Please wait a minute...
, No 4

■ Cover page(PDF 3151 KB) ■  Table of Content, April 2022  (PDF 34 KB)

  • AMC-Based Algorithm for Network Reliability Evaluation of a Manufacturing System with Scrapping and Rework
    Yu-Lun Chao, Yi-Kuei Lin, and Cheng-Ta Yeh
    2022, 18(4): 231-240.  doi:10.23940/ijpe.22.04.p1.231240
    Abstract    PDF (392KB)   
    References | Related Articles
    Focusing on the stochastic flow manufacturing system (SFMS), an Absorptive Markov Chain-based algorithm is proposed to calculate the network reliability as the performance index. The SFMS is a hybrid flow shop with multiple production lines, and each workstation in the production line contains multiple identical parallel machines. Because of machine failure, human factors, maintenance, and other factors, machines may have multiple capacities and be regarded as random variables. In addition, in practice, to avoid the waste of raw materials and reduce manufacturing costs, rework and scrapping mechanisms are often used in manufacturing systems. Therefore, whether the SFMS is meeting the demand is an important issue, and network reliability is an ideal quantitative indicator. However, most of the algorithms proposed for SFMS in the past are established for specific scenarios and are not easy to be used in complex manufacturing systems. The Absorption Markov Chain model is used to calculate the capacity required by all workstations, and then the minimum capacity vectors are obtained for calculating network reliability.
    A Survey of the Analysis of Complex Systems based on Complex Network Theory and Deep Learning
    Dan Lu and Shunkun Yang*
    2022, 18(4): 241-250.  doi:10.23940/ijpe.22.04.p2.241250
    Abstract    PDF (339KB)   
    References | Related Articles
    From the perspective of complex network theory, complex systems can be characterized by the interaction of microscopic units through nonlinear effects, yielding macroscopic emergent behavior. In light of the powerful capability of deep learning in feature extraction and model fitting from large amount of datasets, we try to overview the benefits of combining the complex network analysis with deep learning techniques to investigate complex systems. We first explore the existence of complexity in complex systems. In what followed, we first give a brief description of complex network theory. Then, we present an overview of deep learning technology. Subsequently, we focus on the research advances and applications in the analysis of complex systems based on complex network theory and deep learning. The last section is further discussion and prospects for the combination of these two methods. In a nutshell, the development of deep learning combined with complex network theory allows for exploring the complexity in complex systems at a higher level.
    Automatic Categorization of Software with Document Clustering Methods and Voting Mechanism
    Kai-Wen Chen and Chin-Yu Huang
    2022, 18(4): 251-262.  doi:10.23940/ijpe.22.04.p3.251262
    Abstract    PDF (438KB)   
    References | Related Articles
    Manual software categorization depends on considerable understanding of the categorized software to enable software managers to categorize the software based on certain criteria (e.g., functionality). Unfortunately, the rapid growth of software makes manual software categorization almost impossible and expensive. Therefore, automatic software categorization has become necessary. In this study, we utilized three different unsupervised document clustering methods, namely k-means, non-negative matrix factorization (NMF), and spectral clustering, to analyse source code and to implement automatic software categorization. For evaluation, we selected a well-known unsupervised model, LACT, as our comparison candidate. In general, our proposed methods required at most approximately 1/4 of the execution time of LACT, whereas the fastest method was hundreds of times faster than LACT, achieving at most 26% and 100% better performance based on two criteria: the BCubed F1-measure and the Adjusted Rand Index, respectively. Additionally, we also proposed a voting mechanism inspired by N-version programming and ensemble learning. We selected certain states of NMF and spectral clustering as the referees to improve the average performances of k-means. Our results indicated that combining different clustering techniques to achieve better results is feasible.
    Novel Hybrid Decision Model for Electrical Fire Risk Evaluation of High-Rise Buildings based on Asymmetric Proximity
    Lei Su, Fan Yang, Yu Shen, and Zhichun Yang
    2022, 18(4): 263-274.  doi:10.23940/ijpe.22.04.p4.263274
    Abstract    PDF (1048KB)   
    References | Related Articles
    This paper investigates the risk assessment method of electrical fires of high-rise buildings based on hybrid decision models considering asymmetric proximity. Based on the occurrence mechanism of electrical fires in high-rise buildings, a four-level evaluation index system considering the disaster causing body, fire site environment, affected body, and fire driving factors is established based on FP growth mining association rules, and the risk grade is divided further. An improved DEMATEL+ANP index weight assignment method for balancing the interaction relationship between indexes is proposed, and a hybrid decision model for electrical fire risk assessment in high-rise buildings, taking further account of asymmetric proximity and improved evidence cloud theory. Combined with specific examples, the effectiveness of the proposed building electrical fire risk assessment method is verified. This method can better balance each risk evaluation index, fully considering the fuzziness and randomness in the electrical fire risk assessment of high-rise buildings, and improve the accuracy and applicability of electrical fire risk assessment.
