Search

In modern academic social network, it is very difficult for scholars to find academic papers consistent with their research direction. Time is a critical factor in paper recommendation. As time goes on, the impact of an academic paper would gradually fade. Likewise, the research interests of users may also change. Therefore, we propose a temporal perceptual neural collaborative filtering model that integrates the multi-dimensional features of papers. We conducted our experiments on the dataset from CiteULike, comparing the recommended results by using four time-decay functions and evaluating our model with multiple evaluation indicators. The satisfactory results show that our model is effective in filtering out the expired papers by considering the characteristics of papers and the changes of scholars' interests.

Due to the increasingly growing demand for user identification on cell phones, PCs, laptops, and so on, face anti-spoofing has risen to significance and is an active research area in academia and industry. The detection of the real face then recognize it present an important challenge regarding the techniques that can be used to spoof any recognition system like masks, printed photos. This paper we present an anti-spoofing face method to solve the real-world scenario that learns the target domain classifier based on samples used for training in a particular source domain. Specifically, with the conventional regression CNN, the Spatial/Channel-wise Attention Modules were introduced. Two modules, namely the Spatial-wise Attention Module and the Channel-wise Attention Module, were used at spatial and channel levels to improve local features and ignore the irrelevant features. Extensive experiments on current collections with benchmarks datasets verifies that the recommended solution will significantly benefit from the two modules and better generalization capability by providing significantly improved results in anti-spoofing.

Drowsiness can be dangerous when performing tasks that require constant attention, such as driving a vehicle. Sleepiness is correlated with a variety of physiological variables, such as eye closing, head movements, pulse rate, eye twitch rate, etc. Also, the yawn can be considered as an accurate indicator of drowsiness and fatigue. Yawning detection is very important for the safety purpose of drivers as it will let the driver know if he/she is getting drowsy. Driving at that moment may not be safe. Several automatic yawning detection techniques have been developed for driver's drowsiness monitoring system. Nevertheless, correctly detecting the yawning of the driver and predicting exhaustion in real-time situations is still a crucial challenge. In this paper, we will be reviewing various existing machine learning approaches for driver's yawning detection. In previous approaches, various classical machine learning algorithms such as viola-Jones, contour activation algorithm and SVM have been used for yawning detection, but these approaches failed to predict yawning in realtime situations. Using Deep learning techniques, we can make a real-time yawn detection system with high accuracy. We find that some precious Deep learning algorithms like CNN, RNN, LSTM, Bi-LSTM can detect the patterns with high accuracy. After the comparison of various algorithms and techniques, we find that with the help of Deep learning algorithms the yawning can be detected in real time with high accuracy.

Human recogntion technologies are gaining significant research attention, where the model can be trained to be more precise to recognize the poses performed by objects. Activity identification is a kind of problem, which needs more research consideration and improvement. There is also an increasing need to detect different poses of objects. To tackle the issue, different sensors like Gyroscope and Accelerometer are required to classify the data in the form of images using machine learning algorithms like SVM and CNN. These approaches help us in implementing real-time applicaions such as health monitoring, tracking sports activity and security. The paper also discusses about its benefits, limitations and prominent approach for human activity recognition.

In autonomous driving, 3D object detection is es-sential as it provides basic knowledge about the environment. However, as deep learning based 3D detection methods are usually computation intensive, it is challenging to support realtime 3D object detection on edge-computing devices in selfdriving cars with limited computation and memory resources. To facilitate this, we propose a compiler-aware pruning search framework, to achieve real-time inference of 3D object detection on the resource-limited mobile devices. Specifically, a generator is applied to sample better pruning proposals in the search space based on current proposals with their performance, and an evaluator is adopted to evaluate the sampled pruning proposal performance. To accelerate the search, the evaluator employs Bayesian optimization with an ensemble of neural predictors. We demonstrate in experiments that for the first time, the pruning search framework can achieve real-time 3D object detection on mobile (Samsung Galaxy S20 phone) with state-of-the-art detection performance.

Graph neural networks (GNN) have achieved state-of-the-art performance on various industrial tasks. However, the poor efficiency of GNN inference and frequent Out-of-Memory (OOM) problem limit the successful application of GNN on edge computing platforms. To tackle these problems, a feature decomposition approach is proposed for memory efficiency optimization of GNN inference. The proposed approach could achieve outstanding optimization on various GNN models, covering a wide range of datasets, which speeds up the inference by up to 3×. Furthermore, the proposed feature decomposition could significantly reduce the peak memory usage (up to 5× in memory efficiency improvement) and mitigate OOM problems during GNN inference.

It is difficult to evaluate a person’s ability or performances because there is no common yardstick for measuring individual human abilities. For example, in general, paper based tests have different characteristics (difficulties, accuracies, and so on) because they depend on items of each test. Recently, e-testing technologies from AI approach, for which each test form has equivalent measurement accuracy but with a different set of items, have become popular. Even if different examinees with the same ability take different tests, their test scores should be guaranteed to be equivalent. We introduce state-of the-art e-testing technologies from AI approach.

the real datasets in the world usually are imbalanced; the number of samples for their classes is not equal. Classifying these datasets makes the classifiers pay attention to the class with more samples than the classes with fewer samples. The Qur’anic dataset can be considered an imbalanced dataset because verses of the Qur’anic topics are not equal. Many studies have been performed to classify Qur’anic text using different classifiers. However, few studies classified the Qur’anic verses based on Imbalanced Learning (IL). So, this work aims to classify the Qur’anic text using Ensemble methods, Boosting and Bagging. The base classifiers of these methods were LibSVM, Naive Bayes, KNN, and J48. Three techniques are conducted in this paper based on the standard classifiers. The three techniques are: implementing the base classifiers alone, implementing these classifiers with the Boosting method, and implementing the classifiers with the Bagging method. The results showed that the Quranic classification performance was improved when the ensemble methods were applied for the imbalanced Qur’anic verses in the standard classifiers.

