AI and ML Solutions Driving Modern Farming and Urban Innovation

Early detection of plant diseases is critical in the agricultural sector to reduce the loss caused by the diseases. Paddy crop diseases impact yields in various ways, given that the crop is cultivated globally. The specific disease manifestations and severity, however, often vary with climatic and geographical conditions. The method used to treat the disease in one area may not be suitable in other areas. The manual detection of diseases by farmers is a very costly, time-consuming, and tedious process. To avoid these issues, researchers have developed various automatic methods. Most models utilize image processing and machine learning, or deep learning algorithms, to achieve the designated output. This study analyzed various algorithms used to automate the plant disease detection process, examining their advantages and disadvantages. 

In the rapidly evolving field of agricultural technology, the need for advanced, reliable methods for assessing dry fruit quality has become paramount. This paper presents an empirical review of the rapidly increasing application of Machine Learning (ML) and Deep Learning (DL) techniques in hyperspectral imaging for the qualitative analysis of dry fruits and their subtypes. Traditional methodologies, while having laid a robust foundation, often grapple with constraints such as limited precision, suboptimal accuracy, and scalability challenges. Furthermore, these conventional approaches typically exhibit significant delays in processing and struggle with complexity in handling diverse dry fruit categories. Addressing these limitations, this review comprehensively evaluates ML and DL methodologies tailored explicitly for hyperspectral imaging applications in dry fruit quality analysis. The proposed review process stands out by precisely comparing these methods across a spectrum of critical evaluation metrics, such as precision, accuracy, recall, processing delay, complexity, and scalability. This approach not only bridges the gaps identified in existing literature but also lays the groundwork for a more nuanced understanding of the capabilities and limits of these advanced technologies in practical scenarios. This, in turn, has improved quality control measures and enhanced overall productivity levels. In conclusion, this work extends the existing body of knowledge. It sets a new benchmark for applying deep learning techniques in agricultural hyperspectral imaging, thus marking a significant stride forward in pursuing agricultural innovation and excellence. 

Potatoes are one of the most highly consumed foods in the world. Due to several environmental and climate changes, potato leaf diseases are emerging frequently, causing severe damage to potato plants and drastically reducing potato crop yields, which results in financial losses to farmers. This research presents a deep learning generative approach to enhance the accuracy of identifying classical diseases in potato leaves using a novel data augmentation method, specifically the diffusion model. By employing diffusion models, researchers created images resembling real ones showing diseases such as Early Blight, Late Blight, and healthy leaves on potato leaves. These generated images are then combined with others to strengthen the training of a deep learning model, enabling it to better distinguish between early blight, late blight, and healthy leaves. This combination of synthetic data enhances the model’s robustness. Demonstrates promising advancements in analyzing potato leaf disease. The significance of this study lies in addressing challenges related to data availability while improving the model's ability to generalize across the stages of the disease. The outcomes have implications for agriculture by providing effective diseaseidentification methods for potato farmers. The proposed CNN model delivered classification accuracy (98.99%) with ResNet18 integrating with the diffusion model on the potato leaf disease image dataset. Integrating diffusion models represents a step towards optimizing crop management and maximizing yields in potato cultivation.

Seamless communication between humans and things like cars and other such machines in important industry sectors is constantly evolving as each day passes. In the 21st century, crucial advancements in several fields have implemented IoT in them because of its numerous merits. When IoT is put into practice alongside Cloud Computing and Data Analytics, it can reduce the gap and enhance cooperation between the physical and digital worlds, which is essential for sustainability in this hyperconnected era. The project involves collecting data through sensors in the field, which is then sent to the IoT analytics and cloud-based platform ThingSpeak. This platform enables data ingestion and storage, allowing farmers to remotely visualize and analyze real-time data in the form of data streams and take action accordingly. Fuzzy logic is one such approach to computing that works on degrees of truth rather than single-valued Boolean logic, thus yielding better and more accurate conclusions for an imprecise spectrum of data. Fuzzy logic is implemented in this project to control water pumping time based on user-defined thresholds. With an increasing population and depleting resources, precision farming has proven to be efficient as it focuses on minimizing waste and is cost- and energy-efficient. The objective of this paper is to develop a system that assists farmers in remotely monitoring data, optimizing resource utilization, and increasing productivity. 

For training a deep learning model, one has to get the parameters necessary to satisfy an objective function. In general, the objective is to minimize the loss encountered throughout the learning process. A type of supervised learning involves feeding data samples and their matching outputs to a model. A framework works to get the generated output nearer to the desired result by comparing its results to the planned outcome and taking into account any variations. Optimization algorithms have been utilized to do this. Multiple cycles of optimization are carried out till the process is completed in order to increase the accuracy of the model. Numerous optimization techniques have been developed to address the challenges associated with the learning process. The techniques addressed involve Adadelta, Adagrad, Adam, and stochastic gradient descent. With Adam and Adagrad, Pond1 has a desirable training result of 0.96 at epoch 200, and Pond2 has an optimal training result of 0.96 at epoch 250. For Ponds 1 and 2, the best Adam test results are 0.9936 and 0.9975, respectively. Except for AdaGrad, which yielded the lowest scores, the results show that Adam-based optimizers gave the best results. Depending on the surroundings, an aquaponics system with a well-optimized FNN may help make more accurate predictions of fish development. 

