Transaction Profitability Using HURI Algorithm [TPHURI]

International Journal of Business Information Systems Strategies (IJBISS) Volume 2, Number 1,November 2013
1
TRANSACTION PROFITABILITY USING HURI
ALGORITHM [TPHURI]
Jyothi Pillai1
and O.P.Vyas2
1
Associate Professor, Bhilai Institute of Technology, Durg, Chhattisgarh, India
2
Professor, Indian Institute of Information Technology Allahabad, U.P., India
ABSTRACT
Business intelligence (BI) is formulation of business strategies which help organizations to achieve its
objectives and to predict its future. Data mining is often referred as BI in the domain of business. One of
the major tasks in data mining is Association Rule Mining (ARM). ARM techniques incorporated in BI
systems can be utilized in business decision-making such as retail shelf management, catalog design,
customer segmentation, cross-selling, quality improvement and bundling products marketing.
ARM technique is used for the identification of frequent itemsets from huge databases and then generating
strong association rules by considering each item having same value. But in a large number of real world
applications, items have different values according to their impact on the respective decision making
processes. Traditional ARM techniques cannot fulfil the arising demands from these applications. The data
mining researchers are continuously improving the quality of ARM technique by incorporating the utility of
items. The utility of item is decided by its contribution towards the business profit or quantity of the item
sold, etc. Hence Utility mining focuses on identifying the itemsets with high utilities.
Jyothi et al proposed HURI algorithm in [2] for producing high utility rare itemset according to users’
interest. An algorithm Transaction Profitability using HURI [TPHURI] is proposed in this paper which is
a modified version of HURI. TPHURI finds profitable transactions consisting of high utility rare items and
also finds the share of such items in the overall profit of the transactions.
KEYWORDS
Business Intelligence, Association Rule Mining, Utility, Rare Itemset
1. INTRODUCTION
Association rule mining (ARM) algorithms are used to find frequent itemsets and then generate
association rules. Association rule mining (ARM) algorithms are used to find frequent itemsets
and then generate association rules. In many real world applications such as medical, security,
business, etc, items have got different values according to user’s perspective. Thus Utility Mining
considers different utility values for different items.
Utility Mining is helpful in finding the utility which is a measure to find the usefulness or
profitability of an itemset. The total utility of an itemset depends on internal utility and external
utility. The internal utility or local transaction utility of an itemset is obtained from the
transactional information such as the total quantity of the itemset in a particular transaction. The
external utility of an itemset is obtained also from external information sources other than from
transactions such as the contribution of itemset towards business profit. The external utility is

2
generally assigned to itemsets by considering the user’s preferences. The goal of utility mining is
to identify high utility itemsets which drive a large portion of the total utility [8].
Classical ARM algorithms consider the presence of an item in a transaction to be more important
than its absence. The patterns that are rarely found in a database are known as infrequent patterns.
The itemset whose support value is less than the maximum support threshold is defined as
infrequent itemset or rare itemset [5]. In many business applications, rare itemsets may bring
some unexpected acceptable profits. For expanding the business, the supermarkets or stores may
shortlist high profitable rare items and then their quantity can be increased for increasing the
business profitability.
In this paper, it is proposed that Transaction Profitability using HURI [TPHURI] finds
profitable transactions consisting of high utility rare items from transactional dataset. TPHURI is
used for Transaction Utility Mining to find the share of such items in the overall profit of the
transaction dataset. Profitable or interesting transactions are those in which customers purchase
high utility rare itemsets. The outcome of TPHURI would enable the top management or business
analyst in crucial decision-making such as catalog design, providing credit facility, cross
marketing, finalizing discount policy, analyzing consumers’ buying behaviour, organizing shelf
space, loss-leader analysis and quality improvement in supermarket [3].
The rest of paper is organized as follows. Section 2 presents some related works, Section 3
discusses theoretical definitions and section 4 proposes TPHURI algorithm; an application
algorithm using HURI algorithm. Section 5 presents conclusion and future work.
2. LITERATURE SURVEY
Unlike traditional ARM algorithms, the main aim of utility mining algorithms is to discover
itemsets having high utilities. For addressing the limitations of AMR technique, Yao et al defined
utility mining model [8] for generating high utility itemsets.Yao et al proposed two more utility-
based itemset mining algorithms UMining and UMining_H for generating all high utility itemsets
by quantifying user preferences [6].
In utility mining downward closure property is not applicable, hence lots of time is consumed in
generation of candidate itemsets. In [8], Ying Liu et al presented a Two-Phase algorithm to
generate high utility itemsets efficiently by pruning down the number of candidates. In the first
phase, the transaction-weighted utilization mining model is proposed which applies Transaction-
weighted Downward Closure Property on the search space to expedite the identification of
candidates [8]. High utility itemsets are identified in the second phase by performing one extra
database scan.
In [6], Saravanabhavan et al presented a novel utility frequent-pattern efficient tree structure for
mining high utility itemsets. For mining utility patterns the authors have used the pattern growth
methodology. The authors claim that efficiency of high utility itemsets mining is improved by
using two major concepts: 1) the large database is compressed into smaller data structures and
also the repetition of the database scans can be avoided by using utility FP-tree; 2) by utilizing the
pattern growth method the search space can be reduced by avoiding generation of a large number
of candidate sets.
A new algorithm, named Rarity, is presented by Luigi T et al for mining rare itemsets discovered
from large databases [4]. A High Utility Rare Itemset Mining [HURI] algorithm is proposed by
Jyothi et al for finding high utility rare itemsets according to users’ preferences [6]. Using HURI
algorithm [6], high utility rare itemsets are generated in two phases:-

