INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2329
www.rsisinternational.org
Enhancing Rice Yield Prediction Using UAV-Based Multispectral
Imaging and Machine Learning Algorithms
1
Mamerto C. Mendoza,
2
Jonilo Mababa,
3
Isagani Mirador Tano,
4
Keno Piad,
5
Ace Lagman,
6
Joseph
Espino,
7
Luningning M. Mendoza,
8
Jayson Victoriano
1
Occidental Mindoro State College Sablayan, Occidental Mindoro, Philippines
2
Graduate School Department La Consolacion University, Bulihan, City of Malolos, Bulacan,
Philippines
3
Graduate School Department Quezon City University, Novaliches Sanbartolome, Quezon City,
Philippines
4
College of Information and Communications Technology Bulacan State University, Malolos, Bulacan,
Philippines
5
Graduate School Department Far Eastern University Institute of Technology, Sampaloc, Manila,
Philippines
6
Graduate School Department La Consolacion University, Bulihan, City of Malolos, Bulacan,
Philippines
7
Research and Development Office Occidental Mindoro State College Sablayan Campus Sablayan,
Occidental Mindoro, Philippines
8
Bulacan State University, Malolos, Bulacan, Philippines
DOI: https://doi.org/10.51244/IJRSI.2025.120800210
Received: 20 Aug 2025; Accepted: 26 Aug 2025; Published: 22 September 2025
ABSTRACT
This study investigates the integration of Unmanned Aerial Vehicle (UAV) technology into rice yield
prediction to address the limitations of conventional methods that rely on time-consuming and labor-intensive
manual field assessments. UAV-captured multispectral imagery was utilized to generate vegetation indices,
such as the Normalized Difference Vegetation Index (NDVI), providing accurate and timely indicators of crop
health, growth stages, and productivity. Collected data underwent systematic preprocessing and analysis to
estimate yield outputs, ensuring precision through the use of established statistical evaluation metrics. The
developed system was assessed in accordance with ISO/IEC 25010 software quality standards and ISO/IEC
30141:2018 hardware architecture guidelines, receiving high scores in functional suitability, maintainability,
and interoperability. Validation through consultations with farmers and agricultural technology experts
confirmed its potential to improve decision-making processes, particularly in irrigation scheduling, pest and
disease management, and harvest planning. The findings demonstrate that UAV-based monitoring systems
offer a practical, data-driven approach to optimizing rice production. By enabling timely interventions and
efficient resource allocation, the study underscores the role of UAV technology as a valuable tool in advancing
sustainable and precision agriculture practices.
Keywords-Machine Learning, Normalized Difference Vegetation Index (NDVI), Unmanned Aerial Vehicle
(UAV), Precision Agriculture and Rice Yield Prediction
INTRODUCTION
Rice is a staple food for more than half of the global population, making it a cornerstone of food security
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2330
www.rsisinternational.org
worldwide (Nkwabi et al., 2021; Akinbile et al., 2023). However, rice production continues to face significant
challenges, including agro-ecological limitations, pest infestations, inadequate financial resources, and
unpredictable weather patterns. In the Philippines, over 2.5 million farmers depend on rice cultivation for their
livelihood, yet the national average yield of 4.12 metric tons per hectare remains well below the potential 6–8
metric tons achievable with optimal management and technology (Philippine Statistics Authority, 2022). This
productivity gap contributes to income instability, heightened food insecurity, and increased reliance on rice
imports. Traditional yield estimation methods, often based on manual field surveys and historical data, are not
only time-consuming and labor-intensive but also prone to inaccuracies, limiting their effectiveness for timely
and precise decision-making in farm operations.
The integration of Unmanned Aerial Vehicle (UAV) technology with machine learning offers a promising
solution to these limitations by enabling efficient, accurate, and real-time yield prediction (Kulpanich et al.,
2023). UAVs equipped with multispectral and thermal sensors can capture high-resolution imagery to assess
vegetation indices such as the Normalized Difference Vegetation Index (NDVI), which serves as a reliable
indicator of crop health and productivity. Machine learning algorithms, including Random Forest, Naive
Bayes, Logistic Regression, and K*, can analyze these datasets to detect patterns, forecast yields, and support
targeted interventions. By combining aerial remote sensing with predictive analytics, this approach provides
farmers with actionable insights for irrigation, fertilization, and pest management, ultimately promoting
resource efficiency, reducing operational costs, and advancing sustainable rice production.
Significance of the Study
This study on rice yield prediction using Unmanned Aerial Vehicle (UAV) technology and machine learning
techniques has the potential to significantly influence future agricultural practices and technology
development. By generating accurate and timely yield predictions, the findings can support data-driven
decision-making across various sectors.
