VARIANSI: Journal of Statistics and Its application on Teaching and Research

Penerapan Algoritma Naive Bayes untuk Klasifikasi Penerima Bantuan Program Keluarga Harapan (PKH)

2026-04-27T12:00:29+08:00

Salah satu metode klasifikasi yang umum digunakan untuk menentukan kelayakan penerima bantuan Program Keluarga Harapan (PKH) adalah Algoritma Naive Bayes yang sering disebut juga Naive Bayes Classifier. Metode ini adalah probabilitas untuk mengklasifikasikan data secara cepat dan efisien untuk analisis kelayakan dalam program bantuan sosial. Naive Bayes adalah klasifikasi yang menggunakan pendekatan probabilitas dan statistik untuk mengelompokkan data. Pada penelitian ini, dilakukan penerapan algoritma Naive Bayes dalam mengklasifikasikan penerima bantuan Program Keluarga Harapan serta mengetahui tingkat akurasi, recall dan presisi dari metode Naive Bayes. Hasil dari penelitian ini adalah nilai akurasi yang dihasilkan dari metode Naive Bayes sebesar 90% pada pembagian data training dan testing 60%:40%, akurasi nilai 93% pada pembagian data training dan testing 70%:30%, serta nilai akurasi 90% pada pembagian data training dan testing 80%:20%.

Pengelompokan Kabupaten/Kota di Provinsi Jambi Berdasarkan Indikator IPM Menggunakan Metode K-Medoids Cluster

2026-04-27T12:02:33+08:00

The Human Development Index (HDI) is an index that used to measure the quality of human resources and societal welfare. The HDI value in Jambi Province keeps increasing. Nevertheless, there was a decline in the growth of HLS and RLS in 2024. In addition, the RLS of regencies/cities in Jambi Province just approaching 9 years. Therefore, further analysis of HDI indicators in each regency/city is needed to support the improvement of human development in regencies/cities in Jambi Province. This study aims to group the regencies/cities in Jambi Province based on HDI indicators. The grouping using k-medoids cluster with silhouette coefficient method to determine the optimal number of clusters. The clustering results formed 2 clusters. Cluster 1 consists of 9 regencies/cities with HDI indicators lower than cluster 2, which consists of 2 regencies/cities. Therefore, the results indicate that cluster 2 has better human development quality compared to cluster 1.

Analisis Kepercayaan Informasi Peringatan Dini terhadap Respons Darurat Banjir di Kecamatan Dayeuhkolot dengan Regresi Linier

2026-04-27T12:04:05+08:00

Banjir merupakan bencana hidrometeorologis yang sering terjadi dan berdampak signifikan terhadap kehidupan masyarakat, khususnya di wilayah rawan banjir seperti Kecamatan Dayeuhkolot, Kabupaten Bandung. Respons darurat masyarakat menjadi faktor penting dalam upaya pengurangan risiko bencana, salah satunya dipengaruhi oleh tingkat kepercayaan terhadap informasi peringatan dini. Penelitian ini bertujuan untuk menganalisis pengaruh kepercayaan terhadap informasi peringatan dini terhadap respons darurat banjir masyarakat Kecamatan Dayeuhkolot. Penelitian menggunakan pendekatan kuantitatif dengan desain eksplanatori dan analisis regresi linier. Hasil penelitian menunjukkan bahwa kepercayaan terhadap informasi peringatan dini berpengaruh positif dan signifikan terhadap respons darurat banjir, dengan koefisien regresi sebesar 0,506 dan tingkat signifikansi 0,000 (<0,05). Nilai koefisien determinasi (R²) sebesar 0,213 menunjukkan bahwa kepercayaan merupakan salah satu faktor yang berkontribusi dalam membentuk respons darurat masyarakat. Temuan ini menegaskan bahwa penguatan kepercayaan publik memiliki peran penting dalam mendukung efektivitas sistem peringatan dini yang berorientasi pada masyarakat di wilayah rawan banjir.

