Professional Documents
Culture Documents
Sensors 23 03038
Sensors 23 03038
Article
A Methodology Based on Machine Learning and Soft
Computing to Design More Sustainable Agriculture Systems
Jose M. Cadenas , M. Carmen Garrido and Raquel Martínez-España *
Department of Information and Communication Engineering, University of Murcia, 30100 Murcia, Spain
* Correspondence: raquel.m.e@um.es
Abstract: Advances in new technologies are allowing any field of real life to benefit from using
these ones. Among of them, we can highlight the IoT ecosystem making available large amounts
of information, cloud computing allowing large computational capacities, and Machine Learning
techniques together with the Soft Computing framework to incorporate intelligence. They constitute
a powerful set of tools that allow us to define Decision Support Systems that improve decisions in a
wide range of real-life problems. In this paper, we focus on the agricultural sector and the issue of
sustainability. We propose a methodology that, starting from times series data provided by the IoT
ecosystem, a preprocessing and modelling of the data based on machine learning techniques is carried
out within the framework of Soft Computing. The obtained model will be able to carry out inferences
in a given prediction horizon that allow the development of Decision Support Systems that can help
the farmer. By way of illustration, the proposed methodology is applied to the specific problem of
early frost prediction. With some specific scenarios validated by expert farmers in an agricultural
cooperative, the benefits of the methodology are illustrated. The evaluation and validation show the
effectiveness of the proposal.
Keywords: sustainable agriculture; time series forecast; Soft Computing; machine learning; IoT
1. Introduction
Citation: Cadenas, J.M.; Garrido,
The role of technology in agriculture has particular relevance in the improvement
M.C.; Martínez-España, R. A
of agriculture. A complete scheme of activities related to technology in agriculture is
Methodology Based on Machine
Learning and Soft Computing to
presented in [1], where authors analyse and identify sectors that can take advantage of
Design More Sustainable Agriculture
the latest advances in technology. Using new technologies, farmers can monitor their
Systems. Sensors 2023, 23, 3038. farms remotely and more accurately, and therefore, they are considered an essential tool
https://doi.org/10.3390/s23063038 for sustainable agriculture. Sustainable agriculture is a concept that has become very
important today. To be sustainable, agriculture must meet the food needs of present and
Academic Editor: Xiaoshuan Zhang
future generations at prices that are reasonable for consumers and sufficient to maintain
Received: 6 February 2023 the economy of the agricultural sector without endangering the health of the environment
Revised: 27 February 2023 or the quantity of natural resources. Sustainable agriculture is a system of agricultural
Accepted: 8 March 2023 production that is resource-conserving, environmentally sound, and economically viable.
Published: 11 March 2023 The aim is to produce healthy food with respectful practices for the soil, air, and water and
to respect the rights and health of farmers.
Sustainable agriculture faces problems such as fertility loss, water scarcity, biodiversity
depletion, pesticide pollution, climate change, and how it affects agriculture. In particular,
Copyright: © 2023 by the authors.
climate change is causing major losses in agriculture, both in terms of product quality and
Licensee MDPI, Basel, Switzerland.
farmers’ economies. Agricultural insurance is becoming increasingly restrictive and costly.
This article is an open access article
Thus, farmers have to look for new solutions to mitigate crop damage without investing
distributed under the terms and
conditions of the Creative Commons
large amounts of economic and/or environmental resources.
Attribution (CC BY) license (https://
Sustainable agriculture, and specifically precision agriculture, is considered to be
creativecommons.org/licenses/by/ an area of major interest within the objectives of sustainable development. Precision
4.0/). agriculture is a promising task for economic stabilisation in developing countries [2] and
plays an important role in making agriculture more sustainable. It tries to measure crop
qualities, soil, and climatic factors to apply the best treatment at the right place at the
right time.
