Saeed Iranmanesh, PHD student, Department of Economics, Faculty of Management and Economics, Shahid Bahonar University of Kerman, Kerman, Iran
In this paper macroeconomic data related to the Islamic Republic of Iran have been analyzed and analyzed. These data have been obtained from the World Bank. Raw data related to the Islamic Republic of Iran for the period (1993-2017). The purpose of this paper is to obtain effective exchange rate data in the Islamic Republic of Iran that is affected by green production. For this purpose, a series of indicators is needed to estimate the effective exchange rate data analysis model. The models are indicators of the degree of openness of the economy, the size of the government, and green production as variables that process data. The raw data used in this article are: effective exchange rate, export, import, government expenditure, production. GDP, foreign direct investment, carbon dioxide (CO2) emissions and inflation have been used. Effective exchange rates have the potential to be used in future foreign exchange market research. In addition, effective exchange rates that are influenced by green production can have environmental implications for a countrys foreign exchange market.
Effective Exchange Rate, Green Production, Particle Swarm Optimization Algorithm, Optimization.
Yew Kee Wong, School of Information Engineering, HuangHuai University, Henan, China
Online learning is the emerging technique in education and learning during the COVID-19 pandemic period. Traditional learning is a complex process as learning patterns, approach, skills and performance varies from person to person. Adaptive online learning focuses on understanding the learner’s performance, skills and adapts to it. The use of advanced technology also provides a means to analyse the behavioural learning pattern. As it provides the detailed skill mapping and performance which enables the learner to understand the areas needs to be improved. The information can also be used by assessors to improve the teaching approach. Advanced online learning system using artificial intelligence is an emerging concept in the coming years. In this new concept, the classes are not taken face-to-face in a classroom but through an electronic medium as a substitute. These virtual learning approach are gaining importance every day and very soon they are going to be an integral part of our world. Taking up these virtual learning through an electronic medium is termed as online learning. We proposed two new models which are powered by artificial intelligence (AI) tools. A number of examples of using these new models are presented.
Analysis Algorithm, Artificial Intelligence, Hybrid Integrated Model, Online Learning, Progressive Response Learning.
Mahmoud Alfayoumi, Dania Refai And Saleh Abu-soud, Department of Computer Science, Princess Sumaya University for Technology, Amman, Jordan
One of the most important approaches in machine learning is known as inductive learning, which has the ability to produce general rules from decision tables or from historical examples in the form of a decision tree or a list of rules. In this study, two inductive learning approaches, decision tree, and iterative search, were discussed and compared by comparing two algorithms, one for each approach, ID3, and ILA respectively. The experimental results presented the difference between the two algorithms by explaining the behaviour of each algorithm, the number of extracted rules, the generality in rules induction, and the effect of the number of decision labels in the training set on the accuracy of the extracted rules.
Inductive Learning, Machine Learning; ID3, ILA, Rule Induction, Iterative search, Decision Tree.
Enrico Randellini, Leonardo Rigutini, Claudio Saccà, QuestIT Research Lab, Siena (Italy)
The face expression is the first thing we pay attention to when we want to understand a person’s state of mind. Thus, the ability to recognize facial expressions in an automatic way is a very interesting research field. Because the small size of available train datasets, we apply geometrical transformations and build from scratch GAN models able to generate new synthetic images for each emotion type. On the augmented datasets we fine tune pretrained convolutional neural networks as VGG16, VGG19, InceptionV3 and InceptionResNetV2. To measure the generalization ability of these models, we apply extra-database protocol approach, namely we train the models on the augmented versions of the train dataset and test them on two different databases as CK+ and JAFFE. The combination of these techniques allows to reach mean accuracy values of the order of 85% for the InceptionResNetV2 model.
Computer Vision, Facial Recognition, Data Augmentation, Transfer Learning.
Ahmad Younso, Department of mathematical statistics, Faculty of sciences, Damascus University/Syria
The classical k-nearest neighbor (k-NN) rule of classification for spatially dependent functional data (curves) will be considered. The (strong consistency ) consistency of the classifier will be established based on training data drawn from (β-) α-mixing random field. The results of this paper extend some previous results to the spatial case.
Bayes rule; training data; k-NN rule; random field, smoothing parameter; consistency.
