Generative adversarial networks (GANs) have become popular in medical imaging because of their remarkable performance and ability to translate images across different domains. However, GANs face several issues in image-to-image translation, including training instability, lack of diversity, and mode collapse. These issues become even more complex when using cyclic GANs. Additionally, collecting paired images required for GANs may be costly, especially in the medical domain. Cyclic GANs are a favorable choice for addressing this issue, as they can convert cross-domain images. However, no pre-existing technique or algorithm is comprehensive enough to handle diverse datasets and applications. To address these issues, we propose a novel Quantized Evolutionary Gradient Aware Multiobjective Cyclic GAN (QEMCGAN) that employs evolutionary computation, multiobjective optimization, and an intelligent selection scheme. We use simulated annealing and Pareto ranking selection using three fitness criteria to address local optima stagnation. Additionally, we use model quantization because of its suitability for low-cost IoT-based applications. Extensive trials reveal that EMCGAN and QEMCGAN produces more visually realistic images than other approaches while preserving both background information and salient features. In addition, QEMCGAN performs on par with the baseline approach even when the model size is halved, making it more efficient.

High utility itemsets (HUIs) mining is an emerging area of data mining which discovers sets of items generating a high profit from transactional datasets. In recent years, several algorithms have been proposed for this task. However, most of them do not consider the on-shelf time period of items and negative utility of items. High on-shelf utility itemset (HOUIs) mining is more difficult than traditional HUIs mining because it deals with on-shelf-based time period and negative utility of items. Moreover, most algorithms need minimum utility threshold ( min_util ) to find rules. However, specifying the appropriate min_util threshold is a difficult problem for users. A smaller min_util threshold may generate too many rules and a higher one may generate a few rules, which can degrade performance. To address these issues, a novel top-k HOUIs mining algorithm named TKOS ( T op- K high O n- S helf utility itemsets miner) is proposed which considers on-shelf time period and negative utility. TKOS presents a novel branch and bound-based strategy to raise the internal min_util threshold efficiently. It also presents two pruning strategies to speed up the mining process. In order to reduce the dataset scanning cost, we utilize transaction merging and dataset projection techniques. Extensive experiments have been conducted on real and synthetic datasets having various characteristics. Experimental results show that the proposed algorithm outperforms the state-of-the-art algorithms. The proposed algorithm is up to 42 times faster and uses up-to 19 times less memory compared to the state-of-the-art KOSHU. Moreover, the proposed algorithm has excellent scalability in terms of time periods and the number of transactions.

The growing popularity of online social networks is quite evident nowadays and provides an opportunity to allow researchers in finding solutions for various practical applications. Link prediction is the technique of understanding network structure and identifying missing and future links in social networks. One of the well-known classes of methods in link prediction is a similarity-based method, which uses local and global topological information of the network to predict missing links. Some methods also exist based on quasi-local features to achieve a trade-off between local and global information on static networks. These quasi-local similarity-based methods are not best suited for considering community information in dynamic networks, failing to balance accuracy and efficiency. Therefore, a community-enhanced framework is presented in this article to predict missing links on dynamic social networks. First, a link prediction framework is presented to predict missing links using parameterized influence regions of nodes and their contribution in community partitions. Then, a unique feature set is generated using local, global, and quasi-local similarity-based as well as community information-based features. This feature set is further optimized using scoring-based feature selection methods to select only the most relevant features. Finally, four machine learning-based classification models are used for link prediction. The experiments are performed on six well-known dynamic networks and three performance metrics, and the results demonstrate that the proposed method outperforms the state-of-the-art methods.

Due to the increasing importance of graphs and graph streams in data representation in today's era, concept drift detection in graph streaming scenarios is more important than ever. Contributions to concept drift detection in graph streams are minimal and practically non-existent in the field of toxicology. This paper applied the discriminative subgraph-based drift detector (DSDD) to graph streams generated from real-world toxicology datasets. We used four toxicology datasets, each of which yielded two graph streams - one with abrupt drift points and one with gradual drift points. We used DSDD both with the standard minimum description length (MDL) heuristic and after replacing MDL with a much simpler heuristic SIZE (number of vertices + number of edges), and applied it to all generated graph streams containing abrupt drift points and gradual drift points for varying window sizes. Following that, we compared and analyzed the results. Finally, we applied a long short-term memory based graph stream classification model to all the generated streams and compared the difference in the performances obtained with and without detecting drift using DSDD. We believe that the results and analysis presented in this paper will provide insight into the task of concept drift detection in the toxicology domain and aid in the application of DSDD in a variety of scenarios.

