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

Pemodelan Distribusi Spasial Hotspot di Kabupaten Banjar Menggunakan Log-Gaussian Cox Process

Sigit Dwi Prabowo — 2025-09-30

Forest and land fires are recurring ecological and socio-economic disasters in Banjar Regency, South Kalimantan Province, with complex triggers. A deep understanding of the spatial distribution of fire risk is crucial for effective mitigation efforts. This study aims to model the spatial intensity of hotspots as a proxy for forest and land fires events in Banjar Regency and produce a fire risk surface map. The data used in this study are hotspot data from the siPongi website in Banjar Regency for the period 2013–2024, along with elevation data analyzed using the Log-Gaussian Cox Process (LGCP) spatial statistical model. The analysis results show that elevation has a negative but statistically insignificant effect on hotspot intensity, where fire risk tends to be higher at lower elevations. The LGCP model proved effective in capturing the complex spatial patterns of hotspot occurrences, separating trends driven by covariates and residual spatial clustering. The resulting risk intensity map successfully identified high-risk clusters, particularly concentrated in western districts dominated by peatlands and agricultural activities.

Pemodelan dan Prediksi Pola Musiman Menggunakan Holt-Winters

Thesya Atarezcha Pangruruk — 2025-09-30

Samarinda City, with its tropical climate, experiences significant variations in rainfall throughout the year. This instability has the potential to cause impacts such as flooding, disruptions in the agricultural sector, and damage to infrastructure. This study aims to analyze and forecast the seasonal rainfall patterns in Samarinda City by applying the Holt Winters Exponential Smoothing method based on a multiplicative model. Monthly rainfall data were analyzed to identify stationarity properties in both mean and variance. The results indicate that the data are stationary in the mean but not in the variance, thus justifying the use of the Holt-Winters Multiplicative Exponential Smoothing model. Parameter estimation yielded alpha , beta , and gamma values of 1 each, with a MAPE of 50%, indicating a moderate level of accuracy. Despite the relatively high error rate, the model remains effective in illustrating seasonal patterns, which can be useful for preliminary water resource management planning in the region

Penerapan Metode Kuadratik untuk Peramalan Banyaknya Penduduk Miskin di Sulawesi Selatan Tahun 2008-2025

Nalto Batty Mangiri — 2025-09-30

Masalah kemiskinan adalah masalah yang kompleks dan bersifat multidimensional atau saling berkaitan antara berbagai aspek diantaranya yaitu aspek sosial, ekonomi, dan budaya, serta aspek lainnya. Banyaknya penduduk miskin di Indonesia adalah 23,85 juta pada Maret 2025. Provinsi Sulawesi Selatan pada Maret 2025, terdapat kurang lebih 698,13 ribu penduduk miskin. Sebagai langkah pencegahan meningkatnya angka kemiskinan perlu dilakukan peramalan banyaknya penduduk miskin sehingga pemerintah dapat melakukan perencanaan kebijakan. Data yang digunakan pada penelitian ini adalah data tahun 2008-2025 yang bersumber dari Badan Pusat Statistik Provinsi Sulawesi Selatan. Penelitian ini menggunakan Analisis Trend Nonlinear khususnya Metode Kuadratik untuk melakukan peramalan banyaknya penduduk miskin. Metode ini cocok digunakan untuk data 10 periode atau lebih. Metode Kuadratik memiliki nilai R-Square sebesar 80,24% dan MAPE sebesar 3,28%. Hasil Peramalan selama 6 tahun menunjukkan banyaknya penduduk miskin di Provinsi Sulawesi Selatan mengalami peningkatan.

Penerapan Analisis Regresi Spatial Durbin Model Terhadap Penyakit Tuberkulosis Di Provinsi Sulawesi Selatan Tahun 2022

Muhammad Akhyar Hadi — 2025-09-30

By supplying geographical effects at several sites that serve as the centre of observation, the spatial regression analysis approach assesses the connection between a single variable and multiple other variables. The Spatial Durbin Model is one technique utilised in spatial regression analysis. A special instance of the spatial autoregressive model (SAR) is the spatial Durbin model, which incorporates a spatial lag into the model by adding a lag influence to the independent variables. The goal of this study is to develop a Spatial Durbin model and identify the variables that significantly affect tuberculosis (TBC) in the province of South Sulawesi. The results of this research obtained a Spatial Durbin Model regression model which was significant at a significant level of P-value <α=0.1) using variable influencing factors with a determination coefficient (R2) of 49.74%. Elements that possess a noteworthy impact on the number of Tuberculosis (TB) diseases in South Sulawesi Province are per capita income.

KLASIFIKASI CURAH HUJAN DI KOTA MAKASSAR MENGGUNAKAN GRADIENT BOOSTING MACHINE (GBM)

Hardianti Hafid — 2025-09-30

Rainfall is one of the important parameters in determining the climate of an area. Makassar, as one of the largest cities in Indonesia, has varying rainfall patterns throughout the year. This research aims to classify rainfall in Makassar City using the Gradient Boosting Machine (GBM) method. The secondary data used in this study were obtained from the Meteorology, Climatology, and Geophysics Agency (BMKG), with predictor variables including wind speed, humidity, and air temperature, and the target variable being rainfall category, consisting of no rain, very light rain, light rain, moderate rain, heavy rain, and very heavy rain. To address class imbalance in the data, this study uses the Random Undersampling (RUS) technique. The GBM model with optimal hyperparameter configuration (n_estimators, learning_rate, max_depth, subsample, min_samples_leaf, max_features) achieved a classification accuracy rate of 98.46%, precision of 93%, recall of 98%, and F1-score of 95% with a training and testing data split of 80:20. The research results show that the GBM method is able to classify rainfall very well and can be used as a tool to assist in disaster mitigation planning and water resource management in Makassar City. 95% pada proporsi data pelatihan dan pengujian 80:20. Hasil penelitian menunjukkan bahwa metode GBM mampu mengklasifikasikan curah hujan dengan sangat baik dan dapat digunakan sebagai alat bantu dalam perencanaan mitigasi bencana serta pengelolaan sumber daya air di Kota Makassar.

PENERAPAN ALGORITMA K-NEAREST NEIGHBOR (K-NN) UNTUK ANALISIS SENTIMEN TERHADAP DATA ULASAN APLIKASI E-COMMERCE LAZADA PADA GOOGLE PLAYSTORE

Zulkifli Rais — 2025-09-30

Classification is the process of grouping objects based on their characteristics. Various classification methods have been employed, ranging from manual grouping to using technology as an aid in the process. One commonly used classification method is the K-Nearest Neighbor (K-NN) algorithm. K-NN predicts the class of data based on the majority class of its nearest neighbors. The novelty of this research lies in using the K-NN method on the case of Lazada application user sentiment on the Google Playstore. In this study, the review classification used is positive and negative labels. Additionally, three accuracy comparisons between training and testing data were used: 80% : 20%, 70% : 30%, and 60% : 40%. Based on the research results from the classification process of Lazada application user reviews on the Google Playstore, an accuracy of 87.00% was obtained for the training and testing data comparison of 80% : 20%.