An Improved Frequent Itemset Generation Algorithm Based On Correspondence

Sundarapandian et al. (Eds): ICAITA, SAI, SEAS, CDKP, CMCA, CS & IT 08,
pp. 253–258, 2012. © CS & IT-CSCP 2012 DOI : 10.5121/csit.2012.2522
An Improved Frequent Itemset Generation
Algorithm Based On Correspondence
Ajay R Y1
, Sharath Kumar A2
, Preetham Kumar3
, Radhika M. Pai4
Department of Information and Communication Technology
Manipal Institute of Technology, Manipal University, Manipal-576104, India
ajaycse3@gmail.com,sharathkumara@yahoo.co.in
Abstract
Association rules play a very vital role in the present day market that especially involves generation of
maximal frequent itemsets in an efficient way. The efficiency of association rule is determined by the
number of database scans required to generate the frequent itemsets. This in turn is proportional to the
time, which will lead to the faster computation of the frequent itemsets. In this paper, a single scan
algorithm which makes use of the mapping of the item numbers and array indexing to achieve the
generation of the frequent item sets dynamically and faster. The proposed algorithm is an incremental
algorithm in that it generates frequent itemsets as and when the data is entered into the database.
Keywords
Maximal Frequent Itemset, Support, Data Mining, Mapping.
1. INTRODUCTION
The frequent itemsets involve the generation of the most frequent itemsets from the given set of
transactions with the given support value [1]. The frequent itemsets are the items which occur
frequently in multiple transactions. The number of frequent itemsets need to be generated varies
depending on the application. The frequent itemsets are used to make decisions regarding the
production of the sets of items that are bought more frequently, by benefitting the end users-
retailers. In some of the business applications, the number of transactions may be large, hence
there should be faster way of computing the frequent itemsets.
2. EXISTING ALGORITHM AND DRAWBACKS
The basic algorithm that is being widely used today in association rule for generating the frequent
itemsets is the Apriori algorithm [1] [4]. Algorithm's basic idea is to identify all the frequent
itemsets which exceeds the predefined threshold support. In other words frequent items generates
strong association rule, which must satisfy minimum support and minimum confidence [2]. Even
this algorithm is simple and clear, it has some limitations. It is costly to handle a huge number of
candidate sets. Initially the frequent one itemsets are generated based on the given candidate one
itemsets in the transaction. The candidate 2-itemsets generation is based on the frequent 1-
itemsets. Now again the generation of the frequent 2-itemsets requires the entire scan of the
transactions to count for the required support in the candidate 2-itemsets [3] [7]. This process
goes on until the required number or all the possible number of frequent itemsets are generated
for the given set of transactions. Hence Apriori algorithm needs ‘n’ number of passes for the
generation of frequent n-itemsets. Hence the time required to obtain the maximal frequent

254 Computer Science & Information Technology (CS & IT)
itemsets would be more. Apart from this, the major drawback is to wait until the last transaction.
Hence there is a need for proposing a new algorithm to generate the frequent itemsets as and
when each transaction is entered into the database and also to reduce the number of passes
possible.
To effectively extract information from a huge amount of data in databases, the knowledge
discovery algorithms must be efficient and scalable in large databases [5]. Most of the algorithms
apply only to static databases. That is, when more transactions are added, the process of
generation of the frequent itemsets must start again from the beginning [6]. In the proposed
algorithm, there is no restriction on the number of transactions.
3. PROPOSED WORK
The proposed algorithm involves the generation of the frequent itemsets as and when the
transaction is entered into the database and also reducing the number of scans. In this work, a
single scan algorithm is proposed to generate the frequent itemsets which makes use of mapping
of the item numbers into the array index during computation. In this approach, since algorithm is
incremental, the frequent itemsets are mapped for the particular index in an array during a single
scan.
3.1 Algorithm
The main idea of this algorithm is to keep track of the frequent itemsets as and when a particular
transaction is entered into the database. When a new transaction is added into the database, the
counters of the corresponding itemsets are incremented. After incrementing the counters of the
itemsets obtained are matched with the support value to obtain the frequent itemset.
Pseudocode for the proposed algorithm in generic.
C is an array which counts the itemset frequency
f-set collects all frequent itemsets
Begin
Initialize the elements of array C to zero
Read each transaction
for each transaction read
begin
Consider all items present in the transaction as a set
Generate all the subsets of the above set
Increment the array C element considering the subset as the index
end
for all element of array C
if element is exceeding support value
store the array index in f-set
End
Pseudocode for the generation of the frequent itemsets for the number of items being less than
100.
// Computation for generation of frequent 1 itemset.
for i =1 to n
if a[i] = 1
c[i]++;
// Computation for 2 itemsets
for i = 1 to n

