Indonesian Journal of Statistics and Its Applications

Classification of Rice Growth Phase Using Regression Logistic Multinomial Model and K-Nearest Neighbors Imputation on Satellite Data

2024-12-24T17:24:53+07:00

One of the efforts made by the government to maintain food security is to provide statistical data on rice production through accurate calculation of harvest areas using the area sampling framework approach. Although area sampling framework surveys produce accurate estimates, the costs required are quite high when applying this method. To overcome this problem, one solution that can be applied is to utilize satellite imagery to monitor the greenness index of plants using the enhanced vegetation index. However, in real conditions, the Landsat-8 optical satellite is susceptible to cloud cover, which results in missing data. This study aims to model the phase of rice plants using the regression logistic multinomial model by utilizing Landsat-8 satellites and k-nearest neighbors imputation handling to overcome missing data. The results showed that the model had varying performance in each phase, with an average balanced accuracy of 66.45%. This figure shows that the model can classify the area sampling framework data imputed using the k-nearest neighbors imputation method well. The model shows optimal performance in the late vegetative and generative phases but is less effective in detecting the harvest, puso, and non-rice paddy phases.

Application of Univariate and Multivariate Long Short Term Memory for World Crude Palm Oil Price Prediction

2024-12-06T11:09:42+07:00

Time series analysis is essential for predicting economic and other important factors; it can be done univariately or multivariately. Technological developments created long short term memory that can handle vanishing gradients and long-term dependencies. This research will predict the world price of crude palm oil because Indonesia, as the world's largest crude palm oil producer, is strongly influenced by the world crude palm oil price. This study uses monthly data on crude palm oil, soybean oil, and crude oil prices from January 2002 to May 2024 obtained from the World Bank Commodity Price Data. This research applies univariate and multivariate long short term memory to predicting crude palm oil prices. The use of long short term memory is because the data shows non-linear elements and high volatility. The input used for univariate long short term memory is the crude palm oil price, while multivariate long short term memory uses crude palm oil, soybean oil, and crude oil prices. The univariate long short term memory proved to be more effective in the case of world crude palm oil price prediction. This is proven by the lower mean absolute percentage error of 6,574% compared to the multivariate long short term memory of 6,689%. This univariate long short term memory uses a combination of hyperparameters: neuron 32, epoch 100, time steps 1, batch size 64, and learning rate 0,01.

Missing Value Estimation Using Fuzzy C-Means in Classification of Chronic Kidney Disease

2024-12-07T09:53:15+07:00

Based on World Health Organization (WHO) the cases of death due to Chronic Kidney Disease (CKD) ranked the 10th worldwide in 2020. CKD need to be done prevent early. History data to identify individuals predisposed to CKD in this research. In this research data contains missing values, therefore using Fuzzy C - Means (FCM) method to address it. The percentage of error in clustering CKD using FCM method is 20,25% and balanced accuracy of 84,80%. The result from classification using Classification and Regression Trees (CART) shows that accuracy value of 97,50%; sensitivity of 100,00%; and specificity of 92,86%. Individual suffer from CKD if having (1) hemoglobin more than or equal 13; spesific gravity 1,020 or 1,025; serum creatinine less than 1,3; albumin 1 or 2 or 3 or 4 or 5; and sugar 0 or 2 or 3 or 4 or 5, (2) hemoglobin more than or equal 13; spesific gravity 1,020 or 1,025; and serum creatinine more than or equal 1,3, (3) hemoglobin more than or equal 13 and spesific gravity 1,005 or 1,010 or 1,015, (4) hemoglobin less than 13 and red blood cell count less than 5,5.

Early Preeclampsia Detection Using XGBoost-Cox Proportional Hazard Model

2025-06-11T16:07:38+07:00

Various prognostic models based on survival analysis methods have been proposed to predict the risk of preeclampsia (PE). To develop a more accurate yet interpretable prediction approach, we utilized clinical data from pregnant women collected at a hospital in Jakarta and applied the XGBoost-Cox Proportional Hazard Model (XGB-Cox). This model integrates the predictive power of the XGBoost machine learning algorithm with the Cox Proportional Hazard (Cox-PH) model, which estimates the effect of covariates on event time. Our results show that the XGB-Cox model outperforms the traditional Cox-PH model based on four evaluation metrics: log-likelihood, log-rank test, concordance index (C-index), and Brier score. The XGB-Cox model achieved a higher C-index of 0.8908 compared to 0.7548 for Cox-PH, indicating improved risk discrimination. Kaplan-Meier curves suggest that XGB-Cox provides better separation across risk quartiles. While XGB-Cox generally yields lower Brier Scores, its performance declines at later gestational weeks. The Cox-PH model remains superior in interpretability, offering clear hazard ratios, while XGB-Cox enhances model fitness and still provides meaningful insights into feature importance. Additionally, sensitivity analysis underscores the need to carefully determine the proportion of censored data, as excessive censoring affects model stability. These findings suggest that XGB-Cox provides a robust predictive framework for early PE risk assessment, supporting its potential application in clinical decision-making for maternal healthcare.

