An Integrative Decision Support Model for Smart Agriculture Based on Internet of Things and Machine Learning

The Internet of Things (IoT) has achieved an upset in a considerable lot of the circles of our current lives, like automobile, medical services offices, home automation, retail, education, manufacturing, and many more. The Agriculture and Farming ventures significantly affect the acquaintance of the IoT with the world. Machine learning (ML) is a part of artificial intelligence (AI) that permits software applications to turn out to be more precise at foreseeing results without being expressly customized to do as such. It uses historical data as input to predict new result values. In the event, a specific industry has sufficient recorded information to help the machine "learn", AI or ML can create outstanding outcomes. Farming is likewise one such important industry profiting and advancing from machine learning at large. ML can possibly add to the total lifecycle of farming, at all phases. This incorporates computer vision, automated irrigation, and harvesting, predicting the soil, weather, temperature, moisture values, and robots for picking off the crude harvest. In this paper, I'll work on a smart agricultural information monitoring framework that gathers the necessary information from the IoT sensors set in the field, measures it, and drives it, from where it streams to store in the cloud space. The information is then shipped off the prediction module where the necessary analysis is done using ML algorithms and afterward sent to the UI for its corresponding application.


Introduction
Machine learning is an approach to data evaluation that automates analytical system development. It is a department of artificial intelligence-based totally on the concept that systems can study from data, perceive patterns and make choic- Journal of Informatics Electrical and Electronics Engineering (JIEEE) A2Z Journals es/decisions/predictions with minimal human intervention. Food Security is one of the most necessary for the developing food needs of an ever-increasing population. Due to the increasing populace, we can't produce meals to meet the requirement of 1.38 billion, and nonetheless increasing the populace will put a large burden on the Indian economy. Around the globe, India is having a massive agricultural hub and the mainstream of the Indian populace is established on the agricultural area for meeting their requirement. Agriculture accounts for a fundamental component of GDP (Gross Domestic Product) now not solely of developing countries. However additionally for many developed nations. Thus, improvising and optimizing the current farming technologies is the want of the hour. It will no longer solely assist in flourishing sustainable improvement of mankind, plant life, and fauna but will additionally assist in dealing with the global crisis such as local weather alternate and epidemics such as draught. Internet of things has upgraded due to convergence of greater than one technologies like machine learning, wireless sensors, and embedded systems, real-time analytics. Conventional fields of networking of wireless sensors, embedded systems, automation (including domestic and constructing automation), and many more make contributions to enable the Internet of Things. In the consumer market, IoT technological know-how is most synonymous with product pertaining to the thought of the smart home, gadgets and home equipment (such as lights fixtures, thermostats, domestic protections systems, and cameras, and different domestic appliances) that help one or greater frequent ecosystems, such as smartphones.
The home automation systems are being drastically lookup and developed however, this essential area of Agriculture and especially Smart Agriculture tends to lag at the back of different domains, and require pretty a lot of R&D to gain sustainable goals now not solely at the industrial stage, however, at the root degree of this agriculture industry. Automation of traditional irrigation methods can lead to many folds make bigger in crop yield. To form agriculture based on IoT, there are some needs that are to be fulfilled necessarily. In the first place, the particular sensors (e.g., temperature sensors, moisture sensors, pressure sensors, etc.) required for the IoT utility are to be decided smartly. Secondly, the algorithms must be developed maintaining in thinking all the feasible possibilities required for the quality prediction and machine learning to have to be utilized effectively. Seeing next, the sensors are extra likely to get damage in the fields, so pursuits monitoring of these is required timely. Considering that, the framework of the wireless data transmission has to be as per the needs of the devices to join over a precise land region in the field. Lastly and majorly, the safety, security, and privacy of the device must be ascertained through authenticating the setup system, assuring its concealment, candor, and managed admittance. The model proposed here in this paper is about how the Internet of things and machine learning used in the agricultural system. The proposed system may want to be such there will be no compromise between efficiency, security, performance, and safety of the system, and the favored effect has to produce exquisite precision and accuracy.

Literature Review
A thorough literature review has been done and a fragment of the effective powerful technologies and algorithms based on literature survey and observations are put forward in this paper for the benefit of smart agriculture. In this current digital time, IoT assumes a huge part in each association just as nation advancement, and horticulture is one of the fields in which most of the things should be automated through IoT devices. This new idea unified on farming data has been known with several different names like Smart Farming, Smart Agriculture, Digital Farming, or Agriculture 4.0, and was born when telematics and information management were combined to the already familiar conception of preciseness Agriculture, up the accuracy of operations [13]. The utilization of new technologies to improve crop profitability are shown by [21]. The utilization of new technologies will be useful to measure and break down the information. To store data identified with soil, the authors are using the Blynk application. The data it will save will be its humidity, moisture, temperature, etc. Its uses are shown by [1]. While applying these new technologies, the test for retrieving information from crops is to come out with  [20]. [9], [23][24][25][26][27][28][29][30][31][32][33][34].
A farm vehicle and smart dispatching approach have been researched. [22] provided with associate interconnected ap-   Design of a fuzzy-based automated organic irrigation system for smart farm Automated organic irrigation system in controlling and properly allocating the available water resources for the irrigation system and pumping water for irrigation on right time and use.