3
(i) In first phase, rare itemsets are generated by considering those itemsets which have support
value less than the maximum support threshold.
(ii) In second phase, by inputting the utility threshold value according to users’ interest, rare
itemsets having utility value greater than the minimum utility threshold are generated.
3. PROBLEM DEFINITION
In this section related theoretical concepts of the proposed algorithm TPHURI are presented [3].
DEFINITION 3.1(TRANSACTIONAL DATASET) A transactional dataset is a collection of
transactions where each transaction is a record of items [3]. Let I be a set of quantities of items
I={i1, i2, i3,… , im}and D be a set of transactions {T1,T2,…,Tn} with items, where each item
I
i ∈ (table 1). Each transaction in D is assigned a transaction identifier (T_ID).
For eg. Table 1 is a transactional dataset D consisting of 25 transactions and 20 items.
DEFINITION 3.2 (INTERNAL UTILITY) The internal utility value of item ip in a transaction Tq,
denoted o(ip, Tq) is the value of an item ip in a transaction Tq (Table 2). The internal utility
reflects the occurrence of the item in a transaction database [3]. The set of utilities is defined as
U={u1, u2, u3,… , uk} (table 2).
For e.g., in transaction T19, the quantities of items A0001, B0002, C0003, D0004, E0005, F0006,
G0007,… are 0,0,0,2,0,4,0,… respectively. Internal utility of item G in transaction T6 is
o(G0007, T6) = 0, while internal utility of item G0007 in Transaction dataset D (table 1) is
o(G0007, D) = 8.
DEFINITION 3.3 (EXTERNAL UTILITY) The external utility value of an item is a numerical
value s(ip) associated with an item ip such that s(ip )=u(ip), where u is a utility function assigning
utility values according to user preferences (table 2) [3].
From table 3, external utility of item G0007 is s(G0007) = u(G0007) = 6.
DEFINITION 3.4 (ITEM UTILITY) The utility of an item ip in a transaction Tq, denoted U(ip,
Tq) is product of o(ip, Tq) and s(ip), where o(ip, Tq) is the internal utility value of ip, s(ip) is the
external utility value of ip(table 3) [3].
For e.g. total utility of item A is U(G0007) = s(G0007) * o(G0007) = 6 * 8 = 48 (table 2).
DEFINITION 3.5 (UTILITY TABLE) A utility table UT (table 2) is a table containing items and
their corresponding utility values where each item i has some utility value uj in U={u1, u2,…,uk}
for some k > 0 [3].
DEFINITION 3.6 (TRANSACTION UTILITY) The transaction utility value of a transaction,
denoted as U(Tq) is the sum of utility values of all items in a transaction Tq (table 1, table 2). The
transaction utility reflects the utility of items in a transaction database [3].
For e.g., the transaction utility of the transaction T1,
U(T1) = U(A0001)+U(B0002)+U(C0003)+U(D0004)+ … + U(T0020) = 37
DEFINITION 3.7 (UTILITY MINING) Utility Mining is used to find those itemsets having utility
values greater than user defined minimum utility threshold. The utility of an itemset X, i.e., u(X),

