Recent Advancements in Computational Intelligence: Concepts, Methodologies and Applications (Part 2)

Deep Neural Networks (DNNs) have become a vital technology in the quick development of autonomous cars, enabling vision, decision-making, and control systems. This has created new avenues for exploring intelligent traffic safety, comfortable travel, and smart roadways. Because of its enormous potential to decrease traffic accidents and injury to people, autonomous vehicles have recently become a very attractive study topic. However, object detection, scene identification, lane recognition, and additional challenges are crucial for self-driving cars to solve. The development, difficulties, and probable future directions of DNNs for driverless vehicles are thoroughly examined. Perception tasks, sensor fusion, safety issues, decision-making algorithms, and the incorporation of DNNs into practical autonomous driving systems are just a few of the many areas it addresses. This review attempts to offer insights and suggestions for more breakthroughs in the field through the investigation of the most recent state-of-the-art research and highlighting significant difficulties.

A network that helps us define and dream up interconnections between various components is known as Graph theory. Here we used nodes and edges to represent the data. There are many uses of Graph theory in real life. Tracking down Cybercriminals is also one of them. Nowadays, with the advancement of science and technology, the number of crimes has also increased. The chapter, “Tracking of Cybercrime and Security using Graph Theory” represents how the concepts of Graph Theory can be used in tracking the crime that is happening in the cyber world. One of the great concerns in the era of digital technology is the security of information. Protecting the data and security infrastructure from unethical hackers is an extreme concern. The advances in science and technology have brought both opportunities and challenges to human life. So in today’s world, there is an urgent need to protect data and information such as credit card numbers, personal data, information of national importance, etc. In this chapter, the concept of graphs is intended to be incorporated in cyber security and network security. It needs special mention that the main objectives of cyber security are to ensure the ethics, secretiveness, and accessibility of direction systems. In this chapter, we have tried to showcase the use of graphs in cyber security and networks as it is a matter of growing concern for researchers and scientists. 

Each brain region is crucial for the optimal functioning of the brain, with each region responsible for distinct functions. Our study focuses on analyzing myelination in four essential sections of brain magnetic resonance images, which are critical areas associated with myelination. This process is vital for proper brain development in children and continues for up to two years, making it essential to monitor myelin growth in various brain regions during this phase. We utilized fuzzy cmeans clustering (FCM) to segment myelin from T1-weighted inversion recovery magnetic resonance images (MRI). To validate the segmentation results, we employed several metrics including accuracy, sensitivity, precision, Matthews correlation coefficient (MCC), Dice coefficient, Jaccard index, and specificity. 

Magnetic Resonance Imaging (MRI) plays a crucial role in diagnosing brain tumours. Deep learning offers a promising approach for analysing MRI scans and improving diagnostic accuracy. This research explores the application of deep learning techniques, particularly Autoencoders, for brain tumour detection and classification. The focus lies on utilizing MRI data to train deep learning models to differentiate between tumorous and healthy brain tissue, as well as classify various tumour types. The research uses grayscale MRI and applies Adaptive Histogram Equalization for image enhancement. To learn and extract the features from the images, a simple CNN model is employed. Two feature selection techniques are employed to classify tumours: MRMR and Chi-Square. Multiple autoencoders were used in a Parallel MultiAutoencoder (PMAE) model. In the PMAE model, every autoencoder is trained using a single category of images. For a single input, the PMAE model regenerates several sequences; the class corresponding to the sequence with the least error is designated as the predicted class. When tested independently for each feature selection technique, the suggested model demonstrated an accuracy of 99.90% for MRMR and 99.91% for the Chi-Square method. In the problem domain, the lightweight model performed better than the heavyweight pre-trained models that were already in operation. 

The rise of technology has facilitated access to entertainment media in various formats like audio, images, videos, and memes. This diverse multimedia landscape, however, poses challenges for visually impaired individuals who primarily rely on auditory means and cannot consume the visual content freely available today. InclusiVision addresses this challenge by introducing audio descriptions (AD) generated through advanced technology for images and short videos. These narrated verbal descriptions provide details about visual elements such as people, objects, colors, and settings, making the content more accessible and comprehensible for the visually impaired. To enhance accessibility, InclusiVision offers two essential phases: the Image Description phase, which generates short audio descriptions for images, and the Video Description phase, which employs algorithms to narrate key visual aspects in short videos. Both the image and video captioning generate short captions explaining key points of the visuals. It employs basic encoder-decoder modeling to help achieve the task. Hence, the primary objective of InclusiVision is to promote accessibility and inclusivity in entertainment and educational media by providing contextually relevant audio descriptions. 