Automotive radar sensors offer a promising and effective modality for perception and assessment of the surrounding environment. A key element of environment sensing in automotive radars is the reliable detection, classification and tracking of vulnerable road users such as pedestrians and cyclists. In this paper, we propose an integrated Bayesian framework dubbed BayesRadar, which improves the overall classification accuracy by tracking the embedding vector of a neural network and its prediction uncertainty via recursive Kalman filtering over time. Apart from the classification accuracy of a model, a critical measure includes the analysis of statistical confidence over the target class score. Such measures for predicting the true correctness likelihood of the classification estimates are essential in safety-critical automotive applications. Therefore, in this paper, we present and evaluate the classification, embedding cluster score and statistical confidence performance of the proposed framework in the context of classifying vulnerable road users compared to state-of-art deep learning approaches. Furthermore, we demonstrate superior performance of the BayesRadar for unseen classes compared to long short-term memory based temporal tracking of the embedding vectors.

Power electronic devices such as Insulated Gate Bipolar Transistors (IGBT) are widely used in electronic systems. The failure of IGBT may bring fatal damage to the whole system. The ability to accurately predict the remaining useful life (RUL) of IGBT can provide a basis for intelligent equipment maintenance, thereby minimizing the risk of catastrophe failure at work. The research in this field aims to achieve accurate and timely RUL prediction of equipment. This paper exploited the IGBT accelerated aging data from NASA Ames Research Center, and the monitoring data of IGBT is collected under thermal overstress load condition with square signal bias at gate. This paper developed three machine learning models to predict the RUL of IGBT, including Back Propagation Neural Network(BPNN), Random Forest and Extreme Learning Machine. These models realize real-time prediction of RUL, and the performance of these models is observed and analyzed.

Classification of soil is crucial for the agricultural domain as it is an essential task in geology and engineering domains. Various procedures are proposed to classify soil types in the literature, but many of them consumed much time or required specially designed equipments/applications. Classification of soil involves the accounting of various factors due to its diversified nature. It can be observed that several critical domain-oriented decisions often depend on the type of soil like farmers might be benefitted from knowing the kind of soil to choose crops accordingly for cultivation. We have employed different Convolution Neural Network (CNN) architectures to identify the soil type accurately in real-time. This paper describes the comparative evaluation in terms of performances of various CNN architectures, namely, ResNet50, VGG19, MobileNetV2, VGG16, NASNetMobile, and InceptionV3. These CNN models are used to classify four types of soils: Clay, Black, Alluvial, and Red. The performance of the ResNet50 model is the best with a training accuracy and training loss of 99.47% and 0.0252, respectively compared to other competing models considered in this paper.

Every year, doctors diagnose skin cancer in around 3 million or more patients across the globe. Currently, it is one of the most widely recognized kinds of cancers for human health. Hence, we need an early diagnosis to prevail any critical condition of the infected patients. Apparently, it can treat with topical drugs, if it diagnoses in an early stage. Hence as an outcome, skin cancer is responsible for less than 1% of all cancer deaths. There are two types of tumors in the skin cancer diseases domain, such as benign and malignant. To develop a robust and early screening system to diagnose skin cancer, it requires an efficient algorithm for prediction, trained with a large dataset. The primary aim of this research is to develop an efficient skin cancer screening process using a robust deep neural network with a large dataset. In this paper, we intend to determine considerate and dangerous types of skin cancer tumors using dermoscopic images from a publicly available dataset. We proposed an efficient and fast scaled 2D-CNN based on EfficientNet-B7 deep neural architecture with image preprocessing. This paper also uses two different pre-trained deep neural architectures, such as VGG19, and ResNet-50 to compare the performance with the proposed architecture. The proposed architecture outperformed the other pre-trained CNN models whereas the proposed architecture achieved higher AUC and accuracy compared to other architectures.

This research paper presents the importance of cardiologists and their availability to the public by the usage of various smart devices. These new simplified ECGs may be as complex or sophisticated as those that represent the usage of the medical facility however they are accurately able to observe the heart rate to monitor their health. One of the most commonly overlooked problems in the modern world is Arrhythmia. This is a problem related to the rhythm of the heartbeat and it occurs due to the improper coordination between the electrical impulses with your heartbeat. Heartbeat occurs daily in the lives of several people as they aren’t properly focused upon the monitoring of heartbeat. This research paper attempts to focus on the issue of Heart Arrhythmia as well as the creation and classification model which is capable of identifying the type of heart arrhythmia by an individual how may be suffering from heart issues. To successfully create our model, this research work utilizes the convolution neural networks to train the proposed model with existing ECG data and properly identify the heart rate as well as the type of arrhythmia that they have so that it provides the ability to immediately provide the proper medical attention.

