The Essay On Machine Learning In Agriculture Is Crucial.

The Digitalization of Agriculture

The evolution that has been witnessed in the field of agriculture has majorly been boosted by the introduction of technology in the field of agriculture. (Vishal Meshram) in their article stated that agriculture is considered to be the backbone of every economy, this is because humans depend on food for survival. As much as agriculture is still the major source of food, it has faced several challenges over the past years, and farmers and agricultural scientists are constantly looking for measures that can be taken to ensure that there is food security worldwide. The problems that are associated with poor farming and low yields from crops is because farmers from most parts of the world still lack knowledge in farming. Most farmers lack sufficient knowledge on the crop type, type of soil, fertilizers, climatic patterns and many other things that are associated with farming (Ahmad and Nabi).

Technology has shifted agricultural methods from analogue to digital methods of farming. With digitalization in agriculture farmers are now in a position to carry out large scale farming because they incorporate machine learning algorithms to predict future crop yield, to determine the most suitable soil type for their crops and in other cases to determine trends in rainfall. According to (Amitava Choudhury), machine learning methods have revolutionized crop production in so many ways, and this is due to the ever-growing world population that constantly needs food. Machine learning is very extensive in agriculture because it does not only involve prediction of future yield of crops but rather the planting and harvesting process of the crops. This technique is applied in farming from the onset of farming which include land and seed preparation all the way to the process of harvesting and storage.

(Evagelos D. Lioutas), believe that machine learning algorithms have played a greater part in the digitalization of agriculture. The main reason for the digitalization of agriculture is to try and solve food problem that many people across the world are facing. Scientists and agriculturalists agree that digital farming is one way to ensure food security, and that is why everywhere in the world farmers are adopting the new changes in farming. (Mueller and Massaron), define machine learning as a technique that is used in the data analysis process to automatically get results based on the data that was used. For instance, our case is using data from crops that have been harvested over the past years, the yield on the crop is plotted against soil fertility. Therefore, machine learning will be used to predict future production.

As it has been stated in the previous paragraphs, that machine learning can be used in agriculture to boost crop production, it is important to note that there are four types of machine learning methods. These methods include; supervised machine learning, semi-supervised machine learning, unsupervised machine learning and lastly the reinforcement machine learning (Zhang). These four types of machine learning are crucial but the most commonly used one are the supervised and the unsupervised learning methods. Machine learning is important, and when it comes to data analysis of a given data set, machine learning algorithms are normally used. This study will use four machine learning algorithms to ensure that better results and obtained to allow room for prediction and generation of a conclusion based on the findings. The main machine learning algorithms that will be used in this study, are; Random Forest, Logistics Regression, Support Vector Machine Algorithm (SVM) and Naïve Bayes Algorithm. (Tavasoli), in his document stated that there are several types of machine learning algorithms that someone can choose from whenever they are conducting a test. The choice of the machine learning techniques is dependent of the type of data that an individual has and the type of test that they would like to conduct.

Types of Machine Learning Methods

This study will use the listed algorithm to ensure that suitable results are obtained after the analysis, and interpretation based on each outcome will be made to enhance the process of formulation a general conclusion of the research question. The results will be used to compare the four-machine learning method to determine how they can be used in crop production and determination of nutrients that are present in a given soil. The next section will consist of literature review, methodology and data analysis.

This section will highlight theories and conceptual framework of other scholarly articles that are in relation to the study topic. Literature review is all about showing the ability to use other people’s ideas to support their own ideas, and giving credit to the original authors of the books or articles. According to (Diana Ridley), literature review should bring clarity, improve the research methodology and adjust the focus of the report that is being written.

Computational learning theory, is a theory that deals with the use of mathematical techniques in machine learning techniques (Brownlee). This theory is also known as the statistical theory or machine learning theory. This theory borrows ideas from the computational aspect of a study and the statistical side, and it involves operations such as; creation of mathematical models, proving of algorithms used and the analysis of general issues. (Dutta), says that computational learning theory is applicable in several fields of study, for example, it can be applied in statistics, geometry, calculus, program optimization and many more.

