The Poisson equation is a fundamental partial differential equation that appears in numerous physical and engineering problems, including heat conduction, electrostatics, and fluid flow. This study presents the implementation of a Physics-Informed Neural Network (PINN) model to approximate the one-dimensional Poisson equation. Classical Neural Networks (NN) rely only on data to train a model. However, they often require a large amount of data to achieve an accurate approximation. PINNs incorporate the governing physical laws directly into the learning process, which makes them useful when there is scarce data. This work compares the approximation of the problem using three approaches: data-only, physics-only, and a hybrid approach combining both. The proposed methodology involves formulating the Poisson problem, defining the loss function according to the studied approach, and conducting a statistical analysis of experiments with different combinations of hyperparameters. Results indicate that the hybrid approach (data and physics), requires less computational work to achieve a more accurate approximation. Furthermore, it shows that the physics-based models outperform the data-based models in accuracy. The findings highlight the effectiveness of PINNs in solving differential equations, especially when data availability is limited. This work contributes to understanding the trade-offs between accuracy and computational costs, and opens the possibility for extending of their applicability to more complex real-world problems. By enabling accurate modeling with limited data, PINNs reduce the need for extensive data collection and resource-intensive simulations which can support sustainable engineering practices.