Search

Landslide is one of the most highly frequent natural hazards that can bring serious casualties. One of the most susceptible landslide regions in Indonesia is Bogor area (the Regency and City of Bogor), which records the highest landslide events in the Province of West Java, Indonesia. An assessment of landslide susceptibility is one of the mitigation measures that can spatially model the zone of landslide hazard. Recently, the Landslide Susceptibility Mapping (LSM) model has been developed using Machine Learning (ML) algorithms. However, there is still no agreement yet on which ML technique is the most appropriate for LSM. Accordingly, this paper aims to explore and compare the 7 ML algorithms for generating the most promising LSM. The LSM uses the available 13 landslide causal factors and a dataset consisting of 822 authorized landslide records and 822 prepared non-landslide points. The resulting LSMs are classified into 5 susceptibility levels, and evaluated through the Area Under Curve (AUC) of the Receiver-Operating Curve (ROC) and statistical indices (sensitivity, specificity, precision, F1-score, and accuracy). The resulting LSMs present that: (1) the very high (VH) class has the largest area percentage in all LSM models, (2) generally, the 7 MLs perform excellent for achieving > 90% AUC value, except for the Decision Tree (DT) (87.68%) in model classification, and (3) moreover, the overall accuracy (ACC) reflects that Random Forest (RF) outperforms the other MLs in model prediction. With this promising result, ML-based LSM models can be promoted as one of the mitigation measures for landslide disaster management. [ABSTRACT FROM AUTHOR]

Poverty is one of the main issues in the economic development of a country, including Indonesia. This was exacerbated when the Covid-19 pandemic had an impact on economic activities, so there are educated people who are still below the poverty line. Therefore, strategies in tackling the problem of poverty need to be developed in line with the times that have entered the era of digitalization. The research aimed to examine the importance of Information and Communication Technologies (ICTs) on high-educated poverty in East Java. Data obtained from the National Survey in Indonesia is analyzed with the random forest algorithm as a classification method in machine learning. The analysis shows that social media, laptop ownership, age, higher education, and internet access are the five most important variables for high-educated poverty classification in East Java. Based on this, a number of recommendations can be made to policy makers regarding the effect of digitalization on high-educated poverty in East Java. [ABSTRACT FROM AUTHOR]

The north coast of Java is the center of economic activity in Indonesia. This area is dynamic and sensitive to various geo-bio-physical aspects. Therefore, a vulnerability study in this area is necessary. This study proposes a machine learning tree-based algorithms modeling approach for Coastal Vulnerability Assessment (CVA) and mapping. The tree-based algorithms used are Gradient Tree Boost (GTB), Classification and Regression Trees (CART), and Random Forest (RF). The study utilized the Google Earth Engine (GEE) platform and twelve variables as input. The prediction results of each of these modeling algorithms have been compared and evaluated to determine the most optimal performance and accuracy. Reference data was obtained from the Ministry of Maritime Affairs and Fisheries of the Republic of Indonesia (KKP). Approximately 70% of the reference data was allocated for training, while the remaining 30% was designated for validation. The CVA assessment yielded overall accuracies of 80.22%, 77.40%, and 71.18% based on the RF, GTB, and CART algorithms, respectively. Meanwhile, the Kappa Index for these three algorithms was 0.72, 0.67, and 0.58, indicating that the models have adequately classified the data. The research outcomes are anticipated to offer insights into the potential utilization of machine learning technology for vulnerability assessment and mapping, contributing to the management of coastal environmental issues. [ABSTRACT FROM AUTHOR]

Using regression and classification machine learning algorithms, this study explores audio features on Spotify that contribute to the popularity of songs streamed in Indonesia, and then evaluates the feature importance for prediction. The publicly accessible Kaggle data consists of 92,755 rows and 20 columns. Using multiple model comparisons for regression and classification, this study identifies Extra Trees Regressor and Random Forest Classifier as the two predictive approaches with the highest accuracy. This study contributes to the scientific literature on hit songs by examining the influence of audio features on a song's popularity using both classification and regression machine learning methods, with an emphasis on Indonesia based on consumer culture theory. [ABSTRACT FROM AUTHOR]