4
is the sum of the utilities of itemset X in all the transactions containing X. An itemset X is called a
high utility itemset if and only if u(X) >= min_utility, where min_utility is a user-defined
minimum utility threshold [YH2004]. Identification of the itemsets with high utilities is called as
Utility Mining [3].
DEFINITION 3.8 (RARE ITEMSET MINING) Rare itemsets are those itemsets which occur
infrequently in the transactional dataset. In many practical situations, rare itemsets having high
utilities provide very useful insights to the user. Rare patterns may also indicate the occurrence of
exceptional situations in the data. For e.g. If {Fire=Yes} is frequent but {Fire=Yes, Alarm=ON}
is infrequent, then latter is an interesting infrequent pattern because it may indicate faulty alarm
system [3], [4].
Rare itemset mining is a challenging task where the key issues are: -
(i) Identifying interesting rare patterns and
(ii) Efficiently discovering them in large datasets.

5
4. PROPOSED ALGORITHM
Algorithm TPHURI
Description: Finding High Utility Rare Itemsets of users’ interest and Profitable
Transactions
Ck: Candidate itemset of size k Lk: Rare itemset of size k
For each transaction t in database
do begin
increment support for each item i present in t
End
L1= {Rare 1-itemset with support less than user provided max_sup};
for(k= 1; Lk!=Ø; k++)
do begin
Ck+1= candidates generated from Lk;
//loop to calculate total utility of each item
do begin
Calculate total quantity of each item i in t
Find total utility for item i using following formulae:
Total_utility(item i) = internal_utility(item i) * external_utility(item i)
End
//loop to find rare itemsets and their utility
for each transaction t in database
do begin
increment the count of all candidates in Ck+1 that are contained in t
Lk+1 = candidates in Ck+1 less than max_support
Add Lk+1 to the Itemset_Utility table in database and calculate rare itemset Utility using
formulae:
Utility(R, t) = ∑for each individual item i in R (u(i, t));
End
// loop to calculate profit of each transaction and then to find profitable transactions
do begin
Set profit of each transaction in transaction_utility table as
Profit_transaction_t = utility(item i) * quantity(item i in t);
If (profit_transaction_t > user_provided_transaction_utility)
Then Transaction is a profitable transaction
Else Transaction is a non-profitable transaction
End
//loop to calculate whole database utility
For each transaction t
do begin
Db_utility = ∑(profit of each transaction in transaction_utility table)
End
//loop to calculate share of each rare itemset in whole database using Db_utility
For each itemset iset in itemset_utility table
do begin
Share[iset] = utility[iset] / Db_utility;
If (share[iset] > user_provided_threshold for high_profitable_rare_itemset)
Then iset is a rare_itemset which is of user interest
Else iset is a rare itemset but is not of user interest
End
Return rare_itemsets of user interest, profitable_transactions
END
Figure 1: Pseudo Code for TPHURI

6
A High Utility Rare Itemset Mining [HURI] algorithm is proposed by Jyothi et al for generating
high utility rare itemsets according to users’ interest [2]. HURI can be used in a variety of
business applications for increasing the business profitability. Jyothi et al presented a very
innovative idea for customer utility mining by using HURI as a base. CSHURI algorithm,
Customer Segmentation using HURI, presented in [3] finds those customers who purchase high
profitable rare items and then classify the customers according to some criteria.
Another application of HURI is presented in this paper. The authors propose an algorithm
Transaction Profitability using HURI [TPHURI] which finds profitable transactions consisting
of high utility rare items and also finds the share of such items in the overall profit of the
transactions.
Transaction Profitability using HURI [TPHURI] algorithm uses two-phase HURI algorithm
[2] for finding profitable transactions. Profitable or interesting transactions are those in which
customers purchase high utility rare itemsets.
TPHURI algorithm consists of following three phases:-
(i) In first phase, rare itemsets are generated from data set having support value less than the
maximum support threshold. Rare rules are those rules appearing below the maximum
support value.
By setting the value of maximum support threshold to 40%, the rare itemsets generated from
table 1are listed in table 4.
(ii) In second phase, high utility rare itemsets having utility value greater than the transaction
utility threshold are generated.
If high utility threshold is set as 80, the high utility rare itemsets generated are listed in table
5.
(iii) At last in the final phase, by setting the transaction utility threshold, profitable transactions
consisting of high utility rare items are found.
For e.g.; by setting the user provided transaction utility as 45, transactions can be classified
as Interesting (Profit transaction > 45) or Uninteresting (table 6).
Also the share of high utility rare items in the overall profit of the transactions is found in the last
phase. A concept, itemset share, is proposed in [1] which can be considered as utility because it
reflects the impact of sales of itemset on the itemset cost or profit. Itemset share is defined as a
fraction of some numerical value, such as total quantity of items sold or total profit.
The final outcome of TPHURI is a set of profitable transactions consisting of high utility rare
itemsets which would enable the top management or business analyst in crucial decision-making.
The knowledgeable output generated from TPHURI will be applicable for the following business
decision-making processes –
• catalog designing
• allocation of credit facility
• implementing discount policy
• analyzing customers’ buying behavior and loyalty
• classification of customers
• customer retention management
• organizing shelf space
• quality improvement of products
• demand forecasting

