With three centuries of existence, the study of population's behavior implies the manipulation of large amounts of incomplete and imprecise data with high dimensionality. By virtue of its multidisciplinary character, the work in demography involves at least historicists, statisticians and computer scientists/programmers. Moreover, successful demographic analysis requires qualified experts, who have succeeded in analysing data through many views and relate different sources of information, including their personal knowledge of the epoch or regions under study. In this paper, we present an intelligent system to study demographic evolution (ISSDE). This system has a module based on on-line analytical processing (OLAP), which permits conducting multiple analysis, combining many data dimensions. It has a deductive database system, which allows the execution of elaborated queries through the database. It has another module for date treatment (generalization and/or reduction); and, at last, a data mining module to discover nontrivial relations hidden within data. We discover the data treatment procedure with two phases: data generalization and data reduction. In data generalization, utilizing knowledge about concept hierarchies and relevance of data, aggregation of attribute values is performed. In the data reduction phase, rough set theory is applied to compute the minimal attribute set. We highlight the advantages of combining attribute value generalization with rough set theory, to find a subset of attributes that lets the mining process discover more useful patterns, by providing results from the application of the C5.0 algorithm in a demographic relational database.