Machine learning techniques are widely used for discovering meaningful patterns and classifying real-world data. These datasets may be large and complex, so feature selection is the primary strategy for reducing the dimension of the data, with the general goal of reducing the amount of redundant and disruptive features in a dataset for fast and efficient data analysis without sacrificing significant predictive model performance. Due to exponentially high search space, feature selection is a complex optimization problem. It is practically impossible to evaluate all of the feature subsets manually. Nature-inspired optimization is widely used for this due to its inherent capability, and it solves feature selection tasks as a single objective optimization problem. However, the main issue is their frequent premature convergence, which results in an inadequate contribution to data mining. Even the majority of existing optimizers are not adaptive in nature. As a result, in this paper, we proposed QL-SSA, which combines Reinforcement Learning and the Squirrel Search Algorithm, making it more adaptive and robust for feature selection by maintaining a good balance between exploration and exploitation steps. It is tested on 20 real-world benchmark datasets using two classifiers, and the results show that it outperforms the baseline optimizer in most cases.

Due to advancements in hardware capabilities and computation power, multi-objective optimization has recently been used in many industrial problems. These problems usually have multiple objectives of conflicting nature. To solve such problems, Multi-objective Evolutionary Algorithms (MOEAs) utilize Non-dominated Sorting (NDS) algorithms to rank the viability of the solutions efficiently. Researchers have focused on improving the time complexity of NDS algorithms for a long time due to their wide range of applications. With the increase in the use of GPUs for general-purpose scientific computing, it is now becoming possible to reduce the computation cost of such algorithms. In this paper, we have analyzed one such NDS algorithm, Corner Sort, and highlighted two areas within it with a high scope of parallelism. We propose a highly efficient, parallelized version of Corner Sort, implemented using CUDA framework. Utilizing the thousands of cores in a GPU, our algorithm is able to break the solution set into smaller chunks and simultaneously process them. Furthermore, the comparison between two solutions across all the objectives is done parallelly as well. On benchmark datasets, our algorithm performs up to 10x faster than the serial algorithm, and its performance improves for larger datasets, irrespective of the number of objectives.

Getting real-time information about the operational behavior of industrial or domestic appliances becomes effortless with the availability of sensors. The sensors generate multiple streams of measurements, called as multivariate time series, corresponding to an operation of the appliance. An appliance can go through various types of faults during its lifetime. Such a fault can be identified by classifying the multivariate time series (MTS), which is generated by the sensors corresponding to this fault. As it is also unfeasible to have prior knowledge about all types of faults, the classification approach should also be able to identify an unseen (unknown) fault using its MTS. In this article, we propose a semantic-information-based early classification approach for MTS. The approach uses a concept of zero-shot learning to classify an unseen fault. This work conducts a case study to evaluate the approach by classifying different faults of a washing machine using sensory data.

Activity classification has been an interesting area of research for many years, to better understand human behavior. Recent advancements in embedded computing systems allowed the emergence of several state-of-art solutions for human activity classification using sensors of a smartphone. The sensors generate temporal sequences of observations for human activity, which is called as Multivariate Time Series (MTS). Current state-of-art solutions for human activity classification suffer from two major limitations: first, the length of testing MTS should be equal to the training MTS and second, the MTS should not have any faulty time series. In real-time applications, it is desirable to classify a human activity using an incomplete MTS as early as possible. In this work, we propose a fault-tolerant early classification of MTS (FECM) approach to address these limitations. FECM builds a set of classification models using MTS training dataset. The approach employs Gaussian Process classifier to estimate minimum required length of time series, which is used to predict a class label of new MTS. Further, FECM uses an Auto Regressive Integrated Moving Average model to identify faulty time series in the new MTS. Finally, we conduct an experiment to evaluate the performance of FECM using accuracy and earliness metrics.

