AI DRIVEN SOIL MONITORING AND CROP RECOMMENDATION USING MACHINE LEARNING ALGORITHM

Abstract

AI DRIVEN SOIL MONITORING AND CROP RECOMMENDATION USING MACHINE LEARNING ALGORITHM The method for the development to increase productivity, apply the concepts of precision agriculture to make crop recommendations based on soil characteristics, types, and data-driven insights. The system's output has the potential to improve decision-making, increase input-output efficiency, and decrease inappropriate crop selections, all of which will increase agricultural productivity and benefit India's economy. To get better results, the SMOTE data balancing technique is then used. To further fine-tune the performance as smart factories, the authors also employed optimization techniques. With an accuracy value of 99.5129, F-measure of 0.0916, Precision of 0.0918, and Kappa of 0.8870, Cat Boosting (C-Boost) demonstrated the best performance. The GNB performed better than ROC-0.9569 and MCC-0.9569 in the machine learning algorithms for classification, regression, and boosting. Because multiple classifiers are used, the use of an ensemble of classifiers creates a path to better prediction decisions. To choose the classifier's output, a ranking procedure is also used for decision-making. FIG.1

Specification

Description:AI DRIVEN SOIL MONITORING AND CROP RECOMMENDATION USING MACHINE LEARNING ALGORITHM

Technical Field
[0001] The embodiments herein generally relate to a method for AI driven soil monitoring and crop recommendation using machine learning algorithm.
Description of the Related Art
[0002] The farmers frequently encounter a variety of difficulties in their farming operations, including erratic irrigation, poor soil quality, inexperience with crop selection and fertilization, and large losses from crop diseases. These difficulties have a major influence on crop productivity and economic results, especially in areas like India where a sizable portion of farmers lack the information and tools needed to make wise choices regarding crop selection and soil management. In a nation like India, the majority of people also make their living from farming. One Farmers are unable to predict market prices, choose the best crop for cultivation, identify the crop that will boost productivity and be most environmentally friendly, and more. ML and DL are two examples of the many new agricultural technologies being used to help farmers increase the productivity and profitability of their operations. We tried to suggest the best crop and fertilizer for a given farmland in our study.
[0003] The crop suggestion program allows the user to enter soil data and the kinds of crops they are growing. Agriculturalists are exchanging timely and valuable information with one another, either formally or informally. The open mindset among farmers is referred to as the willingness to share information. The extent and degree of information sharing are determined by this open mindset. We construct the web application using web technologies like HTML and CSS, create a dataset by compiling information from various sources, and then use the results to predict crop prices. The results are then put through a non-linear test, after which priorities are established and rankings are assigned to the crop list. The lack of effectiveness of conventional approaches in tackling these issues is the driving force behind the use of deep learning in agriculture. There is still a lack of use of deep learning techniques to address problems with inconsistent irrigation, poor soil quality, crop selection, and disease prevention, despite recent advances in machine learning and its successful application in crop disease identification. The objective is to use Flask to create an intuitive web application that uses deep learning to provide farmers with solutions. In the absence of harmful substances that could impede crop growth, the soil is the primary component that gives plants the nutrients and water they need to grow and reproduce.
[0004] As soil conditions deteriorate, so does the quality, cost, and capacity to supply the essentials that sustain the ecosystem. For the development of administrative areas like seed sorting, rain farming, water management, and degraded land reclamation, information on soil types, their distribution, size, soil erosion, water installation, etc., is therefore crucial. Information technology use in agriculture has the potential to improve farmers' yields and alter the decision-making process. This compiles the writings of multiple authors in one location, making it useful for experts to learn about the state of data mining systems and applications in relation to the agricultural industry. The IoT SNA-CR model optimizes crop selection and soil nutrient management by combining IoT, cloud, and machine learning. Farmers were able to make well-informed decisions by using an Android app to access real-time data gathered by IoT sensors and stored in the cloud. The MSVM-DAG-FFO algorithm, which famously introduced, outperformed conventional techniques with an accuracy rate of 0.973.