The Mw 7.5 Palu earthquake that occurred on 28 September 2018 (UTC 10:02) on Sulawesi Island, Indonesia, triggered approximately 15,600 landslides, causing about 4000 fatalities and widespread destruction. The primary objective of this study is to perform landslide susceptibility mapping (LSM) associated with this event and assess the performance of the most widely used machine learning algorithms of logistic regression (LR) and random forest (RF). Eight controlling factors were considered, including elevation, hillslope gradient, aspect, relief, distance to rivers, peak ground velocity (PGV), peak ground acceleration (PGA), and lithology. To evaluate model uncertainty, training samples were randomly selected and used to establish the models 20 times, resulting in 20 susceptibility maps for different models. The quality of the landslide susceptibility maps was evaluated using several metrics, including the mean landslide susceptibility index (LSI), modelling uncertainty, and predictive accuracy. The results demonstrate that both models effectively capture the actual distribution of landslides, with areas exhibiting high LSI predominantly concentrated on both sides of the seismogenic fault. The RF model exhibits less sensitivity to changes in training samples, whereas the LR model displays significant variation in LSI with sample changes. Overall, both models demonstrate satisfactory performance; however, the RF model exhibits superior predictive capability compared to the LR model. [ABSTRACT FROM AUTHOR]

The growing utilization of social media platforms enables direct interaction between companies and consumers. However, the expanding range of interactions and real-world data complexities necessitate the development of more sophisticated decision models. To address this, the current research focuses on constructing machine learning models, namely multinomial logistic regression, decision tree, k-nearest neighbor, and random forest, to forecast the engagement level of Twitter posts from three prominent e-commerce platforms in Indonesia: Bukalapak, Blibli, and Tokopedia. The analysis comprises a dataset of 12,786 unique tweets, accumulating 11,870,254 favorites and 2,735,886 retweets over a seven-month period from February 1 to August 31, 2021. The prediction models are built upon three theoretical constructs with seven features, encompassing interactivity (e.g., links, hashtags), vividness (e.g., images, short videos, long videos), and temporal factors (e.g., day of post, last post time). Factors such as post frequency, interactive posting elements, and static visual elements emerge as significant features for predicting the engagement level of Twitter posts. Results demonstrate that the random forest model outperforms singular classifier models, including multinomial logistic regression, decision tree, and k-nearest neighbor models, in terms of precision, recall, and F1 score. [ABSTRACT FROM AUTHOR]

Batik is a work of art from Indonesia that has many types and pattern. One of the batik producing areas is Surakarta, the famous pattern in this area are Sawat, Sementrante, and Satriomanah. The problem that arises is the difficulty of distinguishing the three existing pattern because they have a high level of similarity. Therefore, this research aims to solve these problems using NB and RF methods. As a feature extraction, a Gray Level Cooccurrence Matrix is used as a texture feature extraction. The research phase includes methods for dataset collection, preprocessing, feature extraction, and classification. These two methods, RF and NB, can be used as methods for batik fabric classification. The most accurate result obtained by the RF method was 97.91% accurate in dataset A, while the NB method was 96.66% accurate on the same dataset. According to the research results, it is found that the RF method outperforms the NB method in classifying the types of batik patterns. [ABSTRACT FROM AUTHOR]

Floods in coastal areas occur yearly in Indonesia, resulting in socio-economic losses. The availability of flood susceptibility maps is essential for flood mitigation. This study aimed to explore four different types of models, namely, frequency ratio (FR), weight of evidence (WofE), random forest (RF), and multi-layer perceptron (MLP), for coastal flood susceptibility assessment in Pasuruan and Probolinggo in the East Java region. Factors were selected based on multi-collinearity and the information gain ratio to build flood susceptibility maps in small watersheds. The comprehensive exploration result showed that seven of the eleven factors, namely, elevation, geology, soil type, land use, rainfall, RD, and TWI, influenced the coastal flood susceptibility. The MLP outperformed the other three models, with an accuracy of 0.977. Assessing flood susceptibility with those four methods can guide flood mitigation management. [ABSTRACT FROM AUTHOR]

