Showing papers in "Evolutionary Intelligence in 2021"
TL;DR: A comprehensive survey of applications of CNNs in medical image understanding is presented in this article, where a discussion on CNN and its various award-winning frameworks have been presented, and critical discussion on some of the challenges is also presented.
Abstract: Imaging techniques are used to capture anomalies of the human body. The captured images must be understood for diagnosis, prognosis and treatment planning of the anomalies. Medical image understanding is generally performed by skilled medical professionals. However, the scarce availability of human experts and the fatigue and rough estimate procedures involved with them limit the effectiveness of image understanding performed by skilled medical professionals. Convolutional neural networks (CNNs) are effective tools for image understanding. They have outperformed human experts in many image understanding tasks. This article aims to provide a comprehensive survey of applications of CNNs in medical image understanding. The underlying objective is to motivate medical image understanding researchers to extensively apply CNNs in their research and diagnosis. A brief introduction to CNNs has been presented. A discussion on CNN and its various award-winning frameworks have been presented. The major medical image understanding tasks, namely image classification, segmentation, localization and detection have been introduced. Applications of CNN in medical image understanding of the ailments of brain, breast, lung and other organs have been surveyed critically and comprehensively. A critical discussion on some of the challenges is also presented.
149 citations
TL;DR: The results of simulation showed that chaotic maps (particularly the Tent map) are able to enhance the performance of the VSA and the proposed method has higher percentage of accuracy in comparison to other algorithms.
Abstract: The Vortex Search Algorithm (VSA) is a meta-heuristic algorithm that has been inspired by the vortex phenomenon proposed by Dogan and Olmez in 2015. Like other meta-heuristic algorithms, the VSA has a major problem: it can easily get stuck in local optimum solutions and provide solutions with a slow convergence rate and low accuracy. Thus, chaos theory has been added to the search process of VSA in order to speed up global convergence and gain better performance. In the proposed method, various chaotic maps have been considered for improving the VSA operators and helping to control both exploitation and exploration. The performance of this method was evaluated with 24 UCI standard datasets. In addition, it was evaluated as a Feature Selection (FS) method. The results of simulation showed that chaotic maps (particularly the Tent map) are able to enhance the performance of the VSA. Furthermore, it was clearly shown the fitness of the proposed method in attaining the optimal feature subset with utmost accuracy and the least number of features. If the number of features is equal to 36, the percentage of accuracy in VSA and the proposed model is 77.49 and 92.07. If the number of features is 80, the percentage of accuracy in VSA and the proposed model is 36.37 and 71.76. If the number of features is 3343, the percentage of accuracy in VSA and the proposed model is 95.48 and 99.70. Finally, the results on Real Application showed that the proposed method has higher percentage of accuracy in comparison to other algorithms.
84 citations
TL;DR: An automated system for categorization of the soil datasets into respective categories using images of the soils using Bag-of-words and chaotic spider monkey optimization based method which can further be used for the decision of crops.
Abstract: A proper soil prediction is one of the most important parameters to decide the suitable crop which is generally performed manually by the farmers. Therefore, the efficiency of the farmers may be increased by producing an automated tools for soil prediction. This paper presents an automated system for categorization of the soil datasets into respective categories using images of the soils which can further be used for the decision of crops. For the same, a novel Bag-of-words and chaotic spider monkey optimization based method has been proposed which is used to classify the soil images into its respective categories. The novel chaotic spider monkey optimization algorithm shows desirable convergence and improved global search ability over standard benchmark functions. Hence, it has been used to cluster the keypoints in Bag-of-words method for soil prediction. The experimental outcomes illustrate that the anticipated methods effectively classify the soil in comparison to other meta-heuristic based methods.
64 citations
TL;DR: This paper presents a new population-based metaheuristic algorithm inspired by a new source of inspiration called Giza Pyramids Construction (GPC) inspired by the ancient past that is successful in solving high-dimensional problems, especially image segmentation.
Abstract: Nowadays, many optimization issues around us cannot be solved by precise methods or that cannot be solved in a reasonable time. One way to solve such problems is to use metaheuristic algorithms. Metaheuristic algorithms try to find the best solution out of all possible solutions in the shortest time possible. Speed in convergence, accuracy, and problem-solving ability at high dimensions are characteristics of a good metaheuristic algorithm. This paper presents a new population-based metaheuristic algorithm inspired by a new source of inspiration. This algorithm is called Giza Pyramids Construction (GPC) inspired by the ancient past has the characteristics of a good metaheuristic algorithm to deal with many issues. The ancient-inspired is to observe and reflect on the legacy of the ancient past to understand the optimal methods, technologies, and strategies of that era. The proposed algorithm is controlled by the movements of the workers and pushing the stone blocks on the ramp. This algorithm is compared with five standard and popular metaheuristic algorithms. For this purpose, thirty different and diverse benchmark test functions are utilized. The proposed algorithm is also tested on high-dimensional benchmark test functions and is used as an application in image segmentation. The results show that the proposed algorithm is better than other metaheuristic algorithms and it is successful in solving high-dimensional problems, especially image segmentation.