SUMMARY
[0001] In view of the foregoing, an embodiment herein provides a method for AI driven soil monitoring and crop recommendation using machine learning algorithm. In some embodiments, wherein an innovative crop selection system that uses Bayesian optimization in conjunction with GBRT-based deep learning models to determine the ideal hyperparameter. The impact of parameters was assessed in the modules for data preparation, classification, and performance evaluation using explainable AI. The system achieved a remarkable F1-Score of 1.0 for precise accuracy and recall, classifying soil-specific features into 12 classes. With a remarkable average classification accuracy of 1.00, the DNN-based model demonstrated Bhat's extremely accurate and dependable crop recommendation system. The Agronomy of Chott Meriem (Tunisia) Higher Institute's alluvial-developed soil served as the simulation site The effectiveness of ANN and MLR was verified using metrics like accuracy, correlation coefficient, and RMSE. The association between potato production, tillage, and soil characteristics was ascertained by the ANN model. However, because the ANN only used two hidden layers, it had a higher error percentage and a shorter description than MLR. Over the course of six decades, researchers from a wide range of controls have contributed ideas and tools to demonstrate, a fundamental tool in agrarian framework science. It is essential to review this history and its exercises to ensure that we stay informed as agrarian researchers currently consider the "people to come" models, information, and learning items expected to meet the inexorably mind-boggling frameworks issues looked by society.
[0002] In some embodiments, wherein the precision agriculture is the foundation of the Smart Crop Recommendation System, which targets India's sizable population. The system performed exceptionally well using ML models such as DT, SVM, KNN, LGBM, and RF, with RF reaching an astounding 99.24% accuracy. This system efficiently suggested the best crops by analyzing characteristics like temperature, humidity, pH, rainfall, nitrogen, potassium, phosphorus, and rainfall. Kathiria led this innovation, which maximized land resources for increased productivity and sustainability while providing farmers, the government, and agricultural stakeholders with crucial decision-making support. Ten sets of fertility data were used as predictor variables in the ELM techniques. Additionally, they looked at coffee yield as the objective variable, and the ELM technique was used to solve complex and poorly defined problems. The effectiveness of ELM was verified using established methods like random forest and multiple linear regression (MLR) using parameters like root mean square error (RMSE), means square error (MSE), Legates and McCabe's index, Willmott's index, and Nash-Sutcliffe efficiency coefficient. The verified outcomes demonstrated that ELM was more effective at separating the characteristics of the objective and predictor variables.
[0003] In some embodiments, wherein a major advancement in the use of advanced analytics and the Internet of Things to improve crop productivity. Using sensor-monitored soil data, Parameswari et al. concentrated on using software for sustainable agriculture, giving farmers insights into their crops. For crop recommendations, they created a model based on a number of characteristics, including crop details, soil composition, weather, temperature, pH, and rainfall. Data from the Kaggle repository of the Indian Chamber of Food and Agriculture was used to test machine learning algorithms such as PART, Decision Table, and J Rip. The results showed that agriculture benefited from the implementation. By using data mining with association rules to gather crucial information for forecasting the future impacts of environmental and climatic conditions, this work made a significant contribution. Developing intelligent agro-food systems is essential to tackling the problems of global development.
[0004] These and other aspects of the embodiments herein will be better appreciated and understood when considered in conjunction with the following description and the accompanying drawings. It should be understood, however, that the following descriptions, while indicating preferred embodiments and numerous specific details thereof, are given by way of illustration and not of limitation. Many changes and modifications may be made within the scope of the embodiments herein without departing from the spirit thereof, and the embodiments herein include all such modifications.