This study presents the extreme gradient boosting (XGBoost) and random forest (RF) models to predict tourism demand by incorporating international COVID-19 cases, international tourist arrivals, and the destination's quarantine policy predictors. Unlike other 'black box' machine learning models, those two tree-based models offer better interpretability with explicit feature importance and tree structure representations. This paper evaluates the accuracy of these models in predicting international tourist arrivals in Indonesia during the COVID-19 pandemic using long-range (January 2008–June 2021) and short-range (January 2018–June 2021) training datasets. The performance of these two models is compared with benchmark models, such as the artificial neural network, autoregressive integrated moving average, and seasonal ARIMA models. In general, the tree-based machine learning models outperformed all benchmark models. International COVID-19 cases and tourist arrivals predictors have dominating feature importance scores in XGBoost models. Meanwhile, Google trends keywords on quarantine policies show significant importance in RF models but not in the XGBoost models. Moreover, RF models are better than the XGBoost models in terms of accuracy and overcoming overfitting cases. [ABSTRACT FROM AUTHOR]

Mangroves have been protecting coastlines, nourishing wildlife, and capturing carbon for climate regulation. The decline of mangroves calls for action to rapidly and accurately monitor them. Remote Sensing makes it possible to remotely monitor mangroves from images captured from space. Sentinel-1 and Sentinel-2 are examples of remote sensing satellites and there is extensive research on their land cover mapping capabilities, including mangrove mapping. While machine learning is a popular methodology for mangrove mapping (e.g., the use of Random Forest) there exist simpler techniques, i.e., utilizing threshold segmentation-based indices that only use a formula and a specific threshold to extract mangrove extents from satellite imagery (mostly Sentinel imagery). This study compared the products and the accuracy of different threshold segmentation-based mangrove mapping indices in four study areas in the Philippines and one in Indonesia. Results showed that the Mangrove Vegetation Index (MVI), Automatic Mangrove Map and Index (AMMI), and the Optical and SAR images Combined Mangrove Index (OSCMI) subindex SWIRB (full name of this subindex here) were the superior indices with overall accuracies (OA) greater than 80% in all study areas and reaching a maximum of 90%, 91% and 96%, respectively. By McNemar's test showed that their results have insignificant differences. MVI, SWIRB, and AMMI only used Sentinel-2 optical imagery, which means the addition of Sentinel-1 SAR imagery was unnecessary. Since the validation data is a product of machine-learning classification, this shows that using threshold segmentation-based indices is promising as it is simpler, faster, and requires little skill compared to using classification techniques. [ABSTRACT FROM AUTHOR]

Introduction: Diabetes mellitus type 2 (T2DM) is a chronic-multifactorial disease with a high disease burden. Identifying the risk factors for type 2 diabetes mellitus can help in designing prevention strategies. Surveillance system data can be utilized to accurately predict the prevalence of diseases in a community using machine learning algorithm. The aim of this study was to determine the performance of machine learning and to identify important features in classifying T2DM in the Health and Demographic Surveillance System (HDSS) population of the Sleman region of Yogyakarta, Indonesia. Methods: The first two cycles of the Sleman HDSS database were obtained, and factors such as demographics, risky foods, diet composition, and comorbidity were evaluated. After pre-processing the data, we employed binary classification of T2DM using logistic regression and the random forest method. The performance of the two models was then compared, and the imported features were reported. Results: There were 4,611 subjects included in this study including 463 with self-reported T2DM. Significant differences, such as age, level of education, monthly food expenditure, consumption of coffee or other caffeinated beverages, intake of herbs and instant noodles, and health issues, such as hypertension and stroke, were identified between the T2DM and non-DM groups. Apart from fat, rice is the most predominant food in all types of diet compositions. Conclusion: The random forest machine-learning algorithm shows superior performance, with hypertension being the most important feature in the classification of a self-reported T2DM in the Sleman HDSS population. [ABSTRACT FROM AUTHOR]

