A Laser-Induced Breakdown Spectroscopic Steel Classification Method Using Mixed Feature Selection and LIME

   Laser-induced breakdown spectroscopy (LIBS) technology faces the challenge of redundant or irrelevant features when dealing with high-dimensional data of steel. To enhance the accuracy and interpretability of multivariate classification, this study introduces an innovative hybrid feature selection (FS) method that skillfully combines the filtering characteristics of the select percentile (SP) algorithm with the embedded advantages of the elastic net (EN) algorithm. Under this framework, the support vector machine (SVM) algorithm was applied for classification, demonstrating outstanding performance with an accuracy, precision, and F1 score of 0.9888, 0.9895, and 0.9889 on the test set, respectively. To address the ‘black box’ nature of the SVM algorithm, this paper further introduces the local interpretable model-agnostic explanations (LIME) method. LIME allows for the visualization of the importance of each variable, thereby enhancing the interpretability and credibility of the model. Overall, the model and methods proposed in this study show significant effectiveness in eliminating redundant or irrelevant features and in precise classification, effectively solving most of the challenges faced by LIBS in steel classification issues.

1. K. H. Shah, J. Iqbal, P. Ahmad, et al., Rad. Phys. Chem., 170, 0969–806X (2019).

2. Y. Zhang, T. Zhang, H. Li, Spectrochim. Acta B: At. Spectrosc., 181, 106218 (2021).

3. W. Naixiao, W. Xilin, C. Ping, et al., Sensors, 18, 2623 (2018).

4. T. Feng, X. Zhang, M. Li, et al., Anal. Methods, 13, 3424–3432 (2021).

5. O. Gazeli, D. Stefas, S. Couris, Materials, 14, 541 (2021).

6. L. Brunnbauer, Z. Gajarska, H. Lohninger, et al., TrAC Trends Anal. Chem., 159, 116859 (2022).

7. T. Boucher, C. Carey, M. D. Dyar, S. Mahadevan, S. Clegg, R. Wiens, J. Chemom., 29, 484–491 (2015).

8. T. F. Boucher, M. V. Ozanne, M. L. Carmosino, M. D. Dyar, S. Mahadevan, E. A. Breves, K. H. Lepore, S. M. Clegg, Spectrochim. Acta B: At. Spectrosc., 107, 1–10 (2015).

9. H. Sun, C. Yang, Y. Chen, et al., Appl. Opt., 61, 1559 (2022).

10. M. Yao, G. Fu, T. Chen, et al., J. Anal. At. Spectrom., 36, 361–367 (2020).

11. M. Yuan, Q. Zeng, J. Wang, et al., Opt. Eng., 60, 3286 (2021).

12. S. Kumar, K. Chakravarty, Energy Rep., 6, 343–349 (2020).

13. H. Njah, S. Jamoussi, W. Mahdi, Intelligence, 89, 1012–2443 (2021).

14. I. Guyon, A. Elisseeff, J. Machine Learn. Res., 3, 1157–1182 (2003).

15. Y. Saeys, I. Inza, P. Larrañaga, Bioinformatics, 23, 2507–2517 (2007).

16. T. Mehmood, K. H. Liland, L. Snipen, S. Sæbø, Chemom. Intell. Lab. Syst., 118, 62–69 (2012).

17. A. Larsson, H. Andersson, L. Landstroem, J. Anal. At. Spectrom., 30, 1117–1127 (2015).

18. C. Huffman, H. Sobral, et al., Spectrochim. Acta Part B: At. Spectrosc., 162, 105721 (2019).

19. A. Kumar Myakalwar, N. Spegazzini, C. Zhang, et al., Sci. Rep., 5, 13169 (2015).

20. F. Ruan, L. Hou, T. Zhang, et al., Analyst, 146, 1023–1031 (2020).

21. S. Lu, S. Shen, J. Huang, et al., Spectrochim. Acta B: At. Spectrosc., 150, 49–58 (2018).

22. C. Retracted, J. Electrical and Comp. Eng., 1 (2022).

23. M. T. Ribeiro, S. Singh, C. Guestrin, Proc. 22^nd ACM SIGKDD Int. Conf. Knowledge Discovery and Data Mining, San Francisco, CA, USA, 13–17, 1135–1144 (2016).

24. S. C. Hung, H. C. Wu, M. H. Tseng, Remote Sensing Scene Classification and Explanation Using RSSCNet and LIME, 10, 6151 (2020).

25. Vitaly I. Korepanov, J. Raman Spectrosc., 51, 0377–0486 (2020).

26. H. Bai, P. Liu, X. Fu, et al., Spectrochim. Acta B: At. Spectrosc., 199, 106587 (2023).