Background: Modern agricultural operations collect data from a variety of sources that provide a better knowledge of the constantly shifting conditions of the crop, soil, and environment. This suggests that the processes involved in agriculture will become more and more data-driven. The goal of this study is to demonstrate how to handle diverse data and information from actual datasets that gather physiological in nature, biochemical processes. The agricultural industry only seems to be resistant to digital technological advances, and the "smart farm" concept is becoming increasingly common by using time-series data and the Internet of Things (IoT) paradigm to apply environmental and historical information. In recent years, deep programming has been effectively used for voice recognition, picture recognition, and processing of natural language. Aim: By examining cloud data with crop development trends, examine the potential of deep learning algorithms to optimise the use of agricultural resources, such as water, fertilisers, and pesticides. Method: The present study focuses on the design and implementation of real-world tasks, such as predicting agricultural harvest or recreating data from missing or incorrect sensors, by comparing and using different machine learning algorithms to recommend which way to spend efforts and resources. Results: The results of this study demonstrate the manner in which there are plenty of potential possibilities for innovation to coexist with requests and requirements from businesses who want to establish an optimised and sustainable agriculture industrial use business, making investments not only in technology but also in the expertise and skilled employees necessary to make the most of it. Conclusion: The conclusions presented in this study suggest that better accuracy and faster inference times may be attained by using novel deep learning techniques, and that applications in reality can benefit from the models. Lastly, a few suggestions are made for future study directions in this field.
In recognition of the growing global population, the agricultural sector uses around 85% of
the freshwater that is readily accessible, necessitating a rise in food production. Challenges with
the traditional irrigation management approach include inadequate production and inefficient use
of water. Furthermore, the dynamics of global warming and climate change often have an
influence on the quantity of rainfall that is required to provide plants with water. Similar to this,
the water needs and biological functions of plants are seasonal, vary from plant to plant, and are
impacted by external elements like the weather [
Today, digital agriculture refers to agritechnology and precision agriculture. It is a new
field of study that uses data-intensive methods to increase agricultural output while reducing its
detrimental impacts on the environment. In contemporary agricultural operations, data is gathered
from many different kinds of sensors, photos, and satellite imagery [
Artificial Neural Networks (ANNs) find use in hydrological research such as microclimate
prediction, rainfall-runoff prediction, groundwater level prediction, urban flood forecasting, and
water supply and quality monitoring. Because Artificial Neural Networks (ANNs) can evaluate
tremendous amounts of data fast and effectively [
It has been discovered that adding additional statistical and machine- learning techniques
may increase the predictability of ANNs for agricultural microclimates. A helpful linear method
for assessing multichannel time series statistics with time-varying dynamics and finding similar
patterns across many time series is the Dynamic Factor (DF) model [
Agricultural operations are going to be more and more data-guided as smart tools and sensors multiply on farms and the sheer number and range of agricultural data increase. Conversely, however, the rapid advancement of cloud computing and the Internet of Things (IoT) is driving the growth of Smart Farming. While Smart agricultural takes into account the scenarios generated by occurrences in real-time, Precision Agriculture just relates to managing agricultural variability. With the help of everything mentioned above, farmers are able to respond swiftly to unforeseen events, such disease or weather-related alerts, or to abrupt changes in their operational environment. Typically, such characteristics include astute support throughout the adoption, upkeep, and usage of the technology.
As high-performance bioinformatics technologies, both machine learning and big data have
emerged as new avenues for deciphering, quantifying, and comprehending data-intensive
processes in the context of agricultural operations [
The application of Machine Learning (ML) algorithms to big data has long been a crucial
area of study, therefore assessing the effectiveness and quality of both new and old ML methods
has gained significant importance [
A branch of artificial intelligence called Machine Learning (ML) use computer algorithms
to transform unprocessed data from the actual world into usable models and recommendations for
actions. The system may autonomously acquire information from past events and advance by
using machine learning models. Support Vector Machines (SVM), [
ML incorporates Deep Learning (DL) as a subfield. DL algorithms are more intricate than
those of conventional ML models. The layers of a network that are between the input and the
output are known as hidden layers. A deeper network contains several concealed layers, while a
shallow network just has one [
Implement deep learning algorithms that can recognise symptoms of crop illnesses or stress in cloud photos, enabling farmers to take early action to stop yield loss. Investigate at ways to reduce the cost, increase accessibility to cloud farming technologies for a larger group of farmers, taking into account infrastructure needs, technological know-how, and other variables.