The evolution of marketing strategies has transitioned from traditional approaches to more advanced personalized marketing, driven by the capabilities of artificial intelligence (AI). Traditional marketing methods, such as print advertising, billboards, and broadcast media, provide broad reach but face challenges in targeting precision and interactivity. This chapter analyzes the solution as an integration of AI technologies in augmenting personalized marketing through the direct application of Machine Learning (ML) algorithms. The paper will demonstrate how personalized marketing is implemented with the use of AI and how this integration resolves issues related to generic marketing approaches and limitations of traditional customer engagement methods. Beyond the general characteristics of personalized marketing, several major types that can be implemented with the help of AI will be explained, and the benefits of such, including achieving deeper customer engagement, driving conversions through precise targeting, and fostering loyalty with AI-powered personalization, are going to be discussed. Most importantly, the optimal design principles of AI-driven marketing strategies and the tactics for their execution, as well as the suggested ethical considerations and data privacy measures within personalized marketing environments, will be covered in further sections. For example, a key study by McKinsey & Company highlights that personalized marketing is no longer just a trend, but a crucial strategy for businesses, with 58% of surveyed individuals willing to pay a premium for tailored products and services. The conclusion of the paper underlines future trends and considerations about the possibilities and advancements in the AI sphere and its integration with personalized marketing. The information provided throughout the paper will help to go beyond the limitations of traditional marketing by increasing the effectiveness and customization of marketing efforts through the opportunities presented by AI technologies in the scope of different business sectors.

Artificial Intelligence (AI) has revolutionized various domains, and its integration into exercise mentoring systems holds great promise for optimizing fitness coaching and promoting healthy lifestyles. This chapter provides an in-depth analysis of the advancements and uses of exercise coaching systems that utilize artificial intelligence. It specifically emphasizes the assessment of performance in various age groups and durations of usage. The aim of introducing an AI fitness trainer is to enhance personalized coaching, improve exercise adherence, and achieve better health outcomes by providing real-time, accurate feedback on exercise posture and performance. Through a comparative study of existing literature, we examine the advantages, limitations, and implications for future research and practice. In this study, we select two of the best-performing machine learning models, Random Forest (RF) and Long Short-Term Memory (LSTM) networks, based on their efficacy in exercise classification tasks. Our experimental results demonstrate that the accuracies of the LSTM and Random Forest models are maximum for the age groups 21 to 60. The maximum values obtained are 82% for LSTM and 78% for Random Forest. Also, detailed comparisons showing variations in performance across different age groups and session durations are explained in the chapter. The review highlights the potential benefits, challenges, and future directions of AI-powered exercise mentoring systems in enhancing individualized support, improving adherence, and achieving better health outcomes.

 In our country, one significant public health concern is diabetes mellitus. Thousands of people suffer from this metabolic disease, which is caused by the body accumulating too much glucose. Diabetes can lead to several health problems in humans if it is not identified early on, including kidney issues, nerve damage, vision damage, heart disease, and more. Gestational diabetes mellitus (GDM) is the name for diabetes mellitus that occurs during pregnancy, both the mother and the unborn child or fetus are at serious risk from it. For prompt intervention and better results for the mother and fetus, early and precise GDM prediction is essential. This study aims to predict diabetes mellitus in pregnant women using a real-time diabetes dataset, machine learning algorithms including Random Forest, Gradient Boost, Decision Tree, and XG-Boost classifiers alongside deep learning algorithms like Long Short-Term Memory (LSTM), Recurrent neural network (RNN), Gated Recurrent Unit (GRU), and hybrid GRU+LSTM. In this work, a detailed analysis and methods such as the Synthetic Minority Oversampling Technique (SMOTE) are used to produce the best probable answer for the prediction of diabetes mellitus. The Pima Indian Diabetes Dataset from the National Institutes of Diabetes and Digestive and Kidney Diseases is used in this investigation. All 768 patients are female and at least 21 years old. This dataset contains one target variable and eight predictor variables. It has been discovered that XG-Boost (82.47%) offers the highest accuracy after comparing the accuracies of other algorithms like Gradient Boost (81.82%), Random Forest (81.17%), Decision Tree (66.23%), GRU (79.69%), LSTM (77.08%), RNN (81.25%) and hybrid GRU+LSTM (80.21%) classifiers using imbalanced datasets (without using SMOTE algorithm). This study examines the model accuracies, after applying SMOTE algorithm, Random Forest algorithm, stood top with accuracy such as 88% surpassing the accuracies of remaining models XG-Boost (85.2%), Gradient Boost (84.4%), Decision Tree (83.6%), GRU (80.8%), LSTM (78%), RNN (80.4%) and hybrid GRU+LSTM (80.8%). This paper compares the existing model like ensemble model accuracy of 86% with the proposed model accuracies, and the proposed model has been observed to outperform the accuracy with 88% in Random Forest. The proposed work uses the Application Programming Interfaces using Flas. Flask is like a simple toolbox for making websites with Python.  