60 citations
TL;DR: In this paper, the authors proposed a new variant of GA for FS which uses GDA to strengthen its exploitational ability and application of the proposed method on 15 well-known UCI datasets using KNN, MLP and SVM classifiers.
Abstract: Feature selection methods are used to identify and remove irrelevant and redundant attributes from the original feature vector that do not have much contribution to enhance the performance of a predictive model. Meta-heuristic feature selection algorithms, used as a solution to this problem, need to have a good trade-off between exploitation and exploration of the search space. Genetic Algorithm (GA), a popular meta-heuristic algorithm, lacks exploitation capability, which in turn affects the local search ability of the algorithm. Basically, GA uses mutation operation to take care of exploitation which has certain limitations. As a result, GA gets stuck in local optima. To encounter this problem, in the present work, we have intelligently blended the Great Deluge Algorithm (GDA), a local search algorithm, with GA. Here GDA is used in place of mutation operation of the GA. Application of GDA yields a high degree of exploitation through the use of perturbation of candidate solutions. The proposed method is named as Deluge based Genetic Algorithm (DGA). We have applied the DGA on 15 publicly available standard datasets taken from the UCI dataset repository. To show the classifier independent nature of the proposed feature selection method, we have used 3 different classifiers namely K-Nearest Neighbour (KNN), Multi-layer Perceptron (MLP) and Support Vector Machine (SVM). Comparison of DGA has been performed with other contemporary algorithms like the basic version of GA, Particle Swarm Optimisation (PSO), Simulated Annealing (SA) and Histogram based Multi-Objective GA (HMOGA). From the comparison results, it has been observed that DGA performs much better than others in most of the cases. Thus, our main contributions in this paper are introduction of a new variant of GA for FS which uses GDA to strengthen its exploitational ability and application of the proposed method on 15 well-known UCI datasets using KNN, MLP and SVM classifiers.
60 citations
TL;DR: From the CDTL approach, the RF classifier achieves 89.30% improved prediction accuracy from 76.70% accuracy (without CDTL), and the error rate of RF with CDTL has significantly reduced from 23.30 to 9.70%.
Abstract: In the rural side, there is the absence of centers for cardiovascular ailment. Due to this, around 12 million people passing worldwide reported by WHO. The principal purpose of coronary illness is a propensity for smoking. ML classifiers are applied to predict the risk of cardiovascular disease. However, the ML model has some inherent problems like it’s serene to feature selection, splitting attribute, and imbalanced datasets prediction. Most of the mass datasets have multi-class labels, but their combinations are in different proportions. In this paper, we experiment with our system with Cleveland’s heart samples from the UCI repository. Our cluster-based DT learning (CDTL) mainly includes five key stages. At first, the original set has partitioned through target label distribution. From the high distribution samples, the other possible class combination has made. For each class-set combination, the significant features have identified through entropy. With the significant critical features, an entropy-based partition has made. At last, on these entropy clusters, RF performance is made through significant and all features in the prediction of heart disease. From our CDTL approach, the RF classifier achieves 89.30% improved prediction accuracy from 76.70% accuracy (without CDTL). Hence, the error rate of RF with CDTL has significantly reduced from 23.30 to 9.70%.
47 citations
TL;DR: UNSW-NB15 data set is considered as the benchmark dataset to design UIDS for detecting malicious activities in the network and the performance analysis proves that the attack detection rate of the proposed model is higher compared to two existing approaches ENADS and DENDRON.
Abstract: Intrusion detection system (IDS) using machine learning approach is getting popularity as it has an advantage of getting updated by itself to defend against any new type of attack. Another emerging technology, called internet of things (IoT) is taking the responsibility to make automated system by communicating the devices without human intervention. In IoT based systems, the wireless communication between several devices through the internet causes vulnerability for different security threats. This paper proposes a novel unified intrusion detection system for IoT environment (UIDS) to defend the network from four types of attacks such as: exploit, DoS, probe, and generic. The system is also able to detect normal category of network traffic. Most of the related works on IDS are based on KDD99 or NSL-KDD 99 data sets which are unable to detect new type of attacks. In this paper, UNSW-NB15 data set is considered as the benchmark dataset to design UIDS for detecting malicious activities in the network. The performance analysis proves that the attack detection rate of the proposed model is higher compared to two existing approaches ENADS and DENDRON which also worked on UNSW-NB15 data set.
