TMPA-2021: Meta-heuristic Techniques and Their Applications

Mohamed Elsayed Ahmed Mohamed, TPU

Meta-heuristic Techniques and Their Applications

TMPA is an annual International Conference on Software Testing, Machine Learning and Complex Process Analysis. The conference will focus on the application of modern methods of data science to the analysis of software quality.

To learn more about Exactpro, visit our website https://exactpro.com/

Follow us on
LinkedIn https://www.linkedin.com/company/exactpro-systems-llc
Twitter https://twitter.com/exactpro

5206c19df417b8876825b5561344c1a0?s=128
Exactpro PRO
November 25, 2021

Transcript

1. Metaheuristic Techniques and Their Applications Presented By Mohamed Abd Elaziz Department of Mathematics-Faculty of Science- Zagazig university
2. Mohamed Abd Elaziz  Contacts  Address: Faculty of Science, Zagazig University, Egypt  E-mail: abd_el_aziz_m@yahoo.com  Research gate: https://www.researchgate.net/profile/Mohamed_Elsayed_Abd_El_Aziz2  Google Scholar: https://scholar.google.com/citations?user=70I7KMIAAAAJ&hl=ar
3. Outline  Introduction for Metaheuristics  Global Optimization  Image segmentation  Medical Applications  Cloud computing  Engineering applications  Economic applications  Conclusion and Future work
4. Introduction  Metaheuristic (MH) technique is defined as a computational method which simulate natural process.  We can categorize them into six major families:  Evolutionary algorithms,  Swarm intelligence methods,  Natural phenomena approaches,  Physics law metaheuristics,  Sport algorithms, and  Human inspiration algorithms (as depicted in Figure 1).
6. Introduction  Generally, the main difference between the MH methods and the traditional methods such as the gradient descent and Newton is that the traditional methods are easy to implement.  Why Meta-Heuristics (MH) is suitable?  Traditional optimization techniques can solve such problems, but they struggle in finding global optima when the problem becomes more challenging. Most of such techniques are exact and required gradient information of the problem.  On the other hand, Meta-Heuristics (MH) benefits from stochastic components and do not need gradient information. Therefore, they are better able to solve challenging problems.  These MH methods have different operators which give them high ability to find the suitable solution to the given problem.  Provide many solutions at each iteration, however traditional methods provide only one solution.
7. Introduction  The superiority of these MH techniques over traditional methods in solving various optimization problems has directed many researchers to present it to solve many real- world optimization problems such as  Engineering design problems, and numerical and structural optimization functions [1], task/workflow scheduling issue [2], image segmentation [3], feature selection [4], and data clustering [5].
8. Swarm Techniques
10. 10 Prey? ?????
11. 11 prey Prey? ?????
12. 12 prey Catch
13. 13 Unknown the position of prey
15. 15 Best solution
17. 17 Best solution
19. 19 Best solution
20. Global Optimization  Global Optimization became the way for dealing with many problems in several fields such as parameter estimation, classification, scheduling, feature selection and medical image.  In this part, we improved the performance of sine cosine algorithm, grey wolf optimization, and whale optimization and applied the modified versions to global optimization problem.
21. An improved Opposition-Based Sine Cosine Algorithm for global optimization  The sine-cosine algorithms (SCA) is a global optimization approach based on two trigonometric functions. SCA uses the sine and cosine functions to modify a set of candidate solutions; such operators create a balance between exploration and exploitation of the search space.
22. An improved Opposition-Based Sine Cosine Algorithm for global optimization  However, like other similar approaches, SCA tends to be stuck into sub-optimal regions that it is reflected in the computational effort required to find the best values. This situation occurs due that the operators used for exploration do not work well to analyze the search space.  Also, its accuracy and convergence are affected by the calibration and randomness of some internal parameters this fact is similar to other MH.