Computer Science & Information Technology (CS & IT) 255
for j = i+1 to n
if a[i] = 1 and a[j] = 1
c[i*100+j]++;
// Computation for 3 itemsets
for i =1 to n
for j = i+1 to n
for k = j+1 n
if a[i] =1 and a[j ]= 1 and a[k] = 1
c[i*10000+j*100+k]++;
. . . . . . . . . . .
As an example, consider a transaction containing 3 items, 011(the second and the third items are
added in this transaction). The counter c2 and c3 gets incremented. In correspondence with this,
c23 would be incremented as well. Thus after all the transactions are complete, the counters are
checked for the support value to obtain the frequent itemsets. The counters indicate the indexes to
an array corresponding to the item numbers. Hence the mapping can be done easily. The support
of the itemset ‘i’ can be checked by looking at the content of the array at index i, as each time the
item i occurs the content at that index is incremented.
3.2 Demonstration of the Mapping
As an example for demonstration of the pseudocode for mapping items into array index, consider
the set of transactions as shown in Table 1, Based on the algorithm and the pseudocode specified
above, the mapping of the itemsets into array index takes place as follows.
If ‘a’ is an array keeping track of the count of the itemsets and minimum support (threshold
value) is 3, then the transaction proceeds as follows.
Table1. Transaction set.
Item
1
Item
2
Item
3
Item
4
Item
5
Trans1 1 1 1 0 0
Trans2 0 1 1 0 1
Trans3 0 0 0 1 1
Trans4 1 0 1 1 1
Trans5 0 1 1 0 0
In the scan of the first transaction, a[1], a[2], a[3], a[12], a[13], a[23] and a[123] gets
incremented. Now the array “a” is updated as follows: a[1]=1, a[2]=1, a[3]=1, a[12]=1, a[13]=1,
a[23]=1, a[123]=1.
In the scan of the second transaction, a[2], a[3], a[5], a[23], a[25], a[35] and a[235] gets
incremented. Now array “a” would be updated as -> a[1]=1, a[2]=2, a[3]=2, a[5]=1, a[12]=1,
a[13]=1, a[23]=2, a[25]=1, a[35]=1, a[123]=1 and a[235]=1.

In the third transaction, a[4], a[5], a[45] gets incremented. Now array gets updated as follows,
a[1]=1, a[2]=2, a[3]=2, a[4]=1, a[5]=2, a[12]=1, a[13]=1, a[23]=2, a[25]=1, a[35]=1, a[45]=1,
a[123]=1 and a[235]=1.
In the scan of fourth transaction, a[1], a[3], a[4], a[5], a[13]. a[14], a[15], a[34], a[35], a [45],
a[134], a[145], a[345] gets incremented. Now the array gets updated as follows, a[1]=2, a[2]=2,
a[3]=3, a[4]=2, a[5]=3, a[12]=1, a[13]=2, a[14]=1, a[15]=1, a[23]=2, a[25]=1, a[34]=1, a[35]=2,
a[45]=2, a[123]=, 1[134]=1, a[145]=1, a[235]=1 and a[345]=1.
In the scan of fifth transaction, a[2], a[3] and a[23] gets incremented, the array would updated as
follows, a[1]=2, a[2]=3, a[3]=4, a[4]=2, a[5]=3, a[12]=1, a[13]=2, a[14]=1, a[15]=1, a[23]=3,
a[25]=1, a[34]=1, a[35]=2, a[45]=2, a[123]=1, a[134]=1, a[145]=1, a[235]=1 and a[345]=1. So
frequent itemsets are {2}, {3}, {5} and {2, 3}.
4. ADVANTAGES
Since the algorithm is based on the array index mapping, the algorithm is best suitable when used
for the incremental approach, i.e. as and when the data is entered into the database, the value of
the particular array index is incremented corresponding to the items. Hence it is not required to
explicitly generate the frequent itemsets. In this approach, the frequent itemsets are available at
any point of time. Generation of the n-frequent itemsets is independent of the candidate itemsets
and also on (n-1) frequent itemsets. The algorithm is reliable even if there are millions of
transactions.
5. DISADVANTAGES
Memory is not used efficiently as only some of the array indexes are mapped to the items and the
remaining part of the array would not be utilized. There is a limit on the number of items in the
transactions depending on the availability of memory and also the maximal frequent itemsets to
be generated.
6. RESULT ANALYSIS
Table 2 shows the set of transactions t hat are being entered, while Figure 1 shows the snapshot
of the output generated for given input transactions. For the given example, we considered
minimum support as 3. From Figure 1, we can observe that the frequent itemsets are generated
Table 2. Transactions Sets.