Precision agriculture is a discipline that began to be implemented in the early 1990s
with the initial objective of increasing profitability and reducing environmental impact by
using integrated systems within information technologies, [3]. Every day more and more
agricultural processes are beginning to have and use technology to improve their proce-
dures, reducing economic and environmental costs and increasing profits, achieving a more
sustainable agriculture [4]. In the process of adapting and contributing to sustainability,
farmers need real and reliable sources of information and knowledge to make decisions
more sustainable. Advances in technology and computational paradigms provide farmers
with new methodologies to improve their crops [5].
Among these paradigms, IoT and cloud computing have become very relevant nowa-
days [6], making a large amount of information available to users in any real environment
such as smart homes [7], the industrial market [8], and agriculture [9] among other areas.
In the agricultural world, IoT and computing have revolutionised the sector, applying
technologies to achieve service upgrades, cost savings, increased production, and better
control and balance between economic and environmental sustainability [10]. In traditional
agriculture, farmers have made decisions based on their experience, and although often
right, at other times, both economic and environmental resources are wasted. Precision
farming, coupled with computing and IoT paradigms, has improved these processes and
reduced the economic and environmental impact. For example, in terms of spraying against
pests [11], traditional farmers sprayed in a preventive way with highly polluting products
without knowing the level of pests they had in their crops. Now, there are crop pest predic-
tion techniques that help to spray at the right time and with the right products [12]. Thus,
now with precision farming, the farmer saves on the cost of crop protection products, and
the environment gains in terms of less pollution. Another example of improving sustain-
ability with new technologies is irrigation control. By predicting temperatures, as well as
predicting soil moisture levels, the amount of water used on crops can be reduced. Again,
new technologies help to achieve economic sustainability in the reduction of water costs
and environmental sustainability in the use of water [13].
Thus, new technologies help the agricultural sector to be more sustainable. However,
many of the proposed techniques are not deployed in real environments, and one of the
problems is the lack of a methodology that includes the steps and the way to implement
the new technologies in agricultural problems. Hence, the novel proposal of this paper
develops a generic methodology, which can be applied to agricultural problems, with the
aim of having a final Decision Support System to make decisions with implications for
economic and environmental improvement. To build a Decision Support System, it is
necessary to have data. The IoT poses through all kinds of sensors to collect and send data
related to farmland. This information collected by sensors, and depending on the scope
of the application, can contain numerical values following the trend of a time series. This
indicates that there are dependencies on one value, the next, and the previous one. The
collected information generates significant amounts of data that can be analysed remotely
and in real-time using appropriate Machine Learning techniques. With the use of these
techniques, patterns can be found from the collected data, and predictions can be made,
providing intelligence to the Decision Support Systems. Studies have shown that Decision
Support Systems applied to agriculture contribute significantly to long-term sustainable
development. However, the full benefits of Decision Support Systems are not yet realised,
as they need to be adapted to the needs of farmers [4].
This paper proposes the design of a novel methodology aimed at facilitating the
creation of a Decision Support System to address agricultural problems. The final system
will be embedded in a mobile device where usability is simple, transparent, and flexible for
the farmer. In this way, the farmer can benefit from the use of applications with specialised
information about crops and techniques to improve decision-making in crop management,
Sensors 2023, 23, 3038 3 of 16
changes in their production systems, etc. The choice of the mobile phone as the final
device is based on the fact that there are studies that indicate that farmers are using their
smartphones more as work tools than for entertainment and that they prefer to access the
Internet using their mobile phone rather than other devices [14].
As the main characteristics of the methodology, we can highlight that it is general to
address agricultural problems to create a Decision Support System using computing and
IoT paradigms, is independent of particular techniques and technologies, is designed to
address problems that improve the environmental and economic sustainability of farmers,
and can deal with one of the disadvantages of the IoT, which are precision errors in sensors
and possible delivery failures. The methodology is embedded in the Soft Computing
framework that allows dealing with these problems using a representation of data that
captures the true nature of them.