7
• monitoring movement of high-rated stock and predicting future stock value
• planning for promotion of high utility products
• Sales up-gradation
• Global marketing trend analysis
• High utility product diversification and pricing
Table 3 Transaction utility of the transaction database

8
Table 6 Table showing profitable transactions
5. CONCLUSIONS
Data mining techniques can be used by enterprises for minimizing purchasing costs; ranking
suppliers by scoring the quality of supplied goods and services; identifying the effective
promotions; identifying profitable or high utility itemsets. After identification of high utility rare
itemsets, marketers can do the promotion or advertising of such itemsets to increase the overall
profit of the business. Transaction Profitability using HURI [TPHURI] algorithm first
generates high utility rare itemsets using HURI algorithm. TPHURI then finds profitable

9
transactions consisting of high utility rare items and also finds the share of such items in the
overall profit of the transactions.
The profitability of the companies can be increased by identifying the profitable transactions
consisting of high utility itemsets and accordingly marketing strategies can be developed for
them. The knowledge generated from TPHURI would aid in crucial business decision-making
processes such as catalog design, providing credit facility, cross marketing, finalizing discount
policy, analyzing consumers’ buying behaviour, organizing shelf space, loss-leader analysis and
quality improvement in supermarket. The overall sales will be upgraded by the promotion,
diversification and pricing strategies used for the sale of high utility products. TPHURI can be
efficiently used in other real time applications such as health-care systems, insurance policies,
banking, etc.
REFERENCES
[1] Barber B., Hamilton, H. J. “Extracting share frequent itemsets with infrequent subsets”, Data Mining
and Knowledge Discovery, 7(2) (2003), pp 153-185.
[2] Jyothi Pillai, O.P.Vyas, “High Utility Rare Itemset Mining (HURI): An approach for extracting
highutility rare item sets”, i-manager’s Journal on Future Engineering and Technology (JFET), ISSN
Online: 2230-7184, ISSN Print: 0973 – 2632, Aug.-Oct. 2011.
[3] Jyothi Pillai, O.P.Vyas, “CSHURI – Modified HURI algorithm for Customer Segmentation and
Transaction Profitability”, International Journal of Computer Science, Engineering and Information
Technology (IJCSEIT), Vol.2, No.2, April 2012, pp 79-89.
[4] Luigi T., Giacomo S. and Cosimo B., A Fast Algorithm for Mining Rare Itemsets, Ninth International
Conference on Intelligent Systems Design and Applications, 978-0-7695-3872-3/09, ©IEEE
Computer Society, DOI 10.1109/ISDA.2009.55, pp1149-1155.
[5] Pang N. T., Michael S. and Vipin Kumar, “Introduction to Data mining”, 2009.
[6] Saravanabhavan C. and Parvathi R. M. S., “Utility FP-Tree: An Efficient Approach to Mine
Weighted Utility Itemsets”, European Journal of Scientific Research,© EuroJournals Publishing, Inc.,
http://www.eurojournals.com/ejsr.htm, ISSN 1450-216X , Vol.50, No.4, 2011, pp.466-480.
[6] Yao, H., Hamilton, H. J., “Mining itemset utilities from transaction databases”, Data and Knowledge
Engineering, December 2006, Volume 59, pp 603-626
[7] Yao, H., Hamilton, H. J., and Butz, C. J. “A Foundational Approach to Mining Itemset Utilities from
Databases”, Proceedings of the 4th SIAM International Conference on