This project is focused on enhancing the accuracy of nanopore sensors when applied to DNA-based reads through the development of neural networks. These sensors generate one-dimensional current traces over time. The task is to classify these time series signals with supervised learning. The project is structured into two distinct components: Firstly, we aim to demonstrate the effectiveness of a novel data augmentation technique in enhancing the performance of state-of-the-art neural networks. Preliminary results obtained using in-house GPUs have already indicated the potential of our method to improve performance. To further bolster the credibility of our findings, we intend to conduct extensive experiments by systematically varying hyperparameters. This empirical analysis will substantiate our claim that performance improvements persist even under the optimal tuning parameter configuration. The second phase of the project involves the application of cutting-edge techniques within the field of computer science to refine the performance of existing neural networks. Specifically, we will leverage and fine-tune residual networks in pursuit of superior results compared to the current state-of-the-art, which predominantly relies on convolutional networks in the domain of nanopore sensor analysis. In undertaking these two parallel endeavors, our primary objective is to advance the performance of neural network methods within the field of nanopore-based sensing. This work stands to be highly influential, given the remarkable growth of nanopore sensing, increased data accessibility, and the early stage of machine learning applications within this domain. The ripple effect of our contributions extends beyond the immediate field of nanopore-based sensing, potentially impacting fields such as genomics, molecular biology, proteomics, and others, by virtue of refining and enhancing nanopore-based data analysis methodologies.