Bauchhave and Kersting evaluate the IT professionals’ recent works in employing computational intelligence for agriculture, especially as part of the data mining and pattern recognition process. The authors first survey the papers focusing on their contribution to the data management model, geographic information dependency, and signal translation. Next, two experiments in processing images to recognise drought situations and classify leaf spots are demonstrated in detail to support the author. Finally, the challenges and opportunities are revealed for prior and future research on the matter. The agriculture system has distinctive characteristics and requirements which generate highlighted issues in integrating the data mining due to the off-the-shelf algorithms, distances and devices’ restrictions. The study provides a thorough experimented strategy and its effectiveness in bypassing the claimed limitation. This offers a deep understanding of the current problems, introduces a tactic to overcome them and opens a new path for further research.

The paper describes the development and application of a model for digital farming management in irrigation and fertirrigation based on the PLATEM system. The author first explores the research with similar approaches to gain knowledge and sketch the experimented system. Then, a model is designed from the data mining techniques on the open information network and its backend software framework to the web view for farmers with individual customised features. The result presents the performance of the proposed integrating rule-motivated model in a shared environment with all farming tools in one place. Rather than accuracy rate, the popular measure for success used in other studies, this experiment emphasises the advantages of the public dataset and rule engine generated from the shared information pool to the digital system that enhances and simplifies daily farming activities. It also points out the future research paths in data processing: local data analysis and open community collaboration. This suggests new criteria for evaluating the achievement of a machine learning application in agriculture and a potential research approach in the industry.

This article investigates the adaptation of data mining reinforced by IoT in agriculture, focusing on the resource management system. Gao first studies the current IoT and data mining approaches used in information supervision structure to evaluate the emerging technology adaptation accurately. Thus, the author designs two models with distinct functions to experiment with their performance in various queries and concurrent user scenarios. With those results, two recommendations for further the application of data mining in the farming industry are proposed based on theoretical and experimental analysis. The findings of this paper are explained comprehensively with complete views of the used technology, the process, the input data and testing methods. However, the examined data set is simplified for experimentation purposes and does not present the complexity level of the actual database. Though, it demonstrates the factors affecting the productivity of the developed model, the effectiveness of different techniques and their limitation under the circumstances, which are the primary purpose of this project.

The paper demonstrates the precise automatic irrigation and SMS notification system for Harvey Mudd College lawn, using weather conditions and manual watering patterns as inputs of the core big data analytic algorithm. To overcome problems with rodents damaging the installed underground sensors and pipeline, data are retrieved from online sources and subject matter experts (SMEs) instead. The model was built based on internal factors, such as grass species and campus geometry sections, and external factors, including weather and irrigation patterns by seasons. The algorithm using Maximum Likelihood Estimation, K-Mean Cluster, Gaussian Mixture Model and Principal Component Analysis is the core for the automated lawn irrigation system, which notifies the user through SMS of the water amount used each day. The experiment proved the essential of a customising system in applying to the realistic situation in farming and gardening as the conditions and the purpose of the activity generates unique needs and requirements. Still, generic knowledge and data, including shared online sources, support the development of systems in similar sections. However, a down point of this article is the short and obscure evaluation based on the system performance in a few events without a precise measurement method. Nevertheless, it is a useful example of machine learning in practice that highlights various aspects expected to answer the research question about domains, algorithms, limitations, and requirements of current projects in the field.

The survey’s main topic is using Artificial Intelligence (AI) and machine learning to achieve farming automation. It shows the summarised investigation of AI’s evolution over the past five decades in handling the most significant agricultural problems in crop, storage, pest and disease, weed and irrigation. Next, intensive exploration of automation in farming focusing on using artificial neural networks (ANN) and fuzzy logic, as well as the wireless network, is discussed. Those study findings are used as the base to propose a productive model and further research direction. The authors present a deep analysis of two data learning techniques with the evaluation form on the model precision. While it concentrates on determining the algorithms and the physical network supporting the system structure, the software requirement served as the medium between those two components is not inspected. It is a valuable contribution to the research topic of mining techniques’ comparison and infrastructure requirements.

