Sessions will run in parallel, participants shall attend the session of their choice
Stream A Moderator
Dr. Christos Gatzoulis
Head of School - ICT
Stream A: IoT-Based Access Control with 2-Factor Authentication Using RFID and Facial Recognition
Engr. Herbert Penoso Azuela
Lecturer, University of Bahrain, Bahrain
In a very short amount of time, the Internet has dramatically evolved and changed how we work, live, play and learn. With advancement in technologies, we are now connecting the physical world to the Internet. Internet of Things (IoT) refers to the network of these physical objects able to communicate through the Internet. This IoT research project addresses the design, construction and implementation of an IoT-based secured access control with Two-Factor Authentication (2FA) using RFID and Facial Recognition. Two-Factor Authentication is a method of confirming a person’s identity by utilizing something they have and another factor which is something they are. Access will be granted if the two authentication mechanisms are satisfied. After the user swipes the RFID card on the reader the single-board computer will receive that information and process them. A Raspberry Pi 3B is used in the prototype to interface with the RFID reader and other devices and the Windows IoT Core serves as the Operating System. Windows IoT Core is mainly used for IoT embedded applications and provides the use of cloud stored data in order to store RFID information, user images and names. To facilitate the data storing process, Microsoft Azure is used which features several online cloud services including the facial recognition API and voice API. If there is a match, a voice message will announce the name of the person and grant access by opening the door. If there is no match, the system will deny access to the person. The system is scalable and easy to implement since it is connected to the Internet. Registering new users is also straightforward and can be completed in a couple of minutes. This IoT research project will be very useful in controlling access to highly restricted areas.
Stream A: AI in Banking
Dr. Pradeep Sunkari
Associate Professor, Bhoj Reddy Engineering College for Women, India
Artificial Intelligence, Blockchain, and Internet of Things are emerging technologies that are affecting different parts of human life. These advances have the potential to disturb the manner in which we interface with one another, work our organisations, what's more, even how governments work for their residents. Among these, AI is presumably the most universal and problematic in nature. The utilisation of AI by associations and governments, and its sending in improving client experience, operational proficiency, extortion identification and network safety is on the ascent over the world. Despite the fact that the reception of AI changes fundamentally over geologies, there are pockets of enterprises even inside the created nations that are progressively embracing AI to all the more likely assistance their clients and get efficiencies of scale. One such industry that has grasped AI across topographies is banking. Banks and budgetary foundations remain to profit essentially from AI. Regardless of whether to improve by and large client experience, take more educated choices on layaway guaranteeing, identify cheats and defaults early, improve assortments or increment worker effectiveness, AI can possibly change India's banks. As AI makes advances into a few until now untraversed areas, its definition gets obscured. A significant goal of this report is to demystify the idea of AI and clarify its cozy relationship with information science. Further, this report clarifies the relevance of AI to banks in India and prescribes steps that could be taken to prepare them to grasp the progressions that AI can bring.
Stream A: Robotics with Odor Sensing Capability
Dr. Yousif Albastaki
Chairperson of MIS, Ahlia Uinversity, Bahrain
One of the approaches to mimic the remarkable abilities of the human olfactory system is by the design of computer controlled sensor arrays that are capable of detecting and distinguishing different range of smells and odors with consistent monitoring, referred to as the electronic noses. This paper introduces the opportunity of integrating smell sense in robots by the use of artificial neural networks. The study proposes a structure for integrating electronic noses in robots to add the capabilities of smell related assignments, typically to recognize hazardous substances such as sampling the air, and decide its actions based on this information. The classification of the different odors is accomplished by backpropagation neural network; in which, prior knowledge of the desired odor is fed into the system in order to calculate the error rate and minimize it through iterations. Utilizing the proposed algorithm allows experts in this field to be aware of gas leakage areas, and thus reduce unexpected incidences. The effectiveness of the algorithm is demonstrated by using real-word samples, and the performance is examined via quantitative metrics and analysis. The results show that the proposed algorithmic framework outperforms state-of-the-art methods with an error rate of only 0.0999%.
Stream A: Artificial Intelligence in Digital Marketing: Current Applications and Future Directions
Dr. Maryam Almahdi
Assistant Professor in Marketing, Ahlia University, Bahrain
This presentation approaches artificial intelligence (AI) from a digital marketing perspective, first discussing the benefits and challenges of adopting such human-intelligence-stimulating-processes in digital marketing. The benefits, from a business perspective, include improved customer personalization and customer service, reduced costs, and automation. Challenges include the considerable cost of setting up AI, customer privacy concerns, and the need to constantly adapt to technological change. The presentation also highlights current applications of different artificial intelligence tools in digital marketing, including the use of AI in content creation and curation for social media and websites and its application in digital advertising and online segmentation. Moreover, the role of chatbots and smart search is discussed, specifically in the context of e-commerce marketing, and the use of AI in building websites and personalizing the user experience is highlighted. The presentation discusses future directions for using artificial intelligence in digital marketing, such capitalizing on the Business to Robot to Consumer (B2R2C) model, advances in visual and voice search, and advanced customer segmentation and personalization opportunities. It finally presents an action plan for businesses that strive to set off on the path of adopting AI as a part of their digital marketing practices.
Stream B Moderator
Dr. Rogelio Levardo Jr.