42 citations
TL;DR: The result of this study shows that the AND operation of two classifier output will tend to yield the overall diagnostic accuracy, which outperforms the conventional models.
Abstract: The objective of this study is to frame mammogram breast detection model using the optimized hybrid classifier. Image pre-processing, tumor segmentation, feature extraction, and detection are the functional phases of the proposed breast cancer detection. A median filter eliminates the noise of the input mammogram. Further, the optimized region growing segmentation is carried out for segmenting the tumor from the image and the optimized region growing depends on a hybrid meta-heuristic algorithm termed as firefly updated chicken based CSO (FC-CSO). To the next of tumor segmentation, feature extraction is done, which intends to extract the features like grey level co-occurrence matrix (GLCM), and gray level run-length matrix (GRLM). The two deep learning architectures termed as convolutional neural network (CNN), and recurrent neural network (RNN). Moreover, both GLCM and GLRM are considered as input to RNN, and the tumor segmented binary image is considered as input to CNN. The result of this study shows that the AND operation of two classifier output will tend to yield the overall diagnostic accuracy, which outperforms the conventional models.
41 citations
TL;DR: The effectual performance and comparative analysis prove the stable and reliable performance of the proposed model over existing models.
Abstract: This proposal tempts to develop automated DR detection by analyzing the retinal abnormalities like hard exudates, haemorrhages, Microaneurysm, and soft exudates. The main processing phases of the developed DR detection model is Pre-processing, Optic Disk removal, Blood vessel removal, Segmentation of abnormalities, Feature extraction, Optimal feature selection, and Classification. At first, the pre-processing of the input retinal image is done by Contrast Limited Adaptive Histogram Equalization. The next phase performs the optic disc removal, which is carried out by open-close watershed transformation. Further, the Grey Level thresholding is done for segmenting the blood vessels and its removal. Once the optic disk and blood vessels are removed, segmentation of abnormalities is done by Top hat transformation and Gabor filtering. Further, the feature extraction phase is started, which tends to extract four sets of features like Local Binary Pattern, Texture Energy Measurement, Shanon’s and Kapur’s entropy. Since the length of the feature vector seems to be long, the feature selection process is done, which selects the unique features with less correlation. Moreover, the Deep Belief Network (DBN)-based classification algorithm performs the categorization of images into four classes normal, earlier, moderate, or severe stages. The optimal feature selection is done by the improved meta-heuristic algorithm called Modified Gear and Steering-based Rider Optimization Algorithm (MGS-ROA), and the same algorithm updates the weight in DBN. Finally, the effectual performance and comparative analysis prove the stable and reliable performance of the proposed model over existing models. The performance of the proposed model is compared with the existing classifiers, such as, NN, KNN, SVM, DBN and the conventional Heuristic-Based DBNs, such as PSO-DBN, GWO-DBN, WOA-DBN, and ROA-DBN for the evaluation metrics, accuracy, sensitivity, specificity, precision, FPR, FNR, NPV, FDR, F1 score, and MC. From the results, it is exposed that the accuracy of the proposed MGS-ROA-DBN is 30.1% higher than NN, 32.2% higher than KNN, and 17.1% higher than SVM and DBN. Similarly, the accuracy of the developed MGS-ROA-DBN is 13.8% superior to PSO, 5.1% superior to GWO, 10.8% superior to WOA, and 2.5% superior to ROA.
38 citations
TL;DR: The results prove the superiority of the proposed algorithm compared to the bench mark algorithms and develop the secure framework which restricts the insider attacks.
Abstract: In the present scenario, big data is facing many challenges regarding the data storage, data theft and unauthorized access. Many researchers are concentrated on developing the security mechanism for big data storage. To overcome the above issue, this paper concentrated on developing the encryption algorithm for storing big data in the multi cloud storage. The multi cloud storage environment permits the user to store the data in to different cloud storage services. This paper aims to develop the secure framework which restricts the insider attacks. The proposed framework contains data uploading, slicing, indexing, encryption, distribution, decryption, retrieval and merging process. The hybrid encryption algorithm was developed to provide the security to the big data before storing it in to the multi cloud. The Simulation analysis is carried with real time cloud storage environments. The proposed algorithm recorded around 2630 KB/S for the encryption process. The results prove the superiority of the proposed algorithm compared to the bench mark algorithms.
34 citations
TL;DR: The experimental results confirm the higher performance of the proposed convolutional neural network against existing different machine learning models with the lowest error rate and the introduced classification system is validated on colorectal cancer histology image dataset.
