POSSIBILITY OF USING MACHINE LEARNING ALGORITHMS TO MODERNIZE AGRICULTURAL RESEARCH IN INDONESIA

Even though many researchers and campuses have agricultural technology study programs, various technological and agricultural research in Indonesia still needs to be fully integrated. The author believes that technological advancements, particularly in Artificial Intelligence and Machine Learning, provide substantial advantages in various industries, including agriculture. This paper will examine the opportunities for using machine learning algorithms in precision agricultural research to inspire Indonesian agricultural researchers.


Introduction
Agriculture is essential to the global economy, and the increasing human population greatly strains the agricultural sector's technological development.In line with the advancement of digitalization in all industrial fields, the transformation of the agricultural industry is an issue that must be addressed.The term "digital transformation," as we all know, is defined as follows: Digitization is the process of converting analog information into digital information; digitalization describes how IT or digital technology can be used to change existing business processes; and digital transformation is a phenomenon that occurs across companies with broad organizational implications, in which the core business model of the company can change using digital technology (Verhoef et al., 2019).Digital transformation includes new concepts and technologies such as the internet of things, big data, machine learning (ML), artificial intelligence (AI), and others.
This paper focuses on digital transformation in agriculture, particularly the use of machine learning algorithms to explore information that contributes to precision agricultural research.In this study, the research question is what machine learning algorithms can be adapted to modernize precision agricultural research in Indonesia.This study is qualitative, seeking previous research to obtain a mapping of machine learning algorithms that can be used in precision agricultural research.

• An Overview of Machine Learning
Machine Learning (ML) methodologies typically involve a learning process with the goal of learning to perform a task from "experience" (training data).In machine learning, data is made up of examples.Individual examples are typically described by a set of attributes, also known as features or variables.Nominal (enumeration), binary (i.e., 0 or 1), ordinal (e.g., A+ or B), or numeric attributes are all possible (integer, the actual number, etc.).A performance metric that improves with experience is used to assess the ML model's performance in a specific task.Various statistical and mathematical models are used to calculate the performance of ML models and algorithms.The trained model can then be used to classify, predict, or cluster new examples (testing data) based on the experience gained during the training process.Figure 1 depicts a typical machine-learning approach.
Figure 1.A typical machine-learning approach (Liakos et al., 2018) ML tasks are typically divided into broad categories based on the learning type (supervised/unsupervised), learning models (classification, regression, clustering, and dimensionality reduction), or learning models used to implement the chosen task.In supervised learning, data is presented with example inputs and outputs, and the goal is to build a general rule that maps inputs to outputs.Sometimes, information may be only partially available, with some target outputs missing, or may be provided only as feedback to actions in a dynamic environment (reinforcement learning).The acquired expertise (trained model) is used in the supervised setting to predict the test data's missing outputs (labels).However, in unsupervised learning, there is no distinction between training and test sets, and the data is unlabeled.The learner processes input data to uncover hidden patterns.The goal of machine learning is either prediction or clustering.Prediction is estimating the value of an output variable from a set of input variables.Prediction problems are classified into two types: (1) regression problems, in which the variable to predict is numerical, and (2) classification problems, in which the variable to predict is part of one of a set of predefined categories, which can be as simple as "yes" or "no."One way to remember the difference is that classification predicts a category or class label, whereas regression predicts a quantity.
Clustering is often used in marketing to group users according to multiple characteristics, such as location, purchasing behavior, age, and gender.
Machine Learning in Agriculture There are numerous applications of ML in agriculture.According to a recent literature review, four generic categories were identified from 2004 to 2018 (Liakos et al., 2018).Crop, water, soil, and livestock management are all included in this category (Figure 2).
Figure 2. The four generic categories in agriculture exploit machine-learning techniques (Benos et al., 2021) Crop Management Crop management encompasses diverse aspects resulting from combining farming techniques to manage the biological, chemical, and physical crop environment to meet quantitative and qualitative targets (Yvoz et al., 2020).Using advanced crop management techniques such as yield prediction, disease detection, weed detection, crop recognition, and crop quality helps to increase productivity and, as a result, financial income.

