Applied and Computational Engineering

- The Open Access Proceedings Series for Conferences

Volume Info.

  • Title

    Proceedings of the 2023 International Conference on Machine Learning and Automation

    Conference Date

    2023-10-18

    Website

    https://2023.confmla.org/

    Notes

     

    ISBN

    978-1-83558-285-5 (Print)

    978-1-83558-286-2 (Online)

    Published Date

    2024-01-31

    Editors

    Mustafa İSTANBULLU, Cukurova University

Articles

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230202

    Treating COVID-19 with machine learning

    From 2020 to 2023, SARS-CoV-2 destroyed much of our society, while few treatments were available due to the time required for drug discovery. However, with recent advancements in artificial intelligence, it is now ready to fight viruses such as SARS-CoV-2. Chemprop, a machine-learning backbone for molecular properties prediction, can be used to discover novel antiviral drugs by training a classifier model with hundreds of thousands of data points that include molecular information represented by SMILES strings and the observed efficacy in inhibiting SARS-CoV-2 in laboratory tests. The resulting model predicts the effectiveness of untested molecules, which then can be manually tested, minimizing tedious hunting traditionally done by human scientists. With promising performance, the proposed method pushes the boundary of machine learning’s involvement in drug research. The trained model achieved a high accuracy in predicting the effectiveness of drugs against SARS-CoV-2 with an AUC score of 0.8455. However, the model loses accuracy when predicting the effectiveness of drugs against SARS-CoV, a different strand of coronavirus, with an AUC of 0.7302. The model was then run on one of the data sets to locate the molecule most likely effective against COVID-19, demonstrating its applicability. The result was a molecule with SMILES string CN1CCN(CC1)C(=O)COC=2C=CC(C)=CC2 also called 1-(4-Methyl-piperazin-1-yl)-2-p-tolyloxy-ethanone. Then the model DrugChat was utilized to determine the properties of the molecule. The model’s ability to find likely drugs can hasten drug research drastically, potentially saving countless lives during future pandemics.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230062

    Indian traffic sign detection and recognition using deep learning

    Traffic signs are a fundamental piece of transportation infrastructure and play a vital role in regulating traffic flow, enforcing proper driving behavior, and reducing the risk of accidents, injuries, and fatalities. An Intelligent Transportation System (ITS) must have the ability to automatically detect the sign and then recognise traffic signs which is to be effective. Automatic traffic sign detection is necessary. and is growing in significance with the advent of self-driving cars. This study introduces a brand-new deep learning-based method for identifying traffic signs in India. The proposed system utilizes a region-based convolutional neural network (CNN) to achieve automatic identification and recognition of traffic signs. The authors describe various architectural and data augmentation enhancements to the CNN model and take into account unique and challenging Indian traffic sign types that have not been previously discussed in literature. The system is trained and evaluated using a database of real-time images captured on Indian highways. The deep learning approach is utilized to work on the accuracy and precision of the system, determined to make automated driving automobiles.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230063

    Clinical big data in healthcare: A survey of medical computers

    Big data is a massive amount of information, measurements, and observations, where it has the power to provide a solution to the impossibilities. Recently, it has become the most trending topic in the field of data analysis because of its amazing potentials in extracting the hidden facts. Which attracted various sectors all over the world to collect and analyze the big data in order to improve their services and introduce high valuable products. Specifically, in the healthcare industry, different sources generate big data such as; hospital records, medical records of patients, and results of medical examinations. This type of data is related to the population healthcare, and it requires analysis in order to extract valuable knowledge. Nowadays, with the available high-end computing solutions for big data analysis. It becomes easy for researchers to have solutions that improve the healthcare level of the population. The promising thinking to give new technologies, high services, and big profits for healthcare, can revolutionize the medical solutions and help the community in overcoming the impossible cases. This research discusses essential clinical big data matters related to the healthcare sector by introducing a clear definition and features of the clinical big data in healthcare and its process. Also, by presenting analytics, applications, benefits, challenges, and future of the clinical big data technologies in the healthcare sector. This survey aims to review state of the art for the application of the clinical big data in the healthcare sector, in which it would be an apparent reference, where authors can refer to in their future research.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230064

    Translation from spoken Arabic digits to sign language based on deep learning approach

    Deaf-and-dumb humans make up about 5% of the world's population, and they need special care by providing them alternative methods that help them to communicate with the outside world, whereas the sense of hearing is the main element of human communications, which is indispensable. From the standpoint of introducing helpful applications that help deaf-and-dumb population, the idea of this research aimed used deep learning techniques to create a model based on the principle of converting Arabic spoken digits to sign language images, through a study of two different datasets that were freely taken from open-source websites. The first one contains audio records of Arabic spoken digits that was used to train on-dimensional CNN model to generate a text translation of any Arabic spoken digit record. The second one contains sign language images of Arabic digits, where used to build IF-THEN rules system that can generate the sign language image as a translation of given Arabic digit text. The whole idea conducted through using both systems in one prediction model that can generate the sign language image of any giving spoken Arabic digits’ record, where it had accurate results with 86.85% accuracy value and 0.5039 loss value. The goal of this research is to add a new technology based on deep learning, in order to help this group of people with a simple idea that opens the researchers’ minds to produce a model of all Arabic spoken speech, which in turn can be a complete technology that helps deaf-and-dumb humans’ to easily communicate with the outside world.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230065