Computer Science & Information Technology (CS & IT) 257
implicitly as and when each transaction is updated into the data base. In the time analysis, the
computation of 1-frequent itemset took an average of 40207 ns for a given set of 15 transactions
and 4 items, while apriori took 116343 ns for the same set of transactions and items.
Fig. 1. Correspondence Result.
In the proposed algorithm, the generation of the maximal frequent itemsets is independent of the
generation of previous maximal frequent itemsets. As an example, the generation of maximal
frequent 4-itemsets is independent of the result of frequent 3-itemsets.
7. CONCLUSION AND FUTURE WORK
The proposed algorithm generates the frequent item sets in a single pass in an efficient way which
makes use of the mapping of the item numbers and array indexing. The algorithm also supports
for the incremental approach. Generation of the frequent itemsets are independent of the
candidate itemsets. The scope for the future work includes generation of the large number of
frequent itemsets in an efficient way in a single pass, i.e. making use of the array efficiently,
irrespective of the number of items in the memory.

References
[1] Pujari, A.K. 2001, Data mining Techniques, Universities Press (India) Private Limited, Hyderabad.
[2] Agrawal, R., and Srikant, R. 1994,’Fast Algorithms for Mining Association Rules in Large
Databases’, In Proceedings of the 20th
international conference on Very Large Data Bases, pp 478-
499.
[3] Agrawal, R., Imielinski, T., and Swami, A. 1993,’Mining Association Rules between Sets of Items
in Large Databases’, Proceedings of the 1993 ACM SIGMOD International Conference on
Management of Data, Washington ,DC,pp.207-216.
[4] Wei Yong-qing, Yang Ren-hua, and Liu Pei-yu.2009, "An improved Apriori algorithm for association
rules of mining," IT in Medicine & Education, 2009. ITIME '09. IEEE International Symposium on,
vol.1, pp.942-946.
[5] Chen, M., Han, J., and Yu, P.S., 1996,’Data Mining: An Overview from a Database Perspective’,
IEEE Transactions on Knowledge and Data Engineering, Vol.8, No.6, pp.866-883.
[6] Omiecinski, E., and Savasere, A. 1998,’Efficient Mining of Association Rules in Large Dynamic
Databases’, Proceedings of the 16th British National Conference on Databases: Advances in
Databases, pp.49-63.
[7] Rao, S., and Guptha, P., 2012,’ Implementing Improved Algorithm Over APRIORI Data Mining
Association Rule Algorithm’, IJCST, Vol. 3, Issue 1 Jan. – March.

An Improved Frequent Itemset Generation Algorithm Based On Correspondence

Recommended

Recommended

More Related Content

What's hot

What's hot (19)

Similar to An Improved Frequent Itemset Generation Algorithm Based On Correspondence

Similar to An Improved Frequent Itemset Generation Algorithm Based On Correspondence (20)

More from cscpconf

More from cscpconf (20)

Recently uploaded

Recently uploaded (20)

An Improved Frequent Itemset Generation Algorithm Based On Correspondence