Volume 10, Issue 3, March – 2025 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/25mar1548

Implementation of Random Forest Algorithm for

Air Quality Classification: A Case Study of DKI
Jakarta's Air Quality Index
Mochammad Junus1; Vidorova Nurcahyani2; Rachmad Saptono3;
Nurefa Maulana4; Indra Lukmana Putra5; Zidan Fahreza6
1
Department of Electrical Engineering, State Polytechnic of Malang, Indonesia
2
Departement Policy Analysit, State Goverment of Batu, Indonesia
3
Department of Electrical Engineering, State Polytechnic of Malang, Indonesia
4
Enha Bena Nusantara Ltd, Batu, Indonesia
5
Departement Accounting, State Polytechnic of Malang, Indonesia
6
Department of Electrical Engineering, State Polytechnic of Malang, Indonesia

Publication Date: 2025/04/05

Abstract: Air quality monitoring and classification in urban environments present significant challenges for
environmental management and public health policy. This study implements an optimized Random Forest (RF) algorithm
to classify air quality levels in DKI Jakarta, Indonesia, using the Air Quality Index (AQI) data from 2021. The analysis
incorporates six key pollutants: PM10, PM2.5, NO2, SO2, CO, and O3, with data collected from the Environmental
Management Agency of DKI Jakarta. The RF model was developed using 5000 decision trees with optimized parameters
(mtry=2) and evaluated through stratified sampling with a 70:30 train-test split. The model achieved an exceptional
accuracy of 99.09% with a low Out-of-Bag (OOB) error rate of 2.35%. Feature importance analysis revealed that
particulate matter (PM2.5 and PM10) were the most influential factors, collectively accounting for 78.70% of the model's
decision-making process. The high performance metrics across all air quality categories (Good, Moderate, and Unhealthy)
demonstrate the model's reliability in classification tasks. This research provides insights into environmental monitoring
and policymaking, presenting a framework adaptable to other urban settings. The findings highlight the crucial role of
particulate matter in air quality assessment and suggest targeted strategies for pollution control.. (Abstract)

Keywords: Air Quality Classification, Random Forest, Machine Learning, Air Quality Index, Environmental Monitoring, Jakarta.

How to Cite: Mochammad Junus; Vidorova Nurcahyani; Rachmad Saptono; Nurefa Maulana; Indra Lukmana Putra; Zidan
Fahreza (2025). Implementation of Random Forest Algorithm for Air Quality Classification: A Case Study of DKI
Jakarta's Air Quality Index. International Journal of Innovative Science and Research
Technology, 10(3), 2169-2173. https://doi.org/10.38124/ijisrt/25mar1548

I. INTRODUCTION deteriorating air quality in the region (Kusuma et al., 2019; ,

Syuhada et al., 2023). Several studies have documented that
Air pollution remains one of the most pressing particulate matter (PM2.5) and other pollutants in Jakarta
environmental challenges in urban areas, particularly in frequently exceed national and international safety
rapidly expanding megacities across Southeast Asia (Zuo et thresholds, thereby posing serious health risks to residents
al., 2019). According to the World Health Organization, (Zulfikri, 2023). The Air Quality Index (AQI) in Jakarta has
approximately 99% of the global population breathes air shown concerning trends, with frequent recordings of
with elevated pollutant levels, with developing nations unhealthy air quality levels affecting millions of residents
bearing the most severe consequences (WHO, 2021). (Syuhada et al., 2023).
Jakarta, Indonesia's capital, faces significant air quality
issues driven by rapid urbanization, increasing vehicle Given these challenges, enhancing the accuracy and
emissions, and industrial activities (Amazing Hope Ekeh et efficiency of air quality monitoring systems is essential for
al., 2025). timely policy-making and effective mitigation strategies.
Traditional approaches to air quality monitoring and
Rapid urbanization, increased industrial activities, and classification often lack the predictive capabilities necessary
a surge in vehicular emissions have collectively led to

