This paper introduces probabilistic relational models (PRMs), which extend traditional statistical learning methods to handle relational data. Unlike flat data representations, PRMs allow the properties of an object to depend probabilistically on both its own properties and those of related objects. This makes them more expressive than Bayesian networks, yet they can be learned using similar techniques. The paper describes how to learn PRMs from relational databases, focusing on parameter estimation and structure learning. It shows how standard database retrieval techniques can be used to efficiently learn from large datasets. The approach is validated on both real and synthetic relational databases, demonstrating its effectiveness in capturing complex dependencies. The paper also discusses challenges in learning PRMs, including the need for acyclic dependency structures and the use of Bayesian methods for parameter estimation. Finally, it outlines future directions, including handling missing data and discovering hidden variables. The results show that PRMs can effectively model complex relational data and provide accurate predictions.This paper introduces probabilistic relational models (PRMs), which extend traditional statistical learning methods to handle relational data. Unlike flat data representations, PRMs allow the properties of an object to depend probabilistically on both its own properties and those of related objects. This makes them more expressive than Bayesian networks, yet they can be learned using similar techniques. The paper describes how to learn PRMs from relational databases, focusing on parameter estimation and structure learning. It shows how standard database retrieval techniques can be used to efficiently learn from large datasets. The approach is validated on both real and synthetic relational databases, demonstrating its effectiveness in capturing complex dependencies. The paper also discusses challenges in learning PRMs, including the need for acyclic dependency structures and the use of Bayesian methods for parameter estimation. Finally, it outlines future directions, including handling missing data and discovering hidden variables. The results show that PRMs can effectively model complex relational data and provide accurate predictions.