Today, one of the main concerns that the world faces is climate destruction. The hazardous gases emitted by material waste are a critical component among many that contribute to the constantly changing atmospheric conditions. Waste from agriculture is also included in this group. Controlling the production of these gases largely depends on accurately predicting the amount to be produced. The current study offers new classifiers and prediction models based on machine learning to forecast the generation of hazardous gases from agricultural waste. The Random Forest method achieved 100% classification accuracy after feature optimization and selection were applied to the gathered dataset. Logistic regression, AdaBoost, and Decision Tree ranked second, third, and fourth, respectively, with classification accuracies of 97.19%, 97.18%, and 97.14%. SVM provided the highest prediction accuracy, 98%, in regression models.

The central theme of the chapter was to demonstrate how Ethiopian cooperative coffee farmers can make informed financial management decisions through the mediation of human capital and AI. The investigators employed a cross-sectional examination using the multistage method with the test group. An Exploratory Factor Analysis (EFA) was conducted to verify whether the data were sufficient. Eigenvalues were utilized to explore the principal proxies in the dataset. A Confirmatory factor analysis was used to assess whether the manifested determinants could appropriately describe the validity of the construct. SEM was used to conduct mediation analysis and assess model fitness. According to the results, the investigation revealed a substantial association, as the p-value was below the algebraic threshold for data adequacy, and the sphericity was 0.893. The chi-square was below 3.0, which the investigators found indicative of model adoption. By Tucker Lewis's fitness, the model was manifested as a fit. The predictive ability of farmers has increased by 75% when coffee cooperatives practice human capital mediation. Consequently, there is a fractional conflict between the sustainability of cooperatives and their management of financial methods, brought about by human capital and AI applications. 

The Internet of Things (IoT) can access certain subsets of data for the creation of several digital services, while also integrating a wide range of diverse and heterogeneous end systems in a transparent and seamless manner. This chapter examines innovative solutions to key challenges in smart cities, with an emphasis on ambient noise assessment, air quality monitoring (E-nose system), and smart parking management systems (SPMS). Utilizing servo motors, smartphone apps, and sensors, the SPMS combines Internet of Things technology to maximize parking space distribution, improve user experience, and raise total parking efficiency. An E-nose air quality monitoring system with a sensor array based on the ESP32 can identify gas pollutants, dust, and CO2. Techniques for processing data, such as threshold alarms and median computation, provide precise and prompt air quality monitoring. Additionally, a noise monitoring system with a Gaussian filter and sound level sensors is demonstrated. The proposed analysis demonstrates how IoT-based solutions may be used to enhance sustainability, manage urban infrastructure more effectively, and improve people's quality of life in smart cities.

Traffic congestion is exacerbated by the proliferation of vehicles, ineffective traffic management practices, and inadequate road infrastructure. Especially in cities, people are struggling to have quick mobility. Improper traffic management leads to longer transit times and contributes to fuel waste, air pollution, and increased accident risk. Conventional traffic control systems set the service time for each lane based on its nature. For example, fast lanes may get more service time than slow ones. It is a static system. It uses fixed timing to service all the lanes irrespective of traffic density. It results in increased traffic in a specific lane. In most countries, automated systems are unavailable to service emergency vehicles (ambulances, firefighting vehicles, mobile medical units, blood donation vehicles, and VIP vehicles) due to the enormous implementation and maintenance costs. In this chapter, a low-cost edge device is proposed for an intelligent traffic control system to address this issue. The proposed system focuses on dynamic signaling and automatic green enablers for emergency vehicles. Dynamic signaling controls the timing of the traffic control system based on the lane's traffic density, which is measured by the You Only Look Once (YOLO V3) algorithm. An automatic green light enabler enables the green light in advance for the lane where the emergency vehicle's presence is 750 meters away from the traffic signal. It deals with various types of emergency vehicles, such as ambulances, firefighting vehicles, and mobile medical units. Computer vision and communication technologies are utilized in designing a cost-effective and innovative traffic control system. It assists in efficiently managing traffic. It also services emergency vehicles at road intersections. Decreasing the average waiting time for emergency vehicles at road intersections contributes to society by saving lives.

In the commercial world, customers are paramount, and understanding their behavior patterns can lead to highly impactful business decisions. Customer churn, defined as the rate at which consumers switch from one business to another, presents a significant obstacle, especially for businesses that rely on subscription models, such as telecommunications companies. Accurately forecasting customer turnover enables firms to effectively implement targeted retention initiatives and minimize income loss. The objective of this study is to employ logistic regression in order to forecast customer attrition for a telecoms company. We utilized a dataset that comprised comprehensive client information, encompassing demographics, service usage, and payment records. Through a thorough exploratory data analysis (EDA), we identified important patterns and trends that are directly linked to customer attrition. We performed data preprocessing by addressing missing values, encoding category variables, and normalizing numerical features. A robust statistical technique, specifically designed for binary classification, was employed to construct a predictive model. The discovered key predictive variables included tenure, monthly charges, and contract type. Our data indicate that clients who have been with us for a shorter period have higher monthly rates, and those on month-to-month contracts are more likely to discontinue their services. The regression model attains the highest accuracy, surpassing other machine learning techniques in terms of overall predictive capability.

