Kung-Jeng Wang

Gene selection for cancer identification: a decision tree model empowered by particle swarm optimization algorithm

BackgroundIn the application of microarray data, how to select a small number of informative genes from thousands of genes that may contribute to the occurrence of cancers is an important issue. Many researchers use various computational intelligence methods to analyzed gene expression data.ResultsTo achieve efficient gene selection from thousands of candidate genes that can contribute in identifying cancers, this study aims at developing a novel method utilizing particle swarm optimization combined with a decision tree as the classifier. This study also compares the performance of our proposed method with other well-known benchmark classification methods (support vector machine, self-organizing map, back propagation neural network, C4.5 decision tree, Naive Bayes, CART decision tree, and artificial immune recognition system) and conducts experiments on 11 gene expression cancer datasets.ConclusionBased on statistical analysis, our proposed method outperforms other popular classifiers for all test datasets, and is compatible to SVM for certain specific datasets. Further, the housekeeping genes with various expression patterns and tissue-specific genes are identified. These genes provide a high discrimination power on cancer classification.

Location and allocation decisions in a two-echelon supply chain with stochastic demand - A genetic-algorithm based solution

Research highlights? This study pivots on a formal representation of system configuration design and operations optimization for a two-echelon supply chain. ? The proposed solution algorithm finds near optimal solution while consuming less computational time for large-size problems as compared to an optimization-based tool. ? This study investigates the industrial-cluster effect in a two-echelon supply chain. This study addresses a facility location and task allocation problem of a two-echelon supply chain against stochastic demand. Decisions include locating a number of factories among a finite set of potential sites and allocating task assignment between factories and marketplaces to maximize profit. The study represents the addressed location-allocation problem by bi-level stochastic programming and develops a genetic algorithm with efficient greedy heuristics to solve the problem. The contribution of the study pivots on a formal representation of system configuration design and operations optimization for a two-echelon supply chain. The proposed solution algorithm can find near optimal solution while consuming less computational time for large-size problems as compared to an optimization-based tool. In addition, this study investigates the industrial-cluster effect in a two-echelon supply chain by using the proposed algorithm. Experiments reveal that the proposed algorithm can efficiently yield nearly optimal solutions against stochastic demands.

Modelling and resolving the joint problem of capacity expansion and allocation with multiple resources and a limited budget in the semiconductor testing industry

In the semiconductor testing process, many resources such as testers, handlers, loadboards and toolings are required to be ready simultaneously so that testing tasks can be conducted. A limited budget under depressed economy enhances the need for exploring better solutions of testing capacity expansion and allocation. However, to maximize profit as planning with multiple resources is very challenging. This study focused on the issues pertaining to the decisions of (1) the type and number of testers that should be invested to deal with forthcoming orders at a semiconductor testing facility under a constrained budget and (2) the allocation of tester capacity for the orders so as to maximize company profit with limited, multiple resources. Owing to the high computational complexity of the problem, the study developed a genetic algorithm to resolve the two issues simultaneously. A mathematical model was developed to formalize the problem and serve as a benchmark for comparison with the proposed algorithm that attacked the same problem more efficiently. Taguchi experimental design was employed to find the most appropriate parameters for the proposed genetic algorithm under a variety of budget set-up. Experimental results indicated that the proposed algorithm was robust enough to budget plans, and its performance approximated closely with that of the mathematical model.

A resource portfolio planning model using sampling-based stochastic programming and genetic algorithm

Resource portfolio planning optimization is crucial to high-tech manufacturing industries. One of the most important characteristics of such a problem is intensive investment and risk in demands. In this study, a nonlinear stochastic optimization model is developed to maximize the expected profit under demand uncertainty. For solution efficiency, a stochastic programming-based genetic algorithm (SPGA) is proposed to determine a profitable capacity planning and task allocation plan. The algorithm improves a conventional two-stage stochastic programming by integrating a genetic algorithm into a stochastic sampling procedure to solve this large-scale nonlinear stochastic optimization on a real-time basis. Finally, the tradeoff between profits and risks is evaluated under different settings of algorithmic and hedging parameters. Experimental results have shown that the proposed algorithm can solve the problem efficiently.

A resource portfolio model for equipment investment and allocation of semiconductor testing industry

Profitable but risky semiconductor testing market has led companies in the industry to carefully seek to maximize their profits by developing a proper resource portfolio plan for simultaneously deploying resources and selecting the most profitable orders. Various important factors, such as resource investment alternatives, trade-offs between the price and speed of equipment and capital time value, further increase the complexity of the simultaneous resource portfolio problem. This study develops a simultaneous resource portfolio decision model as a non-linear integer programming, and proposes a genetic algorithm to solve it efficiently. The proposed method is employed in the context of semiconductor testing industry to support decisions regarding equipment investment alternatives (including new equipment procurement, rent and transfer by outsourcing, and phasing outing) for simultaneous resources (such as testers and handlers) and task allocation. Experiments have showed that our approach, in contrast to an optimal solution tool, obtains a near-optimal solution in a relatively short computing time.