Breast cancer is becoming one of the leading causes of mortality in women all over the world. The advanced engineering of Artificial Intelligence techniques and natural image classification approaches is mostly employed to classify breast images. The image classification permits the physicians and the doctor the next chance, and this, in turn, saves the time of the physician and the doctor. Apart from several publications on classifying breast images, only some of the review papers are present that offer a clear definition of the “approaches to breast cancer image classification, feature extraction and selection methods, classification measurement parameterizations, and image classification results” with the help of 3D images. Machine learning (ML) seems to be an important portion of medical imaging research. These are developed over the years from manual seeded inputs to automatic initializations. The enhancements in the area of ML lead to various self-reliant and intelligent "Computer-Aided Diagnosis (CAD) systems" since the learning capability of the ML techniques is increasing in a constant manner. Several automated techniques are developing with deep feature representations and learning. The modern enhancements of ML with extensive and deeper representation techniques called Deep Learning (DL) techniques have created an important effect on enhancing the diagnosing ability of these systems. The main objective is to undergo a critical review on diverse contributions related to breast cancer detection that could be highly advantageous for dealing with 3D mammogram images. This paper plans to make a review on different algorithmic categorizations, datasets used, simulation platforms, and performance metrics. Furthermore, the research gaps and challenges of the breast cancer diagnosis using 3D images are elaborated, encouraging the upcoming researchers to give tremendous concern on new contributions.

In the light of the rapidly growing COVID-19 pandemic, the need for an expeditious diagnosis of COVID-19 infection became essential. The immediate diagnosis will allow the initiation of the isolation process and adequate treatment as well. While the standard test used for the diagnosis of COVID-19 disease (RT-PCR) is usually time consuming (6 hours up to days in some centers); the need for a highly sensitive test became essential. Many studies have illustrated the utility of chest CT scan in the diagnoses of COVID-19. This paper evaluates the value of classical machine learning techniques and the convolutional neural networks in aiding physicians to further classify patients into either COVID-19 positive or negative according to their chest CT findings, and thus facilitating their work. To address this problem, this paper proposes classical neural networks using statistical features and deep CNN models to further classify a dataset of preprocessed chest CT images, using several classifiers and to evaluate the results. This latter showed that the best proposed method was a four layers CNN with SVM classifier with 99.6% accuracy. This demonstrates the potential of the proposed technique in computer-aided diagnosis for healthcare applications, especially for COVID-19 classification.

State-of-the art design methods for a stacker crane rail/wheel system use unrealistic load collectives. Therefore, the results of these design methods are often inaccurate. The BS EN 13001-3-3 introduces a more realistic load collective, called the contact force spectrum factor. However, in practice the contact force spectrum factor is difficult to compute, since all contact forces over the considered stacker crane's life cycle need to be known. One way to overcome this obstacle is to run a simulation. However, this approach is computationally expensive. Therefore, this paper introduces machine learning models for predicting the contact force spectrum factor. For data generation a simulation model is used. The presented machine learning models are compared to the standard approach for computing load collectives. In comparison to this standard approach the machine learning models' predictions are significantly more accurate. The presented results make it possible to predict realistic load collectives for the rail/wheel system of stacker cranes. Hence, with this paper's results the efficient application of the BS EN 13001-3-3 for stacker cranes becomes possible.

Speech Recognition Software is a computer program that is trained to take the input of human speech, interpret it, and transcribe it into text. Most recently, the field has benefited from advances in deep learning and big data. The advances are evidenced not only by the surge of academic papers published in the field, but more importantly by the worldwide industry adoption of a variety of deep learning methods in designing and deploying speech recognition systems. The objective of this paper is to propose an advanced and accurate end-user software system that is able to recognize specific commands to control a robot to perform specified tasks in a hospital. This model will be based on Deep Learning since it is effective in models having huge data as for the two versions of Google TensorFlow and AIY datasets used in our model. Convolutional neural network will be used since it is able to extract features from the dataset instead of traditional methods of feature extraction, thus saving training time and reducing the complexity of the system. With addition to that, NVIDIA CUDA will be also used to train the model with GPU to decrease the training time. During training, some experiments have been done to see the effect of some parameters on the results of the system, and to make sure that the chosen parameters in our model are the best. The results indicate that the training, validation, and testing accuracies of the proposed approach were high, the training duration reached very low values due to the innovation used (CUDA Toolkit) and the commands were successfully recognized by the model. These results outcome the results of the papers that developed similar work which will be presented in the coming sections.

The content of multi-modal medical features representing medical data was researched and many modes of learning were analyzed for the risk assessment of diseases in order to gather information from medical health data and generate intelligent application relevant inquiries. The representation of medical data employs the method of deep learning. To attain model adaptability, the study and extraction of different disease features employ the same procedure. Classification of medical data is an important data mining challenge that has interested various academics throughout the world for a decade. Various data mining algorithms are utilized for health data sorting such as classification, clustering, regression, soft computing and more. Feature selection algorithms play a key part in every challenge in machine learning. The selection of the best method produces an optimal subset of properties, which enhances accuracy and reduces training time. It is also beneficial to remove the useless characteristics for high-dimensional datasets. This paper provides a brief review on medical data classification models to provide a brief assessment of the common feature selection and classification models specifically utilized for medical data classification. Various aspects, such as the medical data sets available, selection of features, choice of classification, problems in the identification and analysis of the key methods of selecting features and detailed mechanisms are represented in this survey paper.

