What is a Deductive Classifier?
Deductive classifiers are an adaptive form of inference engine used to improve the accuracy of a classification label.
Compared to rule-based inference engines, which can only apply triggers like OFF or IF-THEN when a condition is not met, these classifiers seek to mimic human deductive logic.
How do Deductive Classifiers Work?
The first step in these algorithms is to test the unknown data within the class hierarchy by applying the inputs to each class, then sub-class, and following through with any properties or restrictions on allowable values. If the classifier deduces that these declarations are logically consistent throughout the process, then classification can proceed with high confidence. If any errors are found, then the specific inconsistencies can be resolved by another classification technique or by editing existing class information or creating new classes.