Deep learning has gained a huge commitment to the new headway in man-made brainpower. In contrast with conventional artificial intelligence (AI) techniques, for example, choice trees and backing vector machines, profound learning strategies have accomplished considerable improvement in different forecast assignments. Notwithstanding, deep neural networks (DNNs) are similarly powerless in clarifying their derivation cycles and eventual outcomes. In some certifiable applications, for example, business choice, process advancement, clinical determination, and venture suggestion, reasonableness and straightforwardness of our AI frameworks become especially fundamental for their clients, for individuals who are impacted by AI choices, and besides, for the scientists and designers who make the AI arrangements. This chapter gives an insight into explainable AI, the new and trending current technology used for diverse modern-day applications.

Artificial intelligence (AI) is turning out to be an indispensable paradigm for businesses and individuals. AI is automating and accelerating a specific set of everyday problems such as classification, regression, clustering, detection, recognition, translation, etc. AI can classify whether an incoming e-mail is spam or real, recognize a person's face in an image, understand a speech and convert it into text, create an appropriate caption for a scene, etc. The scope of AI is fast expanding. Industry verticals are keenly exploring and experimenting with different things. Business processes are being automated and optimized through the smart leverage of all kinds of noteworthy advancements happening in the AI space. Increasingly AI takes the center stage in business operations across the globe. There is a dazzling array of integrated platforms, frameworks, toolsets, libraries, and case studies and hence the adoption of AI algorithms and models is picking dramatically in the recent past. However, there are a few critical challenges to be surmounted before the widespread usage of AI models in mission-critical domains such as healthcare, security, retailing, supply chain, and infrastructure management. That is, business executives and IT experts insist on trustworthy and transparent decision-making by AI models.

This chapter is to explain the brewing challenges in the AI field and how they can be surmounted through competent technology solutions. Especially how the fast-emerging explainable AI (XAI) is a set of methods and software libraries that allow human users to comprehend and trust the results created by AI models. XAI is to describe an AI model and how it arrived at a particular decision. XAI is to explain the AI model's implications and potential biases. It helps in understanding model accuracy and fairness. XAI turns out to be a crucial cog for mission-critical enterprises to embark on the AI paradigm with all the clarity and confidence. With the maturity of the XAI concept, the AI adoption happens in a responsible manner across industry verticals.

Recent advances in machine learning (ML) strategies have introduced several artificial intelligence (AI)-based systems. These AI systems have the capability to perceive, learn, smartly decide, and act quickly on the given situation. Apparently, this is the requirement from such systems but after witnessing their performance, it has been noticed that these systems are unable to explain their actuation to users (humans). This constraint has been taken into consideration by several researchers later, after all this is the main thing required to make our autonomous systems more intelligent and robust. At this instant, researchers felt the need for explainable AI (XAI) that may make the verifiability of taken decision essential. This will increase the demand for an ability to question, understand, and above all generate a trust level over artificial intelligence systems. There are several models but still there is no consensus on the assessment of explainability. Thus, this chapter presents a comprehensive review of current state-of-the-art over the XAI that have a societal impact. In addition to this, one may find the drivers and tools for XAI. Last but certainly not the least is the complete literature review that provides the future research directions for researchers in this area.

Artificial intelligence (AI) and machine learning (ML) applications are applied in many applications and devices, and it is expected to grow by 15 trillion dollars by 2030. There are more demands for explainable AI (XAI) for its improvement in explainability attributes of the AI quality. Software quality is defined as the product meets its required product specification and is expected to behave as it is expected by the stakeholders. Furthermore, we need a systematic approach to the design, development, implementation, and testing of AI products. Therefore, this chapter proposes a software engineering framework for AI and ML applications (SEF - AI and ML) supporting the complete XAI application development phases including a reference architecture to standardize across XAI applications. The framework has been validated through a case study involving an explainable Chatbot using business process modeling notations (BPMN), modeling, and simulation. The results demonstrate a 98% utilization rate and improved time efficiency, confirming the validation of performance and resource requirements for cloud-driven AI Chatbot services. Therefore, SEF - AI and ML has the potential to be a standard framework for AI and ML applications to achieve the desired quality and certainty of AI products and services.