Implementasi Model Hybrid Autoregressive Fractionally Integrated Moving Average-Neural Network (ARFIMA-NN) pada Peramalan Indeks Harga Saham Gabungan

2026-04-27T12:07:07+08:00

Fenomena fluktuasi ekstrem pada harga penutupan Indeks Harga Saham Gabungan (IHSG) di Bursa Efek Indonesia (BEI) menciptakan ketidakpastian yang sulit diprediksi, sehingga peramalan pada data harga penutupan IHSG dapat membantu investor untuk mengantisipasi risiko investasi dan mempermudah investor untuk menentukan strategi investasi pada periode mendatang. Model hybrid Autoregressive Fractionally Integrated Moving Average-Neural Network (ARFIMA-NN) diimplementasikan karena model ini mampu menangani karakteristik long memory dan memiliki kemampuan menangkap pola non-linier, yang diharapkan dapat meningkatkan akurasi pada peramalan. Berdasarkan hasil analisis, diperoleh hasil peramalan menggunakan model hybrid ARFIMA-NN dengan 1 hingga 3 neuron yang menunjukkan bahwa nilai MAPE berada di bawah 10% atau peramalan sangat baik. Selanjutnya berdasarkan model hybrid ARFIMA(1;0,51;4)-NN 2 menggunakan data IHSG periode Januari 2005 hingga dengan Desember 2024 diperoleh IHSG periode Januari hingga Desember 2025 yang meningkat setiap bulannya.

Comparison of Geographically Weighted Regression (GWR) and Mixed Geographically Weighted Regression (MGWR) Models (Case Study: Crime in South Sulawesi)

2026-04-27T22:45:14+08:00

The Geographically Weighted Regression (GWR) model operates by taking into account how the relationships between different factors change across geographic space. Meanwhile, the Mixed Geographically Weighted Regression (MGWR) model permits certain variables to exhibit spatially varying (local) effects, while other variables are assumed to have constant effects across all locations. Both models are relevant to be applied in crime studies influenced by variations in regional conditions. The objective of this study is to evaluate the GWR and MGWR approaches in selecting the best model to explain factors associated with crime cases in South Sulawesi. The data used include the number of crime cases in South Sulawesi in 2024 along with factors presumed to influence them. The investigation's outcomes suggest the GWR model demonstrates higher appropriateness compared to the MGWR model, evidenced by its reduced Akaike Information Criterion (AIC) score and a 98.44% coefficient of determination . Based on the best-fitting model, population density and the number of poor residents were identified as the main factors influencing criminality in South Sulawesi in 2024.

PEMODELAN JUMLAH KASUS KEMATIAN BAYI DAN IBU MENGGUNAKAN BIVARIATE GENERALIZED POISSON REGRESSION DI PROVINSI JAWA BARAT

2026-04-27T12:10:59+08:00

Angka Kematian Ibu (AKI) dan Angka Kematian Bayi (AKB) tidak dapat dipisahkan karena kondisi kesehatan ibu hamil berdampak langsung terhadap perkembangan dan kesehatan janin. Ini selaras dengan Rencana Pembangunan Jangka Menengah Nasional (RPJMN) 2025-2029 dan target tujuan pembangunan berkelanjutan (SDGs) 2030, dimana Provinsi Jawa Barat ingin menurunkan Angka Kematian Ibu (AKI) dan Angka Kematian Bayi sebesar 4 sampai 5%. Penelitian ini dimaksudkan untuk melakukan memperkirakan jumlah kematian ibu dan bayi dan serta mengkaji faktor apa saja yang terlibat dengan menggunakan pendekatan Bivariate Generalized Poisson Regression (BGPR). Metode BGPR ini dipilih karena mengatasi masalah overdispersi yang ditemukan pada data dan dua data independen yang saling berkorelasi pada data Angka Kematian Ibu dan Bayi di Provinsi Jawa Barat 2024. Penaksir parameter dilakukan dengan metode Maximum Likelihood Estimation (MLE) dan pengujian hipotesis menggunakan metode Maximum Likelihood Ratio Test (MLRT). Pemilihan model terbaik menggunakan nilai AIC terkecil. Hasil penelitian menunjukkan bahwa persentase pemberian zat besi (Fe90), persalinan oleh tenaga Kesehatan, kunjungan ibu hamil, dan persentase komplikasi kebidanan yang ditangani berpengaruh sigfinikan terhadap angka kematian ibu, sedangkan persentase persalinan oleh tenaga kerja dan kunjungan ibu hamil (K4) berpengaruh signifikan terhadap angka kematian bayi.

