神经网络模型,编制了相应的计算机和软件,并将ARTⅡ模型与模糊分类模型和基于BP网络的分类模型进行了对比分析,实例运行结果表明,用ARTⅡ网络进行分类具有分类稳定、结果可靠等特点。     

Data Mining for Network Intrusion Detection: A Comparison of Alternative Methods*

Intrusion detection systems help network administrators prepare for and deal with network security attacks. These systems collect information from a variety of systems and network sources, and analyze them for signs of intrusion and misuse. A variety of techniques have been employed for analysis ranging from traditional statistical methods to new data mining approaches. In this study the performance of three data mining methods in detecting network intrusion is examined. An experimental design (3times2x2) is created to evaluate the impact of three data mining methods, two data representation formats, and two data proportion schemes on the classification accuracy of intrusion detection systems. The results indicate that data mining methods and data proportion have a significant impact on classification accuracy. Within data mining methods, rough sets provide better accuracy, followed by neural networks and inductive learning. Balanced data proportion performs better than unbalanced data proportion. There are no major differences in performance between binary and integer data representation.     

Supply Diversification with Responsive Pricing

We study sourcing and pricing decisions of a firm with correlated suppliers and a price‐dependent demand. With two suppliers, the insight—cost is the order qualifier while reliability is the order winner—derived in the literature for the case of exogenously determined price and independent suppliers, continues to hold when the suppliers' capacities are correlated. Moreover, a firm orders only from one supplier if the effective purchase cost from him, which includes the imputed cost of his unreliability, is lower than the wholesale price charged by his rival. Otherwise, the firm orders from both. Furthermore, the firm's diversification decision does not depend on the correlation between the two suppliers' random capacities. However, its order quantities do depend on the capacity correlation, and, if the firm's objective function is unimodal, the total order quantity decreases as the capacity correlation increases in the sense of the supermodular order. With more than two suppliers, the insight no longer holds. That is, when ordering from two or more suppliers, one is the lowest‐cost supplier and the others are not selected on the basis of their costs. We conclude the paper by developing a solution algorithm for the firm's optimal diversification problem.     

On Sampling Error in Market Share Estimates*

The relative error in the usual estimator of a brand's market share is reformulated in terms of marketing parameters. Such error is shown to be influenced in an important way by market penetration, as well as by variation in brand and product category volume. Of particular interest is the result that the relative error does not depend on the actual share level. Using data from a marketing research firm that supplies share estimates to the health products industry, we find that the relative error may be substantial even when a large sample is available. An upper bound on this relative error is obtained using marketing parameters that can frequently be measured using industry data and a company's internal records, thus reducing the level of judgmental input required in the planning of sample surveys.     

A classification of resources for employee-based value creation and a future research agenda

This conceptual paper aspires to provide a theoretically sound understanding of the value creation process of a specific value network (i.e., firm–employee context). Drawing on service-logic and resource-based frameworks, a classification of four diverse resource types in an organizational context is introduced (i.e., core, augmented, add-on, peripheral resources), based on their exchangeability and their contribution to employees’ value creation and co-creation. This classification enables a better understanding of the nature and the unique features of different firm–employee exchanges in an organizational context, and delineates each type’s distinctive role in employee-based value creation activities. Four propositions derive from this classification; this suggests that not all resource types can be exchanged and that the relative contribution of various firm–employee exchanges to value creation is asymmetrical. A future research agenda is also presented, discussing the potential implications of this classification for contemporary organizations.     

Do Lower Top Marginal Tax Rates Slow the Income Growth of Workers?

Top marginal tax rates are positively correlated with the pretax income growth of the bottom 90 per cent — those who are not subject to the top rates. To explain this correlation, this paper presents and tests a model in which executives can increase firm profitability by (i) increasing the firm's level of technology and (ii) decreasing labor costs. In the model, higher marginal tax rates may reduce pretax inequality by increasing the average income growth of workers. This hypothesis is tested by examining the effect of top marginal tax rates on (unobserved) relative bargaining power between labor and firms and, therefore, on the income growth of workers in the USA. Bargaining power, in both the theoretical and the empirical models, is proxied by private‐sector unionization and use of offshore labor resulting in higher imports.