Predicting energy (fuel and electric) consumption of hybrid vehicles is important on different levels: vehicle industry as a whole, individuals, and can also pave the way towards a more sustainable future. Despite its importance, providing accurate predictions is quite a challenging task. Many essential factors impacting energy consumption, including travel time, average speed, etc., needless to say, these features are not available beforehand. However, these factors are available in our data-set. To use these factors effectively, in this paper, we propose including them as different tasks in a multitask setting to help our main problem of energy consumption. The promise of this approach is that since these tasks are relevant, learning them together would provide a common feature space sharing information about all tasks. More importantly, this shared feature space would carry important information helping energy consumption in particular. In multitask learning, two important issues are task dominance and conflicting gradients of different tasks. Different studies have addressed these two separately. In this paper, we propose a method tackling these two problems simultaneously. We show experimentally the success of this method in comparison to state-of-the-art.

Metrics are the key tools to monitor the safety performance of complex systems like air traffic control. Developing risk metrics is a complex process, which involves prediction of risk and identification of different potential outcomes of undesired safety events. For instance, a metric that helps monitor the risk on the surface in an airport environment needs to detect and identify events such as runway excursions, runway incursions, and taxiway incidents and assign appropriate numerical indices proportional to the outcome of each event. An accident with a fatality, injury and aircraft damage will have a corresponding weight for each outcome and incident that involves no outcome can be assigned a severity weight based on its probability of becoming an accident. Models that support such metrics needs to be able to process data from diverse sources. In this paper, we discuss key aspects of two comprehensive metrics that the office of Safety in the Air Traffic Organization has deployed recently: 1) How to employ an automation to detect all relevant events from different data sources and 2) how to assign severity weights for different types of events (accidents and incidents).1. Detection of Relevant Events: one key element of supporting a comprehensive metrics is an automated detection of relevant events by categories. Unlike in aviation, the use of AI and ML have permeated in most industries. Aviation is a safety-critical domain in which there is very little tolerance for failures. The stringent requirements of aviation have contributed to the slow adoption of AI and ML in aviation, in general. However, many aviation sectors are increasingly adopting AI systems, ranging from automating simple, yet tedious and repetitive tasks to a more sophisticated application of a complex autonomous air traffic control system to de-conflict traffic. In this paper, we outline how machine learning models were employed to identify relevant accidents and incidents to support two metrics, a surface safety metric and an airborne safety metric.2. Severity Weighting Scheme: for a metric to be a comprehensive measure and indicate the overall performance of the system, it needs to account for various types of accidents and incidents (precursors) that occur in the system. This paper shows how a weighting scheme we developed to measure the outcome of accidents, such as injuries to people and damage to property, as well as probabilistically determine the severity of incidents with no outcome. The aggregation of all weights along with the frequency of occurrence in a given period represent the overall safety performance of the system.

Under the light of increasing air traffic congestion and reduced environmental impacts, the growing need for tailored approaches has arisen. To coordinate tailored approaches, accurate predictions of descent trajectories must be available. The ability of ground based automation for such endeavours however must be improved. Specifically in the ability to predict where descents begin, also referred to as the Top of descent (TOD). This paper attempts to provide a machine learning approach to TOD location prediction, considering the operational capabilities of modern-day air traffic management by working within the confines of parameters available to controllers and considerations of stakeholders. To identify the common TOD clusters and the resultant average TOD locations representing the controllers macroscopic perspective, an unsupervised learning approach is utilised. Specifically, an ensemble method of DBSCAN-K-means is used. This was compared to the more microscopic pilot controllers perspective and generic decision-making formulas to determine the proximity to expected TOD. Of which the microscopic option has proven to be more reliable in representing the individual Flight Management System (FMS) recommendations or pilots’ decisions on aircraft. In the prediction of TOD location, operationally available factors such as proximity to expected TOD, current flight level, approach heading, ground speed, fleet and aircraft type are used to train a decision tree model. A Singapore flight information region (FIR) case study is conducted to illustrate the methodology. Results of the methodology employed on the inbound A380 fleet consisting a total of 79 flights, yielded a maximum prediction range of 31NM range from actual TOD occurrence. This paper has demonstrates the possibility of using a machine learning approach to predict TOD of a flight considering both prior cruise characteristics and operational considerations of pilots. Accepted version

Multi-airport, metroplex airspace environments present a challenging air traffic sequencing and scheduling problem, especially during convective weather when the capacity of the metroplex airspace and airports is degraded. In such situations, departure flights often experience long airport surface (taxi) delays waiting in queues for open slots at convective weather-impacted overhead departure fixes and departure routes. During these situations, the available metroplex capacity can be better utilized by rerouting departure flights from overloaded or closed departure fixes/routes to less busy and open fixes/routes. However, when multiple fixes or routes are impacted by convective weather, there could be several candidate departure rerouting strategies that the FAA and airlines can collaboratively adopt to maximally utilize available capacity, and it may not be immediately clear which strategy will work the best. This paper outlines an ATAC-NASA collaborative research effort that developed a machine learning-based Air Traffic Management (ATM) decision support service called the Capacity Finder. The Capacity Finder provides data-driven decision support for determining the departure rerouting strategy that has the best chance of success given the current and predicted demand and capacity situation. The Capacity Finder makes this determination by first identifying historical operational time-periods (i.e., departure banks) that are similar to the time-period under consideration. Similarity is determined by applying machine learning (clustering) to pre-conditioned data sets that include several traffic demand and capacity-related features. Once the similar departure banks are identified, the Capacity Finder identifies the rerouting strategies that worked well during similar historical situations and suggests them to the airline and FAA users. This paper presents preliminary results on machine learning-based similarity identification for departure banks at the Dallas Fort Worth International Airport (DFW).