Knowledge representation and reasoning (KRR) is a key research area in artificial intelligence that deals with the design and development of methods to represent, manipulate, and reason with knowledge in computer systems. KRR techniques are used to develop intelligent systems that can reason about complex domains, make decisions, and provide explanations for their actions. The present research aims to understand the process of representing knowledge and reasoning with different domains, including agriculture, education, healthcare, and business. Different techniques, such as ontologies, semantic networks, frames, rule-based systems, description logics, first-order logic, and Bayesian networks, have been developed to facilitate the representation and reasoning of knowledge. In the present research, different techniques for representing and reasoning with knowledge are used in two different domains: academic knowledge and farmer knowledge. The study examines the strengths and weaknesses of each technique and provides insights into which techniques are most suitable for different types of KRR tasks. The results of this study can help practitioners and researchers to choose the most appropriate technique for their specific KRR needs.

The widely used relational database model organizes data in the format of tables with columns and rows. By checking the tables, the relationship amongst the different data points can be identified. This has been doing well for business operations automation as the data volume is growing slowly. For complicated operations (which involve identifying relationships amongst data points kept in different tables), the relational database model is found inefficient and inadequate. There are other inadequacies in the hugely popular and widely relational database systems (RDBS). Therefore, there is a clarion call for pioneering database solutions for the impending knowledge era.

On the other side, with the rising need for creating and managing intelligent devices and software products for business transformation, the role and responsibility of artificial intelligence (AI) methods go up significantly. Machine learning (ML) and deep learning (DL) algorithms play a very vital role in producing sophisticated AI systems and services. For AI algorithms to do their tasks, they need a lot of correct and cleaned data. Right data is to result in highly accurate predictions/inferences/conclusions.

Data typically gets collected from different sources and cleansed. There are several dissimilar data formats and transmission protocols greatly complicating the goal of data integration. Therefore, there are insistences for competent data integration and virtualization technologies and tools, which play a very vital role in visualizing and realizing profoundly impactful AI systems. That is, there is a need for a fresh and flexible approach to fulfilling the complicated requirement of data integration. The need for data integration has given rise to many research works and thought processes into working out the best possible way to collect, store, manipulate, and maintain digital data. Knowledge graphs (KGs) are the graph databases emerging and getting established as the next-generation data management system for speeding up knowledge engineering and extraction towards digitally transformed businesses and societies. The popularity of KGs is increasing due to their potential and flexibility in dealing with complex and interrelated data.

In a KG, the data is stored and information is depicted in a graphical format. This methodology in KGs can be applied to create a graphical representation of the relationships amongst all of its data points. Hence, even if the data points do not fit neatly together into a table, the associations between the data points can be evaluated fast and with much less computation power which is the explicit advantage over relational database.

Formally, a KG is a directed labelled graph that intrinsically and illustratively represents relations between data points. A node of the KG represents a data point. The entity of this data point could be a person, a place, or a webpage, and an edge represents the relationship between a pair of data points.

The fledgling concept of graph neural network (GNN) has gained a greater acceptance and adoption in the recent past across domains such as social and transport networks, knowledge graphs (KGs), recommendation, expert and question-answering systems, neurons in the brain, and life science that deals with molecular structure. The unique power of GNNs in modeling the intriguing and intimidating dependencies between nodes in a graph has laid down a stimulating environment for envisaging breakthrough results in the graph theory arena. GNN is a special but powerful type of neural networks. GNNs directly operate on the graph-structured data and are capable of assisting in implementing intelligent systems. In short, GNNs are being viewed as an enabling factor and facet of real digital transformation.

This chapter is to explain the distinct characteristics of GNNs and how they contribute to visualizing and realizing a variety of advanced applications for the impending knowledge era.

Machine learning is a branch of artificial intelligence (AI). Machine learning finds its application in different domains like healthcare, travel, and e-commerce, and has enhanced the working of the same. It makes use of statistical prediction and modeling. It takes the raw data as input, analyzes the data, and generates the output according to the analyzed data. This chapter provides an extensive overview of machine learning techniques and a basic conceptual briefing about this extensive topic. The chapter also aims to explain the types of machine learning including supervised learning, unsupervised learning, and reinforcement learning. Machine learning has the potential to produce consistently accurate estimates in this new era. This chapter is to basically understand the fundamental concepts of machine learning and its techniques. It tells about the techniques we generally use, and the places where these types can be applied, profuse algorithms. Here we also get to discuss the various languages that we are going to use, the framework of machine learning, the best tools or the efficient platforms that can support machine learning practices and to implement the concepts.

The excitement around cloud technology often leads people to believe that it can solve all problems, but this is not always true and the complexity of using the cloud is often overlooked by promoters. There is a significant difference between the adoption of cloud technology and the level of innovation that cloud consumers can achieve, which is a major concern for many cloud users. They are questioning the ability of cloud computing to provide continuous service delivery.