Abstract: Histopathological image classification is one of the important application areas of medical imaging. However, an accurate and efficient classification is still an open-ended research due to the complexity in histopathological images. For the same, this paper presents an efficient architecture of convolutional neural network for the classification of histopathological images. The proposed method consists of five subsequent blocks of layers, each having convolutional, drop-out, and max-pooling layers. The performance of the introduced classification system is validated on colorectal cancer histology image dataset which consists of RGB-colored images belonging to eight different classes. The experimental results confirm the higher performance of the proposed convolutional neural network against existing different machine learning models with the lowest error rate of 22.7%.
TL;DR: This paper aims to review various methodologies used and how it has evolved to give better results in the past years, closely moving towards usage of machine learning.
Abstract: Vehicle detection and classification has been an area of application of image processing and machine learning which is being researched extensively in accordance with its importance due to increasing number of vehicles, traffic rule defaulters and accidents. This paper aims to review various methodologies used and how it has evolved to give better results in the past years, closely moving towards usage of machine learning. This has resulted in advancing the problem statement towards helmet detection followed by number plate detection of defaulters. Object detection and Text recognition that are available in various frameworks offer built-in models which are easy to use or offer easy methods to build and train customized models.
TL;DR: An optimization based machine learning algorithm is proposed to classify the twitter data and it is observed that the proposed method i.e., sequential minimal optimization with decision tree gives good accuracy compared to other machine learning algorithms.
Abstract: Sentimental analysis determines the views of the user from the social media. It is used to classify the content of the text into neutral, negative and positive classes. Various researchers have used different methods to train and classify twitter dataset with different results. Particularly when time is taken as constraint in some applications like airline and sales, the algorithm plays a major role. In this paper an optimization based machine learning algorithm is proposed to classify the twitter data. The process was done in three stages. In the first stage data is collected and preprocessed, in the second stage the data is optimized by extracting necessary features and in the third stage the updated training set is classified into different classes by applying different machine learning algorithms. Each algorithm gives different results. It is observed that the proposed method i.e., sequential minimal optimization with decision tree gives good accuracy of 89.47% compared to other machine learning algorithms.
TL;DR: An efficient ant colony optimization (ACO) and particle swarm optimization (PSO)-based framework have been proposed for data classification and preprocessing in the big data environment and shows that the content part can be collaborated and fetched for analysis from the volume and velocity integration.
Abstract: Big data is prominent for the systematic extraction and analysis of a huge or complex dataset. It is also helpful in the management of data as compared to the traditional data-processing mechanisms. In this paper, an efficient ant colony optimization (ACO) and particle swarm optimization (PSO)-based framework have been proposed for data classification and preprocessing in the big data environment. It shows that the content part can be collaborated and fetched for analysis from the volume and velocity integration. Then weight marking has been done through the volume and the data variety. In the end, the ranking has been done through the velocity and variety aspects of big data. Data preprocessing has been performed from weights assigned on the basis of size, content, and keywords. ACO and PSO are then applied considering different computation aspects like uniform distribution, random initialization, epochs, iterations, and time constraint in case of both minimization and maximization. The weight assignments have been done automatically and through an unbiased random mechanism. It has been done on a scale of 0–1 for all the separated data. Then simple adaptive weight (SAW) method has been applied for prioritization and ranking. The overall average classification accuracy obtained in the case of PSO-SAW is 98%, and in the case of ACO-SAW, it is 95%. PSO-SAW approach outperforms in all cases, in comparison to ACO-SAW.
TL;DR: A novel hybrid algorithm developed by merging atom search optimization and simulated annealing algorithms is presented, which clearly demonstrated the superiority of the proposed algorithm over other recently reported best performing algorithms for power system stabilizer design.
Abstract: A novel hybrid algorithm developed by merging atom search optimization and simulated annealing algorithms is presented. The constructed improved algorithm, named as improved atom search optimization algorithm, was proposed for optimizing a power system stabilizer adopted in a single-machine infinite-bus power system. The evaluations were initially performed using several benchmark functions by comparing the results with genetic algorithm, simulated annealing technique, particle swarm optimization, gravitational search algorithm and the original version of atom search optimization algorithm. The obtained results showed the great promise of the developed hybrid algorithm in terms of the balance between exploration and exploitation phases. The performance of the proposed hybrid algorithm was also assessed through designing an optimally performing power system stabilizer for further evaluation. To do so, a power system stabilizer damping controller was formulated as an optimization problem and the improved algorithm was used to search for optimal controller parameters in order to show the applicability and greater performance of the proposed algorithm for such a complex real-world engineering problem. The obtained results for the latter case were compared with the best performing reported approaches of sine cosine algorithm and symbiotic organisms search algorithm. The comparisons clearly demonstrated the superiority of the proposed algorithm over other recently reported best performing algorithms for power system stabilizer design.