23. The Limitation of SI Techniques 23
24. The Limitation of SI Algorithms 24
25. The Limitation of SI Algorithms 25
26. The Limitation of SI Algorithms  Therefore, if this value not determined in optimal value the convergence becomes slow and the accuracy is degraded, the main reason for that limitation is, the global solution may be in opposite direction to the current solution. 26
27. Opposite-based Learning (OBL)  The basics of the OBL method are presented in [22], by considering a real number 𝑥 ∈ [𝑢, 𝑙] (𝑢 and 𝑙 represent the upper boundary and the lower boundary of the problem, respectively), the opposite number ҧ 𝑥 is defined as: ҧ 𝑥 = 𝑢 + 𝑙 − 𝑥 (6)  Finally, in the optimization method, the current solution 𝑥 is selected if 𝑓(𝑥) is better than 𝑓( ҧ 𝑥); otherwise, ҧ 𝑥 is selected. Therefore, the population of the solutions is updated based on the best values of 𝑥 and ҧ 𝑥. 
28. Opposition-based Learning 28 Opposite Solution solution
29. Opposition-based Learning 29 Opposite Solution Solution
30. The proposed OBSCA algorithm Figure 7: The proposed OBSCA algorithm
31. Experimental  The proposed has been tested over several benchmark functions and engineering problems. The comparison results support the efficacy of the proposed approach to find the optimal solutions in complex search spaces (see Figure 8).
32. Figure 8: Convergence curves for fitness function from F1 to F4.
33.  It can be seen that the proposed OBSCA method has a better ability for exploration and exploitation due to the solutions are updated using OBL that gives the proposed method ability to switch between two components, the sine and cosine functions.  Also, the OBL strategy gives the OBSCA ability to skip the local point since it takes the opposite direction into consideration.  Moreover, the other algorithms such as PSO, MFO, ABC, and HS used only one equation to update the solutions and this increase the probability to stagnation in local point. While, the other algorithms such as MVO, SSO and SCA are used two equations, in general, to update their solutions.  However, the last algorithms may be easy to get stuck in local point, since they depend on the current best solution to update the other solutions, this limitation can be avoided by using the OBL strategy. The influence of this strategy appears in the convergence of the OBPSO and OBSCA algorithm, however, the performance of the OBSCA is better than OBPSO since it combined the both strategies (two equations and OBL).
34. Chaotic opposition-based grey-wolf optimization algorithm based on differential evolution and disruption operator for global optimization  Grey-Wolf Optimizer (GWO) was proposed as an optimizer for the global optimization problems, where it simulates the grey wolves’ leadership and the hunting manner in nature.
35. Chaotic opposition-based grey-wolf optimization algorithm based on differential evolution and disruption operator for global optimization  According to these behaviors, the GWO algorithm has been applied to different applications, for example, economic dispatch, the estimation of the parameters of the Multi-layer Feedforward Neural Networks, and feature selection.  The main drawback of the GWO algorithm is that its performance and convergence are affected by several factors.  For example, the initial population, in which, if it is not selected by a suitable way, then the GWO cannot converge to the optimal solution.  Also, the balance between the exploration and the exploitation in the search space has also the main effect on the convergence, in which, the most exploration makes the algorithm leads to be unable to find the global solution,  while the most exploitation makes the algorithm to takes long time and may be stuck in local point.
36. Techniques  In this paper, an improved version of the Grey Wolf Optimizer (GWO) is proposed to improve the exploration and the exploitation ability of the GWO algorithm.  This improvement is performed by using  Chaotic logistic map,  Opposition-Based Learning (OBL),  Differential evolution (DE), and  Disruption operator (DO).
37. Chaotic maps  Instead of generating a random sequence from the uniform distribution, the chaotic sequence provides a more effective searching plan for the heuristic optimization algorithms.  The logistic map adopted in this article to generate the solution sequence is described as follows [25]: 𝐺𝑡+1 = 𝜇 × 𝐺𝑡 1 − 𝐺𝑡 , 𝑡 = 1,2, … , 𝑡𝑚𝑎𝑥 (7) Where 𝜇 ∈ [2,4] is the control parameter, and 𝑡 represents the iteration index.
