International Journal of Bio-Inspired Computation (31 papers in press)
A novel whale optimization algorithm with filtering disturbance and non-linear step
by Jinkun Luo, Fazhi He, Haoran Li, Xian-Tao Zeng, Yaqian Liang
Abstract: As a recent addition of population-based meta-heuristic algorithms, whale optimization algorithm (WOA) has attracted a lot of attention recently. However, WOA still has room for improvement in the accuracy and reliability of solution. In this study, a novel WOA with filtering disturbance and non-linear step (FDNS-WOA) is proposed. Firstly, to enhance the population diversity and the global search ability, we design a weighted Cauchy mutation equation to perturb solution space. Secondly, the excessive disturbance may cause the imbalance between exploration and exploitation, and the quality of solutions will be decreased. Thus, before the disturbance, we propose a filtering mechanism to dynamically select different individuals for disturbance at different evolutionary stages. Thirdly, a weighted local walk mechanism is designed to improve the local exploitation capability. Different from original linear step with large fluctuation, we propose a non-linear step to reduce blind spots in the local exploitation process. Finally, the proposed FDNS-WOA is tested on benchmark functions and applied in a real world problem. The experimental results show that the proposed FDNS-WOA not only outperforms other recent meta-heuristic algorithms in term of accuracy and reliability for most benchmark functions, but also achieves satisfactory results in real world problem.
Keywords: whale optimization algorithm; Cauchy mutation; economic load dispatch; meta-heuristic algorithm; population-based algorithm; filtering disturbance; non-linear step.
An adaptive multi-objective particle swarm optimization algorithm based on fitness distance to streamline repository
by Suyu Wang, Dengcheng Ma, Ze Ren, Yuanyuan Qu, Miao Wu
Abstract: In recent years, multi-objective particle swarm optimization algorithm (MOPSO) has been paid more attention. One of its indispensable structures is the maintenance and update mechanism of the repository. The existing mechanisms are relatively simple, and most of them are based on the crowding distance sorting strategy, and not conducive to the distribution and accuracy of the algorithms. The paper innovated this mechanism and proposed an adaptive multi-objective particle swarm optimization algorithm to streamline repository based on fitness distance (FDMOPSO). Both the concept of fitness distance and the corresponding improve methods of mutation mechanism and adaptive mechanism were proposed. The algorithm itself was tested using benchmarks. The results shown that the proposed application of fitness distance had a better improvement on the convergence and distribution. Compared with other algorithms, the FDMOPSO algorithm had the best overall performance.
Keywords: MOPSO; fitness distance; streamline repository; adaptive; multi-objective optimization.
Artificial Leaf-vein network optimization algorithm for urban transportation network design
by Baozhen Yao, Chao Chen, Wenxuan Shan, Bin Yu
Abstract: The structure of transportation networks determines the traffic flow characteristic of urban roads. In a case where demand and supply are fixed and given, the unevenly distributed traffic flow may lead to traffic congestion. Therefore, designing a rational urban transportation network (i.e., the urban transportation network design problem, UTNDP) has become one of the hottest topics in the transportation field. Due to the property of multi-objects, multi-constraints, and non-convexity, UTNDP is recognized as NP-Hard. Thus, many heuristic algorithms, based on biological laws or swarm intelligence, have been proposed in recent years. However, rare researches paid attention to the perfect topology of the leaf-vein network in nature. In this paper, a definitely new heuristic algorithm is proposed based on the leaf-vein network. By analyzing the growth process of leaf veins, the underlying relationship between urban transportation network and leaf-vein network is first investigated. An evolutionary mechanism of the algorithm or called the artificial leaf-vein generation rule, is then built by simulating the natural selection of biological evolution and genetic transmission. Given the differences between urban transportation networks and leaf-vein networks, a transformation method between the two networks is also designed. Finally, a benchmark instance, the Sioux Falls Network, is set up to demonstrate the performance of the proposed algorithm. The results show that, compared with conventional heuristic algorithms (e.g., GA, SA, and ACO), the proposed algorithm has superior performance and the high application potential in the reduction of the total travel time and length of the entire transportation network. The algorithm may help designers and planners optimize the urban transportation network and alleviate the negative effect of congestion in booming cities.
Keywords: Urban transportation network design problem; artificial leaf-vein network; heuristic algorithm; auxin transport.