Auwal Salisu Yunusa and Cafer Bal, Fırat University, Faculty of Technology, Department of Mechatronics, Machine Learning, TR-23119 Elazığ, Turkey
Plate number recognition is a system designed to read information of vehicle plate number automatically from digital images for many purposes; such as over speed control, parking areas, traffic control, and top governmental agencies etc. This system can be categories into 3 stages: license plate detection, character segmentation, and character recognition. In order to have successful recognition of characters, a license plate number has to be segmented. Segmentation is a process of partition an image into smaller parts. Character recognition is the last stage, were the license plate characters were read from the segmented characters. Three different type of networks were used in this article, (pattern net, perceptron and multi-layer neural network). Simulation result indicated that pattern net has a very good performance in recognizing the license plate image compared to other two types of networks. Also, has an advantage of less training time compared to other types of neural networks.
license plate number, neural network, image processing, machine learning.
Ojo Abosede Ibironke1 and Rahman Mukaila Alade2, 1Department of Computer Science, Ogun State Institute of Technology, Igbesa, Ogun State, 2Department of Computer Science, Lagos State University (LASU), Ojo, Lagos State
The increase amount of applications that require the need for a model that will be able to classify age, gender and smile have made this model a highly required one. This research work is aimed at developing a model where age, gender and smile can be classified using a Deep learning algorithm called Convolutional Neural Network (CNN) with the objectives of implementing the model and carried out several performance evaluations. We present an All-in-One Convolutional Neural Network that can multitask and classify all tasks as a single task which other algorithms cannot perform. The facial images used in this work was a local dataset with the total number of 490 images. All these images were cropped and augmented in other to generate additional 5 images such as blur, lightening, top hat etc. The entire local dataset was divided into two with 70% of the images used for training and the remaining 30% for testing. This was done with the use of random selection algorithm embedded in the CNN. Required 227by227by3 image sizes was pre-trained using AlexNet, a Multi-Task Learning CNN-based method for classifications and MATLAB 2018a was used as the Simulation tool. Several tables and plots were generated, the Confusion Matrix table was generated and the result generated from our CNN shows that it has better performance.
Face Recognition System, Facial Images, Alexnet, Convolutional Neural Network, Age, Gender and Smile.
Souhila Silmi1,2, Zouina Doukha1, Samira Moussaoui1, 1Department of Computer Science, USTHB University, RIMAA Lab., B P N°32 El Alia, 16000 Bab Ezzouar, Algiers, Algeria, 2Department of Computer Science,Higher Normal School Elbachir El-ibrahimi-Kouba,B P N°92 16308 Vieux-Kouba, Algiers, Algeria
The growing need to integrate new technologies into Wireless Sensor Networks (WSNs) has led to the increasing requirement for development and implementation. From the current systems, it can be seen that several research have been made since several years ago, but the WSN deployment still a challenging task. Yet, this is done by real deployment withtestbeds or bysimulation tools when real deployment could be costly and time-consuming. Each kind of deployment has advantages and limits. With this in mind, the purpose of this paper is to review the implementation and evaluationprocess in WSNs. So, we describe relevant testbeds and simulation tools, and their features. Lastly, we conduct an experimentationstudy using these testbeds and simulations to highlight their pro and cons. To display how we obtained our results, we implement a localization protocol as a use case. The aim of this work is to give more clarity to future-work for better implementation in order to improve reliability, accuracy and time consumed. Hence, researchers will be able to choice the appropriate tools in the process of development and deployment for each new proposal in this fields.
Wireless Sensor Network, Testbeds, Experimentation platform, Simulation Tools.
Yew Kee Wong, School of Information Engineering, HuangHuai University, Henan, China
In the information era, enormous amounts of data have become available on hand to decision makers. Big data refers to datasets that are not only big, but also high in variety and velocity, which makes them difficult to handle using traditional tools and techniques. Due to the rapid growth of such data, solutions need to be studied and provided in order to handle and extract value and knowledge from these datasets. Machine learning is a method of data analysis that automates analytical model building. It is a branch of artificial intelligence based on the idea that systems can learn from data, identify patterns and make decisions with minimal human intervention. Such minimal human intervention can be provided using big data analytics, which is the application of advanced analytics techniques on big data. This paper aims to analyse some of the different machine learning algorithms and methods which can be applied to big data analysis, as well as the opportunities provided by the application of big data analytics in various decision making domains.
Artificial Intelligence, Big Data Analysis, Machine Learning.
