06.03.2013 Views

Artificial Intelligence and Soft Computing: Behavioral ... - Arteimi.info

Artificial Intelligence and Soft Computing: Behavioral ... - Arteimi.info

Artificial Intelligence and Soft Computing: Behavioral ... - Arteimi.info

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>info</strong>rmation with their normalized values is then recorded in a database in<br />

the form of a data-tree, to be described shortly.<br />

Collected fuzzy membership<br />

value of <strong>info</strong>rmation<br />

1.0<br />

0.6<br />

0.2<br />

AL=poor<br />

0.2<br />

0.6<br />

Normalized belief of <strong>info</strong>rmation<br />

Fig.23.7: Normalization of fuzzy beliefs using membership functions.<br />

As an illustration of the normalization process, it is observed from fig.<br />

23.7 that an <strong>info</strong>rmation with a fuzzy belief of 0.85, collected from a source<br />

with very poor authenticity level (AL), gives rise to a normalized belief of<br />

0.12 (shown by dotted lines).<br />

23.5.2 The Data-tree<br />

AL=very poor<br />

AL=moderate<br />

AL=good<br />

The database in the proposed ES has been organized in the form of a data-tree<br />

(fig. 23.8) having a depth of 3 levels. The root simply holds the starting<br />

pointer, while the second level consists of all the predicates, <strong>and</strong> the third<br />

level contains the relevant facts corresponding to each predicate of the second<br />

level. The normalized beliefs corresponding to each fact are also recorded<br />

along with them at the third level. Such organization of the data-tree helps in<br />

efficient searching in the database. To illustrate the efficiency of searching, let<br />

us consider that there exist P number of distinct predicates <strong>and</strong> at most L<br />

number of facts under one predicate. Then to search a particular clause, say<br />

has-alibi (jadu) in the data-tree, we require P+L number of comparisons in<br />

the worst case, instead of P* L number of comparisons in a linear sequential<br />

search. Now, we present the algorithm for the creation of the data-tree.<br />

1.0

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!