Data Analytics for IoT Applications

The extraction of optimal parameters of Artificial Neural Networks (ANN) is a tedious task, as it requires continuous efforts. This process also leads to high cost and low efficiency. To overcome these problems, we have proposed a hybrid model of the Particle Swarm Optimization Algorithm (PSO) and Salp Swarm Algorithm (SSA), namely PSO-SSA, which combines three strategies, namely the Weighted Position Strategy, Adaptive Velocity, and Locality Pruning Strategy. To study the effectiveness of the proposed model, the model is applied to seven classification problems varying from low-dimensional to high-dimensional, namely, 3-digit XOR, Iris, Balloon, Breast Cancer, Heart, MicroMass, and Health News on Twitter. The extensive comparative study shows that our model performed best among other state-of-the-art techniques. The performance metrics used are convergence rate, classification accuracy, general performance, and Mean Square Error (MSE). 

Interpretability in machine learning models is explored in this systematic literature review. This review covers a wide range of topics, including healthcare, social media analysis, fake news detection, self-driving vehicles, etc. It examines the importance of interpretability in these domains and highlights the ethical implications associated with black box models. Moreover, the review discusses efforts made to understand and explain complex models using techniques and tools.

Through an analysis of selected research articles from the past decade, the review reveals the impact of interpretability on model performance and decision-making. It identifies highly interpretable models, moderately interpretable models, and black box models, along with their respective applications and outcomes. The review emphasizes the need for ongoing research to maintain a balance between accuracy and interpretability. This systematic literature review demonstrates that interpretability plays a crucial role in domains where accountability, trust, and transparency are paramount. Providing insights for future research, the review describes the benefits and challenges associated with interpretability in machine learning models.

The growing number of children with hearing impairment underscores the urgent need for efficient and accessible screening and diagnosis methods for hearing impairment. Traditional audiometry, while being effective, relies on cumbersome equipment and lengthy testing procedures, hindering early detection and intervention. This paper introduces DhvaniMitra, a novel mobile application-based audiometer designed to simplify hearing health assessment. 

This research paper presents developments regarding Chronic Obstructive Pulmonary Disease (COPD), a lung condition characterized by symptoms that rapidly affect breathing capacity. Patients also suffer from restricted airflow. COPD is a challenge that humankind has faced on a global level. Various measures and studies have been undertaken to combat this respiratory condition. We leveraged studies on the integration of remote healthcare in managing lung diseases to draw insights while constructing our device. We aim to promote the use of the Internet of Things (IoT) to manage COPD effectively and concisely at home. Our device utilizes various sensors, such as a spirometer, to integrate technology for a beneficial purpose. It takes real-time lung capacity measurements, processes the information collected through sensors, and provides appropriate steps for management. Our goal is to implement remote monitoring for Chronic Obstructive Pulmonary Disease.

This research paper talks about the development of an innovative IoT-based smart door lock that also uses facial recognition, Internet of Things (IoT) integration, and the ESP32-CAM microcontroller to increase home security. As locks continue to evolve with more new features, the use of IoT is increasing and enhancing the security and usability. ESP32-CAM was selected for this project because of its built-in Wi-Fi connectivity and camera, which render it exceptionally suited for creating a smart lock system that is able to not only recognize the saved faces but also alert the user if there is an intruder. When compared to other microcontrollers like Raspberry Pi, the ESP32- CAM offers many more advantages, including enhanced security features, a userfriendly interface, and a lower cost, and much more in initial price. This paper also compares and talks about already existing facial recognition-based door locks, providing a detailed analysis of their plus points and limitations to understand the current state of the technology. The research also provides a detailed breakdown including the system's development process, selection of components for the device, and compilation of hardware components for the implementation of the model, The paper also includes the algorithms and user interface design. In this paper, you’ll see how each component works, its usage, and how each component works together to ensure reliable and accurate facial recognition. Also, model testing was done to get the device's performance in a number of different conditions and usage areas. The results of these tests show that ESP32-CAM based smart door lock is a good and powerful solution and it is also cost-effective, offering good amount of savings compared to other smart lock systems in the market. This paper highlights how the project talks about some of the common security problems associated with the old locking system. By offering a more accessible and reliable alternative, this smart lock system has the potential to revolutionize home security. The integration of advanced facial recognition technology with IoT makes the home door lock system a more secure and user-friendly approach to home access control.

