In the last few decades, sensing mechanisms by employing the fiber optics has achieved huge attention owing to their unique characteristics. The machine learning (ML) approach has brought a thoroughgoing rehabilitation in the field of fiber optics-based sensing mechanisms due to its capabilities of extracting a huge chunk of information from a huge datasets that enhance the degree of performance. Various sensing structures including fiber Bragg grating (FBG), multi-single-multi mode (MSM), single-multi-single (SMS) mode have proved their efficacy in these aspects. The main bottleneck of measuring two parameters simultaneously is the crosstalk issue when utilizing the popular transfer matrix method. ML, which is trained with a large dataset collected from the experiment, is a viable solution to overcome this traditional issue. ML has demonstrated its effectiveness by mitigating the crosstalk issue to a higher degree and thereby enhancing the sensing performance. This unique technology has affirmed its potential in several domains to provide an enhanced, reliable sensing performance.