A sales table contains point-of-sale data for a group of products that are sold in several stores to different customers over a period of time. A particular customer buys only a few of the products. The products that the customer does not buy do not appear as rows in the sales table.
If you were to figure out the amount of money a customer has spent for each product, the unpurchased products would have an inferred amount of zero. The value is not random or unknown; it is zero, even though no row appears in the table.
Note that the sales data is dimensioned (by product, stores, customers, and time) and would often be represented as nested data for mining.
Since missing values in a nested column always indicate sparsity, you should ensure that this interpretation is appropriate for the data that you want to mine. For example, when trying to mine a multi-record case data set containing users' movie ratings of a large movie database, the missing ratings would be unknown (missing at random), but Oracle Data Mining would treat the data as sparse and infer a rating of zero for the missing value.