Water Management
The agricultural sector is the world's largest consumer of available fresh water, as plant growth relies heavily on water availability.Given the rapid depletion rate of many aquifers with negligible recharge, more effective water management is required to conserve water better to achieve sustainable crop production (Neupane & Guo, 2019).Water quality can be improved due to effective water management, as can pollution and health risks (el Bilali et al., 2021).
Soil Management Soil, a heterogeneous natural resource, involves exceptionally complex mechanisms and processes.Precise soil information on a regional scale is critical because it contributes to better soil management that is consistent with land potential and, in general, sustainable agriculture (Lampridi et al., 2019).Better soil management is also of great interest due to issues such as land degradation (loss of biological productivity), soil-nutrient imbalance (due to overuse of fertilizers), and soil erosion (as a result of vegetation overcutting, improper crop rotations rather than balanced rotations, livestock overgrazing, and unsustainable fallow periods) (Chasek et al., 2015).Livestock Management Monitoring animal welfare and overall production is critical for improving production systems (Fournel et al., 2017).Precision livestock farming is becoming increasingly important in supporting livestock owners' decision-making processes and changing their roles.It can also help with product traceability and monitoring product quality and animal living conditions, as required by policymakers (Salina et al., 2020).Cameras, accelerometers, gyroscopes, radiofrequency identification systems, pedometers, and optical and temperature sensors are used in precision livestock farming (Li et al., 2020).ML methodologies have become an integral part of modern livestock farming to take advantage of large amounts of data.

Methods
The author performed a preliminary screening by reviewing three previous review studies on the application of machine learning in agriculture (Benos et al., 2021;Liakos et al., 2018;Neupane & Guo, 2019).Furthermore, the authors searched for similar research in Indonesia based on the areas of concern discovered.The study's findings include agricultural mapping and machine learning algorithms that will be used to answer research questions.
In this article, the author limits machine learning models to those shown in Table 1 below.

Results and Discussions
• Crop Management Since disease detection and yield prediction are the subcategories with the most articles, we will only cover these topics.

• Yield Prediction
Previous studies in crop yield prediction research used machine learning models, as shown in  (Kung et al., 2016;Nesarani et al., 2020;Su et al., 2017) • Disease Detection Previous studies in crop disease detection research used machine learning models, as shown in Table 3 below.
Table 3. Crop Management: Disease Detection

ML Model Articles
Image/Color feature/Satellite spectral data 1) Detection and discrimination between healthy and those that are infected 2) Classification of parasites and automatic detection of thrips 3) Detection of plant disease ANN, SVM, DT (Abdulridha et al., 2018;Ebrahimi et al., 2017;Pantazi et al., 2017) Weather data, expert input (disease incidence form visual inspection) 1) Forecasting downy mildew 2) Detection of late blight disease (Chen et al., 2020;Verma et al., 2020) • Water Management Previous studies in water management research used machine learning models, as shown in Table 4 below.• Soil Management Previous studies in soil management research used machine learning models, as shown in Table 5 below.According to the earlier exploration results, the author discovers numerous research opportunities in the field of precision agriculture.The most common models are ANN, BM, DT, and SVM.These models can be found in nearly every industry, including crop management, water management, soil management, and livestock management.
The author discovers that machine learning models have been widely used in agricultural research in Indonesia.Google Scholar revealed 517 articles in 2022, 466 articles in 2021, 320 articles in 2020, 258 articles in 2019, and 216 articles in 2018.According to data on the number of studies, researchers' interest in conducting agricultural research using machine learning has increased yearly.

Conclusion
Indonesia has a very promising agricultural business potential.Various studies in the field of precision agriculture show that there are numerous opportunities for increasing production, managing infrastructure, and controlling agricultural problems.Applying machine learning models will entice a new generation of researchers to revisit the field of technology-based agriculture.
The innovation of data processing technology, data analysis, and artificial intelligence will significantly benefit the advancement of precision agriculture research.The availability of agricultural universities, institutes, and research centers, as well as the involvement of technology-based companies, will encourage the development of agricultural industrialization.

Table 4 .
Water Management

Table 5 .
Soil Management

Table 6 .
Livestock Management