Provide training materials and instructional resources to help farmers and other agricultural professionals use cloud agriculture technology.
(Khan, A., Hassan, M., 2023) [
(Ojo, M. O., 2022) [11] Higher yields, reduced space requirements, and resource efficiency characterise the unconventional production method known as Controlled Environment Agriculture (CEA). Recent advances in CEA have brought Deep Learning (DL) to the field for a variety of purposes, such as microclimate prediction, irrigation, and crop growth prediction, stress both abiotic and biotic detection, and crop monitoring. Nevertheless, no review research evaluates the present situation of the art in DL to address various CEA concerns. In order to close this gap, we thoroughly examined DL techniques employed during CEA. A set of guidelines for inclusion and exclusion were followed in order to create the review framework. Following a thorough screening procedure, we examined 72 paperwork in total to obtain the correct information.
(Pabitha, C., 2023) [12] Agriculture has a significant impact on an economy's growth. The suggested approach investigates how using digital footprints might enhance farming methods and yield. Digital data related to agriculture is becoming more and more accessible due to the advancement of contemporary technology and the proliferation of interconnected gadgets. Digital footprints that capture all aspects of agricultural production lifetime, from planting to harvesting, may be created using this data. After that, farmers may use algorithms that use machine learning educated to analyse these electronic records to identify trends and predict outcomes to determine when to plant, irrigation, fertilise, and harvesting their crops.
(Guillén, M. A., 2021) [14] The digital revolution is being propelled by the Internet of Things (IoT). AL Palliative measures include the fact that almost every economic sector is becoming "Smart" as a result of the Internet of Things' data analysis. Advanced Artificial Intelligence (AI) approaches are used to do this study, yielding insights never previously possible. AIoT is a new trend that is arising from the integration of IoT with AI, providing new avenues for digitalization in the modern day. But there is still a significant difference among AI and IoT, namely in the amount of processing power needed for the former and the deficiency of computing resources provided by the latter type of technology.
(Cubillas, J. J., 2022) [15] In any industry that produces goods, predictive systems are an essential tool for directing and making choices. Knowing ahead of time how profitable a farm is is particularly interesting when it comes to agriculture. In this way, major choices that impact the farm's financial balance may be made based on the season during which this knowledge is accessible. The goal of this project is to create a useful model for anticipating crop yields months in advance that farmers and farm managers may utilise with ease via a web-based application.
(Marina, I., 2023) [16] Encouraging social well-being and meeting the world's food demands depend heavily on agriculture. As a strategically important food crop, soybeans provide vital amounts of protein for both people and animals, and their nitrogen fixation improves soil fertility. However, producers face difficulties due to the increasing demand for soybeans worldwide, especially with regard to cultivation efficiency. Threats from diseases, changes in commodity prices, land usage, and climate change all make these problems worse. Technological developments in the agricultural sector, including Internet of Things, artificial intelligence, remote sensing, and predictive modelling, have great potential to increase both the effectiveness and productivity of soybean farming.
In order to offer advances for the management of data and assessment in small-size manufacturing businesses and, [17], in contingent geographical settings that are often resistant to creative thinking, this effort aims at demonstrating practical and empirical results.
Three separate information sources are taken into consideration during this research (Figure 1), each one of which has distinctive and complimentary qualities that are helpful for designing and testing machine learning techniques: Crp. Type
Apple
Pears Temp. (max) 14.2 14.9
( + )2
agricultural products, and it also shows the proportion of errors for each of the three prediction models. The error mean values for the municipalities of Friuli the Venezia Giulia in Abruzzo, and Calabria is show that the neural network model works best on the linear regression for both the apple plant (9.19% vs. 30.77%) and the pears plant (19.36% vs. 39.11%). dataset, the crop error prediction for apples and pears.
Italian Province
Prediction Error-Apple
Perdition Error-Pears
LR 2.54% 6.4% 14.5% 6.4% 14.25% 44.09% 1
In this work, the polynomial simulation model best matches the prediction of LAI values for the three culture under thought as shown using the predicted errors shown in Table 5.