    Secure method of communication using Quantum Key Distribution

    Secure communication plays a vital role now-a-days. Modern-day secure communication is made possible via cryptography. Modern cryptographic algorithms are based on the process of factoring large integers into their primes, as they are intractable. But the cryptography nowadays is vulnerable to technological advances in computing power like quantum computing and evolution in math to quickly reverse one-way functions like factorization of large integers. Incorporating quantum physics concepts into cryptography is the answer, which results in an assessment of quantum cryptography. A unique type of cryptography known as quantum cryptography makes advantage of quantum mechanics to provide complete protection against the transmitted message. Quantum Key Distribution (QKD), a random binary key distribution used in quantum cryptography, enables communication participants to recognize unauthorized listeners. Quantum Key Distribution (QKD) is likely the most advanced quantum technology currently accessible with full stack systems already in use. This project’s goal is to develop secure and encrypted communication between the parties with the help of a web application using the BB84 protocol.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230066

    Animal detection and classification from camera trap images using residual neural networks

    Using camera traps is common in animal studies. The camera is often activated when the movement is detected to prevent recording when nothing happens. It includes a collection of images of wildlife from Tanzania’s Serengeti National Park. Deep Learning is built on an understanding of the composition and it is the working of behaviour like CPU of the computer. Deep learning model is mainly working as the basic principle of neural networks to analyse any inputs like data or images and videos and make better accurate with predicted value with less loss percentage. With current systems, wind and sunshine may potentially move the plants and start recording, leading to a large number of blank images. Researchers will manually eliminate them from the study, which is a hard way of classification by manually and very much wastage of time. When there is a lot of data accessible, the system has all it needs to train itself. Deep residual neural networks, such as ResNet50, which are very helpful for object detection of many image data and make more viable to the conservation of wildlife are used in this proposed system. It aids in determining if the provided picture data is of an animal or not with better prediction, as well as training on a useful dataset like Serengeti2, where camera trap image collection yields accuracy of 94.64% with better prediction of tested data with greater precision and recall value.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230069

    RFID car using arduino mega 2560 by dijkstra’s algorithm

    This paper introduces a module which is used to transport goods or people from one place to another without any driver assistant. It is mainly used in big industries to save the time and energy. This module is built around an RFID sensor. RFID technology uses fields of electromagnetic waves to track and monitor tags attached to objects. When triggered with a field of electromagnetic waves investigate pulse from an adjacent RFID reader device, this tag delivers digital information back to the reader, which is often an inventory number. This number can be used to keep track of inventories. The sensors collect the data and sends to the main algorithm and then it takes the decision which way to go, we implemented our car to be completely automated and does not required any instructions from human. To achieve this we have chosen the shortest path algorithm know as Dijkstra's algorithm.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230070

    Developing an automated currency transactions forecasting process for global e-commerce and fintech companies

    This paper introduces a groundbreaking automated forecasting process for global currency transactions, specifically designed for e-commerce and fintech companies. Traditional linear models, such as weekly moving averages, ARIMA, and SARIMA, have proven inadequate in capturing non-linearities and complex patterns within the data. To address these limitations, we propose an ensemble of diverse machine learning models. These models, characterized by varying lag periods, integrate regional holiday data, macroeconomic variables, and time-based variables. The proposed process exhibits high scalability, capable of simultaneously predicting forex currency transactions for multiple currencies. The implementation of this forecasting process empowers companies to manage currency exchange risk more effectively, enhance overall financial performance, and increase profits through consolidated transactions. Additionally, the automation of this process eliminates the need for manual forecasting, thereby boosting efficiency, accuracy, and employee morale. The findings of this study carry significant implications for the global e-commerce and fintech companies with operations in multiple currencies. They demonstrate the transformative potential of machine learning models in revolutionizing currency transaction forecasting and assisting in strategic decision-making for finance and treasury teams.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230071

    The application of Python game algorithm in Rouge games

    Our project's topic is algorithms for game design, which mainly describes a series of algorithms in the pygame library in Python, such as collisions, mazes, and a series of algorithms with different functions. The following is mainly about the ideas, processes, problems, and solutions encountered by our group's final project. In addition, we learned how to take part in the joint work and Python programming experience and the algorithms in pygame through this study and group cooperation.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230072