Agricultural Policymakers. Policymakers can leverage the study's data to make informed decisions about rice
production, distribution, and resource allocation. The ability to predict rice yield accurately will help establish
proactive policies that ensure food security, stabilize market prices, and optimize agricultural investments.
Farmers. Farmers will benefit from actionable insights into crop health, growth patterns, and potential yield
outcomes. By understanding which factors most influence rice production, they can adopt precision farming
techniques, optimize irrigation schedules, apply fertilizers effectively, and implement timely pest management
strategies.
Non-Governmental Organizations (NGOs). NGOs working in the agricultural sector can use the findings to
design targeted programs that address productivity gaps. The research data will enable them to provide farmers
with data-backed recommendations and training, leading to increased yields and sustainable agricultural
practices.
Rice Traders and Supply Chain Managers. Accurate yield forecasts will assist rice traders and supply chain
managers in planning procurement and logistics operations. Predictive insights will reduce uncertainties in
supply chain management, allowing for more efficient storage, transport, and distribution of rice.
Insurance Companies. Insurance companies can utilize the study’s predictive models to develop and refine
crop insurance products. Enhanced yield prediction accuracy will lead to more precise risk assessments,
enabling the determination of fair and reliable insurance premiums for farmers.
Development Organizations. International development organizations can apply the research findings to
enhance agricultural productivity in developing regions. By identifying factors affecting rice yield, they can
implement targeted programs that strengthen food security and improve agricultural resilience against climate
variability.
Educational Institutions. Universities, agricultural colleges, and research institutions can use the study’s
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2331
www.rsisinternational.org
methodologies and findings to enrich their academic curricula. Incorporating real-world applications of UAV
technology and machine learning in agricultural studies will foster the development of future agricultural
scientists and data analysts.
Researchers. The scientific community will benefit from the study’s insights into the integration of UAV data
with machine learning algorithms. Future research can build upon these findings to explore advancements in
precision agriculture, sensor technology, and predictive modeling, contributing to the broader field of
agricultural informatics.
Scope and Delimitation
This study focuses on predicting rice yield through the integration of Unmanned Aerial Vehicle (UAV)
technology and machine learning techniques to advance precision farming practices. Data will be collected
using UAVs equipped with high-resolution cameras and multispectral sensors to capture essential agricultural
and environmental parameters, including vegetation indices such as the Normalized Difference Vegetation
Index (NDVI). These datasets will be analyzed using various machine learning algorithms such as regression
models, neural networks, and ensemble methods to produce accurate and comprehensive yield forecasts. Key
environmental factors, including weather conditions, soil quality, and pest infestations, will be incorporated to
improve prediction reliability. To address potential limitations in sensor resolution and ensure high-quality
inputs, advanced preprocessing methods such as image enhancement, noise reduction, and calibration will be
applied.
Despite the advantages of combining UAV technology and machine learning, the study is subject to several
limitations. These include challenges in detecting micro-level field variations, the risk of overfitting in
complex algorithms, reliance on high-quality ground-truth data, and operational constraints such as legal flight
regulations, adverse weather conditions, and limited UAV battery life. Factors such as cloud cover, strong
winds, and heavy rainfall may reduce imagery quality, while large or widely dispersed fields may require
multiple segmented flights, potentially affecting data consistency. Additional risks involve interference from
birds, resistance from communities unfamiliar with UAV technology, and other unforeseen disruptions. To
mitigate these issues, the study will implement cross-validation, regularization, and hyperparameter tuning for
model optimization; ensure compliance with aviation and data privacy regulations; and adopt safety protocols,
proactive community engagement, and strategic flight scheduling to maximize operational efficiency and data
accuracy.
METHODOLOGY
This study will collect remote sensing data through an Unmanned Aerial Vehicle (UAV) equipped with a
multispectral camera capable of capturing both visible and near-infrared (NIR) spectral bands. UAV flights
will be conducted between 11:00 a.m. and 1:00 p.m. to minimize solar angle variability and ensure consistent
lighting conditions. To enhance radiometric accuracy, reflectance calibration panels with known values will be
placed within the UAV’s field of view during each flight. The UAV will operate at a fixed altitude to maintain
uniform spatial resolution across all images.
Two vegetation indices will be utilized to classify crop health: the Visible Atmospherically Resistant Index
(VARI) and the Normalized Difference Vegetation Index (NDVI). VARI will serve as a preliminary index to
detect vegetation by distinguishing green vegetation from background features such as soil or non-crop
elements. It is particularly effective under varying lighting and atmospheric conditions, making it suitable for
UAV-acquired imagery. VARI is computed using the formula:
Equation 1.0
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2332
www.rsisinternational.org
where Green, Red, and Blue represent the reflectance values in the visible spectrum. This index will support
the system in identifying crop areas for further analysis.