Pear 1563.62% 41.65% 10.00% Pacciamata Eggplant 986.6% 256.1% 6.98% The matrix of correlations shown in Table 6 extends a correlation coefficient to a set of a component pairs, which are helpful to detect whether there are additional connected features in addition to the geographical ones, by taking consideration of the previous clusters formed by three monitoring stations.
Big Data makes land mapping for large-scale agricultural production possible via remote sensing. It is crucial to keep an eye on how agriculture is affecting different nations and regions in the context of reaching their targets for ecological responsibility and productivity [19]. It also serves as a foundation for the creation of structures for policy makers, aids in decision-making for the long-term sustainability of ecological ecosystems, and provides highly accurate and precise quantitative examination of the interactions between plants and their surroundings. The cloud's accessibility to satellite picture data makes all of the preceding feasible. Cloud technologies, however, prove suitable for the necessary analytics [20]. This makes it easier to create new frameworks for big data that make appropriate use of machine learning methods.
By understanding the fundamental connections between the data gathered by converting it into information and other resources, Machine Learning (ML) is used to perform categorization and predictive analytics. Additionally, it conducts a variety of computing methods, [21, 22], comprising statistical analysis, image processing, modelling, simulation, prediction, and early warning, and it offers information assistance for novel operations.
For numerous Big Data usage in agriculture, cloud computing offers platform, hardware, software, and infrastructure services [23, 24]. The cloud platform makes it simpler for firms by lowering the cost of storing by providing farmers with inexpensive data storage services for text, photos, videos, and various other agricultural data.
As a result, the DL model may not be universally applicable. For example, if a model has been trained using a dataset from a specific site or an open-source site like ImageNet, it may not be capable of to be effectively used at another site, or its accuracy might decline when applied to the data set collected within the real world [24, 25]. Neither the environment is distinct in the field of agriculture, and every circumstance nor difficulty necessitate its own dataset. Model performance may suffer as a result of the variations in the physical appearance of the pictures in the training and evaluation datasets [26]. Retraining the previously learned model using a tiny dataset from a fresh setting is one method to get around issue [27].
Deep models, sometimes referred to as "black boxes," have intricate designs. One of the difficulties in training deep learning models is the need for a system with a high level of GPU power. Furthermore, the selection of the optimisation technique, loss functions, and hyperparameters that affects how well these models work [28]. Bayesian optimising is one algorithm that may assist in determining the appropriate hyperparameters. Scientists from Google developed the most advanced MobilenetV3 by using the Neural Architecture Search (NAS) method [29, 30]. NAS is a technique that looks for each potential pairing of submodules of that can be continually placed altogether to produce the whole model accurately as feasible.
The research presented in this study deepens the understanding of the smart farm model by introducing beneficial, affordable, and simple-to-develop tasks that can boost an agricultural company's productivity. Technological advancements in fields requiring control and optimisation can actually help preserve the environment, adhere to international and business laws, meet consumer demands, and pursue financial objectives.
Both machine learning and more conventional statistical techniques have been used to leverage the three distinct data sources, with a focus on the IoT sensors dataset. In the initial exercise, a neural network framework with near-ninety percent success rate was able to forecast the total crops of apples and pears on the Istat dataset; in the second task, however, it was found that polynomial anticipatory and regression models were more appropriate for the CNR scientific data due to the dataset's characteristics.
In fact, IoT systems need science and technology and diffusion expenditures that only a wise and imaginative administration can encourage in smart/medium industries; furthermore, the need to invest in knowledge and abilities in order to economically employ the paradigm of the Internet of Things at greater scales emerges from the proposed real cases, which emphasise the necessity of promoting administration and data investigators.
The primary motivation behind the suggested tasks utilising different strategies for machine learning is the use of an experimental and highly hypothetical work; information fusion, along with the corresponding optimisation of methods and outcomes, will be expected as further work, where new tasks and experiments that take advantage of other sensors types and databases will be planned and carried out in order to address the significant diversity of the hardware sensors market and agri-companies.
In further work, we want to further enhance the big data and machine learning framework for agricultural and then apply it to a modest smart enterprise.