38. Disruption Operator  The disruption operator (𝐷𝑜𝑝) has been used to improve the exploration ability of several MH algorithms and it is defined as [27]: 𝐷𝑜𝑝 = ൞ 𝐷𝑖𝑠𝑖,𝑗 𝜓 −0.5,0.5 𝑖𝑓 𝐷𝑖𝑠𝑖,𝑏𝑒𝑠𝑡 ≥ 1 1 + 𝐷𝑖𝑠𝑖,𝑏𝑒𝑠𝑡 𝜓 −10−16 2 , 10−16 2 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 where 𝐷𝑖𝑠𝑖,𝑗 is the Euclidean distance between two wolves, 𝑖th and 𝑗th (where 𝑗 is the nearest neighborhood of the 𝑖th wolf). The 𝜓(𝑎, 𝑏) represents a random number which is uniformly distributed in the interval [𝑎, 𝑏].  From the definition of the operator𝐷𝑜𝑝, we can say that when the two wolves are very close to one another, then 𝐷𝑜𝑝 < 1 and this leads to that the wolves will converge towards the origin.
39. Differential evolution  Mutation operation:  Crossover operation:  Selection operation:
40. Proposed method Figure 9: The framework of the proposed COGWO2D method.
41. Experiments and Discussion  The proposed algorithm, called COGWO2D, is compared with other seven algorithms through a set of experimental series that have been performed over two benchmark functions, the classical CEC2005, and the CEC2014 (as in Table 3).  The performance of the proposed algorithm to improve the classification of the galaxy images is evaluated, where it is used as a feature selection method.
42. Figure : The convergence curves for the CEC2014 fitness functions F1, F2, F5, F6 (F1) (F2) (F5) (F6)
43. Application of the proposed method for galaxy image classification  We apply COGWO2D method as a feature selection method to improve the classification accuracy of a real galaxy image dataset. Since the classification. of different galaxy types is critical to understand the galaxies formation and evaluation of the universe.  The sample of the galaxy images, used in this experimental, is a part of the EFIGI catalog as in Figure 11. Figure 11: Some samples of the images of the galaxy dataset (a) Irregular, (b) Elliptical, (c) Spiral and (d) Lenticular.
44. Application of the proposed method for galaxy image classification  The results of the comparison are shown in Table 4, that shows the accuracy of classification, the optimal subset of the selected features of each algorithm and the needed time.
45. Image segmentation  Image segmentation is one of the most important tasks in the image processing that can be used to separate the components of the image into different classes by combining the same pixels together.  There are several image segmentation methods that have been proposed such as clustering[9], and thresholding [10,11].  However, the threshold image segmentation methods have more attention than the other approaches due to their simplicity and high accuracy against the other methods [12]. Image segmentation Technique
46. Image segmentation  The image threshold methods can be categorized into two groups,  bi-level thresholding  multi-level thresholding  The bi-level thresholding where the histogram of the given, grayscale, image is split into two classes based on the threshold value.  The lower intensity pixels than a threshold are grouped into the first class, while the other pixels (higher than the threshold) belong to the other class.  However, the process of splitting the image into two classes may be insufficient when the image contains more objects with nearly the same levels of gray [13].  Therefore, more threshold values are required to solve this problem, so the bi-level thresholding is extended into multi-level thresholding.