Normally, the long queues and crowd can be seen at the canteens/hotels whenever there is a lunch break in an organization or within a campus. This paper proposes a solution for eliminating the queue system and introduces the facility to remotely place food orders. Further, this paper proposes a real-time food recommendation system to suggest the dishes to users based on their past orders. The solution has been implemented through a mobile application built using Flutter. The mobile application has been empowered with a machine learning model for recommending the items that the user might like. The proposed method has been tested with 2 vendors catering to 60 customers successfully. The accuracy of the recommendation has been satisfactory.

The Publisher's final version can be found by following the DOI link. Data quality plays a crucial role in the context of smart buildings. Meanwhile, missing data is relatively common in acquired datasets from sensors within the smart buildings. Poor data could result in a big bias in forecasting, control and operational services. Despite the common techniques to handle missing data, it is essential to systematically select the most appropriate approach for such missing values. This paper aims to focus on the lift systems as one of the essential parts in the smart buildings by exploring the most appropriate data imputation methods to handle missing data and to provide its service and allow a better understanding of patterns to issue the correct control actions based on forecasted models. The imputed data is not only investigated statistically but also modelled through machine learning algorithm to explore the impact of selecting inappropriate imputation techniques. Seven imputation techniques deployed on datasets with three level of missing values including 10%, 20% and 30% and the performance of methods examined through the normalized root mean square error (NRMSE) approach. In addition, the interaction between imputation techniques and a machine learning algorithm, namely random forest were examined. Findings from this paper can be employed in identifying an appropriate imputation technique not only within the lift datasets, but smart building context.

Our paper aims to build a classification model to discern the typical NEAT (Non-Exercise Activity Thermogenesis) activities done in a home setting. The concept of NEAT is broadly defined as the energy spent in everything which is not sleeping, eating, or a traditional form of physical exercise. We focus on the following NEAT and non-NEAT activities in this paper - cooking, sweeping, mopping, walking, climbing up, climbing down, and non-NEAT activities (e.g., watching television and working on a desk). This aim is to build a classification model which can work with data sampled at a low frequency of 1Hz. However, building such a classifier is non-trivial because the NEAT activities are not easily separable in low-frequency data. The state-of-the-art in the area of human activity recognition either uses multiple physical devices (e.g., accelerometers on arms, waist, and feet) for data collection or use data that is sampled at high frequency (20Hz or above). In contrast, our model performs NEAT activity recognition using data sampled at 1Hz and from a single smartwatch worn on the dominant hand. Thus, making it more energy-efficient and easily usable for widespread use. We evaluate our proposed model using actual data collected on a smartwatch, and we compare it with alternative models. Our results indicate that the proposed model is able to achieve much higher accuracy than the alternative approaches.

Natural experiments are observational studies where the assignment of treatment conditions to different populations occurs by chance "in the wild". Researchers from fields such as economics, healthcare, and the social sciences leverage natural experiments to conduct hypothesis testing and causal effect estimation for treatment and outcome variables that would otherwise be costly, infeasible, or unethical. In this paper, we introduce VAINE (Visualization and AI for Natural Experiments), a visual analytics tool for identifying and understanding natural experiments from observational data. We then demonstrate how VAINE can be used to validate causal relationships, estimate average treatment effects, and identify statistical phenomena such as Simpson's paradox through two usage scenarios., 5 pages, 4 figures, accepted as short paper at IEEE VIS 2021

XSS attacks have become more prevalent in last few decades and thus more challenging to detect their existence. XSS attacks are broadly classified into two categories: server-based XSS attack and client-based XSS attack. Although a lot of research has already been done in this area, still the methods lack in precision and accuracy as per the literature survey. There are ample of methodologies being applied in the detection of XSS attacks using supervised learning, unsupervised learning, reinforcement learning, deep learning and metaheuristic algorithms. We present a survey of the recent approaches being applied by the numerous researchers in their proposed models. Following indexed journals were used for research papers’ collection in order to carry out a survey: Elsevier, Springer, IEEE explore, Hindawi, google scholar, and Web of Science. Moreover, in this paper, we introduce a classification chart of several machine learning algorithms that can be applied to the web-attack detection model.

The dominant causes of visual impairment worldwide are Cataract, Glaucoma, and retinal diseases among patients. The alarming cases of these diseases call for an urgent intervention by early diagnosis. The proposed system is designed and developed to easily facilitate the detection of cataract, glaucoma and retinal diseases among patients. The Logistic Regression, Random Forest, Gradient Boosting and Support Vector Machine algorithms are used for detection. The proposed system will help people to get the proper treatment of the aforementioned diseases at an early stage thus reducing the percentage of blindness being caused. The proposed system evaluates the effectiveness and safety of cataract surgery in eyes with age-related degeneration along with glaucoma and retinal diseases detection. This paper shows the accuracy of algorithms and SVM classifiers based upon the glaucoma, retina, cataract and normal eye’s fundus images. The idea of classifying the images based on its fundus and extracting features is widely known now-a-days and also it plays a vital role in the final outcome. This paper talks about the multiclass built models of these classifiers and on the basis of the ROC curves plotted it predicts the output of the images. As far as the algorithms are concerned, the efficiency of algorithms helps it stand best out of many and in our case Gradient boosting proves to give best results for the eye with cataract with 90% accuracy. Then the supervised algorithms logistic regression and random forest gives the accuracy of 89% and 86% respectively.

