Smartphone Sensor Insights for Pothole Detection and Road Quality Evaluation using Machine Learning
Main Article Content
Abstract
This research addresses the critical challenge of road infrastructure degradation and public safety by developing a robust machine learning framework for automated road condition classification. Utilizing data consolidated from multiple labeled sources, the research classifies road quality into three distinct categories: Potholes, Good Roads, and Bad Roads. To ensure data integrity, the dataset underwent rigorous preprocessing, including the handling of missing values and categorical label encoding. A significant contribution of this work is the application of the Synthetic Minority Over-sampling Technique (SMOTE) to mitigate class imbalance, ensuring the models were trained on a representative distribution of all road conditions. Exploratory Data Analysis (EDA) validated the structural consistency of the data postbalancing. The processed dataset was partitioned using a 70:30 training-to-testing ratio to evaluate the performance of the Extra Trees Classifier (ETC) and the KNeighbors Classifier (KNN). Empirical results demonstrate exceptional predictive power; the Extra Trees Classifier achieved an accuracy of 99%, while the KNN Classifier reached an optimal accuracy of 100%. These results were further validated through confusion matrices and classification reports, confirming near-perfect precision, recall, and F1-scores. This research provides a highly reliable, scalable pipeline for real-time infrastructure monitoring, offering municipal authorities a data-driven tool to prioritize road maintenance and enhance commuter safety.