47. Problem Definition  The mathematical definition of the multilevel thresholding problem is illustrated, by considering a gray level image 𝐼 to be segmented consists of 𝐾 + 1 classes.  Therefore, 𝐾 thresholds, 𝑡1 , 𝑡2 , … , 𝑡𝐾 are required to divided the image into subregions as in Equation (1) [14,15]: 𝐶0 = {𝐼(𝑖, 𝑗) ∈ 𝐼|0 ≤ 𝐼 𝑖, 𝑗 ≤ 𝑡1 − 1} 𝐶1 = {𝐼(𝑖, 𝑗) ∈ 𝐼|𝑡1 ≤ 𝐼 𝑖, 𝑗 ≤ 𝑡2 − 1} (1) … 𝐶𝐾 = {𝐼(𝑖, 𝑗) ∈ 𝐼|𝑡𝑘 ≤ 𝐼 𝑖, 𝑗 ≤ 𝐿 − 1} where 𝐶𝑘 is the 𝑘-th classes of multi-level thresholding, 𝐼(𝑖, 𝑗) is the gray level of the pixel (𝑖, 𝑗), 𝑡𝑘 (𝑘 = 1, … , 𝑛) is the 𝑘th threshold value, 𝑛 is the number of thresholds, and 𝐿 is the gray levels in a image 𝐼, these levels are in the range (0, 1,…,𝐿 − 1).
48. Problem Definition  The essential purpose of multi-level thresholding is to locate threshold values which split pixels into various groups; that can be determined by maximizing the following equation 𝑡1 ∗, 𝑡2 ∗, … , 𝑡𝐾 ∗ = max 𝑡1,𝑡2,…,𝑡𝐾 𝐹 𝑡1 , 𝑡2 , … , 𝑡𝐾 , (2) where 𝐹 𝑡1 , 𝑡2 , … , 𝑡𝐾 is the otsu’s function that defined as: 𝐹𝑂𝑡𝑠 = ෍ 𝑖=0 𝐾 𝐴𝑖 (𝜂𝑖 − 𝜂1 )2 , 𝐴𝑖 = ෍ 𝑗=𝑡𝑖 𝑡𝑖+1−1 𝑃𝑗 , 𝜂𝑖 = ෍ 𝑗=𝑡𝑖 𝑡𝑖+1−1 𝑖 𝑃𝑗 𝐴𝑗 (3) where 𝜂1 is the mean intensity of 𝐼 with 𝑡0 = 0 and 𝑡𝐾+1 = 𝐿. The ℎ(𝑖) and 𝑃𝑖 are the frequency and the probability of the 𝑖th gray level respectively. Moreover, the kapur function can be used as fitness function and it is defined as: 𝐹𝐾𝑎𝑝 = ෍ 𝑖=0 𝐾 𝐻𝑖 , 𝐻𝑖 = − ෍ 𝑗=𝑡𝑖 𝑡𝑖+1−1 𝑃𝑗 σ 𝑗=𝑡𝑖 𝑡𝑖+1−1 𝑃𝑗 𝑙𝑜𝑔 𝑃𝑗 σ 𝑗=𝑡𝑖 𝑡𝑖+1−1 𝑃𝑗 (4)
49. Whale optimization algorithm and moth- flame optimization for multilevel thresholding image  This paper examines the ability of two nature inspired algorithms namely: Whale Optimization Algorithm (WOA) and Moth-Flame Optimization (MFO) to determine the optimal multilevel thresholding for image segmentation.  The MFO algorithm is inspired from the natural behavior of moths which have a special navigation style at night since they fly using the moonlight.  The WOA algorithm emulates the natural cooperative behaviors of whales.
50. Figure 2: The structure of the proposed method based on (a) WOA (b) MFO.
51. Figure 2: The structure of the proposed method based on (a) WOA (b) MFO.
52. Figure 2: The structure of the proposed method based on (a) WOA (b) MFO.
53. Figure 2: The structure of the proposed method based on (a) WOA (b) MFO.
54. Figure 2: The structure of the proposed method based on (a) WOA (b) MFO.
55. Experimental Results  The performance of the proposed algorithms has been evaluated using several of benchmark images and has been compared with five different swarm algorithms.
56. Definition of Measure Where 𝜇1 and 𝜇2 are the mean intensity of the image I and 𝑆𝑒𝑔, respectively. 𝜎1 and 𝜎2 represent the standard deviation of 𝐼 and 𝑆𝑒𝑔, respectively. 𝜎𝐼,𝑆𝑒𝑔 is the covariance of 𝐼 and seg. 𝑐1 = 6.5 and 𝑐1 = 58.5. The results have been analyzed based on the best PSNR, and SSIM measures as represented in Figure 3a and 3b, respectively.