    HR Analytics: Employee Attrition Analysis using Random Forest
    Shobhanam Krishna and Sumati Sidharth
    2022, 18(4): 275-281.  doi:10.23940/ijpe.22.04.p5.275281
    Abstract    PDF (334KB)   
    References | Related Articles
    Employee attrition has indeed been viewed as a crucial concern for businesses due to the negative impact it has on workplace motivation and productivity as well as prolonged growth strategies. Organizations are using machine learning(ml) algorithms to predict employee turnover to address the problem. In this paper, an effort has been made to build up a model for predicting employee turnover rates using HR analytics data provided by IBM Analytics. The actual dataset includes 35 features as well as 1470 samples. Random Forest is being used to make accurate predictions. The model built using Random Forest Classifier is transformed by SMOTE mechanism to improve target class imbalance. After SMOTE mechanism metrics of the training model are improved however, validation metrics are improved slightly specially sensitivity has very little impact. This paper also introduces the factors that influence employee attrition within any organization, giving top management a better perspective when attempting to make major decisions regarding the engagement of the majority of workers in the company. The study may be obligated in prospective research to reduce the prediction error margin.
    Efficacy and Security Effectiveness: Key Parameters in Evaluation of Network Security
    S. Guru Prasad, M. K. Badrinarayanan, and V. Ceronmani Sharmila
    2022, 18(4): 282-288.  doi:10.23940/ijpe.22.04.p6.282288
    Abstract    PDF (225KB)   
    References | Related Articles
    Information is the most critical asset of modern-day organisations. Organisations use various information technology tools, products, and solutions to store the information. Because of its importance to organisations, information is protected using various information protection tools, that would include computer or network hardware and software. This article aims to give the organisations and their Chief Information Security officers a template for technical evaluation of the various network security solutions with Security effectiveness as a key element in their decision making.
    Modified Cat Swarm Optimization for Optimal Assembly Sequence Planning Problems
    Chiranjibi Champatiray, Sonali Samal, M. V. A. Raju Bahubalendruni, R. N. Mahapatra, Debasisha Mishra, and B. K. Balabantaray
    2022, 18(4): 289-297.  doi:10.23940/ijpe.22.04.p7.289297
    Abstract    PDF (517KB)   
    References | Related Articles
    Product manufacturing industries are looking for efficient assembly planners to instantly generate an optimal feasible assembly sequence for multiple product variants. The products with a sizeable part count account for colossal search space, and applying feasibility constraints leads to NP-hard problems. The current study seeks to identify and customize an optimization algorithm for the assembly sequence planning (ASP) problem in order to maximize computational efficiency. Nature has the optimal constructal patterns for several phenomena; three nature-inspired algorithms have been identified, namely Cat Swarm Optimization (CSO), Particle Swarm Optimization (PSO), and Ant Colony Optimization (ACO). Further, they are tailored and enhanced for ASP with a non-linear weight factor and t-distribution. It is found that weighted CSO with t-distribution (TWCSO) is efficient in solving ASP problems with a large number of products with less computational time, and the rate of convergence is better compared with that of weighted ACO with t-distribution (TWACO) and weighted PSO with t-distribution (TWPSO).
    Recommendations based on Integrated Matrix Time Decomposition and Clustering Optimization
    D. R. Kumar Raja, G. Hemanth Kumar, Syed Muzamil Basha, and Syed Thouheed Ahmed
    2022, 18(4): 298-306.  doi:10.23940/ijpe.22.04.p8.298306
    Abstract    PDF (299KB)   
    References | Related Articles
    The prompt progression of web data and the number of web visitors creates a latent information overload problem and complicates data mining to select the right items on the web. E-commerce websites and applications manage information overload using several information filtering techniques such as personalized recommendation systems. The recommendation system creates a list of products to helps users. The proposed NOMINATE methodology offers characteristic values ??for definite elements from the LOD source as input for matrix factorization. This helps NOMINATE to retrieve user-specific functions. Subsequently studying the functions, the proposed approach produces an enhanced cluster of consumers by the weight of every user above the desired values ??of the element characteristics. For this, the NOMINATE methodology uses the particle swarm optimization (PSO) algorithm and k-means clustering algorithm. Using clustered results, the NOMINATE approach estimaties the centrality of proximity among other users in the cluster and identifies an experienced user for each cluster. In addition, the proposed recommendation scheme produces a set of characteristic values for each user based on the adept information of the users in the respective clusters.
ISSN 0973-1318