IJISRT25MAR1548 www.ijisrt.com 2169

Volume 10, Issue 3, March – 2025 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/25mar1548
for effective environmental management and public health By addressing the classification of AQI using Random
protection. Forest, this research not only advances methodological
approaches in air quality analysis but also provides an
Recent advancements in machine learning have offered important tool for local governments and stakeholders. The
promising alternatives to traditional statistical methods for insights derived from this study will support the design of
environmental data analysis. Machine learning techniques, targeted pollution control policies and eventual
particularly Random Forest (RF) algorithms, have emerged improvements in public health outcomes, reaffirming the
as powerful tools for environmental data analysis and critical role of machine learning in environmental science
classification (Beucler et al., 2024). Random Forest has (Jayadi et al., 2024; Vu et al., 2019; Azies, 2023).
gained significant attention in environmental monitoring due
to its ability to handle non-linear relationships, manage II. RESEARCH METHOD
high-dimensional data, and provide robust predictions while
accounting for variable importance (Amazing Hope Ekeh et A. Data Collection and Description
al., 2025). This study utilized air quality monitoring data from
DKI Jakarta collected throughout 2021. The dataset
Unlike other black-box methods, Random Forest comprises 365 daily observations obtained from the
provides insights into the significance of different predictor Environmental Management Agency of DKI Jakarta. Six air
variables and supports a more interpretable decision-making pollutant parameters were measured according to the:
process (Idroes et al., 2023). The algorithm's ensemble
nature, which integrates multiple decision trees, allows it to  Input Variables (Air Pollutants):
effectively capture complex non-linear relationships
between meteorological conditions and pollutant  PM10:Particulate matter with diameter ≤ 10 micrometers
concentrations (Jayadi et al., 2024; , Azies, 2023.). This (μm)
capability is particularly important in urban environments  PM2.5:Fine particulate matter with diameter ≤ 2.5 μm
like DKI Jakarta, where multiple factors interact  NO2: Nitrogen dioxide
dynamically to influence pollutant levels.  SO2: Sulfur dioxide
 CO: Carbon monoxide
Recent studies have demonstrated the algorithm's  O3: Ozone
effectiveness in air quality prediction and classification
across various urban contexts. For example, Natarajan et al.  Output Variable:
(2024) achieved 95% accuracy in classifying air quality in Air quality categories according to AQI standards:
Delhi, while Rakholia et al. (2024) successfully
implemented RF for real-time air quality monitoring in
 Good (0-50)
Mexico City.
 Moderate (51-100)
 Unhealthy for Sensitive Groups (101-200).
Despite numerous studies on air quality prediction
using various machine learning models, there is still a  Unhealthy (151-200)
relative paucity of research applying the Random Forest  Very Unhealthy (201-300)
algorithm to classify the Air Quality Index (AQI)  Hazardous (301 and higher).
specifically for Jakarta. Prior works have applied techniques
such as neural networks and support vector machines; B. Random Forest Model Development
however, they often overlook the advantages of Random
Forest in managing imbalanced datasets and providing  Bootstrap Sampling
feature importance analysis (V. Vu et al., 2019). Moreover,
investigations into the spatial-temporal variability of air  Each tree uses approximately 2/3 of training data
pollutants in Jakarta underscore the need for a more  Remaining 1/3 used for OOB error estimation
adaptable model that can integrate diverse data sources and
yield robust performance in the face of environmental  Node Splitting:
uncertainties (Idroes et al., 2023; Azies, 2023).
 √p features randomly selected at each node (where p=6)
This study attempts to fill this gap by developing a  Gini impurity used as splitting criterion
Random Forest-based classification from AQI (2021 data) in  Minimum samples per leaf = 1
Jakarta. The research investigates six major pollutants,
including: Particulate Matter (PM10 and PM2.5), Nitrogen  Hyperparameter Selection Key Parameters were Chosen
Dioxide (NO2), Sulfur Dioxide (SO2), Carbon Monoxide based on Literature Recommendations:
(CO), and Ozone (O3). Our approach builds upon previous
work by incorporating comprehensive variable importance  n_estimators: 5000 trees for stable performance
analysis and optimizing model parameters for Jakarta's  max_features: 2 (≈√6) following Breiman's
specific context. recommendation
 class_weight: 'balanced' to handle class imbalances.