Hybrid Moth-Flame optimization algorithm with Differential Evolution for visual object tracking
by KUPPIREDDY NARSIMHA REDDY, Bojja Polaiah
Abstract: The improvement of the metaheuristic algorithms is one of the interesting topics to investigators in current years for resolving the optimization problems and constrained engineering problems. One of the population-based search method i.e., Moth-Flame Optimization algorithm (MFO) is eminent by easy execution, low amount of limits and its high speed. On the other hand, the MFO process has shortcomings for instance verdict the local minimum as a substitute of the global minimum and weakness in global pursuit proficiency. In this paper, to resolve these shortages, the MFO algorithm is integrated with Differential Evolution (DE) and proposed a new hybrid method called MFO-DE. The exploration ability of the MFO algorithm is improved and as well, existence trapped in the local minimum is prohibited by via a mixture of the MFO and DE in the MFO-DE algorithm. The proposed algorithm was tested on the set of best known unimodal and multimodal benchmark functions in various dimensions. The obtained results from basic and non-parametric statistical tests confirmed that this hybrid method dominates in terms of convergence and success rate. Furthermore, MFO-DE is applied to visual object tracking as a real-life application. All experimental outcomes, illustrations, and comparative investigation found the MFO-DE algorithm can vigorously track a random target object in many stimulating circumstances than the other trackers successfully.
Keywords: Population-based algorithm; Meta-heuristic; Differential Evolution (DE); Global optimization; Visual object tracking.
A dynamic firefly algorithm based on two-way guidance and dimensional mutation
by Wang Jing, Yanfeng Ji, Hui Chen, William Wei Song
Abstract: As a stochastic optimizer, the firefly algorithm (FA) has been successfully and widely used in the solutions to various optimization problems. Recent related research shows that the standard FA does not sufficiently balance between exploration and exploitation. Especially in high-dimensional problems, it is easy for the standard FA to fall into the local optimum and lead to premature convergence. To overcome the problems as mentioned above, DMTgFA uses three strategies: dynamic step length setting strategy (DS), non-elite two-way guidance model (TG) and elites dimensional mutation strategy (DM). The dynamic step length setting strategy makes the algorithm convergence speed faster. The non-elite two-way guidance model and the elite dimensional mutation strategy cooperate to solve the balance problem between global search and local search. Experimental results show that DMTgFA has stronger optimization ability and faster convergence speed than other state-of-the-art FA variants.
Keywords: firefly algorithm; single-objective optimization; non-elite two-way guidance model; elite dimensional mutation strategy.
A Many-objective particle swarm optimization algorithm based on convergence assistant strategy
by Wusi Yang, Li CHEN, Yanyan LI, Abid Fazeel
Abstract: The multi-objective particle swarm optimization algorithm based on Pareto dominance also has specific dilemmas when dealing with many-objective optimization problems. For example, how to make the algorithm more effectively approach the true Pareto front, and maintain the diversity of solutions. This paper proposes a convergence assistance framework that couples different convergence operators separately to handle many-objective optimization problems. To maintain the convergence and diversity of populations in the environmental selection, random sampling was performed on the population obtained by the shift density estimation and vector angle. The proposed algorithm is compared with several advanced many-objective optimization algorithms on test suites DTLZ and MaF with 4, 6, 8, 10, and 15 objectives. The experimental results show that the proposed algorithm has better convergence and diversity, outperforms most of the comparison algorithms, and verifies the robustness of the algorithm framework.
Keywords: many-objective optimization; particle swarm optimization; self-controlling dominance area of solutions; convergence assistant strategy.
An Adaptive Approach for Cluster-Based Intrusion Detection in VANET
by Muthumeenakshi R., Vanitha Katharine A.
Abstract: Vehicular Ad Hoc Network (VANET) is an external communication system for vehicles assisting intelligent transport. However, the identification of the intruders, who misuse the genuine information of the vehicles are the major concern. Hence, this paper introduces the Adaptive Elephant Fuzzy System (AEFS) algorithm for detecting the intrusion in the environment. The clustering of the vehicles in the environment is done using Sparse Fuzzy C-Means (Sparse FCM) Clustering algorithm, which facilitate the secure communication. The cluster head is provided with the Adaptive Elephant Fuzzy System that detects the intruder in VANET. The proposed algorithm is capable of detecting the intruders, which enables the security for communication. The proposed AEFS algorithm obtained the maximal detection rate of 95.6829, minimal service response of 1.2059, minimal transmission overhead of 354.4174, and minimal verification delay in the absence of the attacks of 1.4593. rnrn
Keywords: Vehicular Ad Hoc Network (VANET); Intrusion detection; Fuzzy; Optimization; Security.