With the proliferation of the digital era, concerns relating to data privacy have emerged with complex implications impacting not just individuals but society at large. In this chapter, the composite aspects and difficulties related to data privacy in India have been analysed. The chapter has been further categorized into parts for a better understanding of the subject. Firstly, it deals with data in general and “Data Protection”, and secondly, the data is divided into two types, namely, “Personal Data” (PD), i.e., data with personally identifiable information (PII), and “Non-Personal Data” (NPD), i.e., data without PII. A framework for the governance of NPD is also suggested by the author, considering that there is no mention of it in the DPDP Act, 2023. Lastly, conclusions and suggestions are presented at the end.

Agriculture is regarded as the principal source of livelihood. Many essential agricultural operations, such as monitoring crops, irrigation systems, applicatio of pesticides, and many others, are carried out using techniques that are labour intensive. A decrease in the Return on Investment (ROI) in this sector may be attributed to a number of factors, including the fact that resources are insufficient, that they are not distributed in accordance with the weather conditions, or that they have not been used to their fullest capacity. There has been an increase in demand for drone technology that is equipped with Artificial Intelligence (AI) and remote sensing capabilities. This is due to the benefits that drone technology offers. To fulfill a wide range of needs at different stages of the crop cycle, drones may be equipped with technologies such as sophisticated cameras, automation, and highly precise GPS navigation systems. The information obtained is useful for regulating crop health, treatment, scouting, and irrigation, as well as for conducting field soil analysis and crop damage assessments at the same time. In addition to reducing the amount of time and money spent, the drone survey helps to increase agricultural yields. These research works present the complete overview of the configuration and implementation of the AI based drones that are used in precision agriculture. An extensive review of the drone technology prerequisites for implementing agricultural drones, including applications as well as hardware factors such as different categories of drones, sensors, and computing units, is also included.

The work aims at designing a Machine learning based highly efficient fresh fish sorting machine that is highly efficient and has high accuracy. The work also claims to follow all ethical practices pertaining to the use of this system for fresh fish sorting. A machine learning algorithm is implemented to categorize fish based on their species and size. This system ensures a higher accuracy when compared with manual sorting and is also less time-consuming. The system is designed using a Machine learning model for Raspberry Pi and ESP32 for real-time communication. The other components used are infrared sensors and servo motors. A Graphic User Interface was designed to enhance the user-friendliness of the sorting mechanism.

Narrowband IoT (NB-IoT) is a Low Power Wide Area Network (LPWAN) wireless technology standardized by the 3rd Generation Partnership Project (3GPP), widely adopted especially by Communication Service Providers (CSPs) in China. It is used in many Internet of Things (IoT) applications for long battery life. Its earmarked Power Saving Mode (PSM) feature enables adjusting the packet ‘detach-attach’ interval as per the requirement of the IoT application. Similarly, PSM feature results are demonstrated through a field trial on the 1800 MHz frequency band using 180 KHz ‘in-band’ mode of operations, and results are obtained for a 4 min (240 Sec) ‘detachattach’ cycle interval. Obviously, the ‘wake-up-cycle’ could be set to every eight hours, once a day, or once a month, etc., thus conserving the battery for a long time. Various applications that are in dire need of such a feature are also listed in this paper, highlighting their demand for long battery life.

Instances of fraud affect numerous industries, leading to the widespread adoption of Artificial Intelligence (AI) for analyzing fraudulent activities. AI can help detect fraud at early stages. Mastercard and Visa have saved billions of dollars using AI. This chapter discusses current trends in how organisations use AI for fraud detection. Data has been collected from the Anti-Fraud Technology Benchmarking Report 2024. The data contains insights from 1187 respondents, mostly accounting professionals, banks, and fraud examiners from public and private sector organizations. Data analysis has revealed that government and private organisations have started using AI in fraud data analytics. This study will help understand the current use of AI in the anti-fraud programs run by various organisations.