TL;DR: In this article, the authors implemented an image processing procedure to extract the tumor section from the clinical-grade MRI slices recorded with Flair and T2 modalities, which integrates thresholding and segmentation procedures.
Abstract: Brain abnormality is a severe illness in humans. An unrecognised and untreated brain illness will lead to a lot of complications despite of gender and age. Brain tumor is one of the severe conditions in humans; begins due to a variety of unavoidable and unpredicted reasons. The clinical level diagnosis of brain tumor is performed with the help of non-invasive imaging procedures, such as Computed-Tomography and Magnetic-Resonance-Imaging. The proposed work implements an image processing procedure to extract the tumor section from the clinical-grade MRI slices recorded with Flair and T2 modalities. This procedure integrates thresholding and segmentation procedures to extract the tumor division from 2D MRI slices with better accuracy. MRI slices with the skull section are considered in this work and the extraction of the tumor is further achieved by implementing the Modified Moth-Flame Optimization algorithm based Kapur’s thresholding and a chosen segmentation technique. Benchmark images of BRAINIX and TCIA-GBM datasets are used in this work for experimental investigation. The outcome establishes the performance values attained with Flair modality images are slightly better compared to T2 modality.
TL;DR: The proposed IGWO is an improved version of the GWO algorithm which uses the hill-climbing method and chaos theory to achieve better results and can outperform other scheduling approaches in terms of metrics such as power consumption, cost, and makespan.
Abstract: The workflow scheduling in the cloud computing environment is a well-known NP-complete problem, and metaheuristic algorithms are successfully adapted to solve this problem more efficiently. Grey wolf optimization (GWO) is a recently proposed interesting metaheuristic algorithm to deal with continuous optimization problems. In this paper, we proposed IGWO, an improved version of the GWO algorithm which uses the hill-climbing method and chaos theory to achieve better results. The proposed algorithm can increase the convergence speed of the GWO and prevents falling into the local optimum. Afterward, a binary version of the proposed IGWO algorithm, using various S functions and V functions, is introduced to deal with the workflow scheduling problem in cloud computing data centers, aiming to minimize their executions’ cost, makespan, and the power consumption. The proposed workflow scheduling scheme is simulated using the CloudSim simulator and the results show that our scheme can outperform other scheduling approaches in terms of metrics such as power consumption, cost, and makespan.
TL;DR: A control structure is designed which weakens the couplings and permits to develop a decentralized control of a quadrotor by PID controller and the efficiency of the proposed strategy where the optimization algorithms achieve good performance with a slight difference between the indicate techniques.
Abstract: This paper aims to investigate the control of a quadrotor by PID controller. The mathematical model is derived from Euler–Lagrange approach. Due to nonlinearities, coupling and under-actuation constraints, the model imposes difficulties to generate its controller by using classic ways. Firstly, we have designed a control structure which weakens the couplings and permits to develop a decentralized control. Secondly, in order to get the optimal path tracking, the controllers’ parameters were tuned by stochastic nature-inspired algorithms; Genetic Algorithm, Evolution Strategies, Differential Evolutionary and Cuckoo Search. A comparison study between these algorithms according to the path tracking is carried out by implementing simulations under MATLAB/Simulink. The results show the efficiency of the proposed strategy where the optimization algorithms achieve good performance with a slight difference between the indicate techniques.
TL;DR: A fuzzy clustering method is proposed by using the strengths of both modified whale optimization algorithm (MWOA) and FCM and it performs better than other compared methods.
Abstract: Fuzzy c-means (FCM) clustering method is used for performing the task of clustering. This method is the most widely used among various clustering techniques. However, it gets easily stuck in the local optima. whale optimization algorithm (WOA) is a stochastic global optimization algorithm, which is used to find out global optima of a provided dataset. The WOA is further modified to achieve better global optimum. In this paper, a fuzzy clustering method has been proposed by using the strengths of both modified whale optimization algorithm (MWOA) and FCM. The effectiveness of the proposed clustering technique is evaluated by considering some of the well-known existing metrics. The proposed hybrid clustering method based on MWOA is employed as an under sampling method to optimize the cluster centroids in the proposed automobile insurance fraud detection system (AIFDS). In the AIFDS, first the majority sample data set is trimmed by removing the outliers using proposed fuzzy clustering method, and then the modified dataset is undergone with some advanced classifiers such as CATBoost, XGBoost, Random Forest, LightGBM and Decision Tree. The classifiers are evaluated by measuring the performance parameters such as sensitivity, specificity and accuracy. The proposed AIFDS consisting of fuzzy clustering based on MWOA and CATBoost performs better than other compared methods.