Computational learning theory is associated with some models that can be used during data analysis, or data preparation. An example of these models include; PAC learning, weak learning model, and online learning model. The use of these model in machine learning has helped improve, and expand the knowledge of individual about machine learning and how they can use it in data analysis. Computational learning theory main goal is to develop better machine learning methods that are automated, and can help individuals to understand important issues that may arise during the learning process (Kearns and Vazirani).  Moreover, computational learning theory is more like a subject of artificial intelligence and it is dedicated to machine learning algorithms. An article by (Sharma) showed that computational learning dictates that sometimes a learner is not supposed to know everything, before they can understand a concept in machine learning. In short computational learning theory is a like a formal technique to learn new projects, and the projects could also include different algorithms that are in a given procedure. (Sharma), furthermore, shows that there are some questions that may arise whenever one is decides to use computational learning theory in their research. The questions that normally arise in computational learning theory include;

• How much does one understand the model that they are using?
• What is the formulated hypothesis?
• How can over fitting be prevented?
• What type of data is being used in the research study?

From all this, it is clear that computational learning theory is very important when considering machine learning. It does help when one wants to tackle computational questions or tasks. It is a wide field that cannot be exhausted in one study because there are many ideas that revolves around this theory, but the good thing is that they all certainly talk about a similar thing. 

Literature Review and Conceptual Framework

The model shows that from machine learning there are two main branches, namely; supervised learning and unsupervised learning. (Molnar) States that in supervised machine learning, the machine is taught through the use of examples. In this case the data set that is available comprises of inputs and outputs, and the algorithm that is used must ensure that it finds a way to get to the inputs or the outputs in the data set. In supervise learning there are three other categories, and they include; classification, regression and forecasting. Classification is all about.

All machine learning problems are optimization problems. There is always a methodology behind a machine learning model, or an underlying objective function to be optimized. The comparison of the main ideas behind the algorithms can enhance reasonings about them.

For instance, the objective of a linear regression model is to minimize the square loss of predictions and the actual value (Mean Square Error, MSE), while Lasso regression aims to minimize the MSE while restricting the learned parameters by adding an extra regularization term to prevent overfitting.

Some taxonomies of machine learning models include a) generative vs discriminative, b) probabilistic vs non-probabilistic, c) tree-based vs non-tree based, etc.

Machine learning (ML) is an important tool for the goal of leveraging technologies around artificial intelligence. Because of its learning and decision-making abilities, machine learning is often referred to as AI, though, in reality, it is a subdivision of AI. Until the late 1970s, it was a part of AI’s evolution. Then, it branched off to evolve on its own. Machine learning has become a very important response tool for cloud computing, and is being used in a variety of cutting-edge technologies.

Before we get started, we must know about how to pick a good machine learning algorithm for the given dataset. To intelligently pick an algorithm to use for a supervised learning task, we must consider the following factors:

1. Heterogeneity of Data:

Many algorithms like neural networks and support vector machines like their feature vectors to be homogeneous numeric and normalized. The algorithms that employ distance metrics are very sensitive to this, and hence if the data is heterogeneous, these methods should be the afterthought. Decision Trees can handle heterogeneous data very easily.  