Policymakers seeking to target health policies efficiently towards specific population groups need to know which individuals stand to benefit the most from each of these policies. While traditional approaches for subgroup analyses are constrained to only consider a small number of pre-defined subgroups, recently proposed causal machine learning (CML) approaches help explore treatment-effect heterogeneity in a more flexible yet principled way. Causal forests use a generalisation of the random forest algorithm to estimate heterogenous treatment effects both at the individual and the subgroup level. Our paper aims to explore this approach in the setting of health policy evaluation with strong observed confounding, applied specifically to the context of mothers' health insurance enrolment in Indonesia. Comparing two health insurance schemes (subsidised and contributory) against no insurance, we find beneficial average impacts of enrolment in contributory health insurance on maternal health care utilisation and infant mortality, but no impact of subsidised health insurance. The causal forest algorithm identified significant heterogeneity in the impacts of contributory insurance, not just along socioeconomic variables that we pre-specified (indicating higher benefits for poorer, less educated, and rural women), but also according to some other characteristics not foreseen prior to the analysis, suggesting in particular important geographical impact heterogeneity. Our study demonstrates the power of CML approaches to uncover unexpected heterogeneity in policy impacts. The findings from our evaluation of past health insurance expansions can potentially guide the re-design of the eligibility criteria for subsidised health insurance in Indonesia. [ABSTRACT FROM AUTHOR]

Flood forecasting based on water level modeling is an essential non-structural measure against compound flooding over the globe. With its vulnerability increased under climate change, every coastal area became urgently needs a water level model for better flood risk management. Unfortunately, for local water management agencies in developing countries building such a model is challenging due to the limited computational resources and the scarcity of observational data. Here, we attempt to solve the issue by proposing an integrated hydrodynamic and machine learning approach to predict compound flooding in those areas. As a case study, this integrated approach is implemented in Pontianak, the densest coastal urban area over the Kapuas River delta, Indonesia. Firstly, we built a hydrodynamic model to simulate several compound flooding scenarios, and the outputs are then used to train the machine learning model. To obtain a robust machine learning model, we consider three machine learning algorithms, i.e., Random Forest, Multi Linear Regression, and Support Vector Machine. The results show that this integrated scheme is successfully working. The Random Forest performs as the most accurate algorithm to predict flooding hazards in the study area, with RMSE = 0.11 m compared to SVM (RMSE = 0.18 m) and MLR (RMSE = 0.19 m). The machine-learning model with the RF algorithm can predict ten out of seventeen compound flooding events during the testing phase. Therefore, the random forest is proposed as the most appropriate algorithm to build a reliable ML model capable of assessing the compound flood hazards in the area of interest. [ABSTRACT FROM AUTHOR]

Throughout Indonesia ecological degradation, agricultural expansion, and the digging of drainage canals has compromised the integrity and functioning of peatland forests. Fragmented landscapes of scrubland, cultivation, degraded forest, and newly established plantations are then susceptible to extensive fires that recur each year. However, a comprehensive understanding of all the drivers of fire distribution and the conditions of initiation is still absent. Here we show the first analysis in the region that encompasses a wide range of driving factors within a single model that captures the inter‐annual variation, as well as the spatial distribution of peatland fires. We developed a fire susceptibility model using machine learning (XGBoost random forest) that characterizes the relationships between key predictor variables and the distribution of historic fire locations. We then determined the relative importance of each predictor variable in controlling the initiation and spread of fires. The model included land‐cover classifications, a forest clearance index, vegetation indices, drought indices, distances to infrastructure, topography, and peat depth, as well as the Oceanic Niño Index (ONI). The model performance consistently scores highly in both accuracy and precision across all years (>75% and >67.5% respectively), though recall metrics are much lower (>25%). Our results confirm the anthropogenic dependence of extreme fires in the region, with distance to settlements and distance to canals consistently weighted the most important driving factors within the model structure. Our results may help target the root causes of fire initiation and propagation to better construct regulation and rehabilitation efforts to mitigate future fires. Key Points: The primary driving factors affecting fires in the ex‐Mega Rice Project area are anthropogenicThe predictive capability of fire models is limited by the influence of human agency on the location of fire ignitionMore comprehensive understanding of the driving social factors affecting fire initiation are required [ABSTRACT FROM AUTHOR]