Nature-inspired optimization is one of the most prevalent research domains with a confounding history that fascinates the research communities. Particle Swarm Optimization is one of the well-known optimizers that belongs to the family of nature-inspired algorithms. It often suffers from premature convergence leading to a local optimum. To address this, several methods were presented using different network topologies of the particles, but either lacked accuracy or were slow. To solve these problems, an improved version of the Directed Weighted Complex Network Particle Swarm Optimization using the Genetic Algorithm (GDWCN-PSO) is presented. This method uses the concept of the Genetic Algorithm after each update to enhance convergence and diversity. Since most of the real-world applications and complex optimization problems involve more than one objective function so to suit this problem, a multiobjective version of GDWCN-PSO is also proposed and validated on standard benchmarks. To demonstrate its applicability in real-world applications, GDWCN-PSO is applied to solve the optimal key-based medical image encryption. It is one of the most challenging problems in health IoTs for protecting sensitive and confidential patient data as well as addressing the major concern of integrity and security of data in today’s advanced digital world.

Summary Online social networks play a pivotal role in the propagation of information and influence as in the form of word‐of‐mouth spreading. The influence maximization (IM) problem is a fundamental problem to identify a small set of individuals, which have a maximal influence spread in the social network. Unfortunately, the IM problem is NP‐hard. It has been depicted that a hill‐climbing greedy approach gives a good approximation guarantee. However, it is inefficient to run on large‐scale social networks. In this paper, a global influence evaluation function is presented for the IM optimization problem. The global influence evaluation function provides a reliable expected diffusion value of influence spread under the traditional diffusion models. To optimize global influence evaluation function, an influence maximization algorithm based on social spider optimization (IM‐SSO) is presented. IM‐SSO redefines the representation and update rule of spider's vibration and performs random walk towards target vibration. The algorithm uses a jump away process to overcome the weakness of premature convergence. The experimental results on six real‐world social networks show that the proposed algorithm is more effective than the state‐of‐the‐art heuristics and more time‐efficient than CELF++, static greedy, and PSO with an approximate influence spread.

Deep networks have escalated the computational performance in the sensor-based high dimensional imaging such as hyperspectral images (HSI), due to their informative feature extraction competency. Therefore in this work, we have extracted the informative features from different CNN models for the benchmark HSI datasets. The deep features have concatenated with spectral features to increase the informative knowledge in the image datacube. The feature concatenation has massively increased the size of datacube. Therefore, we have applied an unsupervised maximum object identification-based salient feature selection to identify the most informative features of datacube and discard the less informative features to reduce the computational time without compromising the accuracy. It is an unsupervised feature selection approach that transforms the data into scale space and achieved robust and strong features. In the previous CNN-based methods, raw features have directly fed to the MLP (multilayer perception) layers for target prediction whereas we have provided our salient features into a multi-core SVM-based set-up and have achieved high accuracy with low computational time as compared to the previous state-of-art techniques.

Supervised classification and target recognition of Hyperspectral images (HSI) is a challenging task due to high dimensionality and spectral mixing. Straightforward cognitive computation and target classification lead to high computation cost and low recognition accuracy. Limited availability of training samples makes the recognition process very slow and inaccurate. The main purpose of this work is to improve the classification accuracy for high-dimensional images by the fusion of posterior probability obtained from the two-stage probabilistic framework. The first stage addresses the issue of high dimensionality and the second stage addresses the spectral mixing problem. Both stages provide the prediction probability of pixels in a particular class. In stage-1, we have addressed the imbalance between dimensionality and training samples problem for which we have integrated the deep CNN based spatial and spectral features in combined data-cube form, using ‘off-the-shelf’ CNN models. Subsequently, a graph-based non-linear manifold embedding has performed to extract and fuse the region-wise external information. A probability of prediction has obtained by using LDA classifier. These probabilistic values have denoted as a global probability, as an outcome of stage-1. In the stage-2, the spectral mixing issue was addressed by computing the regional probabilities of class mixing for each pixel. The regional probabilities have calculated by using a regional subspace regression approach. Subsequently, the probabilistic output, obtained from stage-1 and stage-2, has been combined with a linear decision fusion method using regularizers. The experiments have conducted on three real Hyperspectral images, i.e. Indian pines (IP), Pavia University (PU), Salinas Valley (SV) datasets. The probabilistic fusion of stage-1 and stage-2 yields to the maximum overall accuracy of 97.38%, 95.10%, and 99.88% for IP, PU and SV datasets. The over-all accuracies have compared with past methods, and it has found that the proposed framework is providing higher prediction accuracies than previous state-of-art methods.