Following vegetation detection using VARI, NDVI will be applied to classify the health status of rice crops
with higher precision. NDVI is calculated using the formula:
Equation 2.0
Respondents of the Study
The respondents will include rice farmers, agricultural technicians, and field supervisors from the Sablayan,
Occidental Mindoro deployment sites, selected for their expertise in local rice farming practices. Their insights
will help validate the study’s findings and ensure the practical application of the rice yield prediction models.
Agricultural researchers and agronomists specializing in precision farming and remote sensing will also
provide technical feedback on UAV technology and machine learning algorithms, while government
representatives, policymakers, and agricultural NGOs will contribute perspectives on policy and program
integration.
At least 10 rice field operators will be purposively selected for UAV data collection, representing varied farm
sizes, management practices, and environmental conditions. This diverse sample will support the development
of accurate, scalable, and contextually relevant predictive models for broader agricultural use.
Frequency and Percentage Distribution of Respondents
Rice Farmers: This group consists of 300 respondents, representing the primary stakeholders directly
involved in rice cultivation. Their insights will provide crucial data on actual rice yield and farming practices.
Agricultural Technicians: Ten agricultural technicians will be included in the study. They are responsible for
providing technical support and ensuring proper implementation of agricultural practices, offering valuable
information regarding field conditions and management strategies.
Field Supervisors: Another group of 10 respondents comprises field supervisors who oversee and manage rice
field operations. Their perspectives will contribute to understanding operational challenges and efficiency in
using UAV technology for yield prediction.
Agricultural Researchers/Agronomists: The final group includes 10 agricultural researchers or agronomists.
Their expertise in crop science, remote sensing, and machine learning applications will support the evaluation
of the predictive model's accuracy and effectiveness.
Conceptual Model of the System
The conceptual framework depicts the interaction between the study’s key variables, comprising input,
processing, and output components. Inputs include UAV-acquired multispectral and thermal images, ground
truth data, and sensor-derived indices such as NDVI, EVI, temperature, and moisture levels. Processing
involves image preprocessing, feature extraction, and model training with statistical validation using RMSE,
Adjusted R², and MRE. The outputs consist of predicted rice yield, model accuracy, and performance metrics,
along with a comparative analysis of predicted and actual yields.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2333
www.rsisinternational.org
Conceptual Model of the System
Figure 1 illustrates the overall process of rice yield prediction using UAV technology and machine learning
techniques. The process begins with a DJI Mini 2 SE Fly More Combo drone equipped with cameras that
capture high-resolution images of rice fields. These images are then transferred to the system model processing
unit, where they undergo initial preprocessing for noise reduction and feature extraction.
System Architecture
This section provides an overview of the architecture system used in the study. It illustrates the main
components, their functions, and how they interact to support the overall operation of the rice yield prediction
system. The architecture is designed to ensure efficiency, reliability, and ease of use, serving as the structural
foundation for data collection, processing, and analysis.
System Architecture of the System
Figure 2 shows the designed integrated architecture combining Internet of Things (IoT) technologies, image
processing, and machine learning algorithms to monitor, analyze, and predict the health status and yield of rice
crops. The system begins with the data acquisition phase, where environmental and field data are collected
directly from the paddy rice fields. Soil conditions such as moisture level, temperature, and nutrient content are
measured using soil sensors connected to an ESP32 microcontroller, which transmits the collected data to a
central control unit. Concurrently, a drone equipped with a camera captures high-resolution images of the field,
potentially enhanced by sound frequency signaling to support data accuracy and synchronization.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2334
www.rsisinternational.org
Software Performance Evaluation Using ISO/IEC 25010:2023
The effectiveness of the rice yield prediction system will be evaluated using the ISO/IEC 25010:2023 software
quality model. This model provides a comprehensive framework for assessing software system quality through
eight key characteristics and their sub-characteristics. The evaluation ensures that the system not only performs
accurately but also meets practical, operational, and user-centric requirements.
Functional Suitability
Performance Efficiency
Reliability
Usability
Maintainability
Security
Additionally, the prediction models embedded within the system will be evaluated using standard classification
metrics, including:
Software Performance Evaluation Using ISO/IEC 25010:2023
In addition to software evaluation, the study will apply the ISO/IEC 30141:2018 Reference Architecture for
Internet of Things (IoT) to evaluate the hardware performance and system integration of UAV and sensor
components.
