Model Management Techniques Research Articles

Enhancing thyroid disease prediction with improved XGBoost model and bias management techniques

Enhancing thyroid disease prediction with improved XGBoost model and bias management techniques

Configurable three‐way model merging

SummarySoftware development is a collaborative activity that requires teams of software engineers to cooperate and work in parallel on versions of models. However, model management techniques such as model differencing, merging, and versioning have turned out to be difficult challenges, due to the complexity of operations and graph‐like nature of models. Therefore, a well‐developed support for model merging process, as well as conflict management, is highly desired. This paper presents a novel process for model merging, called the Epsilon‐based Three‐way Merging Process (E3MP) process. Model merging is a significant problem where there are different versions of a system model amongst modeler teams. E3MP includes three components implemented into the Epsilon framework. First, modelers can define domain‐specific rules that customize the merging process. Second, E3MP enables an automated method for syntactic and semantic conflict detection amongst different versions of the system model. Third, E3MP puts forward a pattern‐based approach for conflict resolution. We applied two generic benchmarks to assess conflict detection and resolution capabilities of our approach and carried out an initial scalability evaluation for the model merge with large models and large change sets. The results of our experiments revealed that the proposed process allows generating consistent and semantically correct merged models.

Heterogeneous megamodel management using collection operators

Model management techniques help tame the complexity caused by the many models used in large-scale software development; however, these techniques have focused on operators to manipulate individual models rather than entire collections of them. In this work, we begin to address this gap by adapting the widely used map, reduce and filter collection operators for collections of models represented by megamodels. Key parts of this adaptation include the special handling of relationships between models and the use of polymorphism to support heterogeneous model collections. We evaluate the complexity of our operators analytically and demonstrate their applicability on six diverse megamodel management scenarios. We describe our tool support for the approach and evaluate its scalability experimentally as well as its applicability on a practical application from the automotive domain.

An assessment of different model-management techniques in history-matching problems for reservoir modelling

History matching is a computationally expensive inverse problem. The computation costs are dominantly associated with the optimisation step. Fitness approximation (proxy-modelling) approaches are common methods for reducing computational costs where the time-consuming original fitness function is substituted with an undemanding function known as approximation function (proxy). Almost all of the applied fitness approximation methods in history-matching problems use a similar approach called uncontrolled fitness approximation. It has been corroborated that the uncontrolled fitness approximation approach may mislead the optimisation direction to a wrong optimum point. To prevent this error, it is endorsed that the original function should be utilised along with the approximation function during the optimisation process. To make use of the original function efficiently, a model-management (evolution-control) technique should be applied. There are three different techniques: individual-based, population-based, and adaptive. By using each of these techniques, a controlled fitness approximation approach is assembled, which benefits from online learning. In the first two techniques, the number of original function evaluations in each evolution-control cycle is fixed and predefined, which may result in an inefficient model management. In the adaptive technique, the number is altered based on the fidelity of the approximation function. In this study, a specific adaptive technique is designed, based on heuristic fuzzy roles; then, for the first time, the applications of all the three techniques are investigated in history matching. To deliver an assessment between the four approaches (the uncontrolled approach and three controlled approaches), a framework is developed in which ECLIPSE-E100 is coupled with MATLAB; and an artificial neural network, a genetic algorithm—with a customised crossover—and a Latin hypercube sampling strategy are used as the proxy model, optimiser, and experimental design method, respectively. History matching is carried out using each of the four approaches for the PUNQ-S3 reservoir model, while the same amount of computation time was allowed for each of the approaches. The outcomes demonstrate that the uncontrolled approach cannot deliver reliable results in comparison with the controlled approaches, and among the controlled approaches, the developed adaptive technique is more efficient.