The methodology takes as a starting point the information collected in the IoT ecosys-
tem in the form of time series data, and its ultimate goal is the creation of a Decision Support
System embedded in a mobile phone. In a general way, the methodology is defined based
on the following stages: (1) obtaining data from IoT and cloud services, (2) preprocessing
the data in the Soft Computing framework, (3) extracting knowledge using appropriate
Machine Learning techniques, and (4) defining a set of rules that constitute the main axis of
the Decision Support System.
The study developed in this paper is organised as follows: Section 2 presents a non-
exhaustive review of proposals for solving real problems related to sustainable agriculture
from time series data that can obtain advantages of the design of a general methodology.
Section 3 details the proposed methodology by presenting each of its stages. Section 4
presents a study case where the proposed methodology is applied step by step to the crop
frost problem, illustrating both the environmental and economic benefits obtained in terms
of sustainability. Finally, Section 5 shows the conclusions of the study and future work lines.
2. Background
As mentioned above, the role of technology in agriculture has particular relevance in
improving agriculture. The use of paradigms such as machine learning and Soft Computing
can help the development of sustainable agriculture by providing the information obtained
through IoT with Artificial Intelligence [15]. The joint use of these technologies helps the
design of Decision Support Systems that lead to the development of sustainable agriculture.
Within this field, these technologies have been used, among others, in the prediction of
adverse weather conditions, optimisation of water resource management, optimisation of
crop productivity, etc.
In this section, we review, without being exhaustive, some studies that try to introduce
benefits in agriculture to make it more sustainable from the information provided by
time series datasets collected from the IoT ecosystem and the use of Artificial Intelligence
paradigms. The aim is to show the wide range of possible benefits rather than to go deeper
into any of them.
As a consequence of climate change, the construction of Decision Support Systems that
help farmers to be aware of the arrival of possible adverse weather phenomena such as
frost, floods, etc., is getting more and more important. The problem arises because weather
phenomena are becoming more frequent, according to experts [16]. Thus, to address these
problems, we can find various works in literature. In this way, in [17], the authors use time
series datasets to predict short/long term low temperatures by capturing the dependencies
of environmental factors through causal and associative models. In [18], frost prediction is
carried out in advance using deep learning techniques using a long short/term memory
model. In [19], models of the rainfall are obtained using deep learning architectures and
time series datasets to minimise the risks caused by rainfall and reduce the economic risks
and, therefore, maximising profits.
As it is well known, water is a vitally important resource in agriculture and has been
considered one of the critical issues in the future that may threaten food security. For this
Sensors 2023, 23, 3038 4 of 16
reason, studies have been carried out to measure the quality of the water and optimise
the quantity of water used for crops. For example, in [20], the quality of drained water
that is reused for agriculture is studied using neural network models that will help to
manage the resource and make decisions about irrigation water resources. An important
factor in managing the water resource in crop production is the accurate prediction of
evapotranspiration. The control of this factor can optimize the amount of water needed.
To address this control, in [21], the authors present the development of a computational
method for estimating the monthly mean evapotranspiration of arid and semi-arid areas is
carried out.
Another of the most important issue in agriculture is improving crop management.
In this sense, yield prediction will enable the farmer to improve crop management to match
crop supply to demand and increase productivity. In the same way, the productivity and
quality of crops can be improved by detecting the appearance of diseases at an early stage
and carrying out a more appropriate treatment. Thus, in [22], the authors carry out the
wheat yield production from satellite images to receive measurements of crop and soil
growth. In [23], a classification of the leaves of several plants is carried out to distinguish
between healthy and diseased leaves based on a neural network and images of the leaves.
Other work is presented in [24] where the detection of diseases in wheat is carried out to
optimise the use of fertilisers and fungicides according to the needs of the plant.
Regarding the right estimation of crop soil conditions improves soil management by
the farmer. Nevertheless, the obtaining of soil measurements is generally expensive and
time-consuming, and the process can be improved by using Machine Learning techniques.
Some works in this line are [25,26]. In the first one, soil drying is evaluated from evapotran-
spiration and rainfall data to favour decision-making in crop planning. In the second one,
the authors develop a Machine Learning technique to obtain soil temperature at different
depths with the aim of optimising soil management.