Winsorization for Outliers in Clustering Non-Cyclical Stocks with K-Means and K-Medoids

2025-01-31T12:26:37+07:00

Non-cyclical consumer sector stocks are often chosen by investors because the products in this sector are essential products that always in demand by society. Therefore, the demand for these products tends to be stable and defensive or less affected by economic shocks. However, it does not guarantee that every stock in this sector has good performance, thus it is necessary to group stocks based on their fundamental indicators in the form of financial ratios. This research aims to identify the best method by considering outliers and determining the clusters with the best fundamental performance as a recommendation for investors to make the right investment decisions. The data used in this study is secondary data with observations in the form of 50 non-cyclical consumer sector stocks. The variables used are Earning per Share, Return on Equity, Return on Assets, Debt to Equity Ratio, Price to Earnings Ratio, and Price to Book Value. The clustering results indicated that K-Medoids is the best clustering method, both on the data before and after handling extreme outliers with winsorization approach. However, the optimum number of clusters before and after winsorization are different, with 3 and 6 clusters. Considering the influence of extreme outliers and to get a more informative clustering result, the clustering result after the application of winsorization technique was chosen, which resulted in 6 clusters. Cluster 1, which consists of AALI, GGRM, INDF, and SGRO can be recommended because it has excellent fundamental performance, especially in terms of Earning per Share in 2022.

Comparison of The Singular Spectrum Analysis and SARIMA for Forecasting Rainfall in Padang Panjang City

2024-12-06T11:01:33+07:00

Indonesia is an area with a tropical climate, so it has two seasons, namely the rainy season and the dry season. The rainy season lasts from November to March and during this period rainfall tends to be high in several areas. Padang Panjang City is one of the cities with the smallest area in West Sumatra Province, which has the nickname Rain City. This is because the city of Padang Panjang has cool air with a maximum air temperature of 26.1 °C and a minimum of 21.8 °C, so this city has a fairly high level of rainfall with an average of 300 to 400 mm/year. This article discusses rainfall forecasting for Padang Panjang City by comparing the Singular Spectrum Analysis and Seasonal Autoregressive Integrated Moving Average methods. The data used spans 8 years, from January 2016 to December 2023. Forecasting results are obtained from the best method selected based on the smallest Mean Absolute Percentage Error value. The Singular Spectrum Analysis method has a Mean Absolute Percentage Error value of 5.59% and Singular Spectrum Analysis and Seasonal Autoregressive Integrated Moving Average has a value 7.43%. The best forecasting method is obtained by the Singular Spectrum Analysis method.

Application of Singular Spectrum Analysis in Predicting Rupiah Exchange Yuan

2024-12-24T17:19:22+07:00

The exchange rate between two countries is the price of the currency used by residents of these countries to trade with each other, the relationship between the Rupiah exchange rate and the Yuan is one of the important aspects in the dynamics of international trade. Therefore, forecasting the exchange rate is important as an effort to predict the exchange rate of Rupiah against Yuan in the future. The method used for forecasting is Singular Spectrum Analysis, namely decomposition and reconstruction. The accuracy of the resulting forecast is measured using the Mean Absolute Percentage Error criterion. The exploration results obtained are forecasting accuracy based on the Mean Absolute Percentage Error value of 2.15% with a window length of 23 which identifies that the forecasting results are accurate and effective. Forecasting is said to be accurate if the Mean Absolute Percentage Error value is lower than 10% and close to 10%

Digital Newsworthiness Scores Model Using a Combination of Unsupervised and Supervised Learning Approaches

2024-12-18T15:01:12+07:00

The rapid evolution of digital technology has transformed the media landscape, making news more accessible while also introducing challenges related to content quality and accuracy. The rise of misinformation and fake news has diminished public trust in traditional media. A method for evaluating the quality and potential impact of news articles prior to publication. By adapting credit risk scoring principles, a model was used to predict the suitability of news content based on factors such as title length, number of images, news category, and publication timing. A variable target was firstly formed using three clustering methods: K-Means, K-Modes, and K-Medoids. The results indicated that K-Means outperformed the other methods, leading us to use its outcomes for determining publication suitability. Subsequently, stepwise logistic regression was applied to implement the credit risk scoring approach, allowing for variable selection and assessment of importance. Ultimately, ten variables were identified to generate a newsworthiness score, with minimum and maximum scores of 997 and 1407, respectively. The average scores for articles deemed publishable and not publishable were 1137 and 1110. A cutoff score of 1123 was established based on these averages, categorizing 6708 articles (57.9%) as suitable for publication. These findings aim to assist media organizations in refining their content curation processes, thereby enhancing the overall quality of news consumption.