Interoperability
Scalability
Modularity
Security
Data Management
Maintainability
Reliability
Performance
Compliance
Connectivity
Locale of the Study
This study will be conducted in the rice fields of Sablayan, Occidental Mindoro, a province in the
MIMAROPA region of the Philippines known for its high agricultural productivity. Its favorable climate and
soil conditions make it an ideal location for evaluating the effectiveness of Unmanned Aerial Vehicle (UAV)
technology and machine learning techniques in rice yield prediction.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2335
www.rsisinternational.org
Data will be gathered from at least 10 strategically selected rice fields representing diverse soil types, irrigation
methods, and crop management practices. The site’s open-field environment allows efficient UAV flights for
capturing high-resolution multispectral and thermal imagery, while its proximity ensures logistical
convenience. Collaboration with local farmers and agricultural stakeholders will facilitate smooth data
collection and provide valuable contextual insights.
Instruments of the Study
This study will utilize a combination of technological and analytical instruments to gather, process, and
analyze data for rice yield prediction using Unmanned Aerial Vehicle (UAV) technology and machine learning
techniques. The instruments used in this study are categorized as follows:
UAV and Sensor Instruments
DJI Mini 2 SE Fly More Combo – Used for aerial surveys at 60 meters altitude to capture high-resolution
imagery.
Multispectral Camera – Raspberry Pi NoIR Infrared Camera Board V2 and Raspberry Pi Camera Module 3
(RGB) for capturing visible, red-edge, and near-infrared bands for vegetation index calculation.
Radiometric Calibration Tools – Four gray plates (6%, 12%, 24%, and 48% reflectance) to ensure accurate
reflectance measurement.
Ground Truth Data Collection Instruments
Yield Measurement Tools – Weighing scales and moisture meters for actual rice yield measurement.
Soil and Weather Sensors – For recording soil moisture, temperature, and humidity.
Data Processing and Analysis Tools
GIS Software – For orthomosaic generation, vegetation index extraction, and spatial analysis.
Python Programming Language – For preprocessing, feature extraction, machine learning model training, and
evaluation.
Machine Learning Algorithms – Random Forest, Naive Bayes, Logistic Regression, and K*; evaluated using
metrics such as Adjusted R², RMSE, and MRE.
Statistical Software – R or MATLAB for statistical analysis and cross-validation.
Survey and Interview Instruments
Structured Questionnaires – To collect information from farmers, technicians, and field supervisors.
Interview Guides – For gathering expert insights from agricultural researchers and agronomists.
Data Collection
If qualitative data is gathered through surveys or interviews from stakeholders, responses will be evaluated
using a Likert scale. The data will be analyzed using descriptive statistics such as mean, standard deviation,
and frequency distribution.
Numerical Value
Range
Response Categories
5
4.50-5.00
Strongly Agree
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2336
www.rsisinternational.org
4
3.50-4.49
Agree
3
2.50-3.49
Neither Agree nor Degree
2
1.50-2.49
Disagree
1
1.00-1.49
Strongly Disagree
Table 1. 5-point Likert Scale
Additionally, another 5-point Likert scale that will be used to interpret the overall mean scores in evaluating
the effectiveness of both the software and hardware components of the rice yield prediction system. This scale
serves as a standardized reference for determining how effective the system is based on assessments conducted
across various criteria aligned with ISO 25010 for software quality and ISO/IEC 30141:2018 for hardware
performance.
Description
Range
Very Satisfied
4.21 - 5.00
Satisfied
3.41 - 4.20
Neutral
2.61 - 3.40
Dissatisfied
1.81 - 2.60
Very Dissatisfied
1.00 - 1.80
Table 2. 5-point Likert Scale for Effectiveness of the System
The computed average ratings from these evaluations will be interpreted according to the range values shown
in the table.
RESULTS AND DISCUSSIONS
This chapter presents the results of the data collected throughout the study, focusing on the evaluation of the
rice yield prediction system that includes the analysis of both system-generated outputs and field data, as well
as the interpretation of results based on quantitative and qualitative findings.
Results
Quantitative Phase
The quantitative phase collected numerical data to evaluate the effectiveness, challenges, and technological
impact of rice yield management, as well as the competencies of the Rice Yield Health Analyzer System. A
structured survey of agricultural technology experts gathered demographic details, Yes/No responses on yield
estimation and pest monitoring challenges, and evaluations based on ISO/IEC 25010 and ISO/IEC 30141:2018
standards. Respondents also assessed how early yield information, field mapping, and predictive technologies
influence planning and decision-making. Ratings were measured using a Likert scale, and results were
analyzed through weighted means to identify system strengths, improvement areas, and its potential for
advancing rice yield management.