Technology advancement has developed a shift perception towards better service from internet providers, and the power to move easily to another provider to secure improved quality results in customer churn. Internet service providers must detect the risk of churn at the earliest opportunity if they want to retain their customers. This study aimed to predict churn using recent developments in machine learning approaches, and customer data from one of the biggest fixed broadband companies in Indonesia was selected as a case study. Ensemble learning is the collaboration of meta-algorithms to improve model performance, and two such approaches were performed in this study, namely random forest and extreme gradient boosting (XGBoost). The results show that the ensemble learning models outperform classical technique and XGBoost is the best algorithm for predicting customer churn. Customers are thereby clustered as being at high, medium, or low risk of churn, and the company can specify particular retention strategies towards each customer cluster. [ABSTRACT FROM AUTHOR]

LQ45 is an Indonesia Stock Exchange Index (ISX) incorporate of 45 companies that meet certain criteria to target investors for selecting certain stocks. The prediction of stock price direction in the financial world is a major issue. The implementation of machine learning and other algorithms for market price analysis and forecasting is a very promising field. Different types of classification algorithms were used to predict the stock market. However, when individual studies are considered separately there is no clear consensus that algorithms work best. In this research, a comparison framework is proposed, which aims to benchmark the performance of a wide range of classification models and use them to predict the LQ45 index. The data in this research contains the transaction level and capitalization size are obtained from the Indonesian Stock Exchange (ISX). For analysis purposes, we set out 10 classifiers that can be used to build classification models and test their performance in the LQ45 dataset. The performance criterion chosen to measure this effect is accuracy, recall, and precision. The results showed that the random forest algorithm had the best performance for predicting the LQ45 index. Whilst the classification and regression trees, C4.5, support vector machine, and logistic regression algorithms also perform well. Besides, the models based on traditional statistical-based learners that are Naïve Bayes and linear discriminant analysis seem to underperform for predicting the LQ45 index. These results are not only beneficial to enrichment the machine learning techniques literature but also have a significant influence on the stock market prediction in terms of the ability to predict the LQ45 index. [ABSTRACT FROM AUTHOR]

The random forest (RF) algorithm is among the most commonly applied machine learning algorithms in remote sensing. In this study we tested a new approach to improving the accuracy of RF algorithm when applied to multispectral derived bathymetry by increasing predictor numbers and improving hyperparameter tuning. This approach goes beyond previous work that only applied an auto-tuning hyperparameter and linearized reflectance. We tested our experimental approach on the Gili Islands of Indonesia by comparing the optimized RF to basic RF algorithms used to determine water depth from multispectral imagery. The findings of this study indicate that the optimized RF approach was particularly advantageous in highdimension data: errors in water depth prediction accuracy improved by 46% after optimization. [ABSTRACT FROM AUTHOR]

Coffee is an export commodity that is prone to fraudulent practices. Therefore, this study presents a novel approach to authenticate coffee origins using targeted metabolomics with gas chromatography-tandem mass spectrometry (GC-MS/MS) and machine learning models. A total of 200 coffee samples from different harvest years and areas from Indonesia were extracted using the derivatisation method and then analysed for their metabolite profiles. Several supervised machine-learning models were tested to classify coffee origins and discover their potential markers. The study found various metabolite features spanning diverse chemical classes, encompassing sugar alcohols, carbohydrates, amino acids, organic acids, fatty acids, and phenols. Random forest (RF) and partial least squares discriminant analysis (PLS-DA) were among the most accurate models in predicting the origin of coffee from several classes in the validation dataset. The accuracy of both models is in the range of 91%–100%. Furthermore, this study proposes a new strategy for determining "intersection features" as the set of features that are important and common to both RF and PLS-DA models, thereby providing a robust selection of coffee origin markers. Overall, the approach and findings of this study have far-reaching implications for coffee authentication. • Metabolomics coupled with machine learning approaches can predict the origin of coffee. • Random forest and PLS-DA show high accuracy in classifying the origin of coffee. • Important metabolite features were assigned to each coffee origin. [ABSTRACT FROM AUTHOR]