As a summary, many agricultural domains can obtain benefits from using new tech-
nologies, and the methodology proposed in this work defines a series of steps to follow to
achieve this objective.
grouping
decision model Remote rules Taking a
attributes Interval/fuzzy Pre-procesing data
model
transformation
Lagged
Instance
Local
selection decision
model
attributes
with imprecise data
Instance selection
with imprecise data
1 time
t-1 t t+1
Instance
with imprecise data
Machine learning
grouping
Remote
Interval/fuzzy model
Pre-procesing data
transformation
Lagged Local
attributes model
Instance selection
with imprecise data
therefore, about historical data that will be used, after the appropriate preprocessing,
to obtain Machine Learning models that allow the construction of Decision Support
Systems. This historical data will comprise successive measurements made by the
different context providers in certain time intervals. Therefore, there is a temporal
dependence between the different instances that comprise the dataset. The use of this
data will be carried out offline.
• Obtaining data from context providers in real-time. Once the Decision Support System
has been built, it is necessary to obtain real-time data from the different context
providers to make an appropriate decision according to the scope of the application.
For example, if the application is designed to monitor the quality of a crop soil,
the data obtained in real-time from the soil properties could generate an alert to
farmers indicating that a certain threshold of shortage of fertility properties of their
crops, such as pH, salinity, and available plant nutrients has been reached. The use of
this data will be carried out online.
The designed application will be integrated into the FIWARE architecture. The FI-
WARE architecture consists of several components stored on a server. This server is
responsible for storing the information received from the sensors and different web services
provided by opendata, among others. In addition, the server is also responsible for making
the data available to the visualisation services in the user interfaces as well as providing
data to the intelligent data analysis module. The FIWARE architecture interacts with the
mobile application through a BackEnd service located on the server that connects the appli-
cation with the Orion Context Broker. The Orion Context Broker is a FIWARE component
that is responsible for providing the data corresponding to the mobile application, this
being a transparent process to the user. This component also connects with the framework
services that collect the data, being weather stations or farmers’ own sensors or open
data provided by meteorological services, agricultural companies, or official information
services, among others. The information collected from the different sources is stored
using a database, which can be structured like MySQL or unstructured like MongoDB, or a
combination of both. The information is received using JSON data type protocols.
Original attributes
{x1,x2,…,xn}
Instance grouping
Added attributes
{minx1, maxx1, medx1, stdx1,…, minxn, maxxn, medxn, stdxn}
Fuzzy/Interval transformation
Fuzzy/interval attributes
{fuzzyx1,…, fuzzyxn}/{intervalx1,…, intervalxn}
Lagged attributes
{…,fuzzyx1(t),…, fuzzyxn(t), fuzzyx1(t+1),…, fuzzyxn(t+1), fuzzyx1(t+2),…, fuzzyxn(t+2),…}
/{…,intervalx1(t),…, intervalxn(t), intervalx1(t+1),…, intervalxn(t+1), intervalx1(t+2),…, intervalxn(t+2),…}
as the size of the prediction horizon. For example, in a problem with hourly information
and a prediction horizon of “x” hours, we will have to add “x” delayed attributes that
capture the temporal correlation between hours. . . , h − 1, h, h + 1, . . .. Depending on the
specific application in which the methodology is being applied, the prediction horizon can
be longer or shorter, and we can talk about short-term time series or long-term time series.
In either case, the inclusion of lagged attributes increases the dimension of the prob-
lem, so the reduction of attributes carried out by the fuzzy transformation allows for the
reduction of this dimension while capturing the true nature of the information. The effect
of this step of the methodology is shown in the last step of Figure 3.
The IMIDA’s weather stations have the following ephemeris and sensors: rain gauge,
radiometer, data-logger, weather vane, and thermohygrometer. The obtained measures
are the following: “Radiation”, “Accumulated radiation”, “Rainfall”, “Humidity”, “Tem-
perature”, “Wind direction”, “Wind speed”, “Dew point”, and “Vapor pressure deficit”.
