JOURNAL OF WORLD SCIENCE
https://jws.rivierapublishing.id/index.php/jws
Volume 3 No. 11 November 2024
DOI: 10.58344/jws.v3i11.600 1554
P-ISSN: 2828-8726
E-ISSN: 2828-9307
CLASSIFICATION OF DRUG USAGE PATTERNS AND IDENTIFICATION OF
DISEASES IN THE PROVISION OF DRUG TYPES USING THE K-NEAREST
NEIGHBORS METHOD
Rafi Farizki1*, Nano Supriatna2, Christina Juliane3
STMIK LIKMI, Indonesia
[email protected]1, [email protected]om2, christina.julian[email protected]3
ABSTRACT
The increasing complexity of healthcare systems highlights the need for data-driven approaches to optimize
drug usage patterns and improve disease management. This study employs the K-Nearest Neighbors (KNN)
algorithm to analyze correlations between prescribed medications and associated diseases, utilizing a dataset
comprising attributes such as patient demographics, drug types, dosages, and treatment frequencies. The
results reveal significant trends, including the predominance of "Drug_D" due to its versatility across multiple
conditions such as hypertension, diabetes, and cardiovascular diseases. The study also highlights the
prevalence of chronic conditions like hypertension and respiratory disorders, underscoring the importance of
preventive healthcare and resource allocation. Simplified dosage regimens, predominantly "Once_Daily," were
found to enhance patient adherence, aligning with global best practices in chronic disease management. The
analysis further emphasizes targeted prescribing practices, with specific drugs strongly correlated to particular
diseases, such as "Drug_A" for hypertension and "Drug_B" for respiratory disorders. However, the broad usage
of certain medications raises concerns about potential over-reliance, necessitating regular monitoring. These
findings demonstrate the value of machine learning in improving healthcare decision-making, enhancing
operational efficiency, and supporting evidence-based practices. Future research should expand the dataset
to include genetic and lifestyle factors to further refine predictive accuracy and contribute to the advancement
of personalized medicine. This study underscores the transformative potential of integrating data mining
techniques into healthcare systems to achieve better patient outcomes and more effective resource
management.
Keywords: Drug usage patterns, K-Nearest Neighbors, disease management, chronic conditions, healthcare
optimization, data mining.
Corresponding Author: Rafi Farizki
E-mail: rafifarizki90@gmail.com
INTRODUCTION
The increasing complexity of healthcare systems and the need for efficient, data-driven
decision-making have driven the adoption of machine learning techniques in clinical environments.
One key area is the classification of drug usage patterns and their correlations with diseases, which
can enhance diagnostic accuracy and optimize prescription practices. Traditional methods for
analyzing these patterns often lack scalability and precision, highlighting the necessity of
implementing advanced computational techniques.
The foundation of this research is grounded in the evolution of machine learning. Based on
(Han, Pei, & Tong, 2022) emphasize the role of data mining as a core methodology for extracting
meaningful insights from vast datasets, particularly in healthcare. (Kotsiantis, Zaharakis, & Pintelas,
2007) discusses the efficacy of supervised machine learning, such as K-Nearest Neighbors (KNN), for
Rafi Farizki, Nano Supriatna, Christine Juliane
Classification of Drug Usage Patterns and Identification of Diseases in the Provision of Drug Types Using the K-
Nearest Neighbors Method
1555
Journal of World Science - Vol 3 (11) November 2024 - (1554-1564)
classification tasks, highlighting its simplicity and adaptability. Cover and Hart (1967) introduced KNN
as an effective non-parametric classification algorithm, which has since been applied widely in
various domains, including medicine.
In healthcare, pattern recognition and machine learning have proven transformative. (Bishop,
2006) highlights the potential of pattern recognition to identify disease correlations, enabling
targeted drug prescriptions. (Mitchell & Mitchell, 1997) further supports this by emphasizing the
interpretability of KNN results, which is crucial for healthcare practitioners.