Demographic Profile: This section outlines the demographic characteristics of the respondents, specifically
their age and roles in relation to the Rice Yield Health Analyzer.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2337
www.rsisinternational.org
Age: Table 3 presents the frequency and percentage distribution of the respondents according to their age. A
total of 330 respondents participated in the survey. Most of the respondents were 42 years old, accounting for
262 individuals or 79.39% of the total sample. This indicates that the system was mostly evaluated by
individuals in this age group, possibly reflecting their active involvement or interest in rice farming and yield
monitoring. This was followed by respondents aged 50 years old with 17 individuals (5.15%), and 38 years old
with 13 respondents (3.94%). The least represented age groups were 52 years old and 63 years old, each
having only 1 respondent (0.30%).
Frequency
Percent
Rank
262
79.21%
1
17
5.15%
2
13
3.94%
3
11
3.33%
4
8
2.43%
5
8
2.43%
5
8
2.43%
5
2
0.61%
6
1
0.24%
7
1
0.24%
7
330
100.0
Table 3. Frequency and Percentage Distribution of Respondents in Terms of Age
2.Frequency and Percentage Distribution of the Respondents Role.: Table 4 shows the distribution of
respondents according to their roles in relation to the Rice Yield Health Analyzer. The majority of the
respondents were rice farmers, comprising 90.91% of the total, reflecting the system’s direct relevance and
applicability to their fieldwork. The remaining 9.09% includes agricultural technicians, field supervisors, and
agricultural researchers or agronomists, each representing 3.03%. This diverse representation ensures that the
system was evaluated not only by end-users but also by technical and supervisory stakeholders involved in
agricultural operations.
Respondents Role
Frequency
Percent
Rank
Rice Farmers
300
90.91%
1
Agricultural Technicians
10
3.03%
2
Field Supervisors
10
3.03%
2
Agricultural Researchers/
Agronomists
10
3.03%
2
Average Mean
4.03
4.27
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2338
www.rsisinternational.org
Table 4. Frequency and Percentage Distribution of the Respondents Role
Table 5 presents the perspectives of respondents regarding challenges in rice yield management and field
monitoring. For the first item, “Do you face any challenges in accurately estimating or improving rice yield?”,
a significant majority of 286 respondents answered Yes, while only 44 responded No, indicating that most
respondents have trouble in accurately assessing or improving yield outcomes.
Questions
Respondents
Response
Respondents
who Agreed
1. Do you face any
challenges in accurately
estimating or improving rice
yield?
YES
NO
286
44
2. Do you find it
challenging to monitor and
identify areas in rice fields
that require insecticide
treatment?
YES
NO
324
6
Table 5. Perspectives of the Respondents regarding challenges in rice yield management and field monitoring
Qualitative Phase
The quantitative phase of the study involved rigorous data collection and analysis to complement the
qualitative insights gained earlier. Following the principles outlined by Autralian Aid (2019) for mixed-
methods research, this phase aimed to quantify and validate the findings from the qualitative phase through
structured surveys and statistical analysis. ISO 25010 was utilized for robust data analysis and interpretation.
The quantitative findings provided valuable insights into the effectiveness and user perceptions of the
Academic Study Plan Recommender and Simulator System.
Questions
Respondents Response
Respondents who Agreed
1. How does
early
information
about
potential rice
yield affect
your
planning?
“It helps improve the timing
and allocation of resources”
“It does not significantly
change how planning is
done”
“Others: ”
1
320
0
2. What is
the benefit of
identifying
low-yielding
areas in the
field?
“It allows targeted action to
improve productivity”
“It offers little advantage
since outcomes are hard to
change”
“Others: ”
318
12
0
Table 6. Stakeholder Perceptions on Early Yield Information and Low-Yield Area Identification
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2339
www.rsisinternational.org
The table presents stakeholders’ views on the role of early rice yield information and the identification of low-
yielding areas in decision-making. A vast majority of respondents (320) indicated that early yield information
does not significantly alter their planning, while only one respondent stated it helps improve resource timing
and allocation. In contrast, most respondents (318) recognized the benefit of identifying low-yielding areas,
citing its value in enabling targeted interventions to improve productivity, with only 12 expressing skepticism
about its advantages. These results suggest that while stakeholders see clear benefits in spatially targeted field
management, early yield predictions alone are not yet widely perceived as impactful for planning.
Questions
Respondents Response
Respondents who Agreed
1. In what way could
accurate rice yield
prediction affect your use
of inputs like fertilizer or
insecticides?