TL;DR: A competitive grey wolf optimizer (CGWO) to solve the feature selection problem in electromyography (EMG) pattern recognition and experimental results show that OBCGWO can provide optimal classification performance, which is suitable for rehabilitation and clinical applications.
Abstract: This paper proposes a competitive grey wolf optimizer (CGWO) to solve the feature selection problem in electromyography (EMG) pattern recognition. We model the recently established feature selection method, competitive binary grey wolf optimizer (CBGWO), into a continuous version (CGWO), which enables it to perform the search on continuous search space. Moreover, another new variant of CGWO, namely opposition based competitive grey wolf optimizer (OBCGWO), is proposed to enhance the performance of CGWO in feature selection. The proposed methods show superior results in several benchmark function tests. As for EMG feature selection, the proposed algorithms are evaluated using the EMG data acquired from the publicly access EMG database. Initially, several useful features are extracted from the EMG signals to construct the feature set. The proposed CGWO and OBCGWO are then applied to select the relevant features from the original feature set. Four state-of-the-art algorithms include particle swarm optimization, flower pollination algorithm, butterfly optimization algorithm, and CBGWO are used to examine the effectiveness of proposed methods in feature selection. The experimental results show that OBCGWO can provide optimal classification performance, which is suitable for rehabilitation and clinical applications.
TL;DR: This work presents hybrid topic modeling techniques by integrating traditional topic models with visualization procedures to aid in the visualization of topic clouds and health tendencies in the document collection and believes proposed visual topic models viz., Visual Non-Negative Matrix Factorization (VNMF), Visual Latent Dirichlet Allocation (VLDA), and Visual Probabilistic Latent Schematic Indexing (VPLSI).
Abstract: Social media is a great source to search health-related topics for envisages solutions towards healthcare. Topic models originated from Natural Language Processing that is receiving much attention in healthcare areas because of interpretability and its decision making, which motivated us to develop visual topic models. Topic models are used for the extraction of health topics for analyzing discriminative and coherent latent features of tweet documents in healthcare applications. Discovering the number of topics in topic models is an important issue. Sometimes, users enable an incorrect number of topics in traditional topic models, which leads to poor results in health data clustering. In such cases, proper visualizations are essential to extract information for identifying cluster trends. To aid in the visualization of topic clouds and health tendencies in the document collection, we present hybrid topic modeling techniques by integrating traditional topic models with visualization procedures. We believe proposed visual topic models viz., Visual Non-Negative Matrix Factorization (VNMF), Visual Latent Dirichlet Allocation (VLDA), Visual intJNon-negative Matrix Factorization (VintJNMF), and Visual Probabilistic Latent Schematic Indexing (VPLSI) are promising methods for extracting tendency of health topics from various sources in healthcare data clustering. Standard and benchmark social health datasets are used in an experimental study to demonstrate the efficiency of proposed models concerning clustering accuracy (CA), Normalized Mutual Information (NMI), precision (P), recall (R), F-Score (F) measures and computational complexities. VNMF visual model performs significantly at an increased rate of 32.4% under cosine based metric in the display of visual clusters and an increased rate of 35–40% in performance measures compared to other visual methods on different number of health topics.
TL;DR: A novel tree-based algorithm based on the area under the precision-recall curve (AUPRC) for variable selection in the classification context, which found the proposed PRC classification tree, and its subsequent extension, the PRC random forest, work well especially for class-imbalanced data sets.
Abstract: The classification of imbalanced data has presented a significant challenge for most well-known classification algorithms that were often designed for data with relatively balanced class distributions. Nevertheless skewed class distribution is a common feature in real world problems. It is especially prevalent in certain application domains with great need for machine learning and better predictive analysis such as disease diagnosis, fraud detection, bankruptcy prediction, and suspect identification. In this paper, we propose a novel tree-based algorithm based on the area under the precision–recall curve for variable selection in the classification context. Our algorithm, named as the “Precision–Recall curve classification tree”, or simply the “PRC classification tree” modifies two crucial stages in tree building. The first stage is to maximize the area under the precision–recall curve in node variable selection. The second stage is to maximize the harmonic mean of recall and precision (F-measure) for threshold selection. We found the proposed PRC classification tree, and its subsequent extension, the PRC random forest, work well especially for class-imbalanced data sets. We have demonstrated that our methods outperform their classic counterparts, the usual CART and random forest for both synthetic and real data. Furthermore, the ROC classification tree proposed by our group previously, based on the area under the ROC curve, has shown good performance in imbalanced data. Their combination, the PRC–ROC tree, has also shows great promise in identifying the minority class.
TL;DR: This paper proposes a new multi-objective algorithm, called Multi-Objective Marine-Predator Algorithm (MOMPA), dependent on elitist non-dominated sorting and crowding distance mechanism, inspired by biological interaction between predator and prey.