The IMIDA records these values every 5 minutes and therefore, this stored information
gives rise to the time series datasets that are used as starting point in this study case. A time
series dataset is obtained for each weather station.
Instance Attributes
Minimum relative humidity (%) Mean relative humidity (%)
Maximum relative humidity (%) Minimum radiation (W/m2 )
Mean radiation (W/m2 ) Maximum radiation (W/m2 )
Accumulated radiation (W/m2 ) Wind direction (◦ C)
Minimum wind speed (m/s) Mean wind speed (m/s)
Maximum wind speed (m/s) Rainfall (mm)
Vapor pressure deficit (kPa) Dew point (◦ C)
Minimum temperature (◦ C) Mean temperature (◦ C)
Maximum temperature (◦ C)
In addition, in the datasets obtained after grouping, the instances with the attribute
“Minimum temperature” > 7 are deleted as they are not representative of the problem.
The fuzzy attributes are expressed by a trapezoidal function defined by the values
Version January 19, 2023 submitted to Sensors 10 of 13
[x1 , x2 , x3 , x4 ]. More concretely, the constructed fuzzy attributes are defined by the following
values: x1 = minimum, x2 = mean − 5%mean, x3 = mean + 5%mean, and x4 = maximum,
where minimum, mean, and maximum are the three values available for the same measurement.
the hour h − 1 , those obtained in the hour h , and the value of the minimum temperature 380
As commented above, incorporating fuzzy values in the datasets allows us to incorporate
alsothe
in thehour h + 1 of
treatment . Therefore, fromexpressed
missing values the historical dataset
as fuzzy collected
values from a weather
as [minimum station,
dom , minimumdom ,
381
we will obtain the final dataset by joining the values of two consecutive
maximumdom , maximumdom ] where minimumdom and maximumdom are the extremes of the instances and the 382
attributeof"Min.
domain Temperature"
the attribute. (Tmin ) of the next
This representation of ainstance.
missing The
valuestructure of that
expresses the final
it caninstance
be any 383
This step of the methodology allows us to reduce the set of instances while maintaining 396
the accuracy of the predictions made by the model. This process is particularly important 397
when we want to use the model in local mode and we are going to work with a machine 398
learning technique of the "Lazy learning" category [27] where the model is made up of the 399
where Tmin is the forecast of the temperature value for the next hour. Therefore, the farmer
will receive an alert/alarm that will allow him/her to take the necessary measures to
protect his/her crops one hour in advance.
Table 4. Temperature prediction for the next hour (h + 1) from the current information (h) and the
previous one (h − 1). The information for the time h and h − 1 corresponds to the vector (RH f , R f ,
AR, WS f , WD, RF, VPD, DE, T f ). The time at which the prediction is made, the predicted temperature
one hour in advance, and the real values that occurred, are indicated.
The three scenarios in Table 4 have been validated by two farmers from “Sociedad
Cooperativa Alimer Alimentos del Mediterráneo” (https://alimer.es/ accessed on 30
January 2023) in the Region of Murcia. The farmers have paraffin burners on their plot as
an anti-frost system. On the one hand, if the farmers would not have the prediction system,
and according to the current temperature values in period 1 (Table 4), the farmers indicate
that at 23:00, they would have been alert about the activation of the anti-frost system
(devices, staff, etc.). At 24:00, they would have activated the system and kept it until 10:00,
when they would have deactivated it. On the other hand, with the information provided
by the system on the temperature prediction one hour in advance (Table 4), the farmers
indicate that at 22:00 and also knowing the prediction for 23:00, they would be on alert,
at 23:00 with the prediction for 24:00, they would remain on alert, and so on until 01:00.
At that time they know that at 02:00 the temperature will be below 1◦ positive and therefore,
at 01:00 they would start preparing their anti-frost system to activate it at 2:00. At 09:00
they know that the forecast for 10:00 will be over 2◦ positive and therefore, from 09:00
they would start to deactivate their anti-frost system. Thus, with the Decision Support
System created using the proposed methodology, the farmers have saved two hours of both
economic and environmental resources.