Eliane Maria De Bortoli Fávero and Dalcimar Casanova, Department of Informatics, Technological University of Paraná, Curitiba, Brazil
The application of Natural Language Processing (NLP) has achieved a high level of relevance in several areas. In the field of software engineering (SE), NLP applications are based on the classification of similar texts (e.g. software requirements), applied in tasks of estimating software effort, selection of human resources, etc. Classifying software requirements has been a complex task, considering the informality and complexity inherent in the texts produced during the software development process. The pre-trained embedding models are shown as a viable alternative when considering the low volume of textual data labeled in the area of software engineering, as well as the lack of quality of these data. Although there is much research around the application of word embedding in several areas, to date, there is no knowledge of studies that have explored its application in the creation of a specific model for the domain of the SE area. Thus, this article presents the proposal for a contextualized embedding model, called BERT_SE, which allows the recognition of specific and relevant terms in the context of SE. The assessment of BERT_SE was performed using the software requirements classification task, demonstrating that this model has an average improvement rate of 13% in relation to the BERT_base model, made available by the authors of BERT. The code and pre-trained models are available at https://github.com/...
Word embedding, Software engineering, Domain-specific Model, Contextualized pre-trained model, BERT.
Dinindu Koliya Harshanath Webadu Wedanage and S Thelijjagoda, Faculty of Computing, Sri Lanka Institute of Information Technology, Sri Lanka
Security has become a significant challenge in the modern world. Number science and mathematics opened up a vast path to work with programming models to create innovative mechanisms which improve text encryption. Universal Sinhala Library is a Sinhala language-specific text encryption platform. Platform architecture has been designed to demonstrate every possible combination of the Sinhala alphabet, including comma, space and period. The hypothetical architecture includes each and every book that ever has been written in Sinhala, and every book that ever could be, including every poem, every scientific paper and every piece of document in Sinhala. The main goal of the research is to create an encryption mechanism for the Sinhala language. Linear congruential generator and extended euclidean algorithm have been used along with the Hull–Dobell Theorem to outline the backbone of the encryption platform. At present, it contains all possible combinations of Sinhala characters virtually. Sinhala text to be encrypted should be searched in the platform, and it will return the location of that particular text in the virtual library architecture, which is the encrypted text string for the searched Sinhala text. It is helpful to the people who are using or working with the Sinhala language, moreover for sharing and transferring Sinhala context securely.
Text encryption, Sinhala language, Security, Language specific encryption, Virtual architecture.
Xuannuo Chen1 and Yu Sun2, 1Linfield Christian, 31950 Pauba Road, Temecula, CA,92592, 2California State Polytechnic University, Pomona, CA, 91768
The inspiration for the creation of this app stemmed from the deeply rooted history of eating disorders in sports, particularly in sports that emphasize appearance and muscularity which often includes gymnastics, figure skating, dance, and diving . All three sports require rapid rotation in the air which automatically results in the necessity of a more stringent weight requirement. Eating disorders can also be aggravated by sports who focus on individual performances rather than team-oriented like basketball or soccer . According to research, up to thirteen percent of all athletes have, or are currently suffering from a form of eating disorder such as anorexia  and bulimia . In the National Collegiate Athletic Association, it is estimated that up to sixteen percent of male athletes and forty- five percent of female athletes have been diagnosed with an eating disorder.
Data Mining, Mobile APP, Machine Learning.
Rui Huang, Lianyungang Jierui Electronics Co., Ltd., city: Lianyungang, zip code: 222061
Current trends of autonomous driving apply the hybrid of on-vehicle and roadside smart devices to perform collaborative data sensing and computing, so as to achieve a comprehensive and stable decision making. The integrated system is usually named as C-V2X. However, several challenges have significantly hindered the development and adoption of such systems. For example, the difficulty of accessing multiple data protocols of multiple devices at the bottom layer, and the centralized deployment of computing arithmetic power. Therefore, this work proposes a novel framework for the design of C-V2X systems. First, a highly aggregated architecture is designed with fully integration on multiple traffic data resources. Then a multi-level information fusion model is designed based on multi-sensors in vehicle-road coordination. The model can fit different detection environments, detection mechanisms, and time frames. Finally, a lightweight and efficient identity-based authentication method is given. The method can realize bidirectional authentication between end devices and edge gateways.
Network Protocols, Wireless Network, Mobile Network, Virus, Worms &Trojon.