This paper studies the problem of using machine learning to predict the indicator diagram of pumping unit according to the power of pumping unit in a period of time. In the past, it was troublesome to install a machine to collect indicator diagram data or indirectly collect the size of pumping units. This paper introduces several methods of indirectly predicting indicator diagram based on power. They are two nearest neighbor methods and three DNN methods. Finally, through experimental comparison, it is proved that the DNN network model with embedding layers used to distinguish different pumping unit specifications has better prediction index diagram ability, and its prediction index diagram is basically consistent with the original index diagram, which can replace the traditional method.

This paper investigates the effect of offloading machine learning autonomous vehicle algorithms to the edge of 5G networks. It describes the implementation of a field test experiment on the Canadian ENCQOR 5G network using edge computing to command a 1/10th scaled RC vehicle. The vehicle uses a camera feed to predict steering. Predictions are computed on a server on the edge of the network which contains a pre-trained neural network. It was found that the autonomous vehicle performance are very similar with and without offloading. The paper discusses the limits of this result and provides an overview of future opportunities and obstacles in the use of offloading machine learning functionalities for autonomous vehicles in 5G.

Generating a caption from an image is a challenging task that involves combining the concepts of computer vision and natural language processing which has motivated the researchers to model an automated process and in recent times multitude of approaches have been introduced which specifically target this problem. This paper solves the problem of learning the relationship between image and the language by establishing an entity-object relationship that maintains continuity in the caption. This is achieved by using an attention mechanism which allows the decoder to target specific regions at different time steps. To understand the image content, we use an encoder which extracts information about the different objects and the relationships between them. This paper evaluates and compares two models that incorporate two different encoders i.e., InceptionResnetV2 and InceptionV3, both having attention unit along with LSTM as the decoder. Experiments on the MS-COCO dataset show that the models generate sensible and accurate captions in a majority of cases.

Over the past few years, social media platforms like Twitter, Facebook, Reddit, Instagram, etc. have become a medium for one to share their feelings and experiences. This stands as a significant opportunity for the research community to utilize such data for early diagnosis of psychological issues like “depression” (a mood disorder marked by a sense of despair and detachment). This paper emphasizes comparing the performance of various machine learning models on well-labeled twitter data and deals with a class imbalance problem. This paper outlines various baseline feature engineering methods with different machine learning techniques. Evaluation metrics such as f1-score, precision, and recall have been used to compare the performance of these models. This research work yields better results using Bag of Words with AdaBoost Classifier compared to other machine learning classifiers.

Photovoltaic solar energy requires novel algorithms to ensure suitable maintenance management. Supervisory control and data acquisition system, combined with machine learning techniques, is required to obtain reliable information about the real state of photovoltaic systems. This paper introduces an Internet of Things platform for photovoltaic maintenance management based on classification algorithms to detect patterns, where performance ratio decreases significantly in time series. A real case study is presented with SCADA data from a photovoltaic solar plant located in Spain. The classification algorithms employed are Shapelets and K-nearest neighbors. The results prove the robust performance of both algorithms in pattern recognition, whereas K-nearest neighbors is preferable for implementation on the Internet of Things platform due to the reduced execution time. The application of the platform developed in this paper improve photovoltaic maintenance management detecting performance ratio reductions.

Medical data mining is known as methods of extracting data from human services databases to help clinicians get the best diagnosis. In this field, and among cancer diseases, breast cancer is the highest deadly disease in the world in recent years. Therefore, data mining techniques will be the largest part used in this study. To complement the previous research related to breast cancer detection, this paper proposed a model to help solve the difficulty of determining the degree of risk of the disease and obtain the best results. Aiming of reducing the costs and time used in diagnosing the disease too. The experiment used a dataset computed from a digitized fine needle aspirate (FNA) image of a breast mass dataset; diagnostic (WBCD) available in UCI machine learning repository. The model is the application of classification techniques to the breast cancer data collected, which, in turn, predicts the severity of a patient's breast cancer. In addition, this paper classifies and diagnose cancers using Deep-learning algorithms only, hybrid Machine learning algorithms and feature selection methods (applied correlative rules to find out what traits related to breast cancer severity).

This paper examines the problem of parallelizing neural network training using the back-propagation neural network, as a breakthrough example in the field of deep learning. The challenge of our solution is to twist the algorithm in such a way so it can be executed in parallel, rather than sequentially. In this paper, we test validity of a research hypothesis if the speed can be increased by parallelizing the back-propagation algorithm and keep the same accuracy. For this purpose, we developed a use-case of a handwriting recognition algorithm and conducted several experiments to test the performance, both in execution speed and accuracy. At the end, we examine just how much it benefits to write a parallel program for a neural network, with regards to the time it takes to train the neural network and the accuracy of the predictions. Our handwriting problem is that of classification, and in order to implement any sort of solution, we must have data. The MNIST dataset of handwritten digits provides necessary data to solve the problem.

The effects of the communication of physical and virtual sensors on the performance of the mobile network can be assessed through several Key Performance Indicators (KPI), of which the Average Downlink Throughput (ADT) is very important. The paper presents an approach for measuring this indicator in a mobile network using a Machine Learning Model (MLM) which uses contextual variables related to traffic class, user location, time period, signal strength, base station. The geographical area of the m:tel network on the section of the “9th January” motorway, a key road in the Republic of Srpska, BiH, was selected for the case study in the research. The measured ADT performance reflects the degree of Quality of customer Service (QoS) in the observed space. The proposed model was created in the SPSS software package, and its training and testing were performed based on data collected by field measurements. The research results show that the model can perform measurement and prediction with satisfactory accuracy while increasing the data set for training and testing by the Data Augmentation (DA) method. In addition to the MLM model, the paper proposes a wireless sensor network model for the acquisition of contextual data along the motorway.