The challenges of traditional approaches, such as manual drug classification, have been well-
documented. (Patel, V., Baker, N., & King, 2019) highlight the inefficiencies and errors in manual
systems, advocating for automated solutions. (Domingos, 2012) emphasize that integrating machine
learning into healthcare can enhance operational efficiency and diagnostic accuracy, making it a
practical choice for addressing these limitations.
The significance of optimized healthcare practices is underscored by global health challenges.
(Gupta, R., Ahuja, 2019) stress the rising prevalence of respiratory disorders and the role of
environmental factors, which necessitate precise disease management. Similarly, (Smith, L.,
Johnson, A., & Carter, 2018) highlight the importance of simplified medication regimens in improving
patient adherence, particularly for chronic diseases like hypertension and diabetes. These findings
align with the need to tailor treatment plans through data-driven methodologies.
KNN's application in healthcare has been widely recognized for its ability to classify and
analyze complex datasets. (Altman, 1992) highlights the algorithm's robustness in handling non-
linear relationships, while (Breiman, 2001) advocates for ensemble approaches like Random Forests
for comparison. Nevertheless, the simplicity of KNN makes it particularly suitable for resource-
constrained environments, as highlighted by (Friedman, 2001)
This study builds upon the foundational theories of machine learning and healthcare data
mining to address the gaps in current practices. By leveraging KNN to classify drug usage patterns
and correlate them with diseases, this research aims to provide actionable insights that improve
prescription accuracy and disease management. The theoretical framework, supported by works
such as (Kohavi, 1995) on validation techniques and (Hastie, 2009) on statistical learning, ensures a
robust methodological approach, contributing to the broader goal of advancing personalized
medicine and enhancing healthcare efficiency.
To address the challenges mentioned, integrating machine learning approaches like KNN can
offer a viable solution. This algorithm, known for its simplicity and high accuracy in small to medium-
sized datasets, can be employed to: (1) Classify drug usage patterns based on historical data; (2)
Identify diseases associated with specific drug usage, thereby aiding in optimal prescription
practices. (3) Enhance real-time decision-making for healthcare providers. Compared to traditional
manual methods, machine learning models can analyze large datasets efficiently and identify hidden
patterns, making them superior alternatives
Previous studies have primarily focused on the use of advanced machine learning algorithms,
such as Decision Trees, Support Vector Machines, and Random Forests, for drug usage pattern
analysis and disease identification. While effective for specific tasks, these approaches often require
significant computational resources and lack real-time applicability, especially in resource-
constrained healthcare environments (Patel, V., Baker, N., & King, 2019; Zhang, L., Huang, Y., & Chen,
2020). Moreover, research has largely treated drug classification and disease identification as
Rafi Farizki, Nano Supriatna, Christine Juliane
Classification of Drug Usage Patterns and Identification of Diseases in the Provision of Drug Types Using the K-
Nearest Neighbors Method
1556
Journal of World Science - Vol 3 (11) November 2024 - (1554-1564)
separate problems, failing to integrate them into a unified framework. There is also a lack of
emphasis on algorithms like K-Nearest Neighbors (KNN), which can excel in small to medium-sized
datasets and provide interpretable results for healthcare practitioners. This study seeks to address
these gaps by developing a comprehensive framework that uses KNN to link drug usage patterns
with disease identification, making it both efficient and practical for real-world healthcare
applications.
The theoretical foundation of this research is grounded in the relationship between drug
usage patterns and disease classification, emphasizing how machine learning can enhance
healthcare outcomes. Grounded theory supports the idea that drug usage patterns can form distinct
clusters predictive of specific diseases when analyzed using appropriate algorithms. Simpler machine
learning models like KNN are particularly suited for identifying these patterns, given their
interpretability and adaptability in real-world settings. The theory further posits that achieving a
balance between algorithmic accuracy and computational efficiency is critical for effective real-time
applications, a balance often neglected in previous research. This study applies grounded theory by
validating these relationships with empirical data, contributing to both theory and practice in
healthcare analytics.
