Climate change is here and is a reality in the world; therefore, studying this phenomenon based on its relationship with meteorological parameters is the first step to making informed decisions. With this in mind, the objective of this work was to conduct a comparative analysis of machine learning techniques used in weather forecasting to evaluate their accuracy in weather forecasting in a localized area, Iquique. The methodology used was exploratory, and the design was experimental based on Knowledge Discovery in Databases (KDD). The Transformer network and Arima in distant horizons gave better performance, indicating that Machine Learning techniques, particularly Deep Learning, can contribute to and complement classic weather forecasting techniques. Understanding the contribution of classic techniques such as Machine Learning in climate forecasting opens a range of possibilities to be further investigated.