The effectiveness of academic marketing strategies in higher education institutions is often compromised by cognitive biases, particularly confirmation bias, which affects the objectivity of decision-making processes. This study proposes the design of a predictive recommendation system based on machine learning (ML) workflows aimed at supporting strategic planning for academic program promotion within a higher education institution in Bogotá, Colombia. The system leverages historical enrollment data, SNIES datasets, and institutional marketing records to evaluate past campaign effectiveness and forecast program performance. Using the CRISP-DM methodology, the project integrates supervised learning processes that employ statistical regression models (linear, ridge, and lasso) and exploratory dimensionality reduction techniques such as Principal Component Analysis (PCA). Ridge regression without dimensionality reduction demonstrated the best balance between predictive accuracy and model stability. While PCA did not improve predictive performance, it offered exploratory value for visualizing latent patterns, despite limitations in component interpretability in this phase. The proposed system aims to reduce human bias, optimize resource allocation, and align marketing efforts with the actual behavior of prospective students. This research contributes to the technological modernization of higher education institutions and promotes equity in student recruitment through evidence-based, data-driven decision-making.