Intervention of Machine Learning and Explainable Artificial Intelligence in Fiber-Optic Sensor Device Data for Systematic and Comprehensive Performance Optimization
Jatin Rana;Anuj K. Sharma;Yogendra Kumar Prajapati
{"title":"Intervention of Machine Learning and Explainable Artificial Intelligence in Fiber-Optic Sensor Device Data for Systematic and Comprehensive Performance Optimization","authors":"Jatin Rana;Anuj K. Sharma;Yogendra Kumar Prajapati","doi":"10.1109/LSENS.2024.3445324","DOIUrl":null,"url":null,"abstract":"This letter illustrates the successful application of machine learning (ML) models with explainable artificial intelligence (XAI) to enhance the efficacy of a surface plasmon resonance (SPR)-based fiber-optic sensor device (FOSD). The investigation also examines the correlation between the sensor's figure of merit (FoM) and the following variables: light wavelength (λ), sensing region length, metal layer thickness, and refractive index (RI) of surrounding (i.e., sensing or analyte) medium. The study established that the FoM datasets were consistent with various boosting algorithms, such as XGBoost, CatBoost, etc. Incorporating these algorithms into datasets with a λ-resolution of 1 nm led to enhanced FoM magnitudes. The dataset comprises 32 768 data points, each of which falls within one of 15 distinct thickness values and 25 distinct sensing length values. The selected CatBoost ML model exhibits a high level of consistency with the data in terms of trend matching, with all other evaluation parameters lying within acceptable ranges. Furthermore, we have implemented XAI to gain a more comprehensive understanding of the model's internal mechanism in relation to FoM prediction. The results from the shapley additive explanations (SHAP) method indicate that analyte RI and λ play significantly bigger role in dictating the FoM of the SPR-based FOSD. This study emphasizes that the efficient finalization of sensor design and improved sensing performance can be achieved by selecting an appropriate ML model along with XAI and implementing it on a variety of FOSD datasets.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":null,"pages":null},"PeriodicalIF":2.2000,"publicationDate":"2024-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Sensors Letters","FirstCategoryId":"1085","ListUrlMain":"https://ieeexplore.ieee.org/document/10643268/","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
引用次数: 0
Abstract
This letter illustrates the successful application of machine learning (ML) models with explainable artificial intelligence (XAI) to enhance the efficacy of a surface plasmon resonance (SPR)-based fiber-optic sensor device (FOSD). The investigation also examines the correlation between the sensor's figure of merit (FoM) and the following variables: light wavelength (λ), sensing region length, metal layer thickness, and refractive index (RI) of surrounding (i.e., sensing or analyte) medium. The study established that the FoM datasets were consistent with various boosting algorithms, such as XGBoost, CatBoost, etc. Incorporating these algorithms into datasets with a λ-resolution of 1 nm led to enhanced FoM magnitudes. The dataset comprises 32 768 data points, each of which falls within one of 15 distinct thickness values and 25 distinct sensing length values. The selected CatBoost ML model exhibits a high level of consistency with the data in terms of trend matching, with all other evaluation parameters lying within acceptable ranges. Furthermore, we have implemented XAI to gain a more comprehensive understanding of the model's internal mechanism in relation to FoM prediction. The results from the shapley additive explanations (SHAP) method indicate that analyte RI and λ play significantly bigger role in dictating the FoM of the SPR-based FOSD. This study emphasizes that the efficient finalization of sensor design and improved sensing performance can be achieved by selecting an appropriate ML model along with XAI and implementing it on a variety of FOSD datasets.