The Service Level Agreement (SLA) is a crucial document in cloud computing that outlines the obligations of both the customer and the provider, including details about the expected service delivery and penalties for violations. This document is essential for customers to trust the cloud provider's ability to handle their data and rely on the service. Without strong assurances that their requirements and SLA will be enforced, customers will not outsource their data to cloud infrastructures. Thus, managing the SLA within a cloud-based system is crucial to ensuring service continuity.

Cloud providers typically offer two types of SLA: predefined and negotiated. A predefined SLA is a generic template that applies to all customers. However, some customers may have unique QoS requirements that are not covered by a predefined SLA. In such cases, the customer and provider engage in a negotiation process to establish a mutually agreed-upon SLA before service provision (negotiated SLA). In this context, negotiation refers to the process by which parties arrive at a mutually acceptable agreement on a particular matter.

In a typical scenario, the terms of an SLA are fixed upon construction and remain unchanged throughout the service period. However, this approach conflicts with the dynamic nature of cloud computing, which is characterised by flexibility. The inability of current SLA management frameworks to accommodate this dynamic environment can negatively affect service delivery performance. In this context, a framework refers to an abstract representation of functionality for managing the SLA. Service providers must be capable of accommodating changes in the needs and circumstances of cloud consumers over time. Failure to address these factors can result in service violations that may impact the acceptance of a cloud service.

Existing frameworks for managing SLAs do not include provisions for adjustable SLAs based on customer preferences during operation, nor do they address service violation handling for cloud-based systems that can minimise such violations. The focus of typical SLA renegotiation is limited to the initial phase of service delivery, and service violation handling only reacts after the violation has been detected. Consequently, there is a need for adjusting SLAs during service operation, considering the customer's changing preferences, the provider's current situation, and any occurrences of service violations.

Artificial intelligence (AI) has undoubtedly been a center for the latest advancements and has brought about significant developments in the medical field. However, the healthcare practitioners' and researchers' desire for an interpretation of the system's predictions based on the health statistics acquired through advanced machine learning models did not meet. So, the study of eXplainable AI (XAI) has been pursued through science establishment to give justifications for machine predictions and assure accuracy in the sophisticated medical framework. Because depending on artificial conclusions to rescue the health of an individual without an adequate knowledge of the underlying logic is unacceptable. Before applying the results to the patient, XAI helps the medical domain crew understand the reasons and keep the conclusions in check for a better cause. In this book chapter, we will emphasize the motives for espousing XAI in the medical domain and examine the basic principle behind it, as well as how it might help to dependent AI-based solutions in healthcare.

Breast cancer is the most common cancer not only affecting women but also men. Diagnosis and treatment are crucial stages in the cancer treatment process. However, even after treatment, individuals often experience ongoing challenges, including the regular need for painful procedures such as biopsies, MRIs, and scans as part of their journey towards recovery. We propose that in this case, machine learning (ML) and deep learning analyses may be used to perform longitudinal studies of women with breast cancer. We do a comparative analysis of three situations, in the first situation, we apply ML algorithms just after the primary preprocessing steps, in the second situation, we add balanced class weights hyperparameters, and in the third, we do principal component analysis (PCA). In the first situation, the light gradient boosting machine (LightGBM) gives the best accuracy of 87.87%, and the random forest (RF) gives an accuracy of 87.87% after the hyperparameter of balanced class weights is given. After PCA, logistic regression gives a maximum accuracy of 84.84%.

The reasonableness and logical AI have started expanding considerably by both examination of local area and industry. Clarifying the capacity of artificial intelligence (AI) gives numerous subjects of dynamic examination by the need of passing on well-being and trust to clients in the "how" and "why" of computerized decision-production in various applications like independent driving, clinical determination, or banking and money. In this chapter, we present a chronicled point of view of explainable AI (XAI). The calculations utilized in AI can be separated into white-box and discovery AI [machine learning (ML)] calculations. White-box models are ML models that give results that are reasonable to specialists in the space. Black-box models, then again, are amazingly difficult to clarify and can barely be seen even by space experts. XAI calculations are considered to follow the three standards of straightforwardness, interpretability, and logic. We examine how reasonableness was mostly imagined before, how it is perceived in the present, and how it very well may be perceived later on. We close the chapter by proposing measures for clarifications that we accept will assume an essential part in the advancement of human-justifiable logical frameworks.