Proposed Work (Framework and Algorithm)
The Internet of Things (IoT) has achieved an upset in a considerable lot of the circles of our current lives, like an automobile, medical services offices, home automation, retail, education, manufacturing, and many more. The Agriculture and Farming ventures significantly affect the acquaintance of the IoT with the world. Machine learning is an approach to data evaluation that automates analytical system development. It is a department of artificial intelligence-based totally In this proposed model IoT and machine learning is combined to create an efficient and effective smart faring model, which will increase the produc-   Colab Notebook. This evaluation forms the basis of the importance and needs for automation in the zone to limit cost and enlarge productivity. The bar graph shown is an outcome of the analyzed data. Figure 3 represents the graphical analysis for few crops production. This shows the variety of crops with quantity of their production.

Implementation
The Internet of Things (IoT) portrays the network of physical connected smart devices that are embedded with sensors, other applied sciences technologies for the motive for interfacing and supplanting realities with various units and designs over the web. This proposed model is based on a farming land, each region consists of different kinds of sensors which will explore it's close by climate by which farmers will be able to act accordingly and the necessities can be executed as soon as possible. Journal of Informatics Electrical and Electronics Engineering (JIEEE) A2Z Journals

4-Layer Architecture
The perception layer represents physical layer which has smart IoT devices. It identifies some physical parameters in the environment and how they communicate with one another and with the second layer transport layer. The IoT objects are responsible for gathering information/data, enabling the interaction between smart devices. This should be possible by utilizing sensor nodes or devices embedded with sensors, unmanned aerial vehicle and microcontrollers like Raspberry-Pi, Arduino to create sensor nodes and transmission gateways. Sensors are used to monitor soil moisture, atmospheric pressure, temperature, water level, fire detection, humidity, intrusion detection. The transport layer relates to network layer which is responsible for interconnecting other smart IoT devices, network devices, and servers. It is also used for processing of sensor data and transmitting it. The processing/prediction layer contains information storage, prediction, and visualizing assets. In this specific circumstance, big data helps store data in huge amounts and information handling, extraction of data in the briefest conceivable time. Such data are utilized as models by machine learning that is a data handling strategy through various algo-

Interacting IoT Basic Components:
Some major components for model:

Machine Learning Approach
Machine learning is an approach to data evaluation that automates analytical system development. It is a department of artificial intelligence-based totally on the concept that systems can study from data, perceive patterns and make choic- Journal of Informatics Electrical and Electronics Engineering (JIEEE) A2Z Journals • Support Vector Regression Algorithm: Support Vector Regression is a supervised learning algorithm that is utilized to anticipate distinct values. Support Vector Regression utilizes a similar rule as the Support Vector Machines. The essential thought behind SVR is to locate the best fit line. In SVR, the befitting line is the hyperplane that has the most extreme number of points. Dissimilar to other Regression models that attempt to limit the blunder between the actual and predicted values, the SVR attempts to fit the most effective line inside a threshold value. The threshold value is the interval between the hyperplane and borderline. The analysis of Soil Moisture training dataset taken from kaggle is done and Support Vector Regression is applied on it to get the predicted values. The comparative study of various algorithms suggests that Support Vector Regression gives the accuracy approximately of 90.6%. Support Vector Regression is the correlation of SVM for regression problems. SVR recognizes the existence of non-linearity in the data and anticipates a effective prediction model. Smart farming with this machine learning algorithm will provide smooth flow of work for all kinds of different scales of farming. First soil moisture data will be collected from sensor then this algorithm will be applied on that provided data.
• Implementing Support Vector Regression Support Vector Regression (SVR) is implemented in python using Google Colab Notebook It follows the following steps: Step 1. Picking random K data points from the training set.
Step 2. Put up a decision tree with respect to these K data points.
Step 3. Select the number of trees needed to be created and then repeat the above steps.
Step 4. Smart agriculture with this machine learning algorithm will provide smooth progression of work to all sorts of various sizes of farming. First temperature data will be gathered from sensor then this calculation will be applied on Random Forest Regression algorithm for accurate predictions.
• Implementing Random Forest Regression Random Forest Regression is implemented in python using google colab notebook

Result
An intelligence-based farm monitoring model is developed in this proposed work which will provide precise and sustaining solutions to various epidemics like food shortage, economic crisis, food security etc. Internet of Things and Machine Learning algorithms such as SVM (Support Vector Machine) and SVR (Support Vector Regression) with Radial basis function kernel, Random Forest Regression provides in predicting, classifying of water level, atmosphere pressure, soil moisture, weather forecast, irrigation system. The analysis is done on training datasets obtained from "Kaggle" for Indian Agriculture production about farming information such as crop production in various states, particular crop rice production in Indian states, soil moisture, temperature. The Support Vector Regression algorithm is applied on the soil moisture data as follows are the results:

Conclusion
This paper represents an economically efficient approach towards successful and powerful automated agriculture or we can say agriculture 5.0, Smart Farming, Smart agriculture. It utilizes the Internet of Things and a Machine Learning-based approach to obtain the best effective outcomes of farming production. Sensor nodes are interconnected with each other to monitor the fields completely. Then the data is being collected and transferred to the cloud. From the cloud, the data is fetched, and then machine learning algorithms like SVM (Support Vector Machine) and SVR (Support Vector Regression) with Radial basis function kernel, Random Forest Regression are applied to predict and make accurate decisions for farm fields.
Which results in very effective and efficient outcomes of farming. Therefore, smart agriculture is conceivable to convey a more profitable, sustainable, and manageable type of farming productions, based on ensemble learning giving more exact precise, and resource-efficient methodology.