Information Technology Reference
In-Depth Information
Such vectors are fed to our neural network in figure 3 for supervised learning.
Back-propagation algorithm is used to train network; thus Boolean document
vectors are considered training tuples. We suppose that the weights in neural
network are changed as in figure III.3.9 after training process.
5 Discovering User Interests Based on Document Classification
Suppose in some library or website, user
U
do his search for his interesting topics,
documents… There is demand of discovering his interests so that such library or
website can provide adaptive documents to him whenever he visits in the next
time. This is adaptation process in which system tailors documents to each
individual. Given there is a set of key words or terms {
computer, programming
language, algorithm, derivative
} that user
U
often looking for, the his searching
history is showed in following table:
Table 10
User's searching history
Date Keywords (terms) searched
Aug 28 10:20:01 computer, programming language, algorithm, derivative
Aug 28 13:00:00 computer, programming language, derivative, algorithm
Aug 29 8:15:01 computer
Aug 30 8:15:06 computer
This history is considered as training dataset for mining maximum frequent
itemsets. The keywords are now considered items. A itemset is constituted of some
items. The support of itemset
x
is defined as the fraction of total transaction which
containing
x
. Given support threshold
min_sup
, the itemset
x
is called
frequent
itemset
if its support satisfies the support threshold (
≥min_sup
). Moreover x is
maximum
frequent itemset
if
x
is
frequent itemset and all super-itemsets of x are not
frequent. Note that
y
is super-itemset of x if x ⊂ y. The itemset that has
k
items is
called
k-itemset
. Tabel 9 shows the supports of
1-
itemsets.
Table 11
1-itemsets
1-itemset support
computer 4
programming language 2
algorithm 2
derivative 2
Applying algorithm Apriori, it is easy to find maximum frequent itemsets given
min_sup
= 2. The maximum frequent itemset that user searches are showed in
below table:
