Abstract This paper provides a detailed comparison of Support Vector Machines (SVM) and Neural Networks (NN) in the context of indoor visible light positioning systems, emphasizing positioning accuracy and error correction. The findings reveal that the SVM model substantially outperforms the NN model across several key metrics. SVM achieves a prediction accuracy of 99.87%, notably higher than NN’s 96.94%, and a mean positioning error of just 0.0025, compared to the NN’s 0.0634. The SVM model also demonstrates a superior alignment between predicted and actual positions, with minimal discrepancies, whereas the NN model shows significant discrepancies. Furthermore, the scatter plot observations confirm the SVM’s minimal error distribution, reinforcing its reliability. These results indicate that the SVM model is not only more accurate but also more robust, making it a superior choice for implementing high-precision indoor visible light positioning systems. The SVM’s outstanding performance metrics, including an R² score of 1, highlight its potential as a leading solution in this domain.
Moharam et al. (Tue,) studied this question.