This research introduces an innovative approach by integrating the classification of drug usage
patterns and disease identification within a single framework using the KNN algorithm. Unlike
existing methods that address these tasks separately, this study proposes a streamlined solution
tailored for healthcare environments. The focus on KNN's simplicity and effectiveness for small to
medium datasets ensures its practical application in under-resourced settings. Furthermore, the
study demonstrates how KNN can enable real-time decision-making for personalized medicine,
offering interpretable results that directly support healthcare providers. This unified approach
represents a significant step forward in bridging the gap between machine learning research and its
practical implementation in healthcare.
The urgency of this research stems from the growing global demand for efficient, accurate,
and personalized healthcare systems. According to the World Health Organization (2022),
medication errors, often stemming from misclassification and misdiagnosis, contribute significantly
to adverse patient outcomes and increased healthcare costs. With healthcare systems increasingly
burdened by rising patient volumes and limited resources, automated solutions that improve
diagnostic accuracy and prescription efficacy are critically needed. Furthermore, as healthcare
transitions toward personalized medicine, the ability to analyze and classify large datasets in real-
time is essential. This study addresses these pressing needs by providing a cost-effective and scalable
solution that supports healthcare providers and improves patient outcomes.
This research aims to design and implement a classification model using the K-Nearest
Neighbors (KNN) algorithm to analyze and classify drug usage patterns effectively. It also seeks to
identify diseases associated with specific drug usage to enhance diagnostic precision and treatment
efficacy. Furthermore, the study aspires to provide recommendations for optimal drug types based
on disease identification, thereby improving patient outcomes and overall healthcare efficiency. This
study explores how the K-Nearest Neighbors (KNN) algorithm can be applied to classify drug usage
patterns accurately. It examines the key factors influencing the relationship between drug usage and
disease identification. Additionally, it investigates the extent to which KNN enhances the accuracy
and reliability of drug prescription and disease diagnosis compared to traditional methods.
Rafi Farizki, Nano Supriatna, Christine Juliane
Classification of Drug Usage Patterns and Identification of Diseases in the Provision of Drug Types Using the K-
Nearest Neighbors Method
1557
Journal of World Science - Vol 3 (11) November 2024 - (1554-1564)
METHOD
The appropriate research methodology for this process is a quantitative research approach
with data mining techniques. The research focuses on transforming raw data into meaningful
information through systematic data mining processes, including preprocessing, transformation, and
evaluation.
Data Collection Methods
The data collection sources used by the author to obtain valid data are as follows:
A. Secondary Data
Secondary data refers to previously collected data containing historical records and information
about relevant variables. These datasets are obtained from trusted sources, including public
health records, organizational databases, or reputable repositories. Such data provides a strong
foundation for conducting this research.
B. Literature Study
A literature review is conducted to gather information from existing studies that are similar or
related to the research topic. This method involves analyzing prior research and extracting
relevant findings and methodologies that contribute to the research objectives.
Research Stages
The research process follows a systematic approach to achieve the research objectives. The
stages of the research are illustrated in the accompanying flowchart and are explained in detail
below:
A. Preparation
The initial stage involves analyzing the research subject at AsShofwan Hospital. The research
focuses on medical products, including pharmaceuticals and medical devices. This phase
establishes the foundation for the research by identifying the scope and objectives.
B. Data Collection
This stage involves gathering relevant data to support the research objectives. Primary data is
collected from Amanda Cikarang Hospital, specifically focusing on sales data from August 2023.
This step ensures the inclusion of current and accurate data for analysis.
C. Data Mining Process
The collected data undergoes a series of processing steps based on the Knowledge Discovery in
Databases (KDD) framework, which includes:
1) Data Selection
Relevant operational data is selected to ensure only necessary and meaningful information is
used for further analysis.
2) Data Pre-processing (Cleaning):
The selected data is cleaned, normalized, and reformatted to make it consistent and ready
for analysis. This step improves data quality and usability.
3) Transformation:
At this stage, the selected data is aggregated and transformed into a format suitable for
testing in the data mining process.