BRIEF DESCRIPTION OF THE DRAWINGS
[0001] The embodiments herein will be better understood from the following detailed description with reference to the drawings, in which:
[0002] FIG. 1 illustrates a method for AI driven soil monitoring and crop recommendation using machine learning algorithm according to an embodiment herein; and
[0003] FIG. 2 illustrates a method for data balancing using SMOTE according to an embodiment herein.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
[0001] The embodiments herein and the various features and advantageous details thereof are explained more fully with reference to the non-limiting embodiments that are illustrated in the accompanying drawings and detailed in the following description. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.
[0002] FIG. 1 illustrates a method for AI driven soil monitoring and crop recommendation using machine learning algorithm according to an embodiment herein. In some embodiments, it makes use of a large conventional yield dataset with farming boundaries. Another dataset is used as the element dataset. The datasets are compiled from kaggle.com and government websites. The size of the harvest dataset is 7841 kb. The expectation boundaries for this dataset include relative mugginess, pH level, temperature, precipitation, and region. Wheat, rice, maize, millet, peas, pigeon peas, sugarcane, green grams, and other yields are included in this dataset. Various attributes are available for each estimate boundary for a single harvest. The expectation boundaries can be set to any value from the range of values in the dataset, for instance, if the yield is assumed to be wheat. In order to prognosticate crop recommendations based on N, P, K, temperature, moisture, pH, and rainfall, this exploratory study uses pre-processing techniques, feature extraction, feature selection, and classification techniques. The data is collected from reliable sources. The classification process also makes use of artificial intelligence techniques. As long as the necessary soil, board, environment, and financial information are available, it can identify the organization to arrive managers and transversely over reality. Decision Support Systems (DSSs) are used to create the information that the board and officials need. These systems do not process the data using the compelled strategies. Therefore, make decisions about the problem using the clever system's thoughts. It expects a crucial activity in the comprehension of agronomic results, and their use as decision sincerely steady systems for farmers is extending.
[0003] In some embodiments, prior to training the model, the dataset must be prepared. Reading the acquired dataset is the first step in the data preprocessing process; data cleaning follows. Data cleaning removes some unnecessary attributes from the datasets so that crop predictions can be made. Therefore, it is necessary to remove unnecessary attributes and fill in any missing values in datasets that contain undesired nan values in order to improve accuracy. Next, specify the goal of the model. Once the input data is acquired, pre-processing methods are used to increase the data collection's accuracy. The pre-processing procedure is divided into two stages, with stage I successfully eliminating noise from the input data. Tokenization and normalization have been completed in stage II. Lemmatization and stemming procedures are also included in the normalization process to finish the pre-processing phase. This paper identifies and examines the advantages of IoT and DA as well as open issues. Iot is relied upon to provide the agribusiness division with a few advantages. In any case, there are still a number of issues that need to be addressed in order to make it manageable for small and medium-sized farmers. Cost and security are the main concerns. It is common for the cultivating portion to become more contentious.
[0004] In some embodiments, the predictive analysis strives for accurate forecasts and insights from past events by using data, statistics, and artificial intelligence to forecast future outcomes based on historical data. In our framework, we used regulated AI with the subcategories of relapse and characterization. An order technique will work well for our framework. predicting precipitation using the SVM computation. Crop prediction is done using the choice tree technique. By doing this, the quantity of data is decreased, classification accuracy is increased, the algorithm's running time is decreased, and the classification algorithm's overall quality is enhanced throughout the learning process. Learning accuracy is decreased when unrelated features become overfit, less comprehensible, and computationally complex. The degree of dependence between the associated variable qualities was then established using the chi-square test strategy. It was found that the primary factors influencing product development, yield, and wine quality were the daily absurd climate conditions, such as the most extreme and least fluctuation in temperature, precipitation, dampness, and wind speed.