2.  Redundancy of Data:If the data contains redundant information, i.e., contain highly correlated values, then it’s useless to use distance-based methods because of numerical instability. In this case, some sort of Regularization can be employed to the data to prevent this situation.  
3. Dependent Features: If there is some dependence between the feature vectors, then algorithms that monitor complex interactions like Neural Networks and Decision Trees fare better than other algorithms.  
4. Curse of Dimensionality:If the problem has an input space that has a large number of dimensions, and the problem only depends on a subspace of the input space with small dimensions, the machine learning algorithm can be confused by the huge number of dimensions and hence the variance of the algorithm can be high. In practice, if the data scientist can manually remove irrelevant features from the input data, this is likely to improve the accuracy of the learned function. In addition, there are many algorithms for feature selection that seek to identify the relevant features and discard the irrelevant ones, for instance Principal Component Analysis for unsupervised learning. This reduces the dimensionality.
5. Overfitting: The programmer should know that there is a possibility that the output values may constitute of an inherent noise which is the result of human or sensor errors. In this case, the algorithm must not attempt to infer the function that exactly matches all the data.

This section is going to display the results that were obtained after the data analysis process. The results will be visualized in tables and graphs, and the interpretation of the finding will be done. The dataset used is a sample of environmental and mineral factors that affect the growth of crops in different geographical areas. The following is the definition of the target variables that will be in the data set.

N - Nitrogen content in soil

P - Phosphorus content in soil

K - Potassium content in soil

Temperature – The hotness of an area

Humidity – Concentration of water vapor in the atmosphere

PH   Measure of acidity or basicity in soil

Rainfall – Moisture content in the soil

The dataset contains 2200 vectors, this can be used to develop a statistical model that will be used to help farmers during crop selection and thus, improve yields in the long run. 

Methodology and Data Analysis

Exploratory Data Analysis (EDA), is a significant step in any data analysis or data science project (Patil). EDA is the process where the dataset is studied to discover patterns, and irregularities (outliers), and form hypotheses based on the understanding of the outcome in the dataset. EDA involves generating summary statistics for numerical data in the dataset and creating various graphical representations to understand the data better. In this study, the concept of EDA will be understood better with the results from the analysis. Python analysis tool was used to perform the data analysis process. 

After conducting the analysis using the Python analysis tool, the following is the result that was obtained from the procedure. Five tests were conducted, they include; descriptive statistics, correlation analysis test, decision tree, Naïve Bayes, logistics regression, random forest and SVM’s accuracy.

Descriptive statistics, is used in data analysis to provide a summary of the variables. The data that we used had seven variables, and they are as shown in the following table 

From the table we can see the means and standard deviation of each variable. The total count for the data set was 2200 for each variable that was used. The means show the average of each variable and the standard deviation indicate the extent to which the data set lie apart. From the table temperature and Ph both had low standard deviations, meaning that the data points for the two variables are closer to their means. On the hand, rainfall has a high standard deviation showing that the data points are far from the means. The maximum values of each variable is also seen in the table.

Correlation test according to (Butler and Center) is used to show the association that exist between variables. In other terms, it shows the strength between two or more variables. The variables can be positively or negatively correlated with 1 being a strong positive correlation and -1 being a strong negative correlation. The table that follows shows the correlation analysis results that was obtained after the test. 

The table above shows the Python results for pairwise correlation of all the variable columns that were in the data frame. All the diagonal value in this correlation matrix is 1, which shows the strength of the relationship of the variables. The results showed a correlation coefficient that were both negative and positive, depending on the variable that it is correlated against.

Decision tree is a type of machine learning algorithm, where variables are categorized and the prediction of the future outcome is made based on the previous outcome. This type of machine learning can also show the probability or chances of the next event occurring (Magee). The table below shows the results that was obtained from the decision tree algorithm on each crop that was being tested. The results from the table shows that the decision tree accuracy was 95.9% which is good for this case. Each crop that was being studied had a probability of 1 or 0.9, meaning that all the crops had higher chances of doing well in the soil that was also tested.  Based on the precision and recall outcome the results showed that each crop had a very high f1-score, meaning that they all had higher chances of thriving in a particular soil, with good environmental conditions that favour them. 