Decision making for time-constrained commodity trasportation

“Xenios”is a speci c DSS that was developed to assist the daily VRS activities of Greek transport rms during special events. This system incorporated essential functions of GIS, database systems and model management techniques to support overall routing, scheduling and decision-making processes for Vehicle Routing and Scheduling (VRS) problems encountered during the Athens 2004 Olympic Games (ATHENS 2004). 4500 Z. Manoussaridis, C. Mamaloukas and M. Sagheb-Tehrani During the Athens Olympic Games 2004 in Athens, the Vehicle Routing and Scheduling problem was very intense. Actually, the commodities transportation had to be done within strictly de ned time-periods and under many security and tra¢ c restrictions [2, 8]. Our work aimed at the successful setting up of DSS “Xenios” so that it could cover a wide range of variations of the VRS problem, especially problems with hard time-windows in the route generation [15, 17]. Mathematics Subject Classi cation: 68U35 Keywords: Decision Support Systems, VRS problem dimensions, commodity transportation 1 Standing VRS problems and decision-making Trying to develop a speci c Decision Support System (DSS) for VRS activities of transport rms, during ATHENS 2004, came up with the need to survey the various problem dimensions [14] and create a theoretical framework for the speci c context. This would assist us with the successful setting up of the DSS “Xenios”which covered a wide range of variations of the VRS problem encountered by transport rms during the Olympic Games 2004. The main di¤erence of this system from a previous DSS named “Dromones”[14] was the aspect of hard time-windows in route generation. Therefore, the process of VRS decision-making was considered as a quite rigorous one [17]. The system combined GIS features, database management system and several model management techniques to support routing, scheduling and decision-making processes needed by general transport rms [15]. Dantzing & Ramser [6] were the rst to present the mathematical de nition of the vehicle routing problem. In simple words, the problem focuses on the delivery of certain quantities of commodities to a number of customers who are scattered throughout a geographical area. A certain number of vehicles are available for these deliveries and each vehicle has a given capacity. Our aim is to determine a set of routes that each one would start and nish in the depot so that the overall covered route can be minimized always under the conditions that all orders are being served and the issue of time-windows is taken into account. Datzing & Ramser de ne the following speci cations explicitly [6]: A set of n points Pi (i = 1; :::; n), in which all orders are being delivered from a point P0 (the depot). A quantity of orders q that has to be delivered in Pi (i = 1; :::; n). The capacity of the truck is C (C > qi for i = 1; :::; n) A matrix of routs D = [dij] that de nes the route between each pair (i; j = 0; :::; n). Decision making for time-constrained commodity trasportation 4501 The objective is to nd a set of routes that each would start and end at point P0, in order to minimize the overall covered route from the trucks under the condition that all orders are delivered and the capacity of all trucks is respected. There are a number of variations as far as the denomination of the present problem is concerned. Consequently, the present problem will be called as a classical vehicle routing problem (CVRP). The grouping of the problem dimensions aimed to the formation of dimension groups with common characteristics. Several came from the relevant bibliography [1], [3], [4], [5], [7] while others revealed from the eld research conducted in various rms [13], [15]. Based on this groundwork, the formulation of the conceptual schema for the necessary database and the DSS rules de nition were accomplished without any severe problems (see g1).

Decision-enabled dynamic process management for networked enterprises

Enterprises in today’s networked economy face numerous information management challenges, both from a process management perspective as well as a decision support perspective. While there have been significant relevant advances in the areas of business process management as well as decision sciences, several open research issues exist. In this paper, we highlight the following key challenges. First, current process modeling and management techniques lack in providing a seamless integration of decision models and tools in existing business processes, which is critical to achieve organizational objectives. Second, given the dynamic nature of business processes in networked enterprises, process management approaches that enable organizations to react to business process changes in an agile manner are required. Third, current state-of-the-art decision model management techniques are not particularly amenable to distributed settings in networked enterprises, which limits the sharing and reuse of models in different contexts, including their utility within managing business processes. In this paper, we present a framework for decision-enabled dynamic process management that addresses these challenges. The framework builds on computational formalisms, including the structured modeling paradigm for representing decision models, and hierarchical task networks from the artificial intelligence (AI) planning area for process modeling. Within the framework, interleaved process planning (modeling), execution and monitoring for dynamic process management throughout the process lifecycle is proposed. A service-oriented architecture combined with advances from the semantic Web field for model management support within business processes is proposed.