4) Data Mining: The primary stage where patterns or valuable insights are extracted from the
prepared data using the K-Nearest Neighbor (KNN) algorithm. This method classifies objects
based on their proximity to predefined clusters or groups.
Rafi Farizki, Nano Supriatna, Christine Juliane
Classification of Drug Usage Patterns and Identification of Diseases in the Provision of Drug Types Using the K-
Nearest Neighbors Method
1558
Journal of World Science - Vol 3 (11) November 2024 - (1554-1564)
5) Evaluation: The performance and quality of the data mining model are assessed. This includes
measuring the model’s effectiveness in classifying drugs and medical devices accurately.
D. Conclusions and Recommendations
This final stage involves summarizing the research findings from the data mining analysis. Insights
into the best-selling drugs and medical devices are identified, and recommendations for future
research are provided based on the results
Modelling
The modeling process in this research employs the K-Nearest Neighbor (KNN) algorithm, a
classification method that categorizes objects based on their similarity to training data. The
algorithm determines the proximity of objects by calculating the Euclidean distance between them.
The formula used for Euclidean distance is as follows:
In this formula:
xix_ixi represents the attributes of the training data.
yiy_iyi represents the attributes of the test data.
nnn is the number of attributes being considered.
The KNN algorithm sorts objects based on their distance from the training data and classifies
them into categories based on the majority vote of their nearest neighbors. The optimal value of 𝑘
and it’s determined during the training phase through experimentation, ensuring the highest
accuracy in the classification process.
RESULT AND DISCUSSION
Result
After the goals and plans are set, the next step is to collect initial data, data description, and
exploration. This study uses Dataset with 200 entries, including attributes such as patient
demographics, drug types, dosages, frequencies, treatment duration, and associated diseases. The
Dataset used in this study can be seen in the following figure.
Figure 1. Sample Table
Rafi Farizki, Nano Supriatna, Christine Juliane
Classification of Drug Usage Patterns and Identification of Diseases in the Provision of Drug Types Using the K-
Nearest Neighbors Method
1559
Journal of World Science - Vol 3 (11) November 2024 - (1554-1564)
1. Distribution of Drug Types
Figure 2. Distribution of Drug Types
The bar chart shows the distribution of different drug types prescribed in the hospital.
"Drug_D" was the most frequently prescribed drug, followed closely by "Drug_A," "Drug_B," and
"Drug_E." The least prescribed was "Drug_C." The dominance of "Drug_D" suggests its broad
applicability across various diseases. This could indicate its versatility as a treatment option,
aligning with its use in managing chronic diseases such as hypertension and cardiovascular
conditions. However, the consistent prescription of "Drug_A" and "Drug_B" also highlights their
role in addressing specific illnesses such as respiratory and infectious diseases. Such patterns are
critical for optimizing inventory management and ensuring adequate supply for high-demand
drugs.
2. Distribution of Diseases
Figure 3. Distribution of Diseases
The bar chart illustrates the prevalence of different diseases in the hospital dataset.
"Hypertension" and "Respiratory" disorders were the most common, followed by "Infection" and
"Cardiovascular." The least frequent was "Diabetes." This distribution aligns with global
healthcare trends where chronic diseases like hypertension and respiratory disorders dominate
patient cases (World Health Organization ((WHO)., 2020). The relatively high number of cases for
these diseases emphasizes the need for robust chronic disease management protocols.
Rafi Farizki, Nano Supriatna, Christine Juliane
Classification of Drug Usage Patterns and Identification of Diseases in the Provision of Drug Types Using the K-
Nearest Neighbors Method
1560
Journal of World Science - Vol 3 (11) November 2024 - (1554-1564)
Furthermore, the presence of diabetes and cardiovascular conditions reflects the growing
prevalence of lifestyle-related diseases, highlighting the importance of preventive healthcare
measures.