A novel binary multi-swarms fruit fly optimization algorithm for the 0-1 multidimensional knapsack problem
by Xin Du, Jiawei Zhou, Youcong Ni, Wentao Liu, Ruliang Xiao, Xiuli Wu
Abstract: To improve solution quality and accelerate convergence speed of traditional fruit fly optimization algorithm in solving MKP, a novel binary multi-swarm fruit fly optimization algorithm (bMFOA) is proposed. It comprises four novelties. Firstly, an item frequency tree (IFT) is constructed based on the idea of frequency pattern mining, and a new search strategy is proposed to obtain heuristic information. Secondly, two new heuristic operators of "ADD" and "DROP" are designed according to the obtained heuristic knowledge. Thirdly, a multi-swarm cooperation strategy is presented to strengthen the exploitation capability. To prevent algorithm falling into local optimum prematurely, a swarm location escape strategy is put forward. To verify the efficiency of bMFOA, it is compared with some existing meta-heuristic methods by solving 58 MKPs from ORLIB. The experimental results show that the bMFOA performs better than existing meta-heuristic methods.
Keywords: Fruit fly optimization;Multidimensional knapsack problem; Binary optimization.
A novel rough semi-supervised k-means algorithm for text clustering
by Lei-yu Tang, Zhen-hao Wang, Shu-dong Wang, Jian-cong Fan, Guo-wei Yue
Abstract: In view of the fact that many attribute values of high-dimensional sparse text data are zero, we combine the approximation set of rough set theory with the semi-supervised k-means algorithm to propose a rough set-based semi-supervised k-means (RSKmeans) algorithm. Firstly, the proportion of non-zero values is calculated by a few labeled data samples, and a small number of important attributes in each cluster are selected to calculate the clustering centers. Secondly, the approximation set is used to calculate the information gain of each attribute. Thirdly, the different attribute values are partitioned into the corresponding approximate sets according to the information gain. Then the new attributes are increased and the above process are continued to update the clustering centers. The experimental results on text data show that RSKmeans algorithm can find those important attributes, filter the invalid information, and improve the performances significantly.
Keywords: rough set; approximation set; k-means algorithm; semi-supervised clustering; high dimensional sparse data.
Capacity estimation method of lithium-ion batteries based on deep convolution neural network
by Renwang Song, Lei Yang, Linying Chen, Zengshou Dong
Abstract: In order to ensure the safe and reliable operation of lithium-ion battery packs, the battery management system needs to be able to monitor the health states of each single battery in the packs. In this paper, a deep convolution neural network is used to estimate capacity of batteries. First, the main structure of the deep convolution neural network are introduced, It includes five-layer convolutional structures and three-layer fully connected structures. Due to the local connectivity and weight sharing of deep convolution neural network, the model can accurately estimate the capacity according to the measured values in the charging and discharging process. Then, the model is used to estimate the capacity of four cells of NASA. The average root mean square error is 0.033Ah and the prediction accuracy is more than 95%. It is proved that the proposed capacity prediction model can achieve high accuracy and robustness in application cases.
Keywords: lithium-ion batteries; state of health; capacity estimation; deep learning.
Recognition of crop leaf diseases based on multi-feature fusion and evolutionary algorithm optimization
by Lixia Zhang, Kangshun Li, Yu Qi
Abstract: Crop leaf disease identification refers to automatically recognize crop leaf pictures suffering from disease, to determine the type of diseases, which is important for agricultural production. Much progress has been made in this field, but there are still many challenges. For example, there are not enough ideal schemes for either disease spot area segmentation or feature representation and matching. In order to meet these challenges, a new crop leaf disease recognition method was proposed in this paper. First, disease spot segmentation method combined ultra-green feature and threshold segmentation was presented. Then, feature representation scheme with multiple features was proposed, which combined color, texture, and shape features. Finally, evolutionary algorithm was used to optimize similarity function for feature matching. Experimental results show that the scheme proposed in this paper can effectively improve recognition accuracy and has a certain practical value.
Keywords: Crop leaf disease recognition; Evolutionary algorithm; Disease spot area segmentation; Feature representation; Feature matching.
Coke price prediction approach based on dense GRU and opposition-based learning salp swarm algorithm
by Xuhui Zhu, Pingfan Xia, Qizhi He, Zhiwei Ni, Liping Ni
Abstract: Coke price prediction is critical for smart coking plants to make sensible production plan. The prediction of coke price fluctuations is a time-series problem, and gated recurrent unit (GRU) performs well on dealing with it. Meanwhile, densely connected GRU can improve the information flow of time-series data, but its key parameters are sensitive. Therefore, a novel coke price prediction method, named DGOLSCPP, is proposed using dense GRU (DGRU) and opposition-based learning salp swarm algorithm (OLSSA). Firstly, a model with two layer stacked dense GRU is constructed for capturing deeper features. Secondly, OLSSA is proposed by introducing opposition-based learning, following and stochastic walk operation for enhancing searching ability. Finally, OLSSA is employed to adjust the key parameters of DGRU for winning the accurate predictive results. Experimental results on two real-world coke price datasets from a certain smart coking plant suggest DGOLSCPP outperforms other competitive methods.