The paper explores the implementation of sensing mechanisms and learning from big data in the innovative farming system. Liu reviewed the related works in the domain and proposed a collaborative model of multi-generation genetic algorithms with the backpropagation mechanism. The model is introduced from each algorithm explanation to the combination process. Then, a platform was constructed using the ZigBee model to connect the sensor network and multi-sensor data fusion technology to test the proposed model. The result proved the model’s practical value in the field’s implementation. Even though the new model is validated for its innovation and performance, Liu’s work is mainly dedicated to the software and the model part of the experiment. Their complexity raises uncertainty about the hardware required to support heavy data processing and its practical aspect in installing such a network in a standard farming setting. This experiment is an authentic source for finding a more efficient approach to agricultural domains and revealing the industry’s limitations to data mining evolution.

The study reviews the machine learning application to process big data in agricultural animal management. First, the author provides an overview of big data and its position in the farming industry, especially animal science. Next, the data processing procedure is systematically presented, from the goals, advantages, and steps to the technique options. Then, the implementation of the above technologies is demonstrated through examples specific to animal management, surrounding genetic, health and development issues. The analytics reveals the constraint in infrastructure and tools to support the data mining features and proposes further research needs in the animal science domain. Morota et al. (2018) give a distinctive yet generalised perspective of machine learning adaptation to the animal farming sector. Outlier detection is a specific performance measure for this domain, as genetics and genotype play a crucial role in the evolution of biological enhancement. This work will contribute to investigating the research topic to highlight the distinguishing feature of agriculture, the performance measurement and the limitation of a specific domain’s sector.

The paper describes the experimental projects about IoT and data analysis integration in crop management at 3 Thailand villages. The authors research similar professional’s work in the IoT application and the knowledge discovery procedure based on IoT device data collections. The testing system was designed to include three main components: hardware, web application and mobile application. It was installed in 3 located apart villages with distinguishing crops structure. Association rule is the core data mining technique used in the system to generate rules and models to monitor and manage natural impacts on crop yield. The result shows the advantage of sustainable farming using automation and data science technology in enhancing productivity, avoiding hazards and reducing labour demand. Besides providing a comprehensive view of the association rule method in practical settings, Muangprathub et al. (2019) present the benefit of implementing a cloud server-based data analysing system where the processing is centralised with a low hardware requirement at the field, leading to the low installation cost. However, it is a lack of a result comparison between all three villages’ applications to demonstrate the system performance in various conditions. Still, the impacts of multiple elements on each other and the crop yield as well as the benefits and limitations of hardware in knowledge discovery in this experiment are relatable to the research questions.

In this chapter, the author examines the integration of data mining techniques in modern agricultural practices. As the foundation of the investigation, Muniasamy explains popular terms surrounding the digital farming concept and the systematic model of the Smart Farming system combining all experiments around the field. The evaluation of data mining techniques application in existing frameworks from various domains reveals key issues: scope and complexity of data set; imprecise, inconsistent old input; distributed information network and the approachability of technology. Therefore, employing IoT for better accurate statistics and selecting appropriate analysis techniques are recommended strategies to overcome the issues. As the knowledge is examined thoroughly, from the basic definition to the scientific investigation of the mining methods, it offers valuable support for studying data science’s role in the agricultural firm, especially algorithm effectiveness and their limitation in collecting and processing raw data. However, the research needs further review hardware requirements to support such complex systems

This systematic study reviews 45 academic papers to display an overview of data analytics implementation current situation of smart farming. The research questions encircle the statistics of mentioned agricultural domains, stakeholders, integration goals, data mining technology, information characteristic, application hazard and strategies. The crop takes the lead in popular research domains, with the top identified stakeholders being farmers, researchers and agronomists. The primary purpose of applying data science in farming is to increase field productivity and manage natural elements. Therefore, the favourite techniques used in those setting are descriptive and diagnostic analytics. The paper points out that source data characteristics are the biggest concern for the experts in the technology employed in the farming system. This is a high-quality literature review about the topic with an extensive analysis from many perspectives of the smart farming transformation supported by modern technologies. The research questions are similar and fit the study topic, with statistics being visualised and presented systematically through diagrams, graphs and tables. It proposed worthwhile suggestions on how the study should be conducted and its direction to target the audience efficiently.

The researchers describe the integrated system of IoT, mobile devices and big data analytics on the cloud in monitoring the soil and fertilisation conditions, as well as crop management from planting to stock and market requests. Several previous papers are reviewed to design the best solution suitable for Indian crops. Data collected from IoT sensors is sent to a cloud database server for storage and analysis. Existed cloud-based Big Data methods are selected to preprocess the raw data into formatted versions before MapReduce is used to generate exciting knowledge, patterns and prediction for the farmer. Stakeholders can access that information through notification. Generally, the authors explain the structure of the integration between different technologies in the farming monitoring system. In addition, they point out the necessity of real-time data updates to the quality of model performance in the farming industry. However, since the described structure is at a high-level, information about the implementation of the system is vague without an explanation of hardware and software requirements, as well as the cost balancing.

