Predicting Non-Subsidized Diesel Fuel Sales Volume for Ships at PT. Pertamina Patra Niaga East Java Region: A Comparative Analysis of Linear Regression, Random Forest Regression, and XGBoost Regression.

Azzahra, Feryani (2025) Predicting Non-Subsidized Diesel Fuel Sales Volume for Ships at PT. Pertamina Patra Niaga East Java Region: A Comparative Analysis of Linear Regression, Random Forest Regression, and XGBoost Regression. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5010211046-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Penelitian ini mengevaluasi tiga model machine learning yaitu Linear Regression, Random Forest Regression, dan XGBoost Regression untuk memprediksi volume penjualan
diesel non-subsidi untuk kapal di PT Pertamina Patra Niaga Regional Jawa Timur. Tujuan utama dari penelitian ini adalah untuk meminimalkan kesalahan prediksi dan meningkatkan
akurasi guna mendukung pengambilan keputusan strategi pemasaran berbasis data. Evaluasi model dilakukan menggunakan tiga metrik performa, yaitu Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), dan koefisien determinasi (R²), melalui skema 5-fold crossvalidation pada data pelatihan, serta evaluasi akhir pada data pengujian. Hasil penelitian menunjukkan bahwa model Random Forest Regression memberikan kinerja paling optimal dibandingkan dua model lainnya, dengan tingkat kesalahan prediksi yang paling rendah dan akurasi yang paling tinggi. Pada data pengujian, Random Forest menghasilkan nilai MAE sebesar 56,41 juta, RMSE sebesar 72,92 juta, dan R² sebesar 0,9906, yang mencerminkan performa prediksi yang sangat baik dan deviasi yang minimal terhadap nilai aktual. Dengan
demikian, model ini menjadi pilihan terbaik untuk memprediksi volume penjualan diesel nonsubsidi untuk kapal. Selain itu, hasil analisis feature importance menunjukkan bahwa variabel promosi merupakan faktor paling dominan yang memengaruhi volume penjualan, jauh melebihi variabel lain seperti PDB, inflasi, maupun harga minyak dunia. Temuan ini menegaskan pentingnya kegiatan promosi dalam mendorong minat beli konsumen. Oleh karena itu, optimalisasi strategi promosi dapat menjadi pendekatan utama dalam meningkatkan volume penjualan bahan bakar diesel non subsidi. Pendekatan berbasis data seperti ini diharapkan mampu memperkuat daya saing perusahaan dan meningkatkan responsivitas terhadap dinamika pasar.

========================================================================================================================================

This study evaluates three machine learning models there are Linear Regression, Random Forest Regression, and XGBoost Regression for predicting the sales volume of non-subsidized
marine diesel fuel at PT Pertamina Patra Niaga, East Java Regional. The main objective is to minimize prediction errors and enhance prediction accuracy to support data-driven marketing strategies. The models were evaluated using three performance metrics: Mean Absolute Error
(MAE), Root Mean Squared Error (RMSE), and the coefficient of determination (R²), through 5-fold cross-validation on the training dataset, followed by final evaluation on the testing dataset. The results show that the Random Forest Regression model consistently outperformed the other models, achieving the lowest prediction errors and highest accuracy. On the test set, Random Forest produced a MAE of 56.41 million, an RMSE of 72.92 million, and an R² of
0.9906, indicating excellent predictive performance and minimal deviation from actual values. These metrics confirm Random Forest as the most optimal model for predicting the sales volume of non-subsidized diesel fuel for ships. Furthermore, feature importance analysis revealed that promotion is the most dominant factor influencing sales volume, far exceeding other variables such as GDP, inflation, or world oil prices. This highlights the crucial role of marketing initiatives in driving consumer purchasing behavior. Therefore, prioritizing promotional strategies could be a key approach to increasing fuel sales. Overall, this data-driven model offers valuable insights that can strengthen the company's responsiveness to market changes and support more effective decision-making.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Volume Penjualan, Machine Learning, Regresi Linier, Random Forest Regression, XGBoost Regression, Bahan Bakar Diesel Non Subsidi, Prediksi, Sales Volume, Machine Learning, Linear Regression, Random Forest Regression, XGBoost Regression, Non-Subsidized Diesel Fuel, Prediction
Subjects:	H Social Sciences > HA Statistics > HA31.3 Regression. Correlation. Logistic regression analysis.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Industrial Engineering > 26201-(S1) Undergraduate Thesis
Depositing User:	Feryani Azzahra
Date Deposited:	29 Jul 2025 07:40
Last Modified:	29 Jul 2025 07:40
URI:	http://repository.its.ac.id/id/eprint/123006

Actions (login required)

View Item