Cooperative capacity planning and resource allocation by mutual outsourcing using ant algorithm in a decentralized supply chain

Lumpy demand forces capacity planners to maximize the profit of individual factories as well as simultaneously take advantage of outsourcing from its supply chain and even competitors. This study examines a business model of capacity planning and resource allocation in which consists of two profit-centered factories. We propose an ant algorithm for solving a set of non-linear mixed integer programming models of the addressed problem with different economic objectives and constraints of negotiating parties. An individual factory applies a specific resource planning policy to improve its objective while borrowing resource capacity from its peer factory or lending extra capacity of resources to the other. The proposed method allows a mutually acceptable capacity plan of resources for a set of customer tasks to be allocated by two negotiating parties, each with private information regarding company objectives, cost and price. Experiment results reveal that near optimal solutions for both of isolated (a single factory) and negotiation-based (between the two factories) environments are obtained.

Performance Analysis of Auction-Based Distributed Shop-Floor Control Schemes from the Perspective of the Communication System

A growing level of interest in academia and industry centers on the paradigm of distributed shop-floor control in which task and resource allocation in the manufacturing system is accomplished in a distributed manner through message passing and auction-based decision making among autonomous entities. Due to the prominent role played by the communication system in this paradigm, it is important to consider the requirements and performance characteristics of the communication system during the design and evaluation of distributed shop-floor control schemes. In this paper, we propose a two-phase methodology for analyzing auction-based shop-floor control schemes from the perspective of the communication system. In the first phase, the control scheme is modeled as a closed queueing network and performance measures related to the auctioning process and the communication system are obtained rapidly using asymptotic bounding analysis and mean value analysis. Control schemes identified as attractive in the first phase are then evaluated in greater detail during the second phase, using a discrete event simulation model. We illustrate this methodology using two-class and four-class control schemes and discuss insights learned about the impact of various control-scheme-related factors on the performance of the auctioning process and the communication system.

A hybrid knowledge discovery model using decision tree and neural network for selecting dispatching rules of a semiconductor final testing factory

One of the most challenging production decisions in the semiconductor testing industry is to select the most appropriate dispatching rule which can be employed on the shop floor to achieve high manufacturing performance against a changing environment. Job dispatching in the semiconductor final testing industry is severely constrained by many resources conflicts and has to fulfil a changing performance required by customers and plant managers. In this study we have developed a hybrid knowledge discovery model, using a combination of a decision tree and a back-propagation neural network, to determine an appropriate dispatching rule using production data with noise information, and to predict its performance. We built an object-oriented simulation model to mimic shop floor activities of a semiconductor testing plant and collected system status and resultant performances of several typical dispatching rules, earliest-due-date (EDD) rule, first-come-first-served rule, and a practical dispatching heuristic taking set-up reduction into consideration. Performances such as work-in-process, set-up overhead, completion time, and tardiness are examined. Experiments have shown that the proposed decision tree found the most suitable dispatching rule given a specific performance measure and system status, and the back propagation neural network then predicted precisely the performance of the selected rule.

Applying particle swarm optimization-based decision tree classifier for cancer classification on gene expression data

This study proposes a new method for gene selection utilizing PSO combined with a decision tree.Experiment on 11 gene expression cancer datasets from Taiwan Cancer Registry and GEMS is done.We investigate a variety of cancers: bladder, blood, bone marrow, brain, breast, colon, kidney, liver, and lung.The proposed method outperforms SVM, SOM, ANN, and C4.5 decision tree by two-way ANOVA analysis. BackgroundThe application of microarray data for cancer classification is important. Researchers have tried to analyze gene expression data using various computational intelligence methods. PurposeWe propose a novel method for gene selection utilizing particle swarm optimization combined with a decision tree as the classifier to select a small number of informative genes from the thousands of genes in the data that can contribute in identifying cancers. ConclusionStatistical analysis reveals that our proposed method outperforms other popular classifiers, i.e., support vector machine, self-organizing map, back propagation neural network, and C4.5 decision tree, by conducting experiments on 11 gene expression cancer datasets.

Coordination-based inventory management for deteriorating items in a two-echelon supply chain with profit sharing

This study develops a model for inventory management consisting of a two-echelon supply chain (SC) with profit sharing and deteriorating items. The retailer and the supplier act as the leader and follower, in which the supplier faces a huge setup cost and economic order quantity ordering strategy. The market demand is affected by the sale price of the product, and the inventory has a deterioration rate following a Weibull distribution. The retailer executes three profit-sharing mechanisms to motivate the supplier to participate in SC optimisation and to extend the life cycle of the product. A search algorithm is developed to determine the solutions as using the profit-sharing mechanisms. The outcomes from numerical experiments demonstrate the profitability of the proposed model.