57. The average of algorithms over all images based on (a) PSNR, and (b) SSIM (a) PSNR (b) SSIM
58. ANOVA Test
59. Medical applications
60. COVID-19 Image Classification Using Deep Features and Fractional-order Marine Predators Algorithm  Currently, a new coronavirus, called COVID-19, has spread to many countries, with over two million infected people or so-called confirmed cases. Also, it has killed more than 376,000 (up to 2 June 2020) [1] .  The family of coronaviruses is considered serious pathogens for people because they infect respiratory, hepatic, gastrointestinal, and neurologic diseases.  They are distributed among people, bats, mice, birds, livestock, and other animals[1,2] . In the last two decades, two famous types of coronaviruses SARS-CoV and MERS- CoV had been reported in 2003 and 2012, in China, and Saudi Arabia, respectively[3]. Although outbreaks of SARS and MERS had confirmed human to human transmission [3], they had not the same spread speed 8 and infection power of the new coronavirus (COVID-19).  For diagnosing COVID-19, the RT-PCR (real-time polymerase chain reaction) is a standard diagnostic test, but, it can be considered as a time-consuming test, more so, it also suffers from false negative diagnosing [4]. However, using medical imaging, chest CT, and chest X-ray scan can play a critical role in COVID-19 diagnosis.
61.  In this paper, we used two different datasets. The first one, dataset 1 was collected by Joseph Paul Cohen and Paul Morrison and Lan Dao, where some COVID-19 images were collected by an Italian Cardiothoracic radiologist.  Negative COVID-19 images were collected from another Chest X-ray Kaggle published dataset. The whole dataset contains around 200 COVID-19 positive images and 1675 negative COVID19 images. The data was collected mainly from retrospective cohorts of pediatric patients from Guangzhou Women and Children’s medical center.
62.  Medical imaging techniques are very important for diagnosing diseases. Image segmentation is a necessary image processing task that applied to discriminate region of interests (ROIs) from the area of outsides. Also, image segmentation can extract critical features, including the shape of tissues, and texture [5,6]  In general, feature selection (FS) methods are widely employed in various applications of medical imaging applications.  For example, Lambin et al. [7] proposed an efficient approach called Radiomics to extract medical image features. They showed that analyzing image features resulted in more information that improved medical imaging.  Chong et al. [8] proposed an FS model, called Robustness-Driven FS (RDFS) to select futures from lung CT images to classify the patterns of fibrotic interstitial lung diseases. They applied the SVM classifier with and without RDFS. The evaluation showed that the RDFS improved SVM robustness against reconstruction kernel and slice thickness. In \cite{sohail2011classification}, to classify ultrasound medical images, the authors used distance-based FS methods and a Fuzzy Support Vector Machine (FSVM). Moreover, a multi- objective genetic algorithm was applied to search for the optimal features subset. 67
63. Material and methods (Features extraction using convolutional neural networks)  In this paper, we apply a convolutional neural network (CNN) to extract features from COVID-19 X-Ray images. We adopt a special type of CNN called a pre-trained model where the network is previously trained on the ImageNet dataset  So, transfer learning is applied by transferring weights that were already learned and reserved into the structure of the pre-trained model, such as Inception, in this paper. 68
64. Fractional-order Marine Predators Algorithm (FO-MPA)  Recently, a combination between the fractional calculus tool and the meta-heuristics opens new doors in providing robust and reliable variants.  Hence, the FC memory is applied during updating the prey locating in the second step of the algorithm to enhance the exploitation stage. Moreover, the 𝑅𝐵 parameter has been changed to depend on Weibull distribution as described below. 69
66. Results and Discussion 71
