English
كوردى
عربي
KNN-Based Fuel-Aware Range Prediction and Gas-Station Recommendation in an Android App
(1) Farid Baskoro (State University of Surabaya, Indonesia)
(2) Widi Aribowo (State University of Surabaya, Indonesia)
(3) Hewa Zangana (Duhok Polytechnic University, Iraq)
(4) Wahyu Sasongko Putro (Nanyang Technological University, Singapore)
(5) Rifqi Firmansyah (State University of Surabaya, Indonesia)
(6) Ali Nur Fathoni (State University of Surabaya, Indonesia)
(7) Aristyawan Putra Nurdiansyah (State University of Surabaya, Indonesia)
This study aims to develop and validate a fuel-aware prediction model to estimate the reachable travel distance from remaining fuel and to support feasibility-based gas-station candidate filtering. The proposed approach consists of two stages: (1) a K-Nearest Neighbors (KNN) regressor predicts reachable distance based on remaining fuel and vehicle type, and (2) gas-station candidates are filtered within the predicted range using Haversine-based geodesic distance. The novelty of this study lies in coupling non-parametric KNN model to predict reachable distance, combined with a feasibility filter that retains only stations whose Haversine distance does not exceed the predicted reachable range. Experiments were conducted on three motorcycle categories (110cc, 125cc, and 150cc) using a dataset of 90 samples (30 samples per category) and evaluated through 5-fold cross-validation, with MAE and R² as the primary performance metrics. The performance of KNN (k = 5) was compared against all baseline models, including mean baseline, linear regression, and second-order polynomial regression. The results show that KNN consistently achieved the best performance across all categories, with MAE values of 0.210 km (110cc), 0.189 km (125cc), and 0.159 km (150cc), and average R² values of 0.85 ± 0.02 for 110cc and 125cc, and 0.80 ± 0.03 for 150cc. These findings indicate that the proposed KNN model provides stable reachable-distance estimates suitable for feasibility-based gas-station filtering under fuel constraints.