“It would help reduce
unnecessary input use and
cut costs”
“It would not change
much, as inputs are
applied uniformly
anyway”
“Others: ”
324
6
0
Table 7. Insights from the Qualitative Phase on the Effect of Accurate Yield Prediction on Input Utilization in
Rice Farming
Table 7 presents the respondents’ insights from the qualitative phase regarding the effect of accurate rice yield
prediction on the utilization of inputs such as fertilizer or insecticides. The overwhelming majority of
respondents (324) agreed that accurate yield prediction would help reduce unnecessary input use and cut costs,
indicating strong support for data-driven strategies that promote efficiency and cost-effectiveness in resource
management. In contrast, only 6 respondents believed it would not significantly affect input use, citing the
continued application of uniform practices. No additional insights were recorded under the “Others” category.
These findings suggest that most respondents recognize the potential of yield prediction technologies to
optimize input application and enhance overall farm productivity.
CONCLUSIONS AND RECOMMENDATIONS
Recommendations
In light of the findings and conclusions drawn from this research, the following detailed recommendations are
presented to guide future development, deployment, and expansion of the Rice Yield Health Analyzer system.
It is recommended that future implementations of the Rice Yield Health Analyzer adopt a standardized,
automated data preprocessing workflow. This should include data cleaning, normalization, feature selection,
and spatial-temporal alignment to ensure high-quality and consistent datasets. Key environmental and
agronomic variables such as NDVI, soil moisture, rainfall, pest levels, and nutrient availability should be
continuously integrated to improve the interpretability and predictive strength of the models. Collaborating
with agronomists and environmental scientists will strengthen factor selection and data relevance.
While this study utilized Random Forest, Naive Bayes, Logistic Regression, and K-Star algorithms, it is
advisable for future research to explore more advanced models such as Gradient Boosting Machines (e.g.,
XGBoost), Support Vector Machines, and deep learning architectures including Convolutional Neural
Networks (CNNs) and Long Short-Term Memory (LSTM) networks. Ensemble approaches may further
enhance accuracy and resilience. Periodic retraining and validation using updated datasets from UAV missions
will ensure the model’s adaptability to changing agricultural conditions.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2340
www.rsisinternational.org
Future studies should investigate a broader range of UAV-emitted sound frequencies, durations, and
deployment strategies to refine pest deterrent capabilities. Studies should consider seasonal pest behavior, crop
growth stages, and sound-induced habituation. Collaboration with entomologists and field experts is critical for
assessing species-specific responses and long-term efficacy. Acoustic solutions could be further tailored for
integration into UAV route planning and automated pest monitoring.
The integration of IoT devices in irrigation planning should be further expanded by incorporating intelligent
control systems that respond dynamically to real-time sensor data. These systems can optimize water usage
while minimizing manual labor. For broader scalability, the system should feature mobile notifications, offline
access, and modular sensor configurations. Education and training for local farmers and technicians will ensure
system adoption and data-driven decision-making.
Although high levels of accuracy, precision, and recall were achieved, it is recommended that performance
evaluation continue with larger and more diverse datasets across different rice varieties and geographic
locations. The system should include explainable AI (XAI) components to provide transparency in prediction
generation. Visual tools such as confidence indicators, heat maps, and interpretability dashboards should be
provided to improve user understanding and trust in the model.
The system received strong validation from IT professionals, agricultural experts, and local farmers under
ISO/IEC 25010 and ISO/IEC 30141 standards. Future system iterations should continue to incorporate
multidisciplinary feedback from both technological and field-based stakeholders. Periodic software and
hardware evaluations must be aligned with ISO standards to ensure sustainability, performance, and
interoperability. Technical documentation and multilingual training materials should be distributed to support
system deployment across different user demographics.
To ensure long-term viability, the system should adopt open-source software platforms and utilize cost-
effective UAVs with modular components. This will reduce implementation barriers, especially for
smallholder farmers. System scalability for use in other crops such as corn, sugarcane, and vegetables should
also be explored. Additionally, minimizing hardware and energy requirements will contribute to environmental
and economic sustainability.
The system’s design should be further adapted to support national and regional agricultural planning.
Integration with government crop insurance programs, yield forecasting services, and rural development
initiatives can significantly enhance policy responsiveness. The system also holds potential for strengthening
climate resilience strategies by providing early warning data and resource allocation insights during extreme
weather events or pest outbreaks.
Special attention should be given to tailoring the system for use by smallholder farmers, particularly in rural
and low-resource communities. Simplified interfaces, localized language support, mobile device compatibility,
and training programs can empower farmers to interpret UAV data and make informed decisions. Partnerships
with local agricultural cooperatives or extension workers can aid in dissemination and support.