Natural processes have been known to produce optimal results, which is the very basic reason that these natural processes have inspired many algorithms for solving complex problems. This class of meta-heuristic algorithms derived from natural processes is called Nature-Inspired Algorithms (NIA). Many such algorithms have been used in different processes associated with healthcare, from medical image processing to its analysis for disease prediction. Early detection of diseases is very important, as medical treatment at an early stage may not cure a critical disease fully, but can control and limit its aggressive growth. This chapter focuses on the study of different NIA for various diseases’ prediction while comparing their suitability. A variety of evolutionary algorithms and swarm intelligence algorithms have been used for other important tasks in the healthcare industry. This chapter surveys the use of various nature-inspired algorithms for medical signal and image processing, followed by their application in disease prediction using medical images and other medical data. A wide range of diseases are covered, including heart disease, diabetes, etc. The latest trends to improve the efficiency and accuracy of these algorithms with machine learning techniques, with faster computations, are emphasized for each task. Lastly, the use of these algorithms is studied for various medical devices and systems for optimal data collection. The chapter is expected to help the readers understand the choice of algorithm for each of the required targeted tasks in healthcare. 

Integration of Artificial Intelligence and Robotics has revolutionised the field of Forensic Science. AI can process data swiftly and accurately from various databases like DNA, fingerprint, palm print, voice and pattern recognition in surveillance footage, etc. AI-led tools could be beneficial in the reconstruction of a crime scene by using the available information for creating 3-dimensional models of the location. By automating a number of forensic tasks, AI saves time for the investigators along with a reduction in errors on his / her part, thereby increasing the overall amount of work done manually.

Robots offer a physical hand in doing forensic specimen tests better and without bothering about any errors. They can be used to gather evidence in dangerous environments where forensic scientists struggle to reach physically. It can also reduce the chances of evidence contamination at a crime scene by being present only during collection, hence ensuring accurate results. Additionally, robots are suitable for forensic labs as well, where they perform elaborate tasks like DNA fingerprinting with more precision than any human being.

Although AI and Robotics play a vital role in reducing human efforts and making the entire process of investigation error-free, they face certain barriers regarding privacy and admissibility in the court of law. Hence, the legal system must address these challenges in order to adhere to the concept of fairness and dependability of the evidence.

LPG gas leakage symbolizes a foremost threat in both residential and industrial environments. There is a huge shortcoming in the current system’s ability to detect LPG leakages. Whenever a 1% of gas leakage happens, it takes an average of one hour to be detected. This emphasizes the need for robust detection and prevention mechanisms. The essential component for this detection mechanism is a load cell that meticulously monitors gas consumption, promptly alerts users to reducing levels, thus helping them to avoid unexpected disruptions. Moreover, there is a GSM module within a system that sends real-time text messages about a gas leak or a low level of gas detected by sensors. This feature enables quick responses and interventions, ensuring safety risks are managed proactively. By integrating these components, this not only enhances safety but also promotes peace of mind for homeowners and businesses alike. Through continuous monitoring and timely alerts, it aims to mitigate the risks associated with gas leaks, thereby safeguarding lives and property. As it strives to create a safer environment, it recognizes the importance of preventive measures in minimizing the impact of potential hazards. Hence, this chapter offers a comprehensive solution tailored to detect and prevent gas leaks from LPG cylinders, prioritizing household safety.

This method is an automated, low-cost, simple, stress-free, and non-intrusive solution to fulfil the needs of animal behaviour monitoring. Additionally, uncommon in rural locations is the availability of veterinary physicians. Farmers in remote locations must make frequent, expensive trips with their animals to veterinary hospitals. Cattle that are in poor health are given medical attention. It is also a complete waste of the money paid for both the trip and the therapy if the cattle's condition does not improve after the treatment. An Internet of Things (IoT) system can be used to track a cow's health, identify the primary diseases that affect dairy cows, and determine the types of wearable, non-invasive sensors required for disease monitoring. The Internet of Things (IoT) based system will continuously monitor the health of the cattle to ensure highquality milk, boost productivity, detect anomalous health issues early to stop them from spreading throughout the herd, and significantly impact farm profitability.