68.  Figure 5, shows that FO-MPA shows an efficient and faster convergence than the other optimization algorithms on both 273 datasets. Whereas, the slowest and the insufficient convergences were reported by both SGA and WOA in Dataset 1 and by 274 SGA in Dataset 2. 73
70. Task scheduling in cloud computing
71. Task scheduling in cloud computing based on hybrid moth search algorithm and differential evolution  The cloud task scheduling problem can be defined as how to schedule and allocate various tasks to numerous virtual machines (VMs) practically and make all the tasks accomplished in a short execution time period.  Considering the cloud system (CS) which consists of 𝑁𝑝𝑚 physical machines (PM) and each physical machine consists of 𝑁𝑣𝑚 virtual machines (VMs).  The main objective functions is to reduce the makespan by locating the best set of tasks to be executed on VMs. 𝐸𝐶𝑇𝑙𝑘 = 𝑡𝑎𝑠𝑘𝑙𝑒𝑛𝑔𝑡ℎ 𝑀𝐼𝑃𝑆𝑘 (15)  where 𝐸𝐶𝑇𝑙𝑘 refers to the required execution time of lth task on kth VM where is the number of VMs and is the number of tasks. The fitness value can be defined as [28]: 𝑓𝑖𝑡 = max 𝐸𝐶𝑇𝑙𝑘 , ∀𝑙 ∈ 1, 𝑁𝑡𝑠𝑘 , 𝑘 = 1,2, … , 𝑁𝑣𝑚 (16)
72. Related work  Meta-heuristic approaches can handle massive search space to discover optimal solution for task scheduling problem within polynomial time. Keshanchi et al. [37], proposed an improved genetic algorithm with a heuristic-based HEFT named as N-GA used for static task scheduling in the cloud. Two-Stage-Task Scheduling Problem in Data-Centers solved by Johnson’s-rule-based GA (JRGA) Algorithm.  Akbari et al. [38], improves the performance of genetic algorithm through significant changes and overview of new operators that assure sample diversity and reliable coverage of the whole space.
73. The Proposed Approach  In this section, the moth search algorithm based on DE is proposed for solving the task scheduling problem in cloud computing. The DE algorithm is used as local search method to enhance the exploitation ability of MSA.
74. Figure 13: The proposed MSDE method for cloud Task scheduling.
75. Experiments and Discussion  In order to validated these expectations, a set of experimental series are performed.  In first experiment, we used CloudSim 3.0.3 to implemented the algorithms. We have evaluated the performance of makespan and compared with existing meta-heuristics techniques such as PSO, WOA and MSA, and heuristics techniques such as Shortest Job First (SJF) and Round robin (RR). Figure 14: The average of makespan of the algorithms using synthetic workload models.
76. Second Experiment  We evaluated the meta-heuristic algorithms by simulations using the CloudSim to create a simulated cloud architecture consisted of identical heterogeneous VMs, physical machines, and with thousands of CloudSim's cloudlets.  The information of cloudlets and VMs is extracted from a real log-traces (HPC2N Seth log-trace) to model HPC jobs/tasks and NASA Ames iPSC/860 log in Feitelson's Parallel Workloads Archive (PWA) presented in Table 7.
77.  We have evaluated the performance of makespan of the proposed algorithm and compared with existing meta-heuristic techniques such as PSO, WOA and MSA. To present the scalability of the proposed algorithm in terms of the number of tasks, we vary the number of cloud tasks from 100 to 2000.  The comparison results between the proposed MSDE algorithm and the other three algorithms using HPC2N (NASA iPSC) real trace are given in Figures 15-16. Figure 15: The results of the algorithms using HPC2N real trace. (b) Large tasks (a) Small tasks
78. (a) Small tasks (b) Large tasks Figure 16: The results of the algorithms using NASA iPSC real trace.  Figure 15a and Figure 16a show the comparison of average makespan for smaller tasks (100-600) between MSDE, MSA, PSO and WOA using HPC2N and NASA iPSC, respectively. When the number of tasks ranges from 100 to 300, the makespan of MSDE algorithm with others is not much difference except in the case of NASA iPSC there is large difference between the proposed MSDE and MSA.  As well as, with this small tasks, the proposed algorithm still achieves the better results than the other three algorithm. In addition, when the number of tasks ranges from 400 to 600, the makespan of MSDE was significantly smaller than that of the other algorithms.