With the emergence of the smart grid and smart city trends, smart buildings are gaining attention. Energy management is a key component of a smart building management system, and it necessitates precise forecasting of the building's electrical energy usage profile. In view of this, the paper focuses on the analysis of a data-driven approach for predicting the electricity consumption of a smart building in a model-free environment. For predicting the load profile the dynamic mode decomposition (DMD) and deep learning models such as recurrent neural network (RNN), long short term memory (LSTM), bidirectional LSTM (BiLSTM), and gated recurrent unit (GRU) model are considered in this paper. The paper also proposes the development of a hybrid model using the best features of RNN and BiLSTM. A comparative study is carried out for analyzing the performance of different data-driven approaches and for that various test scenarios were conducted to evaluate the prediction accuracy. Finally, from the prediction results, it can be claimed that the deep learning models especially the proposed hybrid model outperform other deep learning methods as well as DMD due to the consideration of hyperparameters.

With the development in the sector of smart cities, smart buildings are becoming quite popular. For effective energy management of smart buildings, the accurate prediction of the electrical consumption data is required. To increase the accuracy of prediction, the paper proposes deep learning methods for load profile prediction in a model-free environment. In this paper, a commercial smart building load profile is predicted using the re-current neural network (RNN), long short term memory (LSTM), bidirectional LSTM (BiLSTM), and gated recurrent unit (GRU) model. Using the best features of RNN and BiLSTM a hybrid model is developed for the load profile prediction. To compare the performance of different models error metrics such as Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) are considered for performance evaluation. The paper also highlights the significance of hyperparameters in improving the accuracy of prediction. Finally, from the results, it can be claimed that the hybrid model performs better as compared to RNN, LSTM, and GRU models.

The outbreak of the COVID-19 pneumonia in 2019 has caused great damage to the world economy. With the continuous growth of the amount of data, using machine learning algorithm to analyze and predict the economic development of different countries and regions is a hot topic in recent years. In this paper, three machine learning algorithms (XGBoost, AdaBoost and random forest algorithms) are coupled together, and a new algorithm is proposed. Combined with data preprocessing and fine feature engineering processing, GDP values of different countries and regions are predicted. Experimental results show that our coupled method has better performance than each single machine learning algorithm used in this paper. Specifically, the MSE metrics of proposed model is 1.64%, 3.69% and 8.95% lower than XGBoost, AdaBoost and Random Forest algorithm, respectively. In addition, we also study the correlation coefficient between features and get some constructive guidance to improve the accuracy of the algorithm and restrain the further development of the epidemic situation.

The Campus placement is one of the most important objectives for students as well as for educational institutes. The main goal of every student is to get placement in a well-known desired companies. In most of the institute, admission depends on the placement factor which they provide to their under graduate. This paper emphasizes whether a student will get placed or not at the primitive stage, which will give us an advantage and persuade us to take further action for that student to get placed. This will indirectly decrease the workload of the training and placement cell. Our experiment revolve around the forecasting of the under graduate getting placed in the certain companies. We have done suitable data pre-processing for the removal of all the redundant features. Machine Learning models like Logistic regression, SVM classification, Decision Tree classifier, Gaussian Naive Bayes, Random Forest Classifier, k-NN classification, and Ensemble Voting classifier have been used in this paper and it has been found that the Voting classifier achieved the highest accuracy and outperform.

Liver diseases like fatty liver disease, chronic active hepatitis, and cirrhosis are the major cause of mortality in India. Alcohol consumption, inhalation of harmful toxic gases, improper consumption of contaminated pickles, drugs, and foods are the major cause of diseases in the liver. Diagnosis of liver disease needs high accuracy and precise results for predicting whether a person is suffering from liver disease or not. Major disastrous repercussions can be the result of minor errors in the diagnosis of liver diseases. The major goal of this paper is for the detection of liver disease at right time and helping the doctors and combating the increasing number of cases. In this paper, we implemented various machine learning techniques like logistic regression, KNN, XG-Boost, SVM, Gaussian NB, Random forest, Decision tree, Gradient Boosting, CatBoost, AdaBoost, and LightGBM on selected features from the dataset for predicting liver disease and it was found that Random Forest performed best among all the technique and gained high accuracy and performed outstandingly in all metric evaluations.

As an indispensable part of the mechanical manufacturing system, the wear state of the tool will affect the productivity and machining quality. Therefore, it is of great significance to accurately predict its remaining useful life (RUL). This paper proposes a non-autoregressive dual-stage attention-based long-short term memory network (DA-LSTM) for tool RUL prediction. The dual-stage attention mechanism of the proposed method can adaptively assign attention weights to the input features and the context vectors output by the LSTM encoder, thereby improving the performance of the model without manual feature selection. Validation experiments are conducted on a real public dataset, and the experimental results show that the method proposed in this paper has better performance in short-term and medium-term RUL prediction tasks.

Over the years in social engineering, phishing URLs became a significant threat for internet users as most cyber-crime or attacks redirect to individual target users by sending a malicious or crafted URL which allows the attacker to capture sensitive pieces of information of the victim users. According to this research, most of the existing phishing detection tools showed an overall accuracy level of 70% to 92.52%. This paper introduced a random forest (RF) and artificial neural network(ANN) based machine learning (ML) algorithmic classification models for detecting phishing URLs effectively with higher accuracy. The experimental results showed that the proposed RF and ANN models could classify phishing URLs' legitimacy labels with an efficiency rate of up to 99%, much more significant than traditional ways. In this paper, this comparison's algorithmic learning models demonstrate that the proposed ANN classification holds a competent accuracy label of 98.72%.