IJISRT25MAR1548 www.ijisrt.com 2170

Volume 10, Issue 3, March – 2025 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/25mar1548
III. RESULTS AND DISCUSSION

 Model Performance Analysis

The Random Forest classifier demonstrated exceptional performance in categorizing air quality levels. Table 1 presents the
confusion matrix showing the classification results across different air quality categories.

Fig 1 Confusion Matrix of Air Quality Classification

The model achieved an overall accuracy of 99.09% on achieved an accuracy of 98.37% in classifying air quality in
the test dataset, with an Out-of-Bag (OOB) error rate of urban environments.The detailed performance metrics for
2.35%. Based on findings from similar studies, as reported each category are presented in Table 1:
by (Shaziayani et al., 2022), the Random Forest model

Table 1 Classification Performance Metrics by Category

Category Precision Recall F1-score Support
Good 0.94 1.00 0.97 17
Moderate 1.00 0.99 0.99 89
Unhealthy 1.00 1.00 1.00 4

 Variable Importance Analysis Figure 2 illustrates the contribution of each pollutant to the
One of the key advantages of Random Forest is its classification model:
ability to quantify the relative importance of input variables.

Fig 2 Contribution of each pollutant to air quality classification

IJISRT25MAR1548 www.ijisrt.com 2171

Volume 10, Issue 3, March – 2025 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/25mar1548
 The Analysis Revealed the Following Hierarchy of Data Analytics to address urban and ecological
Pollutant Importance: challenges. Gulf Journal of Advance Business
Research, 3(2), 456–482.
 PM2.5 (46.62%) https://doi.org/10.51594/gjabr.v3i2.92
 PM10 (32.08%) [2]. Beucler, T., Gentine, P., Yuval, J., Gupta, A., Peng,
 NO2 (9.77%) L., Lin, J., Yu, S., Rasp, S., Ahmed, F., O’gorman, P.
 O3 (4.46%) A., Neelin, J. D., Lutsko, N. J., & Pritchard, M.
 CO (3.76%) (2024). Climate-invariant machine learning. In Sci.
 SO2 (3.32%) Adv (Vol. 10). https://www.science.org
[3]. Natarajan, S. K., Shanmurthy, P., Arockiam, D.,
The dominance of particulate matter (PM2.5 and Balusamy, B., & Selvarajan, S. (2024). Optimized
PM10) in the model's decision-making process aligns with machine learning model for air quality index
findings from recent studies in other Asian megacities prediction in major cities in India. Scientific Reports,
(Beucler et al., 2024). This result is particularly significant 14(1). https://doi.org/10.1038/s41598-024-54807-1
given that PM2.5 and PM10 are considered the most [4]. Rakholia, R., Le, Q., Vu, K., Ho, B. Q., & Carbajo,
harmful pollutants to human health (WHO, 2021). R. S. (2024). Accurate PM2.5 urban air pollution
forecasting using multivariate ensemble learning
 Model Robustness and Limitations Accounting for evolving target distributions.
While the model demonstrates high accuracy, several Chemosphere, 364.
considerations should be noted: https://doi.org/10.1016/j.chemosphere.2024.143097
[5]. Shaziayani, W. N., Ul-Saufie, A. Z., Mutalib, S.,
 Class Imbalance: The dataset shows an uneven Mohamad Noor, N., & Zainordin, N. S. (2022).
Classification Prediction of PM10 Concentration
distribution of categories, with moderate conditions
being predominant. This was addressed through the use Using a Tree-Based Machine Learning Approach.
of balanced class weights in the model. Atmosphere, 13(4).
https://doi.org/10.3390/atmos13040538
 Spatial Limitations: The current model relies on
[6]. Syuhada, G., Akbar, A., Hardiawan, D., Pun, V.,
aggregated data for DKI Jakarta and may not capture
Darmawan, A., Heryati, S. H. A., Siregar, A. Y. M.,
localized variations in air quality across different city
Kusuma, R. R., Driejana, R., Ingole, V., Kass, D., &
districts.
Mehta, S. (2023). Impacts of Air Pollution on Health
and Cost of Illness in Jakarta, Indonesia.
IV. CONCLUSIONS
International Journal of Environmental Research
and Public Health, 20(4).
This study successfully implemented a Random Forest
https://doi.org/10.3390/ijerph20042916
algorithm for air quality classification in DKI Jakarta using
[7]. WHO. (2021, September 22). WHO global air
2021 monitoring data. The key findings and implications are
quality guidelines: particulate matter (PM2.5 and
as follows:
PM10), ozone, nitrogen dioxide, sulfur dioxide and
 Model Performance
carbon monoxide.
[8]. Zuo, X., Yang, X., Dou, Z., & Wen, J. R. (2019).
 The Random Forest classifier achieved an exceptional RUCIR at TREC 2019: Conversational Assistance
accuracy of 99.09% Track. 28th Text REtrieval Conference, TREC 2019 -
 The low Out-of-Bag error rate of 2.35% demonstrates Proceedings.
the model's robustness https://doi.org/10.1145/1122445.1122456
 High precision and recall values across all air quality [9]. Azies, H. A. (n.d.). Air Pollution in Jakarta,
categories indicate reliable classification performance. Indonesia Under Spotlight: An AI-Assisted Semi-
Supervised Learning Approach.
 Pollutant Importance [10]. Idroes, G. M., Noviandy, T. R., Maulana, A.,
Zahriah, Z., Suhendrayatna, S., Suhartono, E.,
 PM2.5 and PM10 emerged as the most influential Khairan, K., Kusumo, F., Helwani, Z., & Abd
pollutants, collectively accounting for 78.70% of the Rahman, S. (2023). Urban Air Quality Classification
model's decision-making process Using Machine Learning Approach to Enhance
 Secondary contributions came from NO2 (9.77%) and Environmental Monitoring. Leuser Journal of
O3 (4.46%) Environmental Studies, 1(2), 62–68.
 CO and SO2 showed relatively minor influences on air https://doi.org/10.60084/ljes.v1i2.99
quality classification [11]. Jayadi, B. V., Lauro, M. D., Rusdi, Z., Handhayani,
T., & Informasi, F. T. (n.d.). Sistemasi: Jurnal Sistem
REFERENCES Informasi Klasifikasi Indeks Standar Pencemaran
Udara untuk Data Tidak Seimbang menggunakan
[1]. Amazing Hope Ekeh, Charles Elachi Apeh, Pendekatan Pembelajaran Mesin Air Quality Index
Chinekwu Somtochukwu Odionu, & Blessing Classification for Imbalanced Data Using Machine
Austin-Gabriel. (2025). Leveraging machine learning Learning Approach. http://sistemasi.ftik.unisi.ac.id
for environmental policy innovation: Advances in

