CONCLUSION

Incomplete information is endemic to real-world data collections. Often, there are values that are unknown or imprecise, but other kinds of incompleteness, such as probabilistic data, could also be present. Multidimensional databases are not currently engineered to manage incomplete information in base data, derived data, and dimensions. Different techniques are needed to correctly aggregate incomplete information during multidimensional operations such as drill-down and roll-up.

In this chapter we presented several strategies for managing incomplete information in a multidimensional database. The strategy to use depends upon the kind of incomplete information and also on where it occurs in the multidimensional database. A common technique is to replace incomplete information with complete information. The advantage of this technique is that all multidimensional databases can manage complete information. The trick is in replacing the incompleteness with an appropriate, complete value. Other strategies for managing incomplete information require more substantial changes to a multidimensional database. One strategy is to include the incompleteness in a computed aggregate, but this is possible only if the multidimensional database can have incomplete values in the hierarchy. Another technique is to measure the amount of incompleteness in an aggregated value by tallying how much uncertain information was used in its production. The final general strategy is to gerrymander the hierarchy to accommodate incomplete information.