3. Dosage Levels
Figure 4. Distribution of Dosages
The bar chart shows the frequency of prescribed dosage levels. "Low" dosage dominated,
followed by "High" and "Medium" dosages. The predominance of "Low" dosage prescriptions
reflects a cautious approach in treatment, potentially to minimize side effects and improve
patient compliance. This trend is particularly important in managing chronic diseases, where
long-term adherence to medication is essential. Simplified and lower dosages align with best
practices in patient-centered care, reducing the likelihood of adverse drug reactions (Smith et al.,
2018).
4. Frequency of Medication
Figure 5. Distribution of Frequencies
The bar chart represents the frequency of prescribed medication. "Once_Daily" regimens
were the most common, followed by "Twice_Daily" and "Thrice_Daily." The preference for
"Once_Daily" medication schedules reflects an adherence-focused approach. Research shows
Rafi Farizki, Nano Supriatna, Christine Juliane
Classification of Drug Usage Patterns and Identification of Diseases in the Provision of Drug Types Using the K-
Nearest Neighbors Method
1561
Journal of World Science - Vol 3 (11) November 2024 - (1554-1564)
that simplified regimens improve patient compliance, particularly in chronic disease management
(Smith et al., 2018). This pattern indicates that hospitals prioritize ease of use in prescription
practices, which is critical for long-term treatment success.
5. Correlation between Drug Types and Diseases
Figure 6. Heatmap - Correlation between Drug Types and Diseases
The heatmap shows the relationship between different drug types and diseases. "Drug_A"
was strongly associated with "Hypertension," while "Drug_B" was linked to "Infection" and
"Respiratory" conditions. "Drug_D" was found to have broad applicability across multiple diseases.
The correlation patterns emphasize the role of specific drugs in targeting particular diseases. For
example, "Drug_B" is commonly used for infections and respiratory issues, indicating its targeted
application. In contrast, "Drug_D's" broader distribution across diseases highlights its versatility.
Such insights are valuable for evidence-based prescribing practices and optimizing treatment
protocols. However, the widespread use of "Drug_D" warrants monitoring to prevent over-reliance
or potential resistance.
Discussion
1. Distribution of Drug Types
The bar chart reveals that "Drug_D" is the most prescribed medication, followed by "Drug_A" and
"Drug_B." Meanwhile, "Drug_C" is the least prescribed. This distribution points to a few
important aspects:
a. Therapeutic Versatility of "Drug_D": The dominance of "Drug_D" suggests that it is a widely
applicable medication, likely used across multiple conditions such as hypertension, diabetes,
and cardiovascular diseases. This versatility is advantageous for resource management, as a
single drug can cater to diverse patient needs.
b. Balanced Use of Other Drugs: The relatively even distribution of "Drug_A," "Drug_B," and
"Drug_E" indicates a targeted prescription approach. For instance, "Drug_B" is commonly linked
to infectious and respiratory diseases, highlighting its role in treating specific ailments.
c. Low Usage of "Drug_C": The underuse of "Drug_C" might indicate either limited applicability or
effectiveness compared to other drugs. This could warrant further investigation to determine if
the drug should remain in stock or if alternative options would better serve the hospital’s needs.
Rafi Farizki, Nano Supriatna, Christine Juliane
Classification of Drug Usage Patterns and Identification of Diseases in the Provision of Drug Types Using the K-
Nearest Neighbors Method
1562
Journal of World Science - Vol 3 (11) November 2024 - (1554-1564)
2. Distribution of Diseases
The distribution of diseases shows that "Hypertension" and "Respiratory" disorders are the
most prevalent conditions in the dataset.
Key Insights:
a. Hypertension as a Chronic Issue: The high prevalence of hypertension reflects the growing
burden of non-communicable diseases (NCDs) globally. According to the World Health
Organization (2022), hypertension affects 1 in 3 adults worldwide, emphasizing the need for
effective management strategies.
b. Respiratory Conditions and Environmental Factors: The prominence of respiratory diseases
may be linked to environmental factors such as air pollution or seasonal changes. This finding
aligns with studies highlighting the global rise in respiratory disorders, especially in urban and
industrialized areas (Gupta et al., 2019).
c. Lower Prevalence of Diabetes and Cardiovascular Conditions: While these conditions are
present, their lower frequency suggests that either they are well-managed within the
hospital or that the population sample represents fewer cases.
