Virtual Booth

Street lighting system plays an important role in ensuring road safety, urban security 
and public safety. However, current streetlight maintenance system are still relies on 
traditional approaches where maintenance team wait for public report or problem are 
identified during scheduled inspections. This approach is inefficient and ineffective as 
it may result in delayed maintenance action and prolonged streetlight downtime, 
especially when faults are not reported immediately. Therefore, this project proposed 
a system that are no longer relies solely on manual reporting, but is capable of 
predicting and identifying streetlight faults automatically based on sensor readings. 
This project utilizes machine learning models such as Decision Tree (DT), Random 
Forest (RF) and K-Nearest Neighbors (KNN) to classify streetlight fault type. The 
dataset used is obtained from Kaggle and undergoes preprocessing before forwarding 
to the machine learning training module. Hyperparameter optimization is performed 
using Grid Search and 5-Fold Cross Validation to improve model performance and 
determine the optimal model configuration. After the models have classify the 
streetlight fault type, the prediction results are mapped into health status categories, 
namely Healthy, Warning, and Faulty, to support maintenance prioritization and 
decision-making. The final results showed that the optimized Decision Tree with 5
Fold Cross Validation achieved the best overall performance, with an accuracy of 
84.74%, precision of 96.43%, recall of 61.48%, and F1-score of 68.99%. Finally, the 
prediction results, health status assessments, maintenance recommendations, and alert 
notifications are displayed through a dashboard developed using MATLAB App 
Designer.Finally, all outputs, including fault predictions, health status assessments, 
maintenance recommendations, and alert notifications, are displayed through a 
dashboard developed using MATLAB App Designer.