Federated learning is a machine learning approach that eliminates the need to send data to a central location by training the model across a number of dispersed devices, or clients. Customers benefit from increased privacy and security with this strategy as their data is saved on their devices instead of a third party or central server. On the other hand, centralized learning requires users to provide their data to a central server, which raises concerns about privacy and data security. The present work used simulated data to assess and contrast the efficacy of the federated and centralized learning paradigms in the setting of a straightforward regression problem. The results showed that federated learning may achieve accuracy levels comparable to centralized learning while maintaining user privacy protection. Our research also showed that popular machine learning frameworks such as TensorFlow Federated may be successfully used to create federated learning.