Design optimization using support vector regression

Polynomial regression (PR) and kriging are standard meta-model techniques used for approximate optimization (AO). Support vector regression (SVR) is a new meta-model technique with higher accuracy and a lower standard deviation than existing techniques. In this paper, we propose a sequential approximate optimization (SAO) method using SVR. Inherited latin hypercube design (ILHD) is used as the design of experiment (DOE), and the trust region algorithm is used as the model management technique, both adopted to increase efficiency in problem solving. We demonstrate the superior accuracy and efficiency of the proposed method by solving three mathematical problems and two engineering design problems. We also compare the proposed method with other meta-models such as kriging, radial basis function (RBF), and polynomial regression.

A novel multi-objective optimization method based on an approximation model management technique

In this paper, a novel multi-objective optimization method is suggested based on an approximation model management technique. It is a sequential approximation method, in which a multi-objective optimization with approximation models subject to design variable move limits is iterated until convergence. In each iteration step, the approximation models are constructed by the response surface approximations with the samples which are obtained from the design of experiments, and a Pareto optimal set predicted by the approximations is identified through a multi-objective genetic algorithm. According to the prediction of the approximation models, a move limits updating strategy is employed to determine the design variable move limits for the next iteration. At the end of each iteration step, some uniform distributed points chosen from the predictive Pareto optimal frontier are verified by the high fidelity models and the obtained actual Pareto optimal set is stored in an external archive. The high efficiency of the present method is demonstrated by four different test functions and two engineering applications.

Practical aspects of DSS design for commodities transportation during special events

Transport firms that daily pickup or deliver commodities encounter a plethora of problems relating to their Vehicle Routing and Scheduling (VRS) activities. To make decisions about Routing and Scheduling during special events is even more complicated process. During the Athens Olympic Games 2004 in Athens, the Vehicle Routing and Scheduling problem was very keen. Actually, the commodities transportation had to be done within strictly defined time-periods and under many security and traffic restrictions [2], [8]. "Xenios" is a specific DSS that was developed to assist the daily VRS activities of Greek transport firms during special events. This system incorporated essential functions of GIS, database systems and model management techniques to support overall routing, scheduling and decision-making processes for VRS problems encountered during the Athens 2004 Olympic Games (ATHENS 2004).Our work aimed at the successful setting up of DSS "Xenios" so that it could cover a wide range of variations of the VRS problem, especially problems with hard time-windows in the route generation [15], [17].The first step in the development of any specific DSS is the identification of the applying problem dimensions [3], [7], [10]. In order to make easier the development procedure, we tried to classify the most significant VRS problem dimensions that have common characteristics in groups. According to our consideration, this work could accelerate and ease the development of the specific DSS.

Toward a logical/physical theory of spreadsheet modeling

In spite of the increasing sophistication and power of commercial spreadsheet packages, we still lack a formal theory or a methodology to support the construction and maintenance of spreadsheet models. Using a dual logical/physical perspective, we identify four principal components that characterize any spread sheet model: schema, data, editorial, and binding . We present a factoring algorithm for identifying and extracting these components from conventional spreadsheets with minimal user intervention, and a synthesis algorithm that assists users in the construction of executable spreadsheets from reusable model components. This approach opens new possibilities for applying object-oriented and model management techniques to support the construction, sharing, and reuse of spreadsheet models in organizations. Importantly, our approach to model management and the Windows-based prototype that we have developed are designed to coexist with, rather than replace , traditional spreadsheet programs. In other words, the users are not required to learn a new modeling language; instead, their logical models and data sets are extracted from their spreadsheets transparently, as a side-effect of using standard spreadsheet programs.

A model management solution system for multicommodity network flows

There are certain decision problems in the industry for which theoretically reasonable optimization models are not usable in practice, either because the model is cumbersome and difficult, or because the decision information produced is inappropriate and poorly focused on the problem at hand. The class of models known in operations research literature as multicommodity network flows belongs in this category. This paper presents a decision modeling system for multicommodity network flows using model management techniques to generate part-solutions targeted on the decision maker's specific queries. This method avoids the repeated complex solution process of a large model. The decision maker can direct his attention to decision issues, without regard to the technical characteristics of the model. The dynamic character of the decision process is enhanced by direct interaction with the normal business database. The decision modeling system is illustrated here with a problem in the operations of a hypothetical oil company, a typical example of multicommodity network flows.