Keywords: Salp swarm algorithm; GRU; Dense network; Coke price prediction.
Adaptive Surrogate-based Swarm Intelligence Algorithm and Its Application in Wastewater Treatment Processes
by Jing Jie, Rui Dai, Hui Zheng, Miao Zhang, Lu Lu
Abstract: To solve the computationally expensive optimization problems effectively, a surrogate-based swarm intelligence algorithm is presented. A global surrogate model and a local one are respectively built to approximately evaluate the fitness values of the individuals instead of the accurate function during the evolution process. Following that, the control strategy of swarm evolution and the surrogate management are designed in detail to improve the efficiency of the algorithm. The performance of the proposed algorithm is observed based on comparative experiments with notable benchmark problems and computationally expensive wastewater treatment processes (WWTP). The experimental results prove that the proposed algorithm keeps the trade-off between not only the global and local search, but also the evaluation cost and convergence, which is suitable for computationally expensive optimization problems.
Keywords: Swarm Intelligence; Surrogate Model; Particle Swarm Optimization; Gaussian Process; Wastewater Treatment Processes (WWTP).
Adversarial Transformation Network with Adaptive Perturbations for Generating Adversarial Examples
by Guoyin Zhang, Qingan Da, Sizhao Li, Jianguo Sun, Wenshan Wang, Qing Hu, Jiashuai Lu
Abstract: Deep neural networks are susceptible to adversarial examples which can mislead or even manipulate the predictive behavior of deep neural networks. This raises concerns about the safety of deep learning. In this paper, to ensure rapid generation of adversarial examples, we propose an adversarial transformation network with adaptive perturbations by using the framework of a generative adversarial network. For the adversarial training phase, the direction of the adversarial perturbation is adaptively adjusted to generate more adversarial examples with transferability. Besides, the perceptual constraint based on game theory, the pixel-level constraint based on mixed norms, and the target constraint based on the C&W method are introduced to guide the optimization of the generator. Experiments conducted on MNIST, CIFAR-10, and ImageNet show the proposed algorithm can generate adversarial examples with stronger attack abilities in a shorter time. And the proposed algorithm can generate more transferable adversarial examples when attacking models with similar structures.
Keywords: Adversarial examples; Adaptive perturbations; Adversarial Transformation network; Transferability; Mixed norms constraint.
Optimized Coordinated Control of Hybrid ac/dc Microgrids along PV-wind-battery: A Hybrid based Model
by Shobana Shobana, Gnanavel B.K.
Abstract: This paper aims to present a novel harmonized control approach for a MG with hybrid ac/dc loads and energy resources. Initially, a coordinated control scheme of distributed converters was presented, in which a MPPVC technique is used for dc/ac interlinking converter that offers high quality voltages and this would guarantee the optimal transfer of power amongst ac and dc sub grids. In fact, this paper aims to make the enhancement in the control Fly and structure of Energy storage system (ESS), where the optimized PI controller is insisted. Particularly, the gain of PI controller is tuned optimally by a new hybrid scheme named as Index based Rider combined Model (IF-RM) that incorporates both the concepts of Rider Optimization Algorithm (ROA) and FireFly Algorithm (FF). At last, the betterment of developed model is validated over existing models. \r\n\r\n
Keywords: PI controller; Micro Grid; Energy storage; IF-RM model; Steady state measures. \r\n\r\n.
Spatial-Temporal Attention Based Seq2seq Framework for Short-term Travel Time Prediction
by Ningqing Zhang, Fei Wang, Xiong Chen, Tong Zhao, Qi Kang
Abstract: The short-term travel time prediction is an important research direction of the intelligent transportation system. However, due to the complex temporal and spatial correlation of the traffic data, the more accurate and efficient prediction of the short-term travel time is hard to achieve. In order to conquer the problems and accelerate the construction of smart cities, this study first proposes the seq2seq model based on LSTM network with strong time series processing capability and can avoid manual feature extraction. Then, a seq2seq model based on multi-level attention mechanism is proposed. Finally, the method proposed above is evaluated and compared with the traditional prediction methods through experimental simulations based on the real traffic dataset. The results verify the accuracy and effectiveness of the proposed models, demonstrate that the utilisation of multi-level attention mechanism can effectively integrate temporal and spatial features of the dataset, improve the efficiency of the original seq2seq model.
Keywords: attention mechanism; LSTM; seq2seq framework; short-term travel time prediction; intelligent transportation system; ITS.