Exploring a Large Language Model on the ChatGPT Platform for Indonesian Text Preprocessing Tasks

2025-06-24T09:09:29+07:00

Preprocessing is a crucial step in Natural Language Processing, especially for informal languages like Indonesian, which contain complex morphology, slang, abbreviations, and non-standard expressions. Traditional rule-based tools such as regex, IndoNLP, and Sastrawi are commonly used but often fall short in handling noisy, user-generated text. This study explores the capability of Large Language Model, particularly ChatGPT-o3, in performing Indonesian text preprocessing tasks, namely text cleaning, normalization, stopword removal, and stemming/lemmatization, and compares it to conventional rule-based approaches. Using two types of datasets, consisting of a small example dataset of five manually constructed sentences and a real-world dataset of 100 tweets about the Indonesian “Makan Bergizi Gratis” program, both preprocessing methods were applied and evaluated. Results show that ChatGPT-o3 performs equally well in text cleaning and significantly better in normalization. However, rule-based methods like IndoNLP and Sastrawi still outperform ChatGPT-o3 in stopword removal and stemming. These findings indicate that while ChatGPT-o3 demonstrates strong contextual understanding and linguistic flexibility, they may underperform in rigid, token-based operations without fine-tuning. This study provides initial insights into using Large Language Models as an alternative preprocessing engine for Indonesian text and highlights the need for hybrid approaches or improved prompt design in future applications.

K-Prototypes Algorithm for School Indexing in Report Card-Based Student Admissions

2025-06-20T12:11:38+07:00

Institut Pertanian Bogor, also known as IPB University, is a state university that was ranked first as the best university in Indonesia by the Ministry of Research and Technology in 2020. It has three main channels in the new student admission selection system. The selection method is called “Seleksi Nasional Berdasarkan Prestasi”. “Seleksi Nasional Berdasarkan Prestasi” is one of the new student admission pathways at IPB University based on report cards without a test. The selection of new student admissions based on report cards requires creating a school index to assess the quality and commitment of each school by grouping schools among “Seleksi Nasional Berdasarkan Prestasi” applicants. One method that can be used is the K-Prototypes algorithm. K-Prototypes can be used to cluster large and mixed-type data (numeric and categorical) by combining distance measures from two non-hierarchical methods, namely the K-Means and K-Modes algorithms. Based on the analysis, the K-Prototypes algorithm yields three optimal clusters, each with distinct characteristics. Cluster 1 is the lowest cluster because it comprises schools with the lowest quality and commitment to new student admissions at IPB University, as indicated by the report card. Cluster 2 has a quality that is not superior to Cluster 3 but is higher than that of Cluster 1. Cluster 3 is the best cluster because it consists of schools that have high quality and commitment to new student admissions at IPB University through the report card route.

Performance Evaluation of ARDL Model Stacked with Boosted Ridge Regression on Time Series Data with Multicollinearity

2025-06-30T00:44:00+07:00

Time series data plays a vital role in financial and economic study. Two commonly applied models for such data are Vector Autoregression (VAR) and Autoregressive Distributed Lags (ARDL). Nonetheless, interdependence among explanatory variables often leads to multicollinearity, posing challenges for model reliability. This study investigates the effectiveness of the ARDL model integrated with boosted ridge regression as a method to mitigate multicollinearity. Due to limitations in available empirical data, simulation data will be generated to support the analysis. The research consists of two stages: synthetic data generation and analysis on simulated data. Results suggest that ARDL performs well under various multicollinearity conditions, particularly when the training set is sufficiently large and model structure is correctly specified. For smaller training sets, the ARDL Ridge variant demonstrates improved predictive performance.

Forecasting Nonlinear Time Series with ARIMA, ANN, and Hybrid Models: A Case Study on Inflation Rate in Sri Lanka

2024-12-12T11:32:44+07:00

In time series forecasting, hybrid models combining autoregressive integrated moving average (ARIMA) and artificial neural networks (ANNs) have gained prominence due to their ability to capture both linear and nonlinear patterns within data. ARIMA models are effective at modeling linear relationships, while ANNs are adept at handling complex nonlinear structures. However, each model has its limitations when used independently. This study presents a hybrid model that integrates the strengths of both ARIMA and ANN to forecast the monthly inflation rate in Sri Lanka using historical data from 1988 to 2018. Our findings demonstrate that the proposed hybrid model outperforms the standalone ARIMA and ANN models, particularly in terms of Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE). By leveraging the complementary strengths of ARIMA and ANN, this hybrid approach provides a robust forecasting framework for handling the diverse structural complexities of time series data