Utility mining is a key rising field in data mining, meant to reveal High Utility Item sets (HUIs), from a dataset.
engineering
Description
An efficient technique for High Utility Item sets Retrieval
in Multi-level database
Abstract
Utility mining is a key rising field in
data mining, meant to reveal High Utility Item sets (HUIs), from a dataset. The
item sets with utility value greater than a previously mentioned threshold arereferred to as HUIs. Several algorithms were
projected to date, to retrieve HUIs in a single level database. As per the literature survey, no utility mining algorithms
is proposed for multiple level dataset. The main
objective of this work is to retrieve HUIs from a multi-level dataset. In this work, a unique multi-level utility mining
algorithm namely MUMA (Multilevel Utility Mining Algorithm) was proposed to
retrieve HUIs in a multi-level dataset. The MUMA algorithm implements a tree
structure known as MUMT (Multilevel Utility Mining Tree), to store utility
information of the item sets. Candidate item sets are generated from this tree
thereby reduces the space complexity of the
algorithm. The research work also proposed
an enhanced tree-based algorithm namely Multi-level
Utility Mining using Enumeration Tree (MU_ET), Multi-level Utility Mining using
Utility pattern Tree (MU_UP), Multi-level Utility Mining using Lexicographic
Tree (MU_LG). Since there are no existing
multi-level utility algorithms in the literature, MUMA was compared with MU_ET,
MU_UP and MU_LG and performance are analyzed. The experiments are done using
different datasets like transaction datasets, weblogdatasets,and
synthetic datasets. Performance factors like execution time, Memory space,
number of potential high utility items and number of HUIs retrieved in each
level was evaluated in each dataset.
Keywords:
Utility mining, Multi-level utility mining, Multi-level High utility item sets,
Multiple utility threshold, Multilevel Utility
Mining Tree.
1. INTRODUCTION
|
B |
usiness Analytics is the analyses of historical
organization information, for the data-focused
higher cognitive decision-making process
in business. The organizations use business analytics to systemize and enhance
the business processes. Recognizing profitable items is a crucial task in business analytics.
Understanding the foremost profit-making customers and the profitable product is essential in themostdecision-making
scenario like making a lot of profitable offers, equalization the profitable
product sales and characteristic the simplest targets for brand spanking new
product. Utility mining is a recent
rising field of data mining, accustomed fetch high cash creating item sets in
an exceedingly retail business information.[1] [2] .
1.1. Utility
Minining
Utility refers to a linguistics measure and can be applied to
a transaction database [3] [4] [5]. Utility
mining was introduced to beat the deficiency of frequent itemset mining. The
deficiency of frequent itemset mining was that it counts the number of the existence of the items in a database, not the profit of the items[6]. Hence utility mining was
introduced to contemplate these
factors. The intention of utility mining is to fetch high utility item sets
from a database [7]
[8] [9].This is done by shrewd utility values of all item sets in a database and to retrieve the item sets whose utility
value is larger than
utility threshold.
The utility will
be applied to item/item set in a transaction,
item/item set in a database or to a transaction
itself [2] [10] [11].These
calculations will be done based on the following definitions.[9] [10] [12] [13] [14] [15]
Definition 1:The external
utility of item I1, symbolized as exu(I1), is the unit profit of the item I1 mentioned
in the utility table.
Definition 2:The internal
utility of item I1 in transaction T1, symbolized as inu(I1, T1),
is the quantity of item sold in a transaction T1.
Definition 3:The utility of
itemI1 in transaction T1, symbolized as u(I1, T1), is the product of inu(I1, T1) and exu(I1), where u(I1, T1) = inu(I1, T) × exu(I1).
Definition 4:The utility of itemset S in
transaction T1,
symbolized as u(S,T1), is the sum
of the utilities of all the items in S in T1, i.eu(S,T1)=Σi∈S∧S⊆T u(I, T1).
Definition 5:The utility of
itemset S,
symbolized as u(S), is the sum of the
utilities of S in all the transactions containing S in the database(DB), i.e
u (S)=ΣT∈DB∧ S⊆T u(S,T).
Definition 6: The
transaction utility (TU) of a transaction T1, symbolized as TU(T1), is the sum of the utilities of items in T1, i.e TU(T1) =ΣI∈T ∧ T⊆DB u(I,T).
Definition 7: The Transaction weighted utility of itemI1symbolized
as TWU(I1) is the sum of
all transaction utility (TU) in which the item I1 appears.
Definition
8: An itemset S
in a database DB points out as a
high-utility itemset (HUI) if its utility value exceeds minimum utility
threshold.