An optional splitting extraction based Gain-AUPRC balanced strategy in federated XGBoost for mitigating imbalanced credit card fraud detection
by Jiao Tian, Pei-Wei Tsai, Feiran Wang, Kai Zhang, Hongwang Xiao, Jinjun Chen
Abstract: Credit card defaults cost the economy tens of billions of dollars every year. However, financial institutions rarely collaborate to build more comprehensive models due to legal regulations and competition. Federated XGBoost is an emerging paradigm that enables several companies to build a classification model cooperatively without transferring local data to others. The conventional Federated XGBoost suffers from the inverse inference according to splitting nodes selection and the class imbalance problem severely. Utilizing the characteristic of splitting points selection, we propose an Optional Splitting Extraction Model to reduce the leakage risk of raw data statistics. Moreover, an adjusted AUPRC (the area under the precision-recall curve) is introduced into the gain function to alleviate the class imbalance problem. Our experimental results show Recall and AUPRC increased by 7-10% and 4-8%, respectively, without sacrificing other estimations compared to the existing state-of-the-art. Furthermore, communication iterations also decreased significantly in our proposed method.
Keywords: Federated Learning; XGBoost; Gradient Boosting Decision Tree; Optional Splitting Extraction; Gain-AUPRC Balanced Strategy; Credit Card Fraud Detection.
Data-driven artificial bee colony algorithm based on radial basis function neural network
by Tao Zeng, Hui Wang, Wenjun Wang, Tingyu Ye, Luqi Zhang, Jia Zhao
Abstract: Search strategies play an essential role in artificial bee colony (ABC) algorithm. Different optimisation problems and search stages may need different search strategies. However, it is difficult to choose an appropriate search strategy. To tackle this issue, this paper proposes a data-driven ABC algorithm based on radial basis function neural network (called RBF-ABC). Firstly, a strategy pool with three distinct search strategies is built. The radial basis function (RBF) network is applied to evaluate offspring generated by the search strategies. The search strategy with the best evaluation value is used to guide the search. Dimension perturbation is employed to update multiple dimensions simultaneously, and it improves the convergence speed and the accuracy of the surrogate model. A set of 22 classical benchmark problems with 30 and 100 dimensions are utilised to verify the performance of RBF-ABC. Results show RBF-ABC can effectively save computational evaluations and outperform six other ABC algorithms.
Keywords: data-driven evolutionary optimisation; artificial bee colony; ABC; surrogate model; multiple search strategies; RBF network.
Bio-inspired Algorithms for Cybersecurity
by Kwok Tai Chui, Ryan Wen Liu, Mingbo Zhao, Xinyu Zhang
Abstract: It is witnessed that the popularity of the research in cybersecurity using bio-inspired algorithms (a key subset of natural algorithms) is ever-growing. As emergent research area, researchers have devoted efforts in applying and comparing various bio-inspired algorithms to the cybersecurity applications. It is in need to have a systematic review on the bio-inspired algorithms for cybersecurity to fill the gap of missing research study in this topic. The research contributions of this review article are four-fold. It first highlights the foundation of the baseline and latest development of 12 popular bio-inspired algorithms in three categories namely ecology-based, evolutionary-based and swarm intelligence-based algorithms. A systematic review is conducted to synthesise and compare the research methodologies, results and limitations. In-depth discussion will be made on the shortlisted and highly cited articles. The tips to select appropriate algorithm or the combination of multiple algorithms have been reported, along with the pros and cons on the design and formulations. Future research directions will be presented to meet the trends and unexplored research.
Keywords: bio-inspired algorithms; cybersecurity; ecology-based algorithm; machine learning; evolutionary algorithms; multi-objective optimisation; swarm intelligence; trade-off solution.
Diabetic Macular Edema classification using Gradient Adaptive Thresholding integrated Active Contour and Ant Lion Spider Monkey Optimization-based Generative Adversarial network
by Shweta Reddy, Shridevi Soma
Abstract: Diabetic Macular Edema (DME) is an eye disease, which can highly affect the visual activity for the diabetic patients. The imaging tool, Optical Coherence Tomography (OCT) is used for diagnosis by the ophthalmologists for retinal disease identification. A novel Gradient-based Adaptive Thresholding integrated Active Contour Ant Lion Spider Monkey Optimization driven General Adversarial network (G-AT_AC ALSMO-GAN) is introduced for the DME detection. Here, G-AT_AC scheme is applied for layer segmentation process. In addition, texture features, layer specific features, and image level features are mined for effective classification. The DME classification is carried out using GAN classifier, which is trained by developed ALSMO algorithm, which is the integration of the Ant Lion Optimization (ALO) and Spider Monkey Optimization (SMO). During the classification process 12 layers and 13 boundaries are used for the segmentation process. The DME affected region is classified by the GAN classifier and the classified output is normal or affected one. The proposed method obtained the maximal accuracy, specificity and sensitivity of 94.12%, 92.77% and 98% respectively.
Keywords: Macular Edema classification; Generative Adversarial network; Ant Lion Optimization algorithm; Spide Monkey Optimization algorithm; Gradient Active Contour; Optical Coherence Tomography image.