This is a type of machine learning algorithm that is based on Bayes’ theorem with an assumption of independence among predictor variables (Jayant and Safari). Just to define it further, Naive Bayes classifier is based on the assumption that the presence of a particular feature in a class is unrelated to the presence of any other feature. A good example of when Naïve Bayes can be used, is in a scenario where a fruit may be considered to be an apple if it is red, round, and about 3 inches in diameter. Even if these features depend on each other or upon the existence of the other features, all of these properties independently contribute to the probability that this fruit is an apple and that is why it is known as “Naïve”.

Naive Bayes model is easy to construct and particularly useful for very large data sets. Along with simplicity, Naive Bayes is known to outperform even highly sophisticated classification methods. The table shown below was generated after the analysis, and it shows the probability of each variable that was tested. From the results it is evident that both the decisions tree and the Naïve Bayes machine learning algorithm are similar and they can be used to predict future outcome of a given data set. The table shows that the total dataset was 440 and all the crops had higher probabilities meaning that they can do well with the right environmental conditions. The accuracy level for this test was 99%, meaning that this test was very significant for thus research. 

Support Vector Machine or SVM is one of the most popular supervised learning algorithms, which is used for classification as well as regression problems (Wang). However, primarily, it is used for classification problems in machine learning algorithm. The goal of the SVM algorithm is to create the best line or decision boundary that can segregate n-dimensional space into classes so that we can easily put the new data point in the correct category in the future. This best decision boundary is called a hyperplane.

SVM chooses the extreme points/vectors that help in creating the hyperplane. These extreme cases are called as support vectors, and hence algorithm is termed as Support Vector Machine. Consider the below diagram in which there are two different categories that are classified using a decision boundary or hyperplane. The table shows that the accuracy level for this test is at 98% which is good and all the crops have a high probability of success.

Logistic regression according to (STRICKLAND) is a classification algorithm used to assign observations to a discrete set of classes. Some of the examples of classification problems are Email spam or not spam, Online transactions Fraud or not Fraud, Tumor Malignant or Benign. Logistic regression transforms its output using the logistic sigmoid function to return a probability value. Just like the other machine languages, logistics regression is useful because then it can be used to make predictions on variables, based on the probability of each variable. The result showed that all the variables have a high chance of success because they had higher probabilities. The accuracy of this test is at 95% which is good for this test of study.

Random forest is a Supervised Machine Learning Algorithm that is used widely in Classification and Regression problems (Gokhale, Prenger and Van Essen). It builds decision trees on different samples and takes their majority vote for classification and average in case of regression. One of the most important features of the Random Forest Algorithm is that it can handle the data set containing continuous variables as in the case of regression and categorical variables as in the case of classification. It performs better results for classification problems. The results from the analysis showed that this test has an accuracy of 98%, which we can say is good in this research study. The table also has shown the predicted value of probability for each variable.

Decision Tree --> 95.9090909090909

Naive Bayes --> 99.0909090909091

SVM --> 98.4090909090909

Logistic Regression --> 95.9090909090909

RF --> 98.86363636363636

The results that have been displayed above has shown that the four machine learning algorithms have a high accuracy level and are thus very significant for this test.  After comparing the accuracy score on the unseen data, we can conclude that the datasets were prepared well and every classifier were able to perform between 95% to 99% with Naïve Buyes getting the highest score and decision performing at 95%. 

Conclusion

From the study that has been done in this research, it is evident that machine language algorithm is important when we want to make prediction about something. This study used data that had been collected from an agricultural sector, and the type of crops, and environmental factors that are crucial for growth were also considered. Therefore, it is prudent to conclude that with the change in ways of farming it is important that farmers and other agriculturalist can opt to the use of machine learning algorithms to make predictions on the type of crop that is suitable for their region. While deciding on the type of crop to plant it is important to also consider the type of soil fertility and other environmental factors. One machine learning algorithm is good to use but it is best to test the data over a couple of ML algorithms and compare the accuracy level before making a decision.

Ahmad, Latief and Firasath Nabi. Agriculture 5.0 : artificial intelligence, IoT and machine learning. Boca Raton: CRC Press, 2021., 2021.