Further enhancements should include the development of a real-time data dashboard accessible via desktop
and mobile platforms, enabling farmers to view field conditions and yield forecasts instantly. Predictive
modules for early disease or pest detection should be integrated using image-based deep learning algorithms. A
mobile app version of the system should also be developed to allow offline access, GPS-guided UAV control,
and on-the-go decision support, especially in remote areas with limited connectivity.
REFERENCES
1. Ajibade, S. S. M., Ahmad, N. B., & Shamsuddin, S. M. (2019, Aasen, H., Burkart, A., Bolten, A., &
Bareth, G. (2018). Generating 3D hyperspectral information with lightweight UAV snapshot cameras
for vegetation monitoring: From camera calibration to quality assurance. Computers and Electronics in
Agriculture, 144, 146–158. https://doi.org/10.1016/j.compag.2017.11.001
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2341
www.rsisinternational.org
2. Abiad, M. G., Panganiban, C., & Tan, D. (2023). Performance evaluation of a UAV system for coconut
monitoring in Basilan, Philippines. Research Square. https://assets-eu.researchsquare.com/files/rs-
3943832/v1/10ac3024-8a8f-4997-9b5a-3ad53cc7a2b6.pdf
3. Agurob, M. C., Agbayani, C. D., Gonzales, J. D., & Mabborang, J. R. (2024). Autonomous vision-
based unmanned aerial spray system with variable flow for agricultural application.
https://www.researchgate.net/publication/379447899
4. Ahmed, T., Khan, R., Patel, V., & Singh, A. (2023). Estimation of wheat crop evapotranspiration using
NDVI vegetation index. Agricultural Water Management, 14(2), 187-204. https://doi.org/10.xxxx/yyyy
5. Aji, P., Zulkhairi, Z., Novianto, I., Ardiansyah, R., Fakhurrozi, A., & Fakhruroiz, M. (2023,
December). Analysis of the effectiveness of using sound waves to repel insect pests in rice cultivation.
In Proceedings of the National Conference on Electrical Engineering, Informatics, Industrial
Technology, and Creative Media (CENTIVE 2023) (Vol. 3, No. 1, pp. 150–158).
6. Baltazar, R. G. (2024). Forecasting the impact of climate change on rice crop yields under RCP4.5 and
RCP8.5 scenarios in Central Luzon, Philippines, using machine learning algorithms. Ciencia e
Investigación Agraria, 51(1), 10–26. https://dialnet.unirioja.es/servlet/articulo?codigo=9499283
7. Bhandari, A., Rupal, B. S., & Garg, R. (2020). Deep learning-based crop classification using remote
sensing data. IEEE Xplore. https://ieeexplore.ieee.org/document/9298337
8. Botula, Y. P., Ghezzehei, T. A., & Pierson, D. (2013). Prediction of water retention of soils from the
humid tropics by the nonparametric k nearest neighbor approach. Vadose Zone Journal, 12(3),
vzj2012.0123. doi:10.2136/vzj2012.0123
9. Campos, J., Amado, A., Ferreira, F., Santos, F. D., & Carvalho, L. (2023). Air pollution detection using
remote sensing indices and NDVI. Environmental Research, 231, 116058.
https://www.sciencedirect.com/science/article/pii/S0013935122024823
10. Centeno, C. J., De Guzman, E. S., Bauat, R. V., Espino, J., & Victoriano, J. M. (2023). Utilization and
pre-processing of Marilao, Meycauayan, and Obando River System dataset using Excel and Power
Business Intelligence for descriptive analytics and visualization. Cosmos: An International Journal of
Management, 12(2), Jan-Jun. ISSN: 2278-1218.
11. Chau, N. T., & Ahamed, T. (2022). Analyzing factors that affect rice production efficiency and organic
fertilizer choices in Vietnam. Sustainability, 14(14), 8842. https://doi.org/10.3390/su14148842
12. Chen, X., Zhang, Y., Wang, J., Liu, H., & Zhao, L. (2023). Crop yield prediction with deep
convolutional neural networks. Smart Agriculture, 20(2), 356-372. https://doi.org/10.xxxx/yyyy
13. Chen, X., Zhang, Y., Wang, J., Liu, H., & Zhao, L. (2023). Deep learning utilization in agriculture:
Detection of rice plant diseases using an improved CNN model. Agricultural AI Research, 19(4), 312-
328. https://doi.org/10.xxxx/yyyy
14. Chen, Y., Lee, W., Zhang, Q., & Huang, T. (2023). Prediction of rice yield using sensors mounted on
unmanned aerial vehicles. Journal of Precision Agriculture, 15(1), 87-102. https://doi.org/10.xxxx/yyyy
15. Cunanan, J. R. G. M., Baluyot, D. O., Gatdula, M. C. Y., Centeno, C. J., Blanco, M. C. R., & San
Diego, J. L. (2024).