79. Parameter estimation of photovoltaic cells  The using of solar energy has been increased since it is a clean source of energy.  In this way, the design of photovoltaic cells has attracted the attention of researchers over the world.  Considering the design problem involves the solution of the complex non-linear and multi-modal objective functions.  Different algorithms have been proposed to identify the parameters of the photovoltaic cells and panels. Most of them commonly fail in finding the optimal solutions.  We proposed the Chaotic Whale Optimization Algorithm (CWOA) for the parameters estimation of solar cells.  The main advantage of the proposed approach is using the chaotic maps to compute and automatically adapt the internal parameters of the optimization algorithm.
80. Double diode model of solar cell Single diode model of solar cells
83. Copper Prices  An accurate forecasting model for the price volatility of minerals plays a vital role in future investments and decisions for mining projects and related companies.  We proposed a hybrid model to provide an accurate model for forecasting the copper prices.  The proposed model combines the adaptive neuro-fuzzy inference system (ANFIS) and genetic algorithm (GA). Genetic algorithms are used for estimating the ANFIS model parameters.
85. proposed
86. Historical copper prices from September 1987 to August 2017
87. Results Statistical results using Wilcoxon rank sum test
88. Actual and forecasted values during training via ANFIS, SVM, GARCH, and ARIMA for copper price
89. Actual and forecasted values during validation via GA–ANFIS, ANFIS, SVM, GARCH, and ARIMA for copper price
90. Oil Consumption Forecasting  Oil consumption is one of the main factors that affect industry and economy. Therefore, it is very important to estimate and forecast the consumption of oil.  This helps the governments to take the right decisions and avoid the wrong decisions that lead to negative outcomes.  For that reason, there are several methods that have been applied to forecast the oil consumption, such as the adaptive neuro-fuzzy inference system (ANFIS) model. It is one of the most popular data mining methods used to perform the forecast.  However, the ANFIS model may not be accurate (biased) in all data, since its parameters require to be determined and updated and this may lead to stuck in the local point and not convergence to the optimal value.  To this end, this paper presents an alternative oil consumption forecasting method by improving the ANFIS using the sine–cosine algorithm (SCA).
92. The consumption of Canada oil during Sep. 2007 to Aug. 2017. The consumption of Japan oil during Sep. 2007 to Aug. 2017 The consumption of Germany oil during Sep. 2007 to Aug. 2017.
93. Resulst Results of RMSE and MAE for all algorithms overall experiments.
94. The real data against forecasted oil consumptions of Canada. The real data against forecasted oil consumptions of Germany.
95. The real data against forecasted oil consumptions of Japan.
96. Conclusion and Future works  The metaheuristic techniques have more attentions in recent years since they have high ability to find the global solution of different problems.  There are three main major’s points of this review, the first major is to illustrate the performance of the MH methods as an image segmentation method.  The second major of this review is to improve the performance of grey-wolf optimization algorithm, sine cosine Algorithm, whale optimization algorithm as global optimization methods.  Third point is to improve the moth search algorithm and applied it as task scheduling.  The results illustrate the high performance of the developed techniques over the traditional techniques.
97. Future works  In future, we will try to developed a new metaheuristic technique to improve the color image segmentation. Since the color image segmentation is very important in different area including remote sensing, satellite analysis, medical image. However, it is hard problem since they search about the threshold in three channel of the image (i.e., RGB). Moreover, we will try a general framework for feature selection which can determine the optimal classifier, number of features and then applied it to different applications.  Moreover, the Fractional-order calculus will be applied to improve performance of some MH techniques since it will give them high ability to benefit from the memory properties of FC. As FC provides a sufficient description of the memory of the processes during its combination with the evolutionary techniques.  The fractional-order derivative encompasses an infinite number of terms treating the past values, while the integer-order one requires only a finite and a very limited number. Accordingly, the integer-order derivatives are local operators whereas the fractional-order derivative has a memory of all previous events.