COMPARISON OF NEWTON RAPHSON AND SECANT METHODS TO DETERMINE THE OPTIMAL POINT OF TIKTOK APPLICATION

2026-05-01T19:23:28+08:00

The growth of digital application users generally follows a non-linear pattern that can be modeled using the logistics growth function, which has the characteristic of an inflection point, which is a condition when the growth rate reaches the maximum value. Optimal point determination involves solving non-linear equations that cannot always be solved directly, so a numerical approach is required. This study aims to determine the optimal growth point of TikTok application users and compare the performance of the Newton–Raphson and Secant methods in solving non-linear equations in the logistics model. User growth data was obtained from the Google Play Store and simulated using logistics growth parameters that represent the characteristics of applications with a high level of virality, with analytics solutions as an evaluation reference. The calculation results show that the optimal point of growth of TikTok users is around the 6th week. The Secant method yielded an optimal point estimate of 5.972 with an RMSE value of 0.0150 and a relative error of 0.25%, while the Newton–Raphson method yielded an estimate of 5.773 with an RMSE value of 0.2140 and a relative error of 3.57%. The difference in error rate and convergence stability shows that the Secant method provides a more effective approach in determining the optimal growth point of digital application users based on the logistics model.

APPLICATION OF SVM FOR SENTIMENT ANALYSIS REGARDING THE EFFICIENCY OF APBN AND APBD IN 2025

2026-04-27T13:19:03+08:00

The policy on expenditure efficiency in the 2025 APBN and APBD has triggered diverse public responses on social media, necessitating sentiment analysis to identify emerging opinion trends. The analysis employs the Support Vector Machine (SVM) method, a margin-based classification algorithm that constructs an optimal separation between classes through the identification of the best hyperplane, where optimality is achieved when the separating margin is maximized. This study aims to identify sentiment patterns and classify public opinion regarding the budget efficiency policy to provide a measurable quantitative overview beyond subjective assessment. Data were collected from the X platform during the period 15 January–25 March 2025 using the keyword “efisiensi anggaran.” The results indicate that negative sentiment dominates at 53%, while positive sentiment accounts for 47%. The SVM model achieved an accuracy of 99%, indicating strong performance in classifying sentiment related to the 2025 budget efficiency policy

Comparative Analysis of DES-Brown and DES-Holt Methods in Forecasting the Stock Price of PT Telekomunikasi Indonesia Tbk

2026-05-01T19:30:44+08:00

This study aims to dermine the best forecasting method for the stock price of PT Telekomunikasi Indonesia Tbk using the Double Exponential Smoothing (DES) Brown and DES-Holt methods. The data used consist of stock prices from January 2019 to September 2025. The DES-Brown method employs a single parameter, while DES-Holt uses two parameters. Forecasting accuracy is evaluated using Mean Absolute Deviation (MAD), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). The results indicate that the DES-Brown method with a smoothing parameter produces the smallest forecasting errors compared to the DES-Holt method, with MAD, RMSE , and MAPE . Therefore, it can be concluded that the DES-Brown method is the most suitable approach for forecasting the stock price of PT Telekomunikasi Indonesia Tbk.