Sensors 2023, 23, 3038 14 of 16
In period 2, and with the information shown in Table 4, farmers act similarly to period 1,
but in period 3 (Table 4) we must highlighted an interesting situation. The farmers tell us
that if they would not have the forecast, and with the current values, at 04:00, they would
have activated the anti-frost system. However, with the forecasting system, at 04:00 they
know the forecast for 05:00, and, therefore, at 05:00, they would be on alert and would make
their decision to activate the anti-frost system based on the forecast for 06:00. As at 5:00 the
forecast for 06:00 remains unchanged, and because of the night is near to end, they would
decide not to activate the system, waiting for the next hour. Thus, in the end, and with the
information available, the farmers would not activate the anti-frost system.
As we can see, with the information provided by the prediction system in these three
situations, farmers make more appropriate decisions and, therefore, can reduce the activation
time of the anti-frost system, or not activate it at all. This leads to savings both in terms
of the devices used and the staff required to activate them and reduces the environmental
impact involved.
Author Contributions: Conceptualisation and methodology, J.M.C., M.C.G. and R.M.-E.; software
and validation, J.M.C., M.C.G. and R.M.-E.; investigation, J.M.C., M.C.G. and R.M.-E.; resources, R.M.-
E.; writing—original draft preparation, review and editing, J.M.C., M.C.G. and R.M.-E.; supervision,
M.C.G.; funding acquisition, J.M.C. All authors have read and agreed to the published version of
the manuscript.
Funding: This work is part of the project of I+D+i PID2020-112675RB-C44, funded by MCIN/AEI/
10.13039/501100011033.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Acknowledgments: Thanks to the farmers from “Alimer Alimentos del Mediterráneo Sociedad Co-
operativa” in the Region of Murcia for their collaboration in the verification of the Decision Support
System proposed in the study case.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Khan, N.; Ray, R.L.; Sargani, G.R.; Ihtisham, M.; Khayyam, M.; Ismail, S. Current Progress and Future Prospects of Agriculture
Technology: Gateway to Sustainable Agriculture. Sustainability 2021, 13, 4883. https://doi.org/10.3390/su13094883.
2. Pawlak, K.; Kołodziejczak, M. The role of agriculture in ensuring food security in developing countries: Considerations in the
context of the problem of sustainable food production. Sustainability 2020, 12, 5488. https://doi.org/10.3390/su12135488.
3. Basso, B.; Antle, J. Digital agriculture to design sustainable agricultural systems. Nat. Sustain. 2020, 3, 254–256.
https://doi.org/10.1038/s41893-020-0510-0.
4. Lindblom, J.; Lundström, C.; Ljung, M.; Jonsson, A. Promoting sustainable intensification in precision agriculture: Review of
decision support systems development and strategies. Precis. Agric. 2017, 18, 309–331. https://doi.org/10.1007/s11119-016-9491-4.
Sensors 2023, 23, 3038 15 of 16
5. Balit, S.; Acunzo, M. A Changing World: FAO Efforts in Communication for Rural Development. In Handbook of Communication
for Development and Social Change; Servaes, J., Ed.; Springer: Singapore, 2020; pp. 133–156. https://doi.org/10.1007/978-981-15-
2014-3_29.
6. Mekala, M.; Viswanathan, P. A Survey: Smart agriculture IoT with cloud computing. In Proceedings of the International Conference on Mi-
croelectronic Devices, Circuits and Systems, Vellore, India, 10–12 August 2017; pp. 1–7. https://doi.org/10.1109/ICMDCS.2017.8211551.
7. D’Adamo, I.; Di Vaio, A.; Formiconi, A.; Soldano, A. European IoT Use in Homes: Opportunity or Threat to Households? Int. J.
Environ. Res. Public Health 2022, 19, 4343. https://doi.org/10.3390/ijerph192114343.