A Modified Artificial Bee Colony Algorithm for Classification Optimization
by Selcuk Aslan, Sibel Arslan
Abstract: The promising capabilities, easily implementable and customisable structures of the meta-heuristic algorithms have increased the researchers attentions to the well-known problems and their new approximations that are suitable to be solved with the meta-heuristics directly. In this study, a new approximation that defines the classification problem by using a set of linear equations was tried to be solved with an artificial bee colony (ABC)-based technique called classifierABC algorithm. The performance of the classifierABC was investigated in detail by using various datasets and assigning different values to the algorithm specific control parameters. The results obtained by the classifierABC algorithm were also compared with the results of the other meta-heuristics including particle swarm optimisation (PSO), differential evaluation (DE), fireworks algorithm (FWA) and different variants of the FWA. Comparative studies showed that the classifierABC solves the new problem approximation more robustly and its solutions determine the classes of instances in sets with high accuracies.
Keywords: meta-heuristics; ABC algorithm; classification optimisation.
RVEA-based multi-objective workflow scheduling in cloud environments
by Fei Xue, Qiuru Hai, Yuelu Gong, Siqing You, Yang Cao, Hengliang Tang
Abstract: Cloud computing is a major heterogeneous distributed system that can obtain the required resources for the different needs of customers through the network. With the advancement of technology, cloud workflow scheduling has become a widely studied area aiming to utilise cloud resources efficiently. In general, the workflow scheduling problem in a cloud environment is parallel, dependent, and complex. So far, there are many algorithms in the field of workflow resource scheduling in the cloud environment. However, most of these algorithms only consider makespan or cost, and research on multiple targets is still relatively scarce. Considering the characteristics of tasks and users, this paper constructs a workflow scheduling model targeting makespan, cost, and load in the cloud environment. To better address the multi-objective cloud workflow scheduling model, a reference vector-guided evolutionary algorithm (RVEA) is used in this paper. The results show that the algorithm can effectively improve the performance of the proposed model and obtain a suitable workflow scheduling policy compared with existing multi-objective evolutionary algorithms.
Keywords: cloud computing; workflow scheduling; multi-objective; evolutionary algorithms.
Deep Recurrent Neural Network-Based Hadoop Framework for COVID Prediction with Applications to Big Data in Cloud Computing
by Jagannadha Rao Db, Vijayakumar Polepally, Nagendra Prabhu S, Parsi Kalpana
Abstract: This paper proposes a Particle Squirrel Search Optimization-based Deep Recurrent Neural Network (PSSO-based DRNN) to predict the coronavirus epidemic (COVID). Here, the cloud-based Hadoop framework is used to perform the prediction process by involving the mapper and reducer phases. Initially, the technical indicators are extracted from the time series data. Then, the Deep Belief Network (DBN) is employed for feature selection from the technical indicators. After that, the COVID prediction is done by the DRNN classifier, trained using the PSSO algorithm. The PSSO is developed by the integration of Particle Swam Optimization (PSO) and Squirrel Search Algorithm (SSA). The PSSO-based DRNN is compared with existing methods and obtained minimal MSE and RMSE of 0.0523, and 0.2287 by considering affected cases. By considering death cases, the proposed method achieved minimal MSE and RMSE of 0.0010, and 0.0323 and measured minimum MSE of 0.0049 and minimum RMSE of 0.0702 for recovered cases.
Keywords: COVID-19; MapReduce; cloud; Deep Belief Network; Deep Recurrent Neural Network.
Alligator Optimisation algorithm for solving unconstrainted optimisation problems
by Weng-Hooi Tan, Junita Mohamad-Saleh
Abstract: Inspired by cooperative hunting skills and movement patterns of alligators in nature, this research paper proposes a novel bio-inspired meta-heuristic algorithm, named alligator optimisation (AgtrO) algorithm. Upon mathematical modelling, AgtrO emphasises two main phases: the hunting phase that mimics fishing, purse seining and catching mechanisms, and the relocating phase that mimics travelling and homing instinct mechanisms. The hunting phase discovers any promising global optimal area, towards tracking the true global optimal solution. Meanwhile, the relocating phase avoids local optima (traps) through local exploration and conducts in-depth investigations through local exploitation. The proposed AgtrO was tested on 23 classical optimisation benchmark functions and ten modern CEC-C06-2019 benchmark functions, in comparison with eight recently proposed state-of-the-art algorithms. Upon evaluation, AgtrO has been proved to outperform other algorithms in terms of global-best achievement, while being very competitive in terms of convergence speed.
Keywords: bio-inspired; metaheuristic; optimisation; classical benchmark; CEC benchmark.