Ganoderma boninense is a fungus that causes basal stem rot (BSR) disease in oil palm plantations. BSR is a major disease in oil palm plantations in both Indonesia and Malaysia. There is no effective treatment for curing BSR; current treatments only prolong the life of oil palms. One strategy to control BSR is early detection of G. boninense infection. Based on the infection symptoms, many researchers have applied remote-sensing techniques for early detection and mapping of BSR disease in oil palms. The main objectives of this article were to evaluate the potential of machine-learning models for predicting BSR disease in oil palm plantations and to produce maps of the distribution of BSR disease. QuickBird imagery archived on 4 August 2008 was applied in three classifier models: Support Vector Machine, Random Forest (RF), and classification and regression tree models The RF model was best at predicting, classifying, and mapping oil palm BSR in terms of overall accuracy (OA), producer accuracy, user accuracy, and kappa value. Using 75% of the data for training and 25% for testing, the RF classifier model achieved 91% OA. In addition, this model separated the healthy and unhealthy oil palms in the study sites into 37,617 (75%) and 12,320 (25%) individuals, respectively. [ABSTRACT FROM AUTHOR]

This study aims to put a supervised learning method for automatically classifying lithofacies in well-logging dataset, where several machine learning algorithms were compared in this study that took place in the Tarakan Basin, Indonesia. The predicted lithofacies in this study including shale, shaly sandstone, sandstone, and coal, where coal is considered as the unique lithofacies in the study area. As training and testing datasets, we used two separate well log datasets from the Tarakan Basin. The first well, named Omnicron, was used to train the model, while the second well, named Kay, was used to test it. Random Forest and Gradient Boosting outperformed the other models in the experiment, with the accuracy of 87.49% and 87.01%, respectively. When it came to classifying coal, however, both approaches had issues. The Pr-Recall curve revealed that the coal score was under average precision in each facies, with values of roughly 0.52 and 0.38, respectively, which explaining why, even with high accuracy, the machine learning algorithm predicted poorly in one lithofacies class. In order to evaluate this coal misclassification, we used rock physics to analyze the machine learning prediction in this report. As result, we found that each facies is well-differentiated by physical properties, and the predicted lithofacies have a distribution that is close to the original facies however, coal may be potentially misclassified as other lithofacies as some of the coals have similar rock physical properties with the surrounding lithology (e.g. coal with a mixture of shale may have similar DT and GR responses). Based on this research, the use of machine learning in the Tarakan Basin effectively provides lithofacies data with a high degree of precision and accuracy in a much shorter time. • The Tarakan basin lithofacies categorization using random forest and gradient boosting. • Evaluation each multiclass of the supervised learning methods. • Coal misclassification in case of high accuracy result. • Log-facies classification analysis, and rock physics analysis based on machine learning models. [ABSTRACT FROM AUTHOR]

Basal stem rot (BSR) disease occurs due to the most aggressive and threatening fungal attack of the oil palm plant known as Ganoderma boninense (G. boninense). BSR is a disease that has a significant impact on oil palm crops in Malaysia and Indonesia. Currently, the only sustainable strategy available is to extend the life of oil palm trees, as there is no effective treatment for BSR disease. This study used thermal imagery to identify the thermal features to classify non-infected and BSR-infected trees. The aims of this study were to (1) identify the potential temperature features and (2) examine the performance of machine learning (ML) classifiers (naïve Bayes (NB), multilayer perceptron (MLP), and random forest (RF) to classify oil palm trees that are non-infected and BSR-infected. The sample size consisted of 55 uninfected trees and 37 infected trees. We used the imbalance data approaches such as random undersampling (RUS), random oversampling (ROS) and synthetic minority oversampling (SMOTE) in these classifications due to the different sample sizes. The study found that the Tmax feature is the most beneficial temperature characteristic for classifying non-infected or infected BSR trees. Meanwhile, the ROS approach improves the curve region (AUC) and PRC results compared to a single approach. The result showed that the temperature feature Tmax and combination feature TmaxTmin had a higher correct classification for the G. boninense non-infected and infected oil palm trees for the ROS-RF and had a robust success rate, classifying correctly 87.10% for non-infected and 100% for infected by G. boninense. In terms of model performance using the most significant variables, Tmax, the ROS-RF model had an excellent receiver operating characteristics (ROC) curve region (AUC) of 0.921, and the precision–recall curve (PRC) region gave a value of 0.902. Therefore, it can be concluded that the ROS-RF, using the Tmax, can be used to predict BSR disease with relatively high accuracy. [ABSTRACT FROM AUTHOR]