8. Salih, K.O.M.; Rashid, T.A.; Radovanovic, D.; Bacanin, N. A comprehensive survey on the Internet of Things with the industrial
marketplace. Sensors 2022, 22, 730. https://doi.org/10.3390/s22030730.
9. Gagliardi, G.; Lupia, M.; Cario, G.; Cicchello Gaccio, F.; D’Angelo, V.; Cosma, A.I.M.; Casavola, A. An internet of things solution
for smart agriculture. Agronomy 2021, 11, 2140. https://doi.org/10.3390/agronomy11112140.
10. Xu, J.; Gu, B.; Tian, G. Review of agricultural IoT technology. Artif. Intell. Agric. 2022, 6, 10–12. https://doi.org/10.1016/j.aiia.2022.01.001.
11. Schreiner, V.C.; Link, M.; Kunz, S.; Szöcs, E.; Scharmüller, A.; Vogler, B.; Beck, B.; Battes, K.P.; Cimpean, M.; Singer, H.P.; et al.
Paradise lost? Pesticide pollution in a European region with considerable amount of traditional agriculture. Water Res. 2021,
188, 116528. https://doi.org/10.1016/j.watres.2020.116528.
12. Shruthi, U.; Nagaveni, V.; Raghavendra, B. A review on machine learning classification techniques for plant disease detection. In
Proceedings of the 2019 5th International Conference on Advanced Computing & Communication Systems (ICACCS), Coimbatore,
India, 15–16 March 2019; pp. 281–284. https://doi.org/10.1109/ICACCS.2019.8728415.
13. Benyezza, H.; Bouhedda, M.; Rebouh, S. Zoning irrigation smart system based on fuzzy control technology and IoT for water and
energy saving. J. Clean. Prod. 2021, 302, 127001. https://doi.org/10.1016/j.jclepro.2021.127001.
14. Karetsos, S.; Costopoulou, C.; Sideridis, A. Developing a smartphone app for m-government in agriculture. J. Agric. Inform. 2014,
5, 1–8. https://doi.org/10.17700/jai.2014.5.1.129.
15. Liakos, K.G.; Busato, P.; Moshou, D.; Pearson, S.; Bochtis, D. Machine learning in agriculture: A review. Sensors 2018, 18, 2674.
https://doi.org/10.3390/s18082674.
16. FreshPlaza. Frosts in Murcia Will Reduce the Supply of Vegetables in the Coming Weeks. 2021. Available online: https://www.
freshplaza.com/article/9281419/ (accessed on 30 January 2023).
17. Ding, L.; Noborio, K.; Shibuya, K.; Tamura, Y. Frost forecast—A practice of machine learning from data. Int. J. Reason.-Based Intell.
Syst. 2021, 13, 191–203. https://doi.org/10.1504/IJRIS.2021.10038162.
18. Guillén-Navarro, M.; Martínez-España, R.; Llanes, A.; Bueno-Crespo, A.; Cecilia-Canales, J. A Deep Learning Model to Predict
Lower Temperatures in Agriculture. J. Ambient Intell. Smart Environ. 2020, 12, 21–34. https://doi.org/10.3233/AIS-200546.
19. Aswin, S.; Geetha, P.; Vinayakumar, R. Deep Learning Models for the Prediction of Rainfall. In Proceedings of the Inter-
national Conference on Communication and Signal Processing (ICCSP), Tamilnadu, India, 3–5 April 2018; pp. 0657–0661.
https://doi.org/10.1109/ICCSP.2018.8523829.
20. Abdel-Fattah, M.; Mokhtar, A.; Abdo, A. Application of neural network and time series modeling to study the suitability of drain
water quality for irrigation: A case study from Egypt. Environ. Sci. Pollut. Res. 2021, 28, 898–914. https://doi.org/10.1007/s11356-
020-10543-3.