Muzaffar U. Khurram, Senior Principal, Accenture New Jersey, USA
In the world of multiple heterogeneous network access technologies having varying network service quality with ever-increasing demands on bandwidth, capacity and expectation of achieving reduced latencies, an end user expects true user-centric seamless connectivity at an optimally decreasing cost derived from emerging & maturing technologies integration. This is a journey that started long ago but now it seems increasingly possible in coming years.
Seamless connectivity, degrees of freedom, ‘Spectrum as Service’, eSIM, device cloud, UAV/drone, private networks.
Fakoya Johnson Tunde, Ajinaja Mcheal Olalekan, Olaseinde Olayemi Oladimeji and Johnson Olarewaju Victor
Each year, tertiary institutions churn out graduating students who are mandated to get cleared by different unit of the institutions which is usually done using the manual process. The manual tends to be extremely time consuming and difficult as students have to move from one building to another to know where they have problem. It has also been found to be vulnerable to fraud and other vices. The existing automated system available do not provide SMS-notification functionalities and other services such as non-user-friendly interface, lack of adequate information to user, non-prioritization of processes and so on. This study proposes a system that addresses the issues with the existing manual processing while improving on the identified automated ones. The study adopts a case study approach of a complete manual system for a leading tertiary institution of learning in Southwest Nigeria, with a view to evolving a working prototype. First a thorough understanding of the existing procedure is carried out. A new system that is web-based is built using Hypertext Markup Language (HTML) along with PHP for business logic layer, CSS for proper rendering of display pages of the front end and MySQL for the data layer. The new system will reduce the amount of time and efforts wasted on students’ clearance as well as reduce cost incurred on paper by the institution. Another advantage is that students can also initiate and monitor their clearance status from any location they are thereby eliminating the need to travel or be physically present.
SMS, notification, clearance, system.
Mohammad Maroufi1, Reza Abdolee2, and Behzad Mozaffari Tazekand3, 1Department of ECE, University of Tabriz, Tabriz, Iran, 2Department of Computer Science, CSUCI, California, USA, 3Department of ECE, University of Tabriz, Tabriz, Iran
Internet of Things (IoT) emerges to become an essential part of our daily lives and lets intelligent devices collect real-time information from the environment. Increasing the number of IoT devices requires high-speed and secure network connectivity to send large volumes of data through the internet. IoT devices need well-designed infrastructures instead of the current centralized architecture to manage significant data sources. Blockchain distributed technology provides several attractive features to solve IoT security and privacy issues. However, utilizing Blockchain in IoT creates new problems and obstacles due to the tremendous computational process, delays, and bandwidth overhead detrimental to IoT operation. In this regard, Fog and edge computing can bridge the gap between distributed IoT devices and centralized clouds. In this paper, we propose Multistage Blockchain-based Distributed Computing in IoT via Software-Defined Networking which combines fog computing, SDN, and Blockchain to preserve high availability, real-time data delivery, high scalability, security, resiliency, and low latency. This method is designed in three layers: The device layer organizes IoT devices in the private sub-Blockchain, the Fog layer concentrate on SDN to provide essential services, and the Blockchain-based distributed cloud layer.
Reshma Ram, Sanjay Raghavendra, Shruthi Srinarasi, Annapurna P Patil*, Ramaiah Institute of Technology, MSRIT Post, M S Ramaiah Nagar, MSR Nagar, Bengaluru, Karnataka, India 560054
Semantic textual similarity (STS) involves the process of determining the extent of similarity between two pieces of text. A score (usually from 1 to 5) is assigned to each pair of texts as a measure of their semantic similarity. The proposed work is aimed at comparing the state-of-the-art models dealing with sentence similarity, a subpart of semantic textual similarity. These models are categorized based on whether they consider contextual similarity. On fixing the hyper-parameters involved, we provide a comprehensive review of the pre-existing STS models and conclude with how this work can be utilized to optimize the sentence similarity scores.
Semantic textual similarity, Sentence similarity, Natural Language Processing, Pearson coefficient, Cosine similarity.
