Computational Intelligence for Information Retrieval

Dharmender Saini, Gopal Chaudhary, Vedika Gupta, Dharmender Saini, Gopal Chaudhary, Vedika Gupta

• Publisher: CRC Press
• Year: 2021
• Print ISBN:
  9780367680800
• eBook ISBN:
  9781000484724
• Topic: Computer Science
• Subtopic: Neural Networks
• Index: Computer Science

This book provides a thorough understanding of the integration of computational intelligence with information retrieval including content-based image retrieval using intelligent techniques, hybrid computational intelligence for pattern recognition, intelligent innovative systems, and protecting and analysing big data on cloud platforms. The book aims to investigate how computational intelligence frameworks are going to improve information retrieval systems. The emerging and promising state-of-the-art of human–computer interaction is the motivation behind this book.

The book covers a wide range of topics, starting from the tools and languages of artificial intelligence to its philosophical implications, and thus provides a plethora of theoretical as well as experimental research, along with surveys and impact studies.

Further, the book aims to showcase the basics of information retrieval and computational intelligence for beginners, as well as their integration, and challenge discussions for existing practitioners, including using hybrid application of augmented reality, computational intelligence techniques for recommendation systems in big data, and a fuzzy-based approach for characterization and identification of sentiments.

1 Hybrid Computational Intelligence for Pattern Recognition

Abhishek Bhatt

College of Engineering Pune, Pune, India

Vandana Thakur

Technocrats Institute of Technology, Bhopal, India

DOI: 10.1201/9781003134138-1

Contents

1. 1.1 Introduction
   • 1.1.1 Computational Intelligence and Hybrid Intelligence
2. 1.2 Evolution of Computational Intelligence in Health Care
   • 1.2.1 Artificial Neural Network
   • 1.2.2 Machine Learning
   • 1.2.3 Deep Learning
3. 1.3 Hybrid Computational Intelligence for Disease Prediction
   • 1.3.1 Prediction of COVID-19 with Hybrid Computational Intelligence
   • 1.3.2 Analysis of Parkinson’s Disease Using EEG Images
4. 1.4 Areas of Hybrid Computational Intelligence for Future Research
   • 1.4.1 Expert Systems
   • 1.4.2 Neural Nets
   • 1.4.3 Searching Process
   • 1.4.4 E-Learning
   • 1.4.5 Solving the Constraint
5. 1.5 Conclusion
6. References

1.1 Introduction

The emergence of the Information Age has made a profound impact on health sciences. Data from various stages of health care organizations migrate across the different stages of these organizations. Intelligent machines help health practitioners in both the medical and administrative environments [1]. Studies have shown that these methods are rising in popularity because they can manage vast clinical data volumes and ambiguous details. Computational intelligence is based on biologically inspired algorithm computations [2]. Based on this area, there are main pillars such as neural networks [3], genetic algorithms [4], and fuzzy systems [5]. Neural networks can be used for function approximation problems and can classify artifacts. These artificial intelligence (AI) algorithms include supervised, unsupervised, and reinforcement learning. Genetic algorithms are search methods focused on genetic variants in natural systems. They depend on random and non-random genetic mutations. Populations are built over many generations in this case. It takes an evolutionary approach to solve broad complex problems [6]. Fuzzy logic is based on fuzzy set theory to allow reasoning which is fluid or approximate, as opposed to defined and precise. Fuzzy logical variables can accommodate “partial truth” in addition to the “true” and “false” values. Several types of computations have been used for solving world problems. Furthermore, fuzzy and genetic networks may also be used in the area of medical science [7].

1.1.1 Computational Intelligence and Hybrid Intelligence

Some technologies will have their benefits, while others will be sources of grief. In order to gain maximum success, creative and intelligent approaches are also essential. This mutualization should ensure there are no potential problems to worry about. It combines two smart strategies. By combining the neural networks with fuzzy techniques, the final solution is a mixture of both neural and neuro-fuzzy knowledge. When it comes to predicting future behaviors, fuzzy and neural networks are the building blocks of soft computing. AI is a type of computation focused on rationality and thinking. The system is composed of a neural network with a degree of human expert input to produce a neuro-creative system. They train a nation’s citizens to be educated and employees to be prepared. A number of neurons are interconnected by way of running through the vectors. In more simple terms, the architecture consists of two neurons and the third layer consists of two neurons. Similarly, implicit rules were found by artificial neural networks (ANNs). This clarifies the actions of the neural expert system’s thinking process when running on new data. This user-friendly interface is used to connect the neural expert system to the user, being built to allow their cooperation. It finally infers the data processing in the design and passes the knowledge to the neural information structure.

Expert systems rely upon logical reasoning combined with decision trees [10]. Neural networks use parallel processing and human brain structure to extract meaningful patterns.

• The future AI system will undergo a significant impact on the development and control of brain function.
• Knowledge in the rule-based expert system is cached as an “if-then” condition, whereas synaptic connections between neurons denote knowledge in ANNs.
• Expert device awareness can be separated into different laws that the expert can interpret and carry out. There is no distinct segment of information from which the weight of each synapse is learned. Layers of expertise are integrated with the whole network here. If synaptic weights shift in the brain, it may have uncertain and unpredictable effects.
• Neural networks and expert systems are claimed to be more capable than the current system by combining the two into one. A system of connected neural networks is referred to as a neural expert system.