In the era of the Internet of Things (IoT), the sensor-based devices produce the Multivariate Time Series (MTS). A classification approach helps to predict the class label of an incoming MTS. Due to the large dimension and different sampling rate of the sensors in a given MTS, a classifier takes time to predict the class label. Some IoT applications may require early prediction of the class label where the classifier starts the prediction once the minimum number of data points are collected. In this article, we address the problem of early prediction of the class label of an MTS in IoT. This work considers the sensors with different sampling rate to generate the MTS. Each sensor generates a time series (component) of the MTS. We propose a Divide-and-Conquer–based early classification approach for classifying such MTS. The approach constructs an ensemble classifier using a probabilistic classifier and hierarchical clustering. The ensemble classifier employs a Divide-and-Conquer method to handle the different sampling rate components during the prediction of class label. The experimental results show that our approach significantly outperforms the existing approaches on real-world datasets using various evaluation metrics.

An important issue of research in the transportation system is timely classification of the inside-outside environment of the vehicle using sensors. The sensors generate the multivariate time series data, which requires a classification technique to classify it in real-time. Road surface classification is an example, where multivariate time series data can be used for early identification of the type of road surface. The challenge is to maintain the accuracy of the classification using a minimum number of data points of the multivariate time series. This work proposes an early classification approach for multivariate time series with a desired level of accuracy. It is assumed that the number of samples in the time series are not equal for a given period of time due to different type of sensors. Gaussian Process learning method is used to first estimate the minimum required length of the time series which helps to build an ensemble classifier with a desired level of accuracy. The ensemble classifier is used to predict the class label of an incoming multivariate time series. This work demonstrates a road surface classification system using the built ensemble classifier. Finally, the ensemble classifier is also evaluated on the various existing datasets from other domains. The results demonstrate the significance of early classification approach using accuracy, earliness, and confusion matrix, with the minimum required data points.

Abstract High‐utility itemset (HUI) mining is an important tasks during data mining. Recently, many algorithms have been proposed to discover HUIs. Most of the algorithms work only for itemsets with positive utility values. However, in the real world, items are found with both positive and negative utility values. To address this issue, we propose an algorithm named Efficient High‐utility Itemsets mining with Negative utility (EHIN) to find all HUIs with negative utility. EHIN utilises 2 new upper bounds for pruning, named revised subtree and revised local utility. To reduce dataset scans, the proposed algorithm uses transaction merging and dataset projection techniques. An array‐based utility‐counting technique is also utilised to calculate upper‐bound efficiently. EHIN utilises various properties and pruning strategies to mine HUIs with negative utility. The experimental results show that the proposed algorithm is 28 times faster, and it consumes up to 10 times less memory than the state‐of‐the‐art algorithm FHN. Moreover, a key advantage is that EHIN always performs better for dense datasets.

In this paper, a fuzzy agglomerative (FuzAg) approach is proposed for community detection that iteratively updates membership degree of nodes. Earlier approaches assign membership degree to nodes based on communities only. We introduce the notion of self-membership in addition to the membership of different communities. The essence of self-membership is to give opportunity to all nodes in growing their own community. Nodes having higher self-membership degree are referred as anchors, and they get a chance to expand their associated community. Meanwhile, some new anchors may emerge in successive iterations, whereas false or redundant anchors get removed. The time complexity of the proposed algorithm is shown to be O(n2). We compare the results of the proposed FuzAg algorithm with those of state-of-the-art fuzzy community detection algorithms on ten real-world datasets as well as on synthetic networks. Results indicated by various quality and accuracy metrics show impressive performance of FuzAg in identifying both disjoint communities and fuzzy communities.