VAR Strategic Planning for Reactive Power using Hybrid Soft Computing Techniques
by SWETHA SHEKARAPPA G, Sheila Mahapatra, SAURAV RAJ
Abstract: This article proposes a meta-heuristic nature inspired hybridisation of oppositional-based marine predators algorithm (OMPA) hybridised with Harris Hawks optimisation (HHO) which is together implemented as OMPA-HHO on standard IEEE 57 bus system. The algorithms in quest space which is hybridised is mutated by incorporating MPA with oppositional-based learning (OBL) technique in addition to obtaining improved evaluation for the dominant solution. The validation of the proposed algorithm has been with compilation on 23 standardised benchmark functions. Reactive power planning is a complex issue for power system researchers and engineers and proposed OMPA-HHO is applied on standard test system to verify its efficacy and simulation results reflect the improved performance. The results validate the scalability, repeatability and sturdiness of the proposed algorithm which can be considered as a superior one in complex optimisation problems.
Keywords: transmission loss; reactive power planning; RPP; Harris Hawks optimisation; HHO; marine predators algorithm; voltage profile enhancement; oppositional-based learning; OBL.
Special Issue on: Bio-inspired Evolutionary Computations and Their Applications
Brushless Direct Current Motor Design Using A Self-adaptive JAYA Optimization Algorithm
by Li Yan, Chuang Zhang, Boyang Qu, Kunjie Yu, Caitong Yue
Abstract: This paper proposes a self-adaptive JAYA (SAJAYA) to maximize the efficiency of the brushless direct current (BLDC) motor by optimizing the design parameters. In SAJAYA, an adaptive weight strategy is introduced into the original JAYA to control the degree of approaching the best solution and avoiding the worst solution at different evolution stages. Moreover, to maintain the diversity and enhance the local search ability, a new hybrid learning strategy is developed. Further, a self-adaptive selection mechanism is designed in order to dispatch the original learning strategy and the new hybrid learning strategy to each individual automatically. Based on this framework, the exploration and exploitation abilities of the population are expected to be balanced. Besides, a chaotic elite search is introduced to further refine the best solution of each generation. Experimental results show that the proposed SAJAYA shows a superior performance in solving the BLDC motor optimization problem compared with other well-established algorithms.
Keywords: Evolutionary computation; JAYA algorithm; Optimization; Brushless DC wheel motor.
Genetic regulatory network-based optimization of master production scheduling and mixed-model sequencing in assembly lines
by Youlong Lv, Jie Zhang, Liling Zuo
Abstract: The integration of master production scheduling and mixed-model sequencing ensures just-in-time production of orders and workload balancing between stations for assembly lines. However, such integrated optimization is complicated because of the high interdependence between these two problems. Based on mathematical model of the integrated optimization problem, a two-level genetic regulatory network is constructed by representing decision variables with gene states and describing multiple objectives and various constraints with gene regulation equations. The solutions are generated through gene expression procedures in which some gene states are activated based on regulation equations, and the optimal one with minimum objective function value is obtained via parameter optimization in regulation equations. The genetic regulatory network-based method is validated under sixteen testing scenarios and applied to the case study of a diesel engine assembly line. The results demonstrate the effectiveness of this method over other ones in realizing integrated optimization.
Keywords: integrated optimization; diesel engine assembly line; master production scheduling; mixed-model sequencing; genetic regulatory network.
A parallel algorithm to solve the multiple traveling salesmen problem based on molecular computing model
by Zhaocai Wang, Tunhua Wu
Abstract: The Multiple Traveling Salesmen Problem (MTSP), which includes m salesmen starting and ending their tours at a same fixed node (m>1), is an extension of Traveling Salesman Problem (TSP) and has more applications and significance in the field of optimal control. As a classic NP-hard static combinatorial optimization problem, its efficient solution has always been the direction of scholars efforts. In this work we propose biocomputing algorithms to solve the MTSP using Adleman-Lipton model. We make use of DNA chains to appropriately represent the nodes, edges and corresponding weights, then efficiently generate all traveling salesmen tours combinations by biochemical reactions. Combining with the nature of the problem, we exclude the infeasible solution strands to get the optimal solutions, and reduce the algorithm computational complexity to O(n^2) level. Meanwhile, the feasibility and practicability of DNA parallel algorithms are verified by theoretical proof and simulation experiments. The proposed algorithms are also helpful to better understand the nature of computing and can be further applied to the study of extended problems.
Keywords: DNA computing; Multiple Traveling Salesmen Problem; Adleman-Lipton model; NP-hard problem.