[11] Ojo, M. O., & Zahid, A. (2022). Deep learning in controlled environment agriculture: A review of recent advancements, challenges and prospects. Sensors, 22(20), 7965. [12] Pabitha, C., Benila, S., & Suresh, A. (2023). A digital footprint in enhancing agricultural practices with improved production using machine learning. [13] Guillén, M. A., Llanes, A., Imbernón, B., Martínez-España, R., Bueno-Crespo, A., Cano, J. C., & Cecilia, J. M. (2021). Performance evaluation of edge-computing platforms for the prediction of low temperatures in agriculture using deep learning. The Journal of Supercomputing, 77, 818-840. [14] Cubillas, J. J., Ramos, M. I., Jurado, J. M., & Feito, F. R. (2022). A machine learning model for early prediction of crop yield, nested in a web application in the cloud: a case study in an olive grove in southern Spain. Agriculture, 12(9), 1345. [15] Marina, I., Sujadi, H., & Indriana, K. R. (2023). Optimizing Soybean Cultivation Efficiency through Agricultural Technology Integration in Plant Monitoring System. Greenation International Journal of Engineering Science, 1(2), 115-127. [16] Boulard, T.; Baille, A.; Lagier, J.; Mermier, M.; Vanderschmitt, E. Water vapour transfer in a plastic house equipped with a dehumidification heat pump. J. Agric. Eng. Res. 1989, 44, 191–204. [17] Jolliet, O. HORTITRANS, a Model for Predicting and Optimizing Humidity and Transpiration in
Greenhouses. J. Agric. Eng. Res. 1994, 57, 23–37. [18] Al Fahoum, A.S.; Abu Al-Haija, A.O.; Alshraideh, H.A. Identification of Coronary Artery Diseases Using Photoplethysmography Signals and Practical Feature Selection Process. Bioengineering 2023, 10, 249. [19] Al Fahoum, A.; Ghobon, T.A. Performance Predictions of Sci-Fi Films via Machine Learning. Appl.
Sci. 2023, 13, 4312. [20] Zheng, W.; Zhao, P.; Chen, G.; Zhou, H.; Tian, Y. A Hybrid Spiking Neurons Embedded LSTM Network for Multivariate Time Series Learning Under Concept-Drift Environment. IEEE Trans.
Knowl. Data Eng. 2023, 35, 6561–6574. [21] Zhu, C.; Ma, X.; Zhang, C.; Ding, W.; Zhan, J. Information granules-based long-term forecasting of time series via BPNN under three-way decision framework. Inf. Sci. 2023, 634, 696–715. [22] Abu-Qasmieh, I.; Fahoum, A.-A.; Alquran, H.; Zyout, A. An Innovative Bispectral Deep Learning
Method for Protein Family Classification. Comput. Mater. Contin. 2023, 75, 3971–3991. [23] García, I.F.; Lecina, S.; Ruiz-Sánchez, M.C.; Vera, J.; Conejero, W.; Conesa, M.R.; Domínguez, A.; Pardo, J.J.; Léllis, B.C.; Montesinos, P. Trends and challenges in irrigation scheduling in the semi-arid area of Spain. Water 2020, 12, 785. [24] Chlingaryan, A.; Sukkarieh, S.; Whelan, B. Machine learning approaches for crop yield prediction and nitrogen status estimation in precision agriculture: A review. Comput. Electron. Agric. 2018, 151, 61– 69. [25] Celicourt, P.; Rousseau, A.N.; Gumiere, S.J.; Camporese, M. Hydro-informatics for sustainable water management in agrosystems. Front. Water 2021, 3, 119. [26] Maduranga, M.W.; Abeysekera, R. Machine learning applications in iot based agriculture and smart farming: A review. Int. J. Eng. Appl. Sci. Technol. 2020, 4, 24–27. [27] Goap, A.; Sharma, D.; Shukla, A.K.; Rama Krishna, C. An IoT based smart irrigation management system using machine learning and open source technologies. Comput. Electron. Agric. 2018, 155, 41–49. [28] Koech, R.; Langat, P. Improving irrigation water use efficiency: A review of advances, challenges and opportunities in the Australian context. Water 2018, 10, 1771. [29] Patil, S.S.; Thorat, S.A. Early detection of grapes diseases using machine learning and IoT. In Proceedings of the 2nd International Conference on Cognitie Computing and Information Processing (CCIP), Mysore, India, 12–13 August 2016; pp. 1–5. [30] Truong, T.; Dinh, A.; Wahid, K. An IoT environmental data collection system for fungal detection in crop fields. In Proceedings of the IEEE 30th Canadian Conference on Electrical and Computer Engineering (CCECE), Windsor, ON, Canada, 30 April–3 May 2017; pp. 1–4.