A rule-based chase algorithm (called Chase ₂), presented in this paper, provides a strategy for predicting what values should replace the null values in a relational database. When information about an object is partially incomplete (a set of weighted values of the same attribute can be treated as an allowed attribute value), Chase ₂ is decreasing that incompleteness. In other words, when several weighted values of the same attribute are assigned to an object, Chase ₂ will increase their standard deviation. To make the presentation clear and simple, we take an incomplete information system S of type λ as the model of data. To begin Chase ₂ process, each attribute in S that has either unknown or incomplete values for some objects in S is set, one by one, as a decision attribute and all other attributes in S are treated as condition attributes. Assuming that d is the decision attribute, we take a subsystem S ₁ of S by selecting from S any object x such that d(x) ≠ NULL. Now, the subsystem S ₁ is used for extracting rules describing values of attribute d. In the next step, each incomplete slot in S which is in the column corresponding to attribute d is chased by previously extracted rules from S ₁, describing d. All other incomplete attributes in a database are processed the same way. This concludes the first loop of Chase ₂. The whole process is recursively repeated till no more new values can be predicted by Chase ₂. In this case, we say that a fixed point in values prediction was reached.