Machine learning is the most important part of artificial intelligence, and the difference between artificial intelligence and ordinary big data analysis lies in its characteristics of machine learning. It is the only way to promote the development of theory and practice of technology and art in their respective fields to study the dynamic development relationship between technology and art in the specific era background and humanistic environment. Based on the machine learning algorithm, this paper analyzes the influence of artificial intelligence on art design, and demonstrates that the combination of artificial intelligence and art design can provide a more intelligent, stylized and commercialized development path for future art design with the promotion of innovation of ideas and tools. Combining with the development trend of artificial intelligence technology in the future, this paper predicts the future development of artificial intelligence art from the aspects of art content concept, art form and art media, and obtains a series of characteristics of artificial intelligence art in the future, such as micro-art form, strong ethical concept and new art media.

Forest fire weather index (FWI), empirical model for fire risk estimation based on temperature, humidity and wind speed and precipitation, is used by European Forest Fire Information System (EFFIS) and by Croatian Meteorological and Hydrological Service (DHMZ). EFFIS also estimates fires based on satellite images, which is not precise enough for Croatia forest fires because the number actual fires were several times higher number of real fires in southernmost part of Croatia (Dalmatia). The aims of the paper is to examine the possibility of Machine Learning (ML) in identification of forest fire risk by using only meteorological measurements and to compare ML in forest fire risk identification to widely used Fire Weather Index (FWI). Three different ML models have been built and compared - logistic regression, random forest and artificial neural network on two different sets of features on period 2017-2018. The contribution of the paper is the process of ML and FWI models comparison based on actual fires and generated no-fires. The actual fire database have been manually collected from Croatian Fire Brigade website. Our results show that the best model according to F-score is multiple layer perceptron (MLP) with temperature, humidity, wind speed and ultra violet index (UVI) (F-score is 0.78), while FWI model that seems to be good according to F-score is FWI > 11.2 (F- score is 0.7). The disadvantage of both FWI and ML models is that they report ’Fire’ in high percentage of situations in Dalmatia during fire season.

With the development of artificial intelligence, the question answering system has penetrated into various industries and has become an important production factor. In the electricity field, problems such as the diversification of power equipment failures and the complicated terminology of the power industry are challenging the traditional power question answering system solutions. Therefore, it is of great significance to construct a question answering system based on the knowledge base in the electricity field. However, there are two problems to be solved in the question answering system in this field: (1) How to accurately segment the vocabulary (2) How to effectively match the sentence similarity. To solve the above problems, this paper proposes an algorithm model of cosine similarity combined with TF-IDF. First, add a custom electricity power dictionary in the word segmentation stage, secondly use the space vector model (VSM)-based TD-IDF algorithm for vectorization, and finally, use cosine similarity degree to perform similarity comparison. This method is verified on the electricity power question answering data set, and compared with the LDA model, TF -IDF algorithm and LSI model respectively. The experimental results show that the accuracy of the method proposed in this paper reaches 75.8%, which is significantly better than the other three. It proves that the research model can accurately match user questions, effectively reduce labor costs, and help electric power workers better solve the problems encountered in their work.

In this paper, we present a comprehensive, in-depth survey of the literature on reinforcement learning approaches to ridesharing problems. Papers on the topics of rideshare matching, vehicle repositioning, ride-pooling, and dynamic pricing are covered. Popular data sets and open simulation environments are also introduced. Subsequently, we discuss a number of challenges and opportunities for reinforcement learning research on this important domain.

This paper surveys deep-learning-based pedestrian behavior prediction algorithms from the ego-vehicle's perspective. To safely deploy autonomous vehicles, pedestrian behavior must be understood and predicted for safe and efficient vehicle-pedestrian interactions. With more and more algorithms proposed in the past 2–3 years, it is vital to summarize how the state-of-the-art algorithms estimate pedestrian behavior. To achieve a full view of current algorithms in this domain, in this paper we review (1) prediction output types, (2) network input features, (3) network architecture, and (4) the datasets available for training/testing. As pedestrian behavioral studies show many factors impact a pedestrian's willingness to cross the street, prediction algorithms are evolving to include more rich visual annotations accordingly. Networks architecture is changing from temporal to spatio-temporal networks to account for the influences traffic agents have on one another as well. More innovative benchmark datasets are also published to support more research efforts. The survey depicts the current research frontier in predicting pedestrian behaviors to build the foundation for future research in the area.

We know neither what the future holds nor what will occur in the future. It is impossible to anticipate what the next instant may bring, and it could be awful. A natural disaster is an unforeseen event, which can have a tremendous impact on human life and the environment. The internet provides many sources that generate vast amounts of news articles daily. As the amount of news stories online increases, it's becoming more difficult for people to access disaster-relevant news, which makes it necessary to extract and classify news to be easily accessed. This paper presents an automated system that scraps news from various online sources and identifies disaster-relevant news. This paper also states the performance evaluation for classifying the natural disaster types based on machine learning in Indonesia's online news. Our results show that relevant Indonesian's online news about natural disasters can be the accuracy around 96% using the Support Vector Machine for three classes of natural disasters. In the Indonesian news data set, the machine learning algorithm that gets the highest value out of all the parameters is the best.