Techno Economic Analysis Of Novel Multiobjective Quasi-Oppositional Grey Wolf Optimizer
by Gourab Das
Abstract: The optimal DG placement in radial distribution system is an important way for techno-economic analysis. The maximum technical benefits can be extracted by minimizing the distribution power loss as well as bus voltage deviation, whereas the maximum economical benefits can be procured by minimizing the total yearly economic loss which includes installation, operation and maintenance cost. So for the maximum techno-economic benefits, all three objectives should be simultaneously minimized by considering a multiobjective optimization technique. For optimal results, a Pareto optimal concept based a novel multiobjective quasi-oppositional grey wolf optimizer (MQOGWO) algorithm has been proposed. The performance of the proposed algorithm has been tested on IEEE-33 bus radial distribution system. In this analysis, various voltage dependent load models such as constant power, constant current, constant impedance, residential, industrial and commercial load models have been considered at different loading conditions like light load, full load and heavy load. The effects of DG type on the system performance has also been analysed to find the best optimal solution.
Keywords: Distributed Generation; Yearly Economic Loss; Power Loss; Voltage Deviation; Multiobjective Quasi-Oppositional Grey Wolf Optimizer.
Filter-based feature selection: a comparison among binary and continuous cuckoo optimisation algorithms along with multi-objective optimisation algorithms using gain ratio-based entropy
by Ali Muhammad Usman, Umi Kalsom Yusof, Syibrah Naim
Abstract: Feature selection is one of the data mining processes that is used to eliminate both irrelevant and redundant features while maintaining the datasets standard. Filter-based used the intrinsic statistical characteristic to select the high- rank features which make it computationally faster and scalable to high dimensional data. On the other hand, it affects the classification performance due to lack of feature interaction among the selected high-ranked features. In this study, entropy is used as the fitness functions of the cuckoo optimisation algorithm (COA) as well as its binary (BCOA) counterpart to enhance feature interaction. Now, that, FS is considered as multi-objective optimisation problem (MOP). Some of the prominent multi- objective optimisation algorithms, especially, non-dominated sorting genetic algorithm (NSGAIII), multi-objective evolutionary algorithm based on decomposition (MOEA) and evolutionary algorithm of non dominated sorting with radial basis (ENORA). The essence of using entropy is to improve the relationship between and among different features, respectively. Besides, information gain (IG) of the entropy is commonly used, to find the relevance and redundancy of the features. However, IG was reported to be problematic on features with many values, that leads to biased, fragmentation and selection of non- optimal features for the prediction. Thus, gain ration (GR) which is a modification of the IG is employed to correct the issues by considering the intrinsic information of a split. The results achieved showed the superiority of the proposed entropy with GR over the existing entropy with IG in most of the datasets. Besides, the proposed BCOA with the proposed entropy outperformed the standard COA as well as the existing studies of binary particle swarm optimisation algorithm and genetic algorithm. On the other hand, the proposed multi-objective algorithms perform better than the proposed single-objective in the majority of the datasets, with NSGAIII-E being the best among them.
Keywords: Multi-ob jective optimisation; cucko o optimisation algorithm; feature selection; entropy; information gain; gain ration; NSGAIII; ENORA; MOEA/D.
Special Issue on: Intelligent Simulation Optimisation for Complex Systems
Efficient Optimization Methods in a Distributed Memory Storage System Using Data Compression
by Xiaoyang Yu, Songfeng Lu, Tongyang Wang, Xinfang Zhang, Shaohua Wan
Abstract: The storage of data in a memory-centric storage system benefits from great improvements in read and write I/Ornperformance at the price of consuming a large amount of valuable memory resources. As the amount of data grows, a distributedrnsystem must scale accordingly, resulting in an increased number of nodes and greater communication overhead that may impact thernoverall performance. We believe that achieving an appropriate trade-off among computation, storage and network transportation will bernbeneficial for a distributed memory storage system, leading to higher overall efficiency. The challenges of achieving such a trade-off arerndetermining how and when this trade-off should be applied. In this paper, we explore a method to achieve this trade-off by introducingrndata compression technology in a transparent manner. Instead of focusing on specific compressed data structures, we target blockrnlevel compression for a general purpose storage system to incorporate a wide range of existing data analysis frameworks and usagernscenarios, especially with big data. A prototype is implemented and evaluated based on the memory-centric distributed storage systemrnAlluxio to provide transparent compression and decompression during write/read operations. A series of experiments for data withrndifferent types of compression ratio is conducted to determine the proper situation in which this technology can help to improve thernsystem efficiency. The experimental results prove that our approach can achieve huge write/read throughput boosts for highlyrncompressible data and performs better for moderately compressible data when there are remote operations. From the results, wernsuggest that our approach can be applied for data with a medium up to high compression ratio in practical data processing pipelinesrnand provides useful insights for optimizing future memory-based storage systems.
Keywords: System performance optimization; Data compression; Distributed memory storage.