21. Mehdizadeh, S.; Behmanesh, J.; Khalili, K. Using MARS, SVM, GEP and empirical equations for estimation of monthly mean
reference evapotranspiration. Comput. Electron. Agric. 2017, 139, 103–114. https://doi.org/10.1016/j.compag.2017.05.002.
22. Pantazi, X.E.; Moshou, D.; Alexandridis, T.; Whetton, R.L.; Mouazen, A.M. Wheat yield prediction using machine learning and
advanced sensing techniques. Comput. Electron. Agric. 2016, 121, 57–65. https://doi.org/10.1016/j.compag.2015.11.018.
23. Ferentinos, K. Deep learning models for plant disease detection and diagnosis. Comput. Electron. Agric. 2018, 145, 311–318.
https://doi.org/10.1016/j.compag.2018.01.009.
24. Pantazi, X.E.; Moshou, D.; Oberti, R.; West, J.; Mouazen, A.M.; Bochtis, D. Detection of biotic and abiotic stresses in crops by
using hierarchical self organizing classifiers. Precis. Agric. 2017, 18, 383–393. https://doi.org/10.1007/s11119-017-9507-8.
25. Coopersmith, E.J.; Minsker, B.S.; Wenzel, C.E.; Gilmore, B.J. Machine learning assessments of soil drying for agricultural planning.
Comput. Electron. Agric. 2014, 104, 93–104. https://doi.org/10.1016/j.compag.2014.04.004.
26. Nahvi, B.; Habibi, J.; Mohammadi, K.; Shamshirband, S.; Al Razgan, O.S. Using self-adaptive evolutionary algorithm to improve
the performance of an extreme learning machine for estimating soil temperature. Comput. Electron. Agric. 2016, 124, 150–160.
https://doi.org/10.1016/j.compag.2016.03.025.
27. Buza, K. Time Series Classification and its Applications. In Proceedings of the 8th International Conference on Web Intelligence,
Mining and Semantics, Novi Sad, Serbia, 25–27 June 2018; Association for Computing Machinery: New York, NY, USA, 2018;
pp. 1–4. https://doi.org/10.1145/3227609.3227690.
28. Esling, P.; Agon, C. Time-series data mining. ACM Comput. Surv. 2012, 45, 1–34. https://doi.org/10.1145/2379776.2379788.
29. Pratama, I.; Permanasari, A.; Ardiyanto, I.; Indrayani, R. A review of missing values handling methods on time-series data. In
Proceedings of the International Conference on Information Technology Systems and Innovation (ICITSI), Bandung, Indonesia,
24–27 October 2016; pp. 1–6. https://doi.org/10.1109/ICITSI.2016.7858189.
30. Cadenas, J.M.; Garrido, M.C.; Martínez, R.; Muñoz, E.; Bonissone, P.P. A fuzzy k-nearest neighbor classifier to deal with imperfect
data. Soft Comput. 2018, 22, 3313–3330. https://doi.org/10.1007/s00500-017-2567-x.
Sensors 2023, 23, 3038 16 of 16
31. Olvera-López, J.A.; Carrasco-Ochoa, J.A.; Martínez-Trinidad, J.F.; Kittler, J. A review of instance selection methods. Artif. Intell.
Rev. 2000, 34, 133–143. https://doi.org/10.1007/s10462-010-9165-y.
32. Bonissone, P.; Cadenas, J.M.; Garrido, M.C.; Díaz-Valladares, R.A. A fuzzy random forest. Int. J. Approx. Reason. 2010, 51, 729–747.
https://doi.org/10.1016/j.ijar.2010.02.003.
33. Snyder, R.; Melo-Abreu, J.; Matulich, S. Frost Protection: Fundamentals, Practice and Economics; FAO: Rome, Italy, 2005.
34. IMIDA. Murcia Agricultural Information System, Meteorological data for the Region of Murcia. Available online: http://siam.
imida.es/ (accessed on 30 January 2023).
35. Aha, D. Lazy Learning; Springer Science and Business Media: Dordrecht, The Netherlands, 2013. https://doi.org/10.1007/978-94-
017-2053-3.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.