This document analyzes and compares several algorithms for frequent itemset mining on different datasets, including Apriori, FP-Growth, H-mine, and P-Hmine. It describes applying these algorithms to the mushroom dataset, which contains descriptions of mushroom species, and the chess dataset. It finds that FP-Growth performs best on dense datasets due to its compact FP-tree structure, while H-mine is better for sparse datasets. P-Hmine and H-mine outperform Apriori by requiring fewer scans of the database.

1. To: Sir Altaf Hussain
Topic
Analysis of Frequent Item set
Mining on Variant Datasets
Summery By:
ISHTIAQ HUSSAIN BANGASH(15-S-06)
And
FARHAN AKRAM(15-S-27)
Class: BSIT-VI

2. Contents
• Introduction
• Association rule mining
• Frequent itemset mining and Algorithms for data model
• Algorithms:
• Apriori
• FP-Growth
• H-mine
• P-Hmine
• Conclusion

3. Introduction
• In this paper a complete description of the dataset mushroom is
described on hypothetical samples corresponding to different
species of mushrooms.
• The dataset consists of 8124 instances of 119 attributes which are
derived from 24 species.
• So this is checked by different algorithms which discussed the
datasets of mushroom.

4. Association rule mining
• Process of discovering
relationship among the data
items in large data base.
• It is one of the most important
problem in the data mining.
• Finding frequent itemset is one
of the most computationally
expensive tasks in association
rule mining.

5. Frequent itemset mining representations
Follows are the methods of
representation of databases:
1. Horizontal representation
2. Vertical representation
3. Bit-vector representation

6. Algorithms:
• Apriori
• FP-Growth
• H-mine
• P-Hmine

7. Apriori

8. Apiori
• In preprocessing of apriori algorithm the scane of database is
performed to find out support count of each item then all these
whose minimum is less are removed from the database.
• Aprori follows two step method to find out frequent itemset that
is :
• Join step
• Prune step

9. FP-Growth

10. FP-Growth
• FP-Growth is known as one of the fastest algorithm of frequent set
mining.
• it uses a compact Data Structure called a FP-tree.
• FP-growth approach first represent the frequent itemset in the
form of frequent pattern tree fp-tree which is compressed
structure

11. H-mine

12. H-mine
• H-mine is another pattern growth method for frequent pattern
mining in Sparse data H-mine is better than it FP-growth.
• H-mine uses divide and conquer strategy to mine all the frequent
pattern

13. P-Hmine
• The general idea of P-Hmine is that is a represent the database in
the form of a new structure called P-Hstruct. which is similar to
H-struct.
• In P-Hmine struct we represent the database as a set of queues.
Experimental Analysis and Result
• We analyze the running time of algorithm running on both
synthetic and actual data, synthetic data sets generator is taken
from IDM Almanden website.

14. Datasets
• The data set mushroom is a description of hypothetical sample
was corresponding to different species of Mushrooms.
• The dataset consists of 8124 instances of 119 attributes which are
derived from 24 species.
• The chess data set is also a dense datasets that is consist of 3196
instances and 74 itemset.

15. Conclusion
• Conclusion in this paper h-mine for uncertain data. Finally we
have analyzed the performance of frequent pattern mining
algorithm on few benchmark metrics.
• In case of binary dense data model FB-growth performs better
than other algorithms because the dense dataset result in a very
compact FP-tree which requires less amount of data.

16. Continue…
• In case of sparse data sets H-mine performs better than FP-
growth. The reason is that the FP-tree is bigger and spend a lot of
time in building and transversing the conditional FP-trees.
• The Hmine and P-Hmine saved a lot of scans of the database and
achieve better performance than Apriori on all tested datasets.
• The P-Hmine is also scalable for both large number of data items
and large number of transactions.