1.2 Evolution of Computational Intelligence in Health Care

Modern health care technology is spreading across the world, helping people get healthier and live longer [11]. At first, the advancements have been driven by mobile devices’ advent and the rising need for clinical record keeping due to medical progress. Computers today have a little more autonomous and advanced programming and applications. The advances in technology, especially ML [12] and AI applications, accelerate the speed of health technology change [12].

AI is one of the most effective technologies used in health care. Because medical data are becoming increasingly accessible and due to advances in big data, diagnostic technologies have complemented the potential of AI’s existing use in the health care sector. The close relationship between medical problems and future AI techniques will enable many relevant data to be presented to decision-making [13] in health care. The applications of machine learning (ML) and AI in health care have allowed the field to tackle one of its key issues, which is the discovery of new drugs. Any other technical approach still faces problems. Most of the difficulties encountered in health care have been tackled by using AI technology, including regulatory aspects, patient and provider understanding, and data sharing. AI never reaches out to any of the listed challenges as it has shifted away from performing any of these tasks. The aim of AI and ML in the health care industry is to reshape the industry and make it possible. AI needs access to enough data to be useful in medicine. AI has made advancements in classifying complex variables faster and more accurately. Technologies of AI, including AI wellness apps, may allow people to assess their symptoms and take care of themselves whenever possible.

AI can help people become more independent and feel more dignified and comfortable at home. As AI is bound by data availability and data quality, it has some limitations. Also, a lot of computing power is required in the study of large and complex datasets. The clinical profession still requires social expertise that cannot be done by AI. The ML method assists in investigating such unstructured data as DNA, electro-physics, and medical images. ML uses knowledge-based computational methods to conclude. The data of the patients and their healing processes will be taken into consideration in ML algorithms.

A patient’s “essence” essentially includes all information crucial to diagnosing their ailment. Simplifying complex data is ML support for AI. Significant groundbreaking advancement has occurred in this sub-field of neural networks. This has raised the overall interest in different fields of health and medication across substance and study area. Deep neural networks may diagnose complex health problems. When using AI techniques in health care-based applications, you can use any new modern mobile phone. AI can be used to combat essential health problems efficiently. AI will make a wrong judgment, and that will bring another critical issue. Who should be accountable when AI gets it wrong?

1.2.1 Artificial Neural Network

ANNs are among the most common ML models known in recent history. In a neural network, several computation nodes are organized in layers. Data flows through the network, being processed along the way. The network is trained to produce useful and predictable guesses by recognizing trends in a collection of labeled training data. During the neural network training, the network’s parameters – the strength of neuron – are modified to recognize trends resulting in the training data. If the pattern is mastered, it can be used to guess data that are not seen before, such as generalizing to new data. It has long been recognized that ANNs are very versatile, but also that they are computationally difficult to program. This has, in turn, led researchers to concentrate on other ML models. ANNs are one of the leading techniques in ML nowadays. Since the growth of big data, several strong parallel processing devices (e.g., GPUs) have appeared, making the algorithms for neural networks faster. There is tremendous growth in ANN research as well, which pushes other functions of ML to progress.

1.2.2 Machine Learning

One develops and studies methods that help the machine learn from experiences. The aim is to construct mathematical models that can be given good input data and generate useful outputs. ML models are obtained by analyzing sets of training data and modified to generate accurate predictions by an optimization algorithm. The objective of the models is to generalize and produce accurate forecasts for unseen data. A model’s generalization potential is usually estimated in a separate dataset, the validation dataset, and used to construct the model. After several iterations of training, the final model is checked on a test set to assess its accuracy. ML is loosely classified according to how a model processes the input data during preparation. A typical approach in reinforcement learning is to provide an agent that learns from their environment through trial and error while enhancing some objective function. AlphaGo and AlphaZero were recent implementations of reinforcement learning in ML [14]. In unsupervised learning, the machine is involve...

Table of contents

• Cover
• Half Title Page
• Series Page
• Title Page
• Copyright Page
• Contents
• Preface
• Editors
• Contributors
• 1 Hybrid Computational Intelligence for Pattern Recognition
• 2 Secure Image Transmission Using Nested Images
• 3 Accist Automatic Traffic Accident Detection and Notification with Smartphones
• 4 Emotion Prediction through EEG Recordings Using Computational Intelligence
• 5 Finger Vein Feature Extraction Using Contrast Enhancement Dynamic Histogram Equalization for Image Enhancement
• 6 Song Recommendation Using Computational Techniques Based on Mood Detection
• 7 Deep Learning Classification of Retinal Images for the Early Detection of Diabetic Retinopathy Disease
• 8 Protecting and Analyzing Big Data on Cloud Platforms
• 9 Using Flutter to Develop a Hybrid Application of Augmented Reality
• 10 Computational Intelligence Techniques for Recommendation System in Big Data
• 11 Predicting Melanoma Tumor Size through Machine Learning Approaches
• 12 A Fuzzy-Based Approach for Characterization and Identification of Sentiments
• 13 Fingerprint Alterations Type Detection and Gender Recognition Using Convolutional Neural Networks and Transfer Learning
• 14 Content-Based Image Retrieval Using Intelligent Techniques
• Index