Medical X-rays are a preferred option for first diagnosis due to their low radiation dosage, cost-effectiveness, and ability to identify previously unnoticed pathological abnormalities. Identifying and classifying these abnormalities accurately might significantly enhance the diagnostic process in real-world scenarios. Automated algorithms and technologies might soon be available to assist with the reading process and enhance productivity. Chest X-ray (CXR) imaging is an imaging option that is not only inexpensive but also approachable and simple to be employed. It can be utilized to diagnose lung disorders that are caused by infectious illnesses such as Pneumonia and COVID-19. In order to identify issues, this study suggests an approach based on CXR image processing. In order to enhance the image's quality for better analysis, a series of steps are taken. These include resizing the image, grayscale transformation, removing noise, thresholding, and performing morphological operations. To extract features, techniques like LBP, GLCM, and HOG are employed. Classification is then carried out using DL classifiers like CNN and transfer learning methods ResNet50, DenseNet201, InceptionV3, MobileNetV2, and VGG16. The first set of steps is for distinguishing abnormal CXRs from normal and abnormal ones, and the second set of steps involves identifying COVID-19 and PNEUMONIA in abnormal ones. Following that, DL model performance will be evaluated using f1-score, recall, accuracy, and precision, which are parameters of confusion matrices. To assess and anticipate infection and recovery rates, as well as to protect individuals from aberrant situations brought on by malnutrition, poor diets, and abnormal lifestyles, these DL models can make use of picture preprocessing techniques. The ability to quickly identify and uncover issues using DL methods is a huge assistance for both us and physicians when it comes to starting therapy early on.

With the emergence of advancements in computing technologies, versatile and distinguished data sources of huge volume are taken into account for developing the healthcare system. Although database processing and data mining systems are available, the diverse, highly complex, and heterogeneous nature of healthcare information resources necessitates the use of innovative techniques and computational models to uncover valuable patterns from the existing healthcare data. This chapter proposes a disease prediction model for analyzing the multiple sources of healthcare big data on the basis of MapReduce-based Fuzzy Inference System. To begin with, various sources of healthcare records are preprocessed using the MapReduce framework, and the resulting key-value pairs are provided as input into the Fuzzy Inference System. The output values obtained from the Fuzzy System are used to predict the disease details. With the introduction of the proposed framework, large volumes of multi-sourced healthcare data can be efficiently processed, leading to improved analysis and disease prediction. The efficiency of the proposed disease prediction framework has been validated with the aid of Hadoop-based experimental setup. 

In this chapter, we present a novel ensemble technique for the classification of corn leaf diseases using leaf images. Leveraging the strengths of ensemble learning, our method employs a voting mechanism to effectively identify various diseases afflicting corn leaves. This approach integrates three advanced deep learning models—VGG19, ResNet152, and DenseNet121—to enhance the robustness and accuracy of disease classification. We utilize the PlantVillage leaf disease dataset for both the training and testing phases of our proposed detection system. Through comprehensive experimentation, our ensemble model achieves a remarkable classification accuracy of 99.48% on the test dataset. This significant performance underscores the efficacy of our approach in accurately diagnosing corn leaf diseases, thereby enabling timely interventions to reduce crop losses. The proposed ensemble technique not only improves classification accuracy but also enhances the model's reliability and generalization capability, making it a promising solution for automated crop disease detection systems. This chapter will delve into the methodology behind our ensemble approach, detailing the integration of the deep learning models and the implementation of the voting technique. Additionally, we will discuss the results of our experiments, highlighting the superior performance of our ensemble model compared to individual models. The chapter will also explore the implications of our findings for the development of more reliable and efficient crop disease detection systems, emphasizing the potential benefits for agricultural practices and crop management.

In today's world, computers are essential for communication, education, business, healthcare, and technical innovation, all of which improve productivity, connectedness, and general quality of life. They play a vital role in promoting and supporting human endeavors in a variety of fields. For those with impairments, using computers can be quite difficult because of a variety of physical, sensory, and cognitive limitations. For people with physical disabilities, such as those who have restricted mobility or motor impairments, vision-based interfaces offer vital solutions that let them communicate with technology in ways that are not possible with traditional input techniques. This study project fosters equality of opportunity and inclusivity by making sure that technology is available to all people, irrespective of their physical capabilities. This endeavor aims to compensate people, especially those who are physically unable to operate a mouse because of their disabilities. Without the need for further hardware, users can click and move the pointer using the suggested interface merely by looking at it. In this work, the face and eye are detected using a Haar-Cascade classifier. Twenty individuals participated in the study test. The subject's face and eyes are captured and mouse events like left and right clicks and the cursor movements are mapped to the eye movements. The experimental sections showcase the results of the experiments. Each subject's accuracy will be computed. The right-click event yields a greater accuracy of 85.7% in this work.