Based on the prior exploration of rice and shrimp aquaculture on the Mekong Delta using the Bayesian Belief Network (BNN), the paper examines the association between agricultural conditions to the development and yield of the crops to suggest the optimal approach for the decision-making process.  The foundation data is the developed optimising knowledge BNN model of both crops in the same pond from the Southern Mekong Delta, Vietnam. The main objective of the research is to extract potential interactions between recorded factors with the growing and harvesting life cycle of two crops in each phase, including pre-planting, planting, and post-planting, with a priority of avoiding failure in production. For improvement, the authors employ 3 keys approach: turn the discrete values to root nodes to present case scenarios, trim the built decision tree for a better focus model, and transfer the model to a visualised format for a comprehensive report to farmers. Furthermore, the advantages and disadvantages of the BNN algorithm are fully considered in applying it to the specific agricultural study. Thus, the technique is used for its best fit with acknowledging potential downside.  Although the hardware and software aspects are out of the scope of the study, the data mining method is well explained and give a supporting argument to answer the research question.

The paper investigates the integration of data mining and statistic techniques to resolve time-related agriculture information with missing data and noises and conducts experiments to evaluate the time-series algorithms’ performance. First, the authors thoroughly explained different sequencing algorithms to work with time-dependent datasets, such as ridge regression, lasso regression, elastic network and dynamic time warping. Next, the proposed system in practice provides the physical framework, data sources structure, processing layers, and workflow design. In the end, 92 evaluation experiments are recorded with accuracy and statistical significance to prove the potential performance success. Its result offers valuable insight into employing the data mining techniques in handling big data with time series characteristics, from hardware, software and current algorithms limitation. However, it requires a heavy investment in resources and workflows to process and analyse constantly updated data with high volume and needs instant knowledge discovery.

The paper systematically explores the Internet of Things (IoT) employment in the agricultural sector and its integration with data mining techniques or tracking and monitoring purposes. IoT benefits farming through procedure monitoring, records of natural factors, tracking agricultural activities, and automation-enabling steps. Therefore, the sensors are separated into subtypes based on their goals, including biosensor, environment, mass measurement, motion and position. Furthermore, by integrating with data analytics algorithms, IoT is used in estimation, product protection, storage management, precise application, decision-making and machine learning support. Composed of 3 elements: application, software and hardware, together with a complex communication network, IoT enhances the advantages of farming procedures but also generates several issues, such as security, installation, and standardised communication. IoT technology in agriculture is explained thoroughly, from the development, backend mechanism, and structure to benefits, usages, advantages and limitations. This information answers the study regarding IoT integration with data mining algorithms in farming.

The study examines accuracy as the primary performance measurement in agricultural systems applying data mining.  The examination starts with exploring the term “accuracy” from various perspectives and other surrounding definitions, including big data, algorithms, sensors devices and maps. Then, the potential technical sources of the inaccurate outcomes are analysed, considering GPS technology, sensor data set, mapping, and algorithms’ error term. Finally, the effects of inaccuracy in system opacity, prediction model, rural location and farming gaps are recognised, which require further investment for continuing research. The authors present valuable perspectives on performance measurement and highlight its importance in conducting sector research. It is the foundation for this study to evaluate the performance of the previous projects correctly and investigate the technical limitation of the current research and implementation approaches.

The author studies the strategic approach to integrating technology, specifically data mining, in China’s rural agriculture to promote economic development. To support his proposal, Wang first outlines the main focus of China’s agricultural evolution and structural reform and their objectives in favour of technological advances. He then presents a summary of data mining techniques with further explanation in clustering, especially the outliner detection algorithm expected to benefit the farming prediction best. Hence, a scheme with four targets to elevate the technological level and standard in rural areas is suggested. A short experiment is demonstrated at the end of the report to illustrate the performance of different techniques in a similar dataset. This conference paper provides a meaningful explanation of the movement of the agriculture industry over time and its positive and negative impact. With the proposed plan, the author reveals the constraint of applying IT to most farming activities in regional areas and how to mitigate them. However, the description and evaluation of data mining techniques and their usefulness are limited. The obtained knowledge from this study can support this research on understanding the industry’s requirements and its difficulty in realising the modern farming formation.