Abstract: This paper proposes a new multi-objective algorithm, called Multi-Objective Marine-Predator Algorithm (MOMPA), dependent on elitist non-dominated sorting and crowding distance mechanism. The proposed algorithm is based on the recently proposed Marine-Predator Algorithm, and it was inspired by biological interaction between predator and prey. The proposed MOMPA can address multiple and conflicting objectives when solving optimization problems. The MOMPA is formulated using elitist non-dominated sorting and crowding distance mechanisms. The proposed method is tested on various multi-objective case studies, including 32 unconstrained, constraint, and engineering design problems with different linear, nonlinear, continuous, and discrete characteristics-based Pareto front problems. The results of the proposed MOMPA are compared with several well-regarded Multi-Objective Water-Cycle Algorithm, Multi-Objective Symbiotic-Organism Search, Multi-Objective Moth-Flame Optimizer algorithms qualitatively and quantitatively using several performance indicators. The experimental results demonstrate the merits of the proposed method.
TL;DR: The improved optimization algorithm for resource allocation is proposed by considering the objectives of minimizing the deployment cost and improving the QoS performance by considering different customer QoS requirements and allocates the resources within the given budget.
Abstract: In the recent years, cloud computing has emerged as one of the important fields in the information technology. Cloud offers different types of services to the web applications. The major issue faced by cloud customers are selecting the resources for their application deployment without compromising the quality of service (QoS) requirements. This paper proposed the improved optimization algorithm for resource allocation by considering the objectives of minimizing the deployment cost and improving the QoS performance. The proposed algorithm considers different customer QoS requirements and allocates the resources within the given budget. The experimental analysis is conducted on various workloads by deploying into the Amazon Web Services. The results shows the efficiency of the proposed algorithm.
TL;DR: A new multi-objective sine cosine algorithm is proposed for optimal DG allocation in radial distribution systems with minimization of total active power loss, maximization of voltage stability index, minimized of annual energy loss costs as well as pollutant gas emissions without violating the system and DG operating constraints.
Abstract: The integration of distributed generators (DGs) is considered to be one of the best cost-effective techniques to improve the efficiency of power distribution systems in the recent deregulation caused by continuous load demand and transmission system contingency. In this perspective, a new multi-objective sine cosine algorithm is proposed for optimal DG allocation in radial distribution systems with minimization of total active power loss, maximization of voltage stability index, minimization of annual energy loss costs as well as pollutant gas emissions without violating the system and DG operating constraints. The proposed algorithm is enhanced by incorporating exponential variation of the conversion parameter and the self-adapting levy mutation to increase its performance during different iteration phases. The contradictory relationships among the objectives motivate the authors to generate an optimal Pareto set in order to help the network operators in taking fast appropriate decisions. The proposed approach is successfully applied to 33-bus and 69-bus distribution systems under four practical load conditions and is evaluated in different two-objective and three-objective optimization cases. The effectiveness of the algorithm is confirmed by comparing the results against other well-known multi-objective algorithms, namely, strength Pareto evolutionary algorithm 2, non-dominated sorting genetic algorithm II and multi-objective particle swarm optimization. The quality of Pareto fronts from different multi-objective algorithms is compared in terms of certain performance indicators, such as generational distance, spacing metric and spread metric (
$${\varDelta }$$
), and its statistical significance is verified by performing Wilcoxon signed rank test.
TL;DR: This work presents water wave optimization (WWO) based metaheuristic algorithm for clustering task and shows that the proposed WWO clustering algorithm achieves a higher accuracy and F-score rates with most of clustering datasets as compared to existing clustering algorithms.
Abstract: Data clustering is an important activity in the field of data analytics. It can be described as unsupervised learning for grouping the similar objects into clusters. The similarity between objects is computed through distance measure. Further, clustering has proven its significance for solving wide range of real-world optimization problems. This work presents water wave optimization (WWO) based metaheuristic algorithm for clustering task. It is seen that WWO algorithm is an effective algorithm for solving constrained and unconstrained optimization problems. But, sometimes WWO cannot obtain promising solution for complex optimization problems due to absence of global best information component and converged on premature solution. To address the absentia of global best information and premature convergence, some improvements are inculcated in WWO algorithm to make it more promising and efficient. These improvements are described in terms of modified search mechanism and decay operator. The absentia of global best information component is handled through updated search mechanism. While, the premature convergence is addressed through a decay operator. The performance of WWO algorithm is evaluated using thirteen benchmark clustering datasets using accuracy and F-score parameters. The simulation results are compared with several state of art existing clustering algorithms and it is observed proposed WWO clustering algorithm achieves a higher accuracy and F-score rates with most of clustering datasets as compared to existing clustering algorithms. It is also showed that the proposed WWO algorithm improves the accuracy and F-score rates an average of 4% and 7% respectively as compared to existing clustering algorithm. Further, statistical test is also conducted to validate the existence of proposed WWO algorithm and statistical results confirm the existence of WWO algorithm in clustering field.