[0005] FIG. 2 illustrates a method for data balancing using SMOTE according to an embodiment herein. In some embodiments, the pre-processing procedures described in the Data Preprocessing section are followed by the initial dataset, which includes precipitation data from previous years. The model is trained using an RBF kernel in an SVM classifier. SVM, a popular tool for regression and classification, uses an ideal hyperplane to divide data into discrete classes. The training set is then used to modify the classifier. One of the most often suggested methods for handling an unbalanced dataset is data resampling. We often leave out events from data that might contain significant information when we under sample. SMOTE is one of several specific data augmentation oversampling techniques used in this study. For the minority class, this oversampling technique creates artificial samples. Next, we divided the data sets into training (80%) and testing (20%) categories. Following feature extraction from the training set's data, the training and testing sets were divided for classification. We will initialize all of the crop names with the numbers in this metadata. We can use the data in the algorithm with ease thanks to this data. Listen to each crop's metadata, which is assigned a specific number. This number is unique; that is, when one crop receives a number, the other crop does not receive the same number.
[0006] In some embodiments, the most crucial step in the yield figure process is stacking external crop datasets. Building a model that can forecast which crop or crops would be best to grow based on a variety of variables, including soil properties, climate, and other pertinent features, is the goal of machine learning-based crop prediction. One popular method for this kind of prediction task is the Decision Tree algorithm. dataset that includes details about various crops and the circumstances in which they grow. Features like soil type, temperature, humidity, rainfall, sunlight duration, pH level, and the amounts of nitrogen, phosphorus, and potassium (NPK values) in the soil should all be included in the dataset. A family of machine learning algorithms, including decision trees, Gaussian Naive Base, Multinomial Naive Bays, Complementary Naive Bays, and linear regression, are also included in the process. They use Bernoulli naive Baye's, SVM, ridge, RF, and boosting algorithms like bagging, stochastic gradient descent, XG boost, and CB boost, among others. Training the data is simple. Listen to all the information. During the pre-processing step, we load the metadata first, attach it to the data, and then replace the converted data with the metadata. After that, this data will be further processed, the unwanted data will be eliminated from the list, and the data will be separated into train and test data.
[0007] In some embodiments, a yield dataset from kaggle.com that covers a variety of crops, including sugarcane, wheat, rice, maize, millets, and others, is used in the analysis. Predictive characteristics like pH, temperature, humidity, precipitation, and area are all included. To build a predictive model, AI and deep learning algorithms need two sets of data: a test set and a training set. Classifier prediction accuracy is confirmed by metrics such as Mean Outright Blunder (MAE), Root Mean Squared Mistake (RMSE), and Mean Outright Mistake (MAE), Accuracy, Precision, Recall, and F1-score. Word counts and other discrete values are the best fit for the multinomial classification algorithm. Consequently, we expect it to be the most accurate. Here, each case's probability distribution is calculated as follows: Nyi is the count of each feature, n is a smoothing agent, and Ny is the total number of features of the event that belong to y. The Laplace parameter is employed in order to remove the impact of non-vocabulary words. DFD illustrates the flow of information through the system and the various transformations that alter it. It is a visual method that illustrates the flow of information and the changes made as data passes from input to output. Another name for DFD is a bubble chart. Any level of abstraction can be used to represent a system using a DFD. DFD can be divided into levels that correspond to increasing functional detail and information flow.
, Claims:I/We Claim:
1. A method for AI driven soil monitoring and crop recommendation using machine learning algorithm, wherein the method comprises;
collecting soil data such as pH, moisture content, temperature, and nutrient levels, as well as environmental data like humidity, rainfall, and sunlight exposure;
data preprocessing is essential to clean, normalize, and organize the data for accurate analysis;
supervised learning helps predict the suitability of certain crops for specific soil types, while unsupervised learning identifies natural groupings in the data;
evaluating soil health by analyzing pH, organic content, and nutrient levels;
using the results from soil quality analysis, the system recommends crops best suited for the soil's current state;
AI system continuously monitors soil and environmental conditions, providing real-time alerts to farmers if conditions deviate from optimal levels for the selected crop; and
collecting a new data, the model undergoes continuous retraining, enhancing its accuracy.

Documents