This study introduces an innovative approach to analyzing economic phenomena by integrating news and social media data as external sources to forecast the values of commodities such as LBMA GOLD and Brent oil, as well as the USD/COP exchange rate. Over twelve months, data from 166 news sources were collected through RSS and Twitter. Techniques including linear regression and ensemble machine learning, such as XGBoost and Random Forest, were employed to predict daily changes. Furthermore, a multi-agent system inspired by the socio-economic framework was developed, capable of evolving using external information and identifying characteristic patterns of complex systems.