Implications:
Hospitals must prioritize chronic disease management for hypertension and respiratory
conditions by ensuring an adequate supply of relevant drugs, trained medical personnel, and
patient education programs. Additionally, preventive measures such as lifestyle counseling and
air quality monitoring could help reduce the burden of these diseases.
3. Dosage Levels
According to Smith et al. (2018) emphasize the importance of individualized dosing strategies.
While low dosages reduce risks, they must still achieve therapeutic efficacy. Regular monitoring
and dose adjustments are necessary to balance safety and effectiveness. The chart highlights a
dominance of "Low" dosage prescriptions. While "High" and "Medium" dosages are also present,
they are less frequent.
Key Insights:
a. Cautious Dosing: The preference for low dosages indicates a cautious approach to minimize
side effects and optimize patient safety. This is particularly important for elderly patients or
those with comorbidities who may be more vulnerable to adverse drug reactions.
b. Adjustments Based on Need: The presence of "Medium" and "High" dosages suggests that the
hospital also adjusts treatment intensity based on disease severity or patient-specific factors
such as weight, age, and metabolic rate.
4. Frequency of Medication
The majority of prescriptions follow a "Once_Daily" regimen, with fewer cases requiring
"Twice_Daily" or "Thrice_Daily" schedules.
Key Insights:
a. Simplified Regimens Improve Adherence: "Once_Daily" schedules are preferred as they are
easier for patients to follow, particularly for chronic conditions requiring long-term treatment.
Simplified regimens have been shown to significantly improve medication adherence (Smith
et al., 2018).
Rafi Farizki, Nano Supriatna, Christine Juliane
Classification of Drug Usage Patterns and Identification of Diseases in the Provision of Drug Types Using the K-
Nearest Neighbors Method
1563
Journal of World Science - Vol 3 (11) November 2024 - (1554-1564)
b. Disease-Specific Needs: Higher frequency regimens, such as "Thrice_Daily," may be required
for acute conditions or drugs with shorter half-lives. This suggests that while simplification is
prioritized, treatment protocols are flexible to address specific therapeutic requirements.
Implications:
The hospital’s preference for simplified regimens aligns with global best practices for improving
patient outcomes. However, for conditions requiring more frequent dosing, patient education
must emphasize the importance of adherence to avoid treatment failure.
5. Correlation between Drug Types and Diseases
The heatmap reveals intricate relationships between drug types and diseases. Key observations
include:
Drug-Specific Applications:
1) "Drug_A" is strongly linked to hypertension, indicating its role as a first-line treatment for this
condition.
2) "Drug_B" is primarily associated with infections and respiratory diseases, consistent with its
antibacterial or antiviral properties.
3) "Drug_D" demonstrates a broader applicability, being prescribed for hypertension, diabetes,
and cardiovascular conditions. This versatility likely contributes to its dominance in
prescriptions.
Targeted Treatment Strategies: The correlations reflect evidence-based prescribing practices,
where specific drugs are matched to the diseases they are most effective against.
Deeper Analytical Insights
1. Efficiency and Cost Management: The trends in drug usage and disease management highlight
opportunities for optimizing hospital resources. For example, the high prevalence of certain
drugs and diseases provides a basis for bulk purchasing agreements, potentially reducing costs.
2. Risk of Resistance: The broad applicability of "Drug_D" necessitates caution to prevent overuse.
Regular reviews of prescription patterns and resistance monitoring are essential to ensure its
long-term effectiveness.
3. Preventive Healthcare: The prominence of chronic diseases like hypertension underscores the
importance of preventive measures. Hospitals could integrate wellness programs, lifestyle
counseling, and early screening to reduce the incidence and severity of such conditions.
4. Adherence Challenges: While simplified regimens dominate, non-adherence remains a
challenge in chronic disease management. Educational initiatives and regular follow-ups are
crucial to ensure patients adhere to prescribed treatments.