Robust Panel Data Regression Model of Standard Error in Firm Size, Leverage, and Profitability on Firm Value (Case Study: The Indonesian Mining Sector, 2022–2024, Listed on the Indonesia Stock Exchange)

2026-05-01T07:33:41+08:00

Corporate financial information such as firm size, leverage, and profitability sends signals to the market that are reflected in firm value. However, previous studies have yielded inconsistent results, likely due to differences in estimation methods and the disregard of violations of classical assumptions in panel data. This study aims to analyze the effects of firm size (Size), leverage (DER), and profitability (ROA) on firm value (PBV) by applying panel data regression with robust standard error correction. Data were collected from 21 mining sector companies listed on the Indonesia Stock Exchange (IDX) during the 2022–2024 period, yielding 63 observations. The model selected based on the Chow Test (p=1.46E-09) and the Hausman Test (p=0.002) is the Fixed Effects Model (FEM). The results of the classical assumption tests indicate violations of heteroscedasticity (p=0.029) and autocorrelation (p=0.005), so the estimation was continued using cluster-robust standard errors (clustering by time). After adjusting for the model, it was found that all three variables simultaneously had a significant effect on firm value (F-statistic, p = 0.0538). Partially, firm size had a significant negative effect (coefficient -0.481; p=0.038), leverage had a significant positive effect (coefficient 0.672; p=0.018), and profitability had a marginally significant negative effect (coefficient -0.796; p=0.092). An R-squared value of 17.6% indicates that there are still other factors outside the model that influence firm value. The conclusion of this study confirms that in the context of the Indonesian mining sector in the post-pandemic period, the market responds negatively to companies with large assets and high profitability, but responds positively to increased debt. These findings imply that investors should not focus solely on short-term profitability, and that company management should determine the optimal capital structure to increase firm value.

FB Prophet Algorithm Based on Clustering for Stock Price Prediction

2026-05-01T19:26:52+08:00

Extreme volatility in banking stocks like PT Bank Central Asia Tbk (BBCA) decreases single forecasting model accuracy due to high data heterogeneity. This study aims to analyze BBCA stock price prediction accuracy using the FB Prophet algorithm mediated by K-Means Clustering preprocessing. A quantitative time-series method was applied to monthly data from 2014–2025. Results show that K-Means integration (k=3) effectively resolves data heterogeneity. Globally, the FB-Prophet model yielded a Mean Absolute Percentage Error (MAPE) of 20.34%. However, cluster-based evaluation demonstrated superior accuracy during transition phases (MAPE 9.83%) and low-price phases (MAPE 10.13%), dropping the average cluster error to 16.22%. Accuracy decreased only during highly volatile peak price phases (MAPE 28.70%). The 12-month projection for 2026 indicates a stable, conservative linear growth trend, closing at Rp8,532.34. Conclusively, this hybrid Clustering-Forecasting approach provides a more comprehensive and accurate prediction mapping based on distinct market phases.

Trend, Cyclical, and Forecasting Analysis of Indonesia’s Monthly Inflation Using the Hodrick–Prescott Filter and ARIMA

2026-04-30T22:37:29+08:00

This study aims to analyze the structure of inflation and forecast monthly inflation in Indonesia using a time series approach. The method used is the Hodrick–Prescott Filter to decompose data into trend and cycle components, and the ARIMA model to forecast inflation. The data used is monthly inflation data for the period 2010–2025. The decomposition results show that inflation has a relatively stable long-term trend with short-term fluctuations reflecting the presence of economic shocks. Based on model identification, the best model is ARIMA(2,0,1)(1,0,1)[12] which is able to capture past influences, seasonal components, and short-term shocks. The evaluation results show that the model meets the white noise assumption and is suitable for use in forecasting. The forecasting results show that inflation tends to be stable with a moderate increasing tendency, although uncertainty increases over longer periods. This study shows that the combination of structural analysis and time series modeling provides a more comprehensive understanding of inflation dynamics and produces relevant predictions to support decision making.