Aduwati Sali

Abstract Terahertz (THz) metasurface absorbers have emerged as promising platforms for biomedical sensing due to their high-quality (Q) resonances and strong sensitivity to dielectric variations in biological tissues. Conventional single-band designs are effective only over a narrow frequency range, reducing detection reliability and increasing false positives.

In this sub-section, we present the overall performance comparison of our DFA-GPE approach compared to other methods. We begin with the Hikari dataset, as illustrated in Table 7. Our approach achieves the highest accuracy with a mean value of 99.09%, followed by NSGA-II with a mean accuracy of 98.79%. The third RF with a mean accuracy of 98.73%. Followed by LR with a mean accuracy of 98.62%, and the last RFE with a mean accuracy of 98.39%.

3. Results and Discussion In this section, the FWI published by the Malaysian meteorological department and 80% of the data measured by the IoT system in Section 2.1 are imported into the two neural network structures proposed in Section 2.2 for training (i.e., 37,844 samples). Then, the remaining 20% (i.e., 9462 samples) dataset is used to test the performance of the two structural training models.