    Predicting stock prices through deep learning techniques

    Stock price movements are linked to lots of factors, whether it is a statement by a celebrity or an event of the magnitude of Covid-19. This paper will mainly focus on CNN (Convolutional neural network) and RNN (recurrent neural network), which are two ways to do research on stock prediction. It will discuss the challenges and limitations associated with using neural networks for stock prediction, including data preprocessing, model training, and generalization to different market conditions. The results of this study will provide insights into the potential of CNNs and RNNs for stock prediction.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230073

    A review of motion generation technology

    Nowadays, deep learning and neural network-related research play a very important role in the widely use of artificial intelligence -related technologies, Among them, the hot development in the direction of generative adversarial networks (GAN) has given birth to many generation-related techniques. For example, MoCoGAN is based on the implementation principle of GAN, which enables video generation of different actions of the same character or the same action of different characters, through the method that decompose video into actions and content. This paper introduces the history and principles of MoCoGAN, starting from the prospect of using MoCoGAN in artificial intelligence (AI) industry and the technical challenges that need to be overcome in the future application of action generation. Besides, this paper also discusses the two main issues of how to improve the quality of video generation using MoCoGAN and the input conditions that are the most central problem to GAN networks. By summarizing the optimization solutions of other researchers in these two areas in recent years, this paper searches the core problems need to be solved and propose a broad prospect for future video generation techniques that can be implemented by using MoCoGAN in human-computer interaction (HCI) area.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230075

    Research on space attitude recognition and control method for soft robotics based on neural networks

    In the modern field of biomedical robotics, there is increasing attention on the use of soft robots as carriers for surgical devices. Compared to rigid robots, soft robots exhibit more complex postures and work environments, requiring new methods for recognizing and controlling their postures within biological organisms. This paper investigates the recognition and control of the posture of soft robots using a neural network learning approach, utilizing a soft robot equipped with resistive sensors. By establishing the relationship between changes in resistance values and posture variations, the study successfully achieves the identification and control of the soft robot's posture. The obtained posture data based on resistance values are validated for reliability. Thus, by combining resistive sensors with soft robots and employing neural network analysis, the recognition and control of soft robots’ postures within biological organisms can be achieved.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230076

    Literature review on the application of numerical model in improving face recognition performance

    As an important biometric technology, the applications of face recognition technology have been applied in many fields. With the development of related technology, the application of different algorithms has improved the face recognition technology in different situations. This paper mainly reviews three mathematical models of face recognition based on neural networks, namely genetic algorithm and artificial neural network (GA-ANN), principal component analysis method and feed-forward neural networks (PCA-FNN) and Hidden Markov Models (HMMs) and State-Action-Reward-State-Action (SARSA). In this paper, three models are introduced and analysed in detail. Through analysis, it has been concluded that the GA-ANN improves classification accuracy, the PCA-FNN is used in face recognition with constant posture and the use of HMM-SARSA improves the accuracy of face recognition.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230080

    The advantages and comparison of resistive touch screens and capacitive touch screens

    ATouch screens are an indispensable component in electronic products nowadays, and resistive and capacitive touch screens are the most common and widely used touch screens in touch screens. However, there is still a lack of intuitive literature to explain the differences and advantages between capacitive and resistive touch screens. This article provides a partial overview of the principles and structures of resistive and capacitive touch screen technologies and finally analyzes their respective advantages and disadvantages. The resistive touch screen is mainly divided into four-wire resistive touch screens and five-wire resistive touch screens. Its advantage over capacitive touch screens is that the process is more mature, and the product cost is lower. However, capacitive touch screens, due to their different working principles from resistive touch screens, can achieve multi-point touch, which is a significant feature. In capacitive touch screens, mutual capacitive projection capacitive touch screens have become the main favored technology in the touch screen industry due to their low power consumption, stable signal, and fast data collection speed. They are also an essential direction for the future development of touch screens.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230081

    Application of CNN in computer vision

    Today's deep learning continues to be hot, and the application of machine learning can be seen in more and more fields. A neural network model called a Convolutional Neural Network (CNN) was created to imitate the structure of the human brain. It is a convolution operation that maps the relationship between input features and output features to a two-dimensional in the vector space of , the network can effectively process the input data. CNN emerged to solve the computational bottleneck problem faced by traditional networks. This paper discusses the application of the deep learning model CNN in image classification, target detection and face recognition. In these fields, models are continuously proposed, and architectures in each field are constantly emerging. Among them will be the classic architecture of CNN in this field. These classic architectures have their advantages, but there will also be improvements brought about by the shortcomings of the classic architecture. Through the application of these different fields, we can see that CNN-based deep learning can help various fields, and the efficiency will be improved, but it is not perfect and needs continuous improvement.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230082

    Application and comparison of decision tree algorithm and K-Nearest Neighbors algorithm in heart disease prediction

