Prediksi Jumlah Titik Panas (Hotspot) Menggunakan Random Forest dan Extreme Gradient Boosting (XGBoost) di Provinsi Nusa Tenggara Timur

Sari, Nadifa Permata (2026) Prediksi Jumlah Titik Panas (Hotspot) Menggunakan Random Forest dan Extreme Gradient Boosting (XGBoost) di Provinsi Nusa Tenggara Timur. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2043221108-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Kebakaran hutan dan lahan (karhutla) merupakan permasalahan lingkungan yang kerap terjadi di Indonesia. Hingga tahun 2024, peristiwa karhutla masih sering ditemukan, terutama disebabkan oleh praktik pembukaan lahan dengan cara pembakaran serta kondisi lahan gambut yang mengering pada musim kemarau. Provinsi Nusa Tenggara Timur (NTT) tercatat sebagai wilayah dengan luas kebakaran terbesar dalam enam tahun terakhir, sehingga menjadikannya sebagai daerah prioritas dalam upaya mitigasi kebakaran hutan di tingkat nasional. Sehubungan dengan hal tersebut, diperlukan suatu pendekatan prediktif yang mampu mengestimasi potensi karhutla secara lebih akurat dengan memanfaatkan pola dan hubungan kompleks antarvariabel. Penelitian ini bertujuan untuk memprediksi jumlah titik panas (hotspot) menggunakan algoritma Random Forest dan Extreme Gradient Boosting (XGBoost), serta memberikan interpretasi terhadap hasil model melalui metode SHAP (Shapley Additive Explanations). Variabel yang digunakan dalam analisis ini mencakup jumlah titik panas, curah hujan, temperatur, kecepatan angin, dan kelembapan udara. Seluruh proses analisis dilakukan secara eksploratif dan prediktif menggunakan pendekatan machine learning. Hasil analisis menunjukkan bahwa model Random Forest memberikan kinerja prediksi terbaik di NTT dengan nilai RMSE yang paling rendah yaitu sebesar 12,125 dan stabilitas generalisasi yang lebih unggul dibandingkan XGBoost. Melalui analisis SHAP, ditemukan bahwa kelembapan rata-rata menjadi faktor paling dominan dalam memengaruhi jumlah titik panas, di mana penurunan kelembapan berkorelasi kuat dengan peningkatan frekuensi kejadian hotspot dengan ditujukan nilai mean absolute sebesar 4,572. Sementara itu, variabel temperatur rata-rata, curah hujan, dan kecepatan angin turut berkontribusi membentuk pola sebaran titik panas, namun dengan pengaruh yang relatif lebih kecil dibandingkan kelembapan.
=====================================================================================================================================
Forest and land fires are environmental problems that often occur in Indonesia. Until 2024, forest and land fires will still be found, mainly caused by the practice of land clearing by burning and the condition of peatlands that dry up in the dry season. East Nusa Tenggara Province (NTT) was recorded as the region with the largest fire area in the last six years, making it a priority area in forest fire mitigation efforts at the national level. In this regard, a predictive approach is needed that is able to estimate the potential for forest and land fires more accurately by utilizing complex patterns and relationships between variables. This study aims to predict the number of hotspots using the Random Forest and Extreme Gradient Boosting (XGBoost) algorithms, as well as provide interpretation of the model results through the SHAP (Shapley Additive Explanations) method. The variables used in this analysis include the number of hot spots, precipitation, temperature, wind speed, and air humidity. The entire analysis process is carried out exploratively and predictively using a machine learning approach. The results of the analysis showed that the Random Forest model provided the best prediction performance in NTT with the lowest RMSE value of 12,125 and superior generalization stability compared to XGBoost. Through SHAP analysis, it was found that average humidity was the most dominant factor in influencing the number of hot spots, where the decrease in humidity was strongly correlated with an increase in the frequency of hotspot events with an absolute mean value of 4,572. Meanwhile, the variables of average temperature, rainfall, and wind speed also contribute to forming the distribution pattern of hot spots, but with a relatively smaller influence than humidity.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kebakaran Hutan, Nusa Tenggara Timur, Random Forest, Titik Panas, XGBoost, East Nusa Tenggara, Forest Fires, Hotspot, Random Forest, XGBoost
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HA Statistics > HA31.3 Regression. Correlation. Logistic regression analysis. Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Vocational > 49501-Business Statistics
Depositing User:	Nadifa Permata Sari
Date Deposited:	20 Jul 2026 03:24
Last Modified:	20 Jul 2026 03:24
URI:	http://repository.its.ac.id/id/eprint/135475

Actions (login required)

View Item