Atheer Alamri, Daad Albarrak, Emtenan Alawad, Ethar Suwaimil, Hend bin Hussain, Razan bin zoba, Sadeem Aleid and Sarah AlHumoud, Computer Science Department, Imam Mohammad Ibn Saud Islamic University (IMSIU), Riyadh, Saudi Arabia
Arabic is one of the most widely used languages in the world; however, the current state of Arabic machine translation (AMT) systems is still scarce compared with systems in other languages, such as English. This survey aims at compiling and analysing research done in this area. Specifically, we explore the AMT studies by using deep learning conducted from 2009 to 2021, covering 25 studies. In addition, the corpora available for the MT tasks were summarised. The AMT is still an under-researched topic with several gaps and challenges. This study could provide some insights for future research in AMT.
Machine Translation, Deep Learning, Natural Language Processing.
Lulu Dong1, 2 and Lin Li1, 2 and HongChao Ma3 and YeLing Liang1, 2, 1The State Key Laboratory of Tibetan Intelligent Information Processing and Application, Qinghai, Xi Ning, China, 2Department of Computer Science, Qinghai Normal University, Xi Ning, China, 3Beijing Language and Culture University, Beijing, China
Automated Essay Scoring (AES) aims to assign a proper score to an essay written by a given prompt, which is a significant application of Natural Language Processing (NLP) in the education area. In this work, we focus on solving the Chinese AES problem by Pre-trained Language Models (PLMs) including state-of-the-art PLMs BERT and ERNIE. A Chinese essay dataset has been built up in this work, by which we conduct extensive AES experiments. Our PLMs-based AES models acquire 70.70% in Quadratic Weighted Kappa (QWK), which out-perform classic feature-based linear regression AES model. The results show that our methods effectively alleviate the dependence on manual features and improve the portability of AES models. Furthermore, we acquire well-performed AES models with a limited scale of the dataset, which solves the lack of datasets in Chinese AES.
Chinese Automated Essay Scoring, Neural Network, Pre-trained Language Model, Quadratic Weighted Kappa.
AyseKok Arslan, Silicon Valley researcher, Oxford Alumni- Northern California
Given the impact of Machine Learning (ML) on individuals and the society, understanding how harm might be occur throughout the ML life cycle becomes critical more than ever. By offering a framework to determine distinct potential sources of downstream harm in ML pipeline, the paper demonstrates the importance of choices throughout distinct phases of data collection, development, and deployment that extend far beyond just model training. By covering examples ranging from historical context to the process of benchmarking models application-appropriate solutions are suggested for being used instead of merely relying on generic notions of what counts as fairness.
fairness in machine learning, societal implications of machine learning, algorithmic bias, AI ethics, allocative harm, representational harm.
Dele Fei1 and Yu Sun2, 1St. Margaret’s Episcopal School, 31641 La Novia Avenue, San Juan Capistrano, CA 92675, 2California State Polytechnic University, Pomona, CA, 91768
This is a data science project for a manufacturing company in China . The task was to forecast the likelihood that each product would need repair or service by a technician in order to forecast how often the products would need to be serviced after they were installed. That forecast could then be used to estimate the correct price for selling a product warranty . I built the underlying forecast model in the R Programming language for all of the companies products. I developed an interactive web app using R Shiny so the business could see the forecast and recommended warranty price for each of their products and customer types . The user can select a product and customer type and input the number of products and the web app displays charts and tables that show the probability of the product needing service over time, the forecasted costs of service, along with potential income and the recommended warranty price.
Operation Management, Machine Learning, Data Mining.
Saleh Abdel-Hafeez, Sanabel Otoom and Muhannad Quwaider, Jordan University of Science and Technology, Dept. Computer Engineering, Irbid 22110, Jordan
Memory Alias Table exploits a major role in Register Renaming Unit (RRU) for maintaining the translation between logical registers to physical registers for the given instruction(s). This work presents the design of the memory Alias Table based on the 8T-Cell with multiport write, read, and content-addressable operation for 2-WAY three operands machine cycle. Results show that four read ports operate simultaneously within a half-cycle, while two-write ports operate simultaneously within the other half-cycle. The operation of content-addressable with two parallel ports is managed during the half-cycle of the read phase; thus, the three operations occur within a single cycle without latency. HSPICE simulations conduct 32-rows x 6-bit with 21T-Cell memory Alias Table that has 4-read ports, 2-write ports, and 2-content-addressable ports using a standard 65 nm/1V CMOS process. Simulations reveal that our design operates within a one-cycle of 1 GHz consuming an average power of 0.87 mW.
Content-Addressable, 8T-Cell SRAM, 2-WAY Instructions Cycle, Memory Alias Table, Register Renaming Unit.