    In the past two decades, rapid industrialization and urbanization have led to tremendous economic growth and an improvement in people's living standards. However, the impact of people's irregular lifestyles and habits on their health has gradually emerged. Among them, cardiovascular diseases have become particularly prominent, with increasing incidence and mortality rates, especially in developing countries. Heart disease is a major cause of the rising death rates. Early-stage prediction of heart disease poses a major challenge in clinical analysis. Today, the adoption of appropriate decision support systems to achieve cost reduction in clinical trials has become a future development trend for many hospitals. This study compares decision tree classification and K-nearest neighbors (KNN) classification algorithms to seek better diagnostic performance for heart disease. The existing dataset of heart disease patients from the Cleveland database is used to te3st and demonstrate the performance of all algorithms, providing support for the establishment of a heart disease prediction system. This, in turn, can assist doctors in making more accurate diagnoses and timely interventions before the onset of heart disease, thereby reducing the mortality rate of heart disease from the source.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230083

    Comparative analysis of voice denoising using machine learning and traditional denoising

    Noise often affects the content of an audio signal, and noise reduction techniques can help retrieve the original speech content. In recent years, AI-based noise reduction has witnessed rapid development. This article provides a brief introduction to the background and principles of several AI-based noise reduction methods. One of the mentioned methods is an end-to-end time-domain deep learning speech division algorithm, which utilizes a multi-layer CNN network framework. Due to the need for deep network architectures to extract features, it involves a higher computational load. Traditional noise reduction algorithms, on the other hand, are based on researchers' understanding of noise patterns and modeling. Traditional methods may not perform well on non-stationary noise, but they are relatively simple in terms of algorithmic implementation. Through a comparison from various perspectives, AI-based noise reduction demonstrates superior performance in known environments compared to traditional methods. However, in unknown environments, AI-based noise reduction may encounter performance anomalies. Combining AI-based and traditional noise reduction techniques can provide better stability and higher performance in certain scenarios.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230084

    Advancing real-time close captioning: blind source separation and transcription for hearing impairments

    This project investigates the potential of integrating Blind Source Separation (DUET algorithm) and Automatic Speech Recognition (Wav2Vec2 model) for real-time, accurate transcription in multi-speaker scenarios. Specifically targeted towards improving accessibility for individuals with hearing impairments, the project addresses the challenging task of separating and transcribing speech from simultaneous speakers in various contexts. The DUET algorithm effectively separates individual voices from complex audio scenarios, which are then accurately transcribed into text by the machine learning model, Wav2Vec2. However, despite their remarkable capabilities, both techniques present limitations, particularly when handling complicated audio scenarios and in terms of computational efficiency. Looking ahead, the research suggests incorporating a feedback mechanism between the two systems as a potential solution for these issues. This innovative mechanism could contribute to a more accurate and efficient separation and transcription process by enabling the systems to dynamically adjust to each other's outputs. Nevertheless, this promising direction also brings with it new challenges, particularly in terms of system complexity, defining actionable feedback parameters, and maintaining system efficiency in real-time applications.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230085

    An optimized approach to speech transcription using blind source separation and speech-to-text machine learning models

    The use of speech-to-text transcription has a multitude of applications in various industries, including accessibility support, language processing, and automatic subtitling. In recent years, there has been greater interest in incorporating automatic speech source separation features to improve the accuracy and efficiency of transcription mechanisms. This paper aims to design a transcription mechanism that utilizes DUET algorithm to separate speech sources in a stereo setup. The separated sources are then transcribed into text using a machine learning model. The study evaluates the effectiveness of this approach using a dataset of speech recordings. The results of the study indicate high accuracy in speech separation and transcription, highlighting the potential of this approach for practical applications. However, the study also revealed potential issues with the mechanism, indicating the need for further exploration and refinement. These findings indicate the potential of the proposed approach for practical applications, and propose insight for further development and researches in this area.

  • Open Access | Article 2024-01-31 Doi: 10.54254/2755-2721/30/20230086

    Application and analysis of machine learning in handwritten digit recognition

    It appears from the information that Character recognition research is currently focused on handwritten digit recognition, a significant subfield of optical character recognition, i.e. the use of computers to recognise and process digital information. In today's increasingly mainstream computer and data era, handwritten numeric recognition can simplify the process of paper-based offices, reduce the intensity of work when analysing data statistics afterwards and improve work efficiency. There are many algorithms to achieve recognition, each with different recognition accuracy, implementation efficiency and application scope. Based on the basic concepts described above, this thesis investigates the efficiency and accuracy of three algorithms - template matching, SVM and deep learning - in recognising handwritten digits with different sample sizes. The models or kernel functions currently used to process data of varying complexity and restricted scenarios also require continuous improvement to ensure accuracy, making it all the more important to discuss them in detail.

Copyright © 2023 EWA Publishing. Unless Otherwise Stated