Amitava Choudhury, Arindam Biswas, Manish Prateek, Amlan Chakrabarti. Agricultural Informatics: Automation Using the IoT and Machine Learning. S.I: John Wiley & Sons, 2021, 2021.

Brownlee, Jason. "Machine Learning Mastery." 07 August 2020. machinelearningmastery. <https://machinelearningmastery.com/introduction-to-computational-learning-theory/#:~:text=Computational%20learning%20theory%2C%20or%20statistical,a%20wide%20range%20of%20problems.>.

Butler, T G and Goddard Space Flight Center. Statistical correlation analysis for comparing vibration data from test and analysis. Washington, D.C: National Aeronautics and Space Administration, Scientific and Technical Information Branch ; Springfield, Va. : For sale by the National Technical Information Service, 1986., 1986. Book.

Diana Ridley, Dr. The literature review : a step-by-step guide for students. London: SAGE Publications Ltd, 2012., 2012.

Dutta, Bhumika. "analyticsteps." 27 July 2021. analyticsteps.com. <https://analyticssteps.com/blogs/introduction-computational-learning-theory-ai-and-colt-models>.

Evagelos D. Lioutas, Chrysanthi,Charatsari, and Marcello De Rosa. ELSEVIER. 11 September 2021. <https://www.sciencedirect.com/science/article/abs/pii/S0160791X21002190>.

Gokhale, M, et al. Accelerating a random forest classifier : multi-core, GP-GPU, or FPGA?. Washington, D.C : United States. Dept. of Energy ; Oak Ridge, Tenn. : Distributed by the Office of Scientific and Technical Information, U.S. Dept. of Energy, 2012., 2012. Document.

Jayant, Advait and an O'Reilly Media Company. Safari. Data Science and Machine Learning Series Naive Bayes Classifier Advanced Concepts. Erscheinungsort nicht ermittelbar: Technics Publications Boston, MA Safari 2020, 2020. Book.

Kearns, Michael J and Umesh Virkumar Vazirani. An introduction to computational learning theory. Cambridge, Mass: MIT Press, ©1994., 1994.

Magee, John F. Decision trees for decision making. Boston, MA: Graduate School of Business Administration, Harvard University, ©1964., 1964. Book.

Molnar, Christoph. Interpretable machine learning : a guide for making Black Box Models interpretable. Morisville, North Carolina: Lulu], [2019] ©2019, 2019.

Mueller, John and Luca Massaron. Machine learning. Hoboken, New Jersey : John Wiley & Sons, [2021] ©2021, 2021.

Patil, Prasad. "towards data science." March 23 2018. towardsdatascience.com. Document. 14 March 2022.

Sammut, Claude and Geoffrey I Webb. Encyclopedia of machine learning. New York ; London: Springer, 2010., 2010. Book.

Sharma, Manika. "DEEP TECH BYTES." 18 February 2021. deeptechbytes.com. <https://deeptechbytes.com/introduction-to-computational-learning-theory/>.

Strickland, Jeffrey. Logistic Regression Inside And Out. Lulu Com, 2017., 2017. Print book.

Tavasoli, Simon. "simplilearn." 03 March 2022. simplilearn.com. <https://www.simplilearn.com/10-algorithms-machine-learning-engineers-need-to-know-article>.

Vishal Meshram, Kailas Patil, Vidula Meshram, Dinesh Hanchate, S.D.Ramkteke. "Science Direct." 01 December 2021. sciencedirect.com. <https://www.sciencedirect.com/science/article/pii/S2667318521000106>.

Wang, Lipo. Support Vector Machines: Theory and Applications. New York]: Springer-Verlag Berlin/Heidelberg, 2005., 2005. book.

Zhang, Yagang. New Advances in Machine Learning. Erscheinungsort nicht ermittelbar: IntechOpen 2010, 2010.