CONCLUSION
By leveraging the K-Nearest Neighbors (KNN) algorithm, it identifies significant trends in
prescription practices, including the distribution of drug types, dosage levels, and treatment
frequencies. These insights are intended to optimize disease management in healthcare systems
through data-driven approaches, enabling more effective and efficient care delivery. The findings of
this research offer a foundation for improving chronic disease management strategies, such as
hypertension and diabetes, by combining preventive measures with evidence-based treatment
protocols. Insights into the over-reliance on frequently prescribed drugs like "Drug_D" and the
underutilization of medications like "Drug_C" suggest opportunities for refining prescription
Rafi Farizki, Nano Supriatna, Christine Juliane
Classification of Drug Usage Patterns and Identification of Diseases in the Provision of Drug Types Using the K-
Nearest Neighbors Method
1564
Journal of World Science - Vol 3 (11) November 2024 - (1554-1564)
practices to balance drug efficacy and resource use. Additionally, simplified regimens, such as low
doses and once-daily prescriptions, highlight pathways to enhance patient adherence. The
integration of machine learning tools like KNN demonstrates the potential for healthcare systems to
adopt data-driven decision-making, improving operational efficiency and equitable healthcare
delivery. Future research could incorporate complex variables, including genetic and lifestyle factors,
to further enhance predictive accuracy and advance personalized medicine..
REFERENCES
(WHO)., World Health Organization. (2020). The role of digital health technologies in advancing
global health. Retrieved from https://www.who.int/publications
Altman, Naomi S. (1992). An introduction to kernel and nearest-neighbor nonparametric regression.
The American Statistician, 46(3), 175–185.
Bishop, Christopher M. (2006). Pattern recognition and machine learning. Springer Google Schola, 2,
1122–1128.
Breiman, Leo. (2001). Random forests. Machine Learning, 45, 5–32.
Domingos, Pedro. (2012). A few useful things to know about machine learning. Communications of
the ACM, 55(10), 78–87.
Friedman, Jerome H. (2001). Greedy function approximation: a gradient boosting machine. Annals
of Statistics, 1189–1232.
Gupta, R., Ahuja, R. (2019). The rising prevalence of respiratory disorders and the impact of
environmental factors: A global analysis. Journal of Pulmonary Medicine, 24(3).
Han, Jiawei, Pei, Jian, & Tong, Hanghang. (2022). Data mining: concepts and techniques. Morgan
kaufmann.
Hastie, Trevor. (2009). The elements of statistical learning: data mining, inference, and prediction.
Springer.
Kohavi, R. (1995). A study of cross-validation and bootstrap for accuracy estimation and model
selection. Morgan Kaufman Publishing.
Kotsiantis, Sotiris B., Zaharakis, Ioannis, & Pintelas, P. (2007). Supervised machine learning: A review
of classification techniques. Emerging Artificial Intelligence Applications in Computer
Engineering, 160(1), 3–24.
Mitchell, Tom M., & Mitchell, Tom M. (1997). Machine learning (Vol. 1). McGraw-hill New York.
Patel, V., Baker, N., & King, C. (2019). Errors in manual drug classification systems: A systematic
review. Journal of Pharmaceutical Practice, 32(2), 145–153. Retrieved from
https://doi.org/10.1177/0897190018822451
Smith, L., Johnson, A., & Carter, R. (2018). Simplified dosage regimens and their impact on patient
adherence in chronic disease management. Journal of Pharmacology & Therapeutics, 56(7),
403–415.
Zhang, L., Huang, Y., & Chen, X. (2020). Machine learning applications in healthcare: Challenges and
opportunities .2020.104028. Computers in Biology and Medicine, 126(7), 1–12. Retrieved from
https://doi.org/10.1016/j.compbiomed
© 2024 by the authors. It was submitted for possible open-access publication under the terms
and conditions of the Creative Commons Attribution (CC BY SA) license
(https://creativecommons.org/licenses/by-sa/4.0/).