TL;DR: Multiple leader salp swarm algorithm (MLSSA) is introduced, which has more exploratory power than SSA and is tested on several mathematical optimization benchmark functions and compared with some well known metaheuristics.
Abstract: Metaheuristics are one of the most promising techniques for solving optimization problems. Salp swarm algorithm (SSA) is a new swarm intelligence based metaheuristic. To improve the performance of SSA, this paper introduces multiple leader salp swarm algorithm (MLSSA), which has more exploratory power than SSA. The algorithm is tested on several mathematical optimization benchmark functions. Results are compared with some well known metaheuristics. The results represents the capability of MLSSA to converge towards the optimum. In recent studies many metaheuristic techniques are applied to train feed-forward neural networks. In this paper MLSSA is also applied for neural network training and is analysed for thirteen different datasets. Performance is compared with SSA, particle swarm optimization, differential evolution, genetic algorithm, ant colony optimization and evolution strategy.
TL;DR: The goal of this work is to implement a velocity adaptation based particle swarm optimization for localization method to achieve minimum error and the results reveal that the proposed approach works better for obtaining improved location accuracy.
Abstract: Wireless sensor networks are a network of sensors interconnected through a wireless medium. Wireless sensor networks are utilized for many array of applications where determining precise location of the sensors are treated to be the crucial task. The prime job of localization is to determine the exact location of sensors placed at particular area as it makes the reference of anchor nodes to determine the location of remaining nodes in the network. Position information of sensor node in an area is useful for routing techniques and some application specific tasks. The localization accuracy is affected due to the estimations in anchor node placements. Localization information is not always easy as it varies with respect to the environment in which the sensors are deployed. Ranging errors occur in hostile environments and accuracy effects as there are signal attenuations in sensors when deployed underwater, underground etc. Efficiency can be enhanced by reducing the error using localization algorithms. Particle swarm optimization is one approach to overcome the localization problem. Results are considered for localization algorithms like Particle swarm optimization, Social group optimization and Velocity adaptation based Particle swarm optimization. The goal of this work is to implement a velocity adaptation based particle swarm optimization for localization method to achieve minimum error. The results reveal that the proposed approach works better for obtaining improved location accuracy.
TL;DR: The objective of this paper is to estimate the threshold values from software metrics by using novel evolutionary intelligence techniques to prevent software aging by using machine learning (evolutionary algorithms).
Abstract: The software metrics play the important role in the software industry. As the software industry growing in size and complexity enhanced support is mandatory for computing and managing the software quality. Quality measurement is one of the key features of the manager in the software industry; where threshold plays the crucial role. Software measurement is necessary by means for evaluating different quality attributes and characteristics, such as size, complexity, maintainability, and usability. Instead of that effective and efficient software system is straightforward dependent on the meaning of suitable thresholds. The objective of this paper is to estimate the threshold values from software metrics by using novel evolutionary intelligence techniques. The threshold and aging software design optimization algorithms and models to prevent software aging by using machine learning (evolutionary algorithms). Apart from the above-mentioned techniques, this paper also proposed a novel threshold estimation, aging, and survivability aware (sensitive) reusability optimization model of an object-oriented software system. To expand firmness, aging and survivability aware (sensitive) optimization threshold scheme aging prediction and software rejuvenation model and algorithms proposed.
TL;DR: An IVS system design is proposed using convolution neural networks, which achieves very low false alarm rates and in case of emergency like fire, thieves’ attacks, Intrusion Detector, the proposed system sends an alert for the corresponding services automatically.
Abstract: In Smart cities surveillance is an extremely important feature required for ensuring the safety of citizens and also for deterring the crime. Hence, intelligent video surveillance (IVS) frameworks are by and large increasingly more famous in security applications. The investigation and acknowledgment of anomalous practices in a video succession has step by step attracted the consideration in the field of IVS, as it permits sifting through an enormous number of pointless data, which ensures the high productivity in the security assurance, and spare a great deal of human and material assets. Techniques are proposed in the literature for analyzing the IVS systems. Existing systems for video analysis, suffer with some limitations. The one of the major limitation is lack of real time response from the surveillance systems. In order to overcome this limitation, an IVS system design is proposed using convolution neural networks. In case of emergency like fire, thieves’ attacks, Intrusion Detector, the proposed system sends an alert for the corresponding services automatically. Experimentation has done on the number of datasets available for video surveillance testing. The results show that the proposed surveillance system achieves very low false alarm rates.