Data warehouses integrate data from different source systems to support the decision process of users at different management levels. Data warehouses rely on a multidimensional view of data usually represented as relational tables with structures called star or snowflake schemas. These consist of fact tables, which link to other relations called dimension tables. A fact table represents the focus of analysis (e.g., analysis of sales) and typically includes attributes called measures. Measures are usually numeric values (e.g., quantity) used for performing quantitative evaluation of different aspects in an organization. Measures can be analyzed according to different analysis criteria or dimensions (e.g., store dimension). Dimensions may include hierarchies (e.g., month-year in the time dimension) for analyzing measures at different levels of detail. This analysis can be done using on-line analytical processing (OLAP) systems, which allow dynamic data manipulations and aggregations. For example, the roll-up operation transforms detailed measures into aggregated data (e.g., daily into monthly or yearly sales) while the drill-down operations does the contrary. Multidimensional models include a time dimension indicating the timeframe for measures, e.g., 100 units of a product were sold in March 2007. However, the time dimension cannot be used to keep track of changes in other dimensions, e.g., when a product changes its ingredients. In many cases the changes of dimension data and the time when they have occurred are important for analysis purposes. Kimball and Ross (2002) proposed several implementation solutions for this problem in the context of relational databases, the so-called slowly-changing dimensions. Nevertheless, these solutions are not satisfactory since either they do not preserve the entire history of data or are difficult to implement. Further, they do not consider the research realized in the field of temporal databases. Temporal databases are databases that support some aspects of time (Jensen & Snodgrass, 2000). This support is provided by means of different temporality types1, to which we refer in the next section. However, even though temporal databases allow to represent and to manage time-varying information, they do not provide facilities for supporting decision-making process when aggregations of high volumes of historical data are required. Therefore, a new field called temporal data warehouses joins the research achievements of temporal databases and data warehouses in order to manage time-varying multidimensional data.
Mining association rules is typically decomposed into three sub-problems: 1) prepare the data for analysis, 2) find frequent patterns and 3) generate association rules from the sets representing those frequent patterns (Agrawal et al., 1993).