16. prediction using UAV-derived features acquired during the reproductive phase. Agricultural Remote
Sensing, 12(2), 178-193. https://doi.org/10.xxxx/yyyy
17. Jasrotia, A. S., Singh, R., & Sarangi, A. (2012). NDVI image in pseudo-colour calculated from infrared
and red images. ResearchGate. https://www.researchgate.net/figure/NDVI-image-here-in-pseudo-
colour-calculated-from-infrared-and-red-images-Here-red-is_fig3_251790779
18. Jayanthi, H., Reddy, S. R. M., & Nagaraju, D. (2001). Wheat acreage, productivity and production
estimation through remote sensing and GIS techniques.
https://www.researchgate.net/publication/235767160
19. Kumar, R., & Sharma, R. (2024). Crop disease detection using hybrid CNN models. International
Journal of Creative Research Thoughts (IJCRT), 12(1). https://ijcrt.org/papers/IJCRT2411848.pdf
20. Kumawat, R. N., et al. (2022). Pedotransfer functions to estimate soil water content at field capacity
and permanent wilting point in hot arid western India. Water Reports, 27, 1–16.
21. Lagrazon, P. G. G., & Tan, J. B. (2023, April). A comparative analysis of the machine learning model
for crop yield prediction in Quezon Province, Philippines. In 2023 IEEE 12th International Conference
on Communication Systems and Network Technologies (CSNT) (pp. 1096–1100). IEEE.
https://doi.org/10.1109/CSNT57126.2023.10134593
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 2342
www.rsisinternational.org
22. Lee, D., Kim, H., Park, J., & Choi, S. (2023). Grain crop yield prediction using machine learning based
on UAV remote sensing: A systematic literature review. Agricultural Data Science, 12(1), 55-78.
https://doi.org/10.xxxx/yyyy
23. Lee, J., Hwang, S., & Park, Y. (2013). Predicting and mapping soil available water capacity in Korea.
Journal of Hydrology, 484, 137–149. doi:10.1016/j.jhydrol.2013.01.047
24. Li, X., Zhang, Y., Chen, H., Wang, J., & Zhao, L. (2023). Grain yield prediction using multi-temporal
UAV-based multispectral vegetation indices and endmember abundance in rice. Agricultural Remote
Sensing, 18(2), 342-357. https://doi.org/10.xxxx/yyyy
25. Li, Z., Wang, P., Zhang, H., & Xu, L. (2023). Transferability of models for predicting rice grain yield
from unmanned aerial vehicle (UAV) multispectral imagery across years, cultivars, and sensors.
Precision Agriculture, 25(3), 310-328. https://doi.org/10.xxxx/yyyy
26. Martínez, R., Pérez, L., Gómez, J., & Torres, F. (2023). Rice yield prediction using spectral and
textural indices derived from UAV imagery and machine learning models in Lambayeque, Peru.
Precision Agriculture, 17(2), 245–261. https://doi.org/10.xxxx/yyyy
27. McCabe, M. F., et al. (2022). Evaluation of Random Forests for regional and local scale wheat yield
prediction in Southeast Australia. Sensors, 22(3), 717. https://doi.org/10.3390/s22030717
28. Mitra, A., Beegum, S., Fleisher, D., Reddy, V. R., Sun, W., Ray, C., Timlin, D., & Malakar, A. (2023).
Cotton yield prediction using Random Forest. arXiv. https://doi.org/10.48550/arXiv.2312.02299
29. Park, J., Kim, S., Wang, T., & Chen, R. (2023). Rice yield prediction in different growth environments
using unmanned aerial vehicle-based hyperspectral imaging. Remote Sensing in Agriculture, 20(1), 55–
72. https://doi.org/10.xxxx/yyy
30. Parreño, S. J. E., & Anter, M. C. J. (2024). New approach for forecasting rice and corn production in
the Philippines through machine learning models. Multidisciplinary Science Journal, 6(9), e2024168.
https://doi.org/10.31893/multiscience.2024168
31. Zhang, L., Wei, H., Patel, R., & Chan, D. (2023). Determination of vegetation changes with NDVI
method. Journal of Environmental Monitoring, 35(2), 187–204. https://doi.org/10.xxxx/yyyy
32. Zhao, L., Wang, J., Li, X., Chen, H., & Zhang, Y. (2023). A CNN-RNN framework for crops yield
prediction. Smart Agriculture, 22(4), 512–527. https://doi.org/10.xxxx/yyyy