Acting Manager, Teaching and Learning Unit
Stream B: Artificial Intelligence (AI) and Big data Framework in Higher Education
Dr. Hasan Kadhem
Assistant Professor, American University of Bahrain, Bahrain
Higher education institutions have larger amount of data than ever before and they are constantly searching for actionable insights from their data to develop strategies, programs, and policies. Artificial Intelligence (AI) and big data analytics are the technologies that will enable institutions to address their present challenges, as well as predict possible future outcomes. Currently, institutions use AI applications in a variety of ways in learning analytics such as student’s performance prediction, and course recommendation. However, AI and big data can be used in other areas to obtain information, support instructors, students and policymakers to facilitate data-driven decision making based on better evidence. In this research, we first investigate different big data resources for higher education, AI algorithms, and different usage scenarios. Secondly, we propose a general big Data framework for higher education institutions to improve programs, policies and decision making. Thirdly, a use-case scenario is presented to demonstrate the benefits of the proposed framework.
Stream B: Finding the Best Model to Predict Student's Marks using Weka Machine Learning Classifiers
Dr. Rito Opol
Lecturer, Bahrain Training Institute, Bahrain
Educators like many other professionals has to make decisions based on available relevant data. There should be an effective methodology to diagnose, evaluate and predict students' learnings based on their performances in regular tests.
Classification is a popular and effective technique to construct functions or sets of functions to predict certain behavior. Discovered insights or knowledge through classification algorithms are usually presented in the form of hierarchical trees that are easy to understand yet powerful in predicting the response variable or classes. In this study, Weka, a tried and tested open-source machine learning software is used. This software has numerous classification algorithms to model the performance of students.
This study is designed to find the best model that predicts the students’ marks in test. There were 38 attributes (input variables) considered from recent OMA200 (Basic Math for Non-Engineers) Test used to make predictions. Several variables were transformed or discretized to achieve more accurate predictions. Five different classifiers are used to model students’ test results using default settings of Weka (like 10-folds cross-validation). These are REPTree, RandomTree, Linear Regression, M5P, and M5Rules classifiers. Their correlation coefficients were respectively 0.6674, 0.6904, 0.8726, 0.8837, and 0.8441. Hence, Linear Regression Model has the highest linear relationship between input and output variables. Considering the size of tree, REPTree has 11, RandomTree has 301, and M5P has only 4 and M5Rules had 4 rules generated. In this criterion, M5P gets the optimal tree representation. The best two models in terms of mean absolute errors are M5P and Linear Regression which are respectively 2.8458 and 2.4786. Based on the above performance criteria, Linear Regression and M5P Models achieved the best predictions of student marks. These two models can be used help students and teachers in determining important topics to revise in order to improve their success rate in succeeding tests. M5P Model has identified 3 important attributes that split the data into homogenous clusters that can be predicted by 4 respective Linear Models.
Stream B: How Good is Automated Sentiment Analysis in Approximating Human Sentiments? A Case Study of Airline Companies
Ms. Lea Catapang
Senior Instructor, Bahrain Training Institute, Bahrain
One of the challenges faced by airlines today is the unpredictable consumers’ behavior and their indecisiveness. Twenty-first century travelers often used online review platforms in decision making. Reviews and star ratings have greatly affected todays travelers’ decisions in choosing which airline to book. Are these human reviews and star ratings really worth believing?
In this paper, human reviews and comments for Gulf Airlines and Emirates Airlines are generated from Skytrax Ratings, an international air transport rating organization. The author uses the built-in function of Mathematica for Sentiment Analysis (SA), a sentiment classifier that infer a snippet of text conveyed, to examine the human reviews and comments for the abovementioned airlines. Furthermore, this paper tests the correlation between the human sentiments and machine sentiments.
The results indicate that for positive reviews, there is a strong positive correlation (r), with computed r = 0.725598 (GulfAir) and r = 0.642149 (Emirates), between human sentiments and machine sentiments. There is a strong negative correlation with r = -0.762764 for negative reviews of GulfAir and a moderate negative correlation with r = -0.544458 for negative reviews of Emirates.
These results support many studies that shows machine generated SA can be a good way to estimate of the overall image of a company.
Stream B: A Model to Detect the Student's Type of Understanding Based on Emotion Recognition
Dr. Raiza Borreo
Academy Instructor, Nasser Vocational Training Center, Bahrain
Understanding is an important factor for both the students and the lecturer in every discussion of a lesson. Discussion comprises the learning foundation of an important topic. If the students missed to pick up the lesson, there is no other way but for the teacher to repeat and alter the discussion in a manner that is better understood by the students. In a typical classroom setup, the teacher conducts verbal questioning to assess whether the students well understood the topic. Moreover, the facial expression that conveys emotion and non-verbal communication is highly considered as an alternative way to detect whether the students understand the topic or not. This paper presents a model of detecting the emotions facial expressions of the students after the teacher asked who among them understand the discussion. With an accuracy level of 65%, results showed that the model can be used for detecting the level of student’s understanding of a topic.
Stream B: Artificial intelligence (AI) in antibody design and drug discovery
Dr. Dana Naeem Ashoor
Biotechnology and Molecular Medicine Specialist, Arabian Gulf University, Bahrain
The “artificial intelligence (AI)” idiom was first introduced in 1956 by John McCarthy to describe human-like intelligence exhibited by machines. Currently, (AI) systems have overcome human performance in several tasks, such as game playing and image recognition. This leads to an adaptation of the AI-based technology by several biopharmaceutical companies and start-ups focusing on drug design and discovery. These companies include Pfizer and IBM Watson, Sanofi Genzyme and Recursion Pharmaceuticals, AstraZeneca, Abbvie, Merck, Novartis, GSK and Exscientia. Today, AI becomes an integral part in speeding up drug development and discovery, where advanced AI algorithms now are replacing the classical expensive experimental methods especially in molecular design of monoclonal antibodies and vaccines (Cancer monoclonal antibodies and COVID-19 vaccines as an example). However, the vast increase of AI used in healthcare raised some challenges that need to be carefully regulated. These challenges includes AI software validation, patient’s data security and patent laws.