IJISRT25MAR1548 www.ijisrt.com 2172

Volume 10, Issue 3, March – 2025 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/25mar1548
[12]. Kusuma, W. L., Chih-Da, W., Yu-Ting, Z., Hapsari,
H. H., & Muhamad, J. L. (2019). Pm2.5 pollutant in
asia—a comparison of metropolis cities in indonesia
and taiwan. International Journal of Environmental
Research and Public Health, 16(24).
https://doi.org/10.3390/ijerph16244924
[13]. Syuhada, G., Akbar, A., Hardiawan, D., Pun, V.,
Darmawan, A., Heryati, S. H. A., Siregar, A. Y. M.,
Kusuma, R. R., Driejana, R., Ingole, V., Kass, D., &
Mehta, S. (2023). Impacts of Air Pollution on Health
and Cost of Illness in Jakarta, Indonesia.
International Journal of Environmental Research
and Public Health, 20(4).
https://doi.org/10.3390/ijerph20042916
[14]. V. Vu, T., Shi, Z., Cheng, J., Zhang, Q., He, K.,
Wang, S., & M. Harrison, R. (2019). Assessing the
impact of clean air action on air quality trends in
Beijing using a machine learning technique.
Atmospheric Chemistry and Physics, 19(17), 11303–
11314. https://doi.org/10.5194/acp-19-11303-2019
[15]. Zulfikri, A. (2023). Effects of Pollution and
Transportation on Public Health in Jakarta. In West
Science Interdisciplinary Studies (Vol. 1, Issue 04).

IJISRT25MAR1548 www.ijisrt.com 2173