This study highlights the efficacy of this instrument as a valuable tool for forecasting customer turnover, enabling firms to enhance customer satisfaction and reduce churn rates by making informed decisions based on data. 

Ensuring safe and clean water availability is vital to the health of not only human beings but also all species. Additionally, it is crucial for the sustainability of the environment. With the emergence of advanced technologies like machine learning, predictive models can significantly contribute to assessing and managing water quality. Current research proposes a methodology that predicts water quality using several machine learning classifiers on a dataset comprising diverse parameters, such as pH levels, dissolved oxygen, turbidity, and other pollutants, collected from multiple water sources. Initially, the data were preprocessed to remove missing values and outliers. Feature engineering was employed to identify the most relevant parameters that contribute to water quality. Several popular machine learning classifiers, including Random Forest, Support Vector Machines, Decision Trees, and XGBoost, were evaluated and compared for their performance in predicting water quality. The trained models were validated and tested using cross-validation techniques to ensure generalizability and resilience. The research findings demonstrated that the proposed method is effective in accurately forecasting water quality levels. The XGBoost, in particular, exhibited superior performance with high accuracy and minimal overfitting. Additionally, feature importance analysis revealed key factors influencing water quality, providing valuable insights for policymakers and environmentalists.

In the 21st century, the internet has become a ubiquitous platform for exchanging information, contributing to an extreme daily influx of approximately 2.5 Exabytes of data. Statistics show that a substantial 1/5th of the data transmitted over the internet comprises videos. With 30,000 hrs. of videos uploaded on YouTube every hour, the need to evaluate and derive valuable insights from this massive amount of video data becomes apparent in a dynamic landscape. Particularly for professionals engaged in internet exploration for various purposes, such as prominent data analysts, data scientists, and cybercrime analysts, it becomes apparent that they need to have an understanding of it. Though traditional machine learning-based systems exist, they have become obsolete due to modern times’ issues, including decreased performance, greater computing complexity, and resource-intensive processes. GenAI has effectively addressed these problems, which is considered a transformative force with enhanced accuracy and semantic comprehension. This chapter focusses on the task of video summarization, which has been poised to supplant antiquated methodologies with the cutting-edge capabilities of GenAI. We have also applied the concepts of hybridization and fine-tuning of models in machine learning to create an improved version of generative models. The proposed methodology has demonstrated better performance in obtaining information, understanding computational complexity, etc. 

Working in tandem, the Hadoop Distributed File System (HDFS) plays a crucial role in managing large-scale data across distributed clusters. The basic structure of HDFS comprises the NameNode, which manages the file system namespace and client interactions, and the Data Nodes, which handle storage and data arrangement duties. Although HDFS faces constraints, particularly regarding scaling the NameNode, which is responsible for managing metadata, these constraints impact namespace scalability and performance. Therefore, this paper discusses the two architectures, the Quorum Journal Manager (QJM) components and the active and standby NameNodes, which enable the sharing of edit logs between them. To mitigate the risk of metadata corruption from split-brain scenarios, the system ensures that only a single NameNode writes to the JournalNodes. Through the utilization of the QJM, HDFS attains High Availability (HA), enabling swift failover in the event of machine crashes or scheduled maintenance. Dynamic Federated Metadata Management (DFMM) addresses the limitations of the current Hadoop architecture by distributing metadata management across multiple federated components. It entails dispersing metadata management among various federated components. HDFS can overcome the problems of connected block storage, namespace scalability, and performance bottlenecks by dynamically managing metadata on an exabyte scale. This work summarizes the key findings and presents a comparative analysis of both architectures. QJM performs a vital function that ensures the high availability, reliability, and fault tolerance of the NameNode in Hadoop’s HDFS. DFMM enhances the fault tolerance, scalability, and performance of distributed file systems by distributing metadata across multiple servers or nodes.

Gold has long been regarded as a valuable asset and a reliable haven for investors, particularly during periods of economic uncertainty. Gold price forecasting is difficult due to the market's intrinsic volatility and sensitivity to various factors, including global economic trends, inflation rates, currency value fluctuations, and geopolitical events. This study presents a novel strategy for predicting gold prices that combines Long Short-Term Memory (LSTM) neural networks and Particle Swarm Optimization (PSO). Traditional models struggle with these complications of uncertainty, but the PSO-LSTM technique improves LSTM hyperparameters for greater accuracy. The PSO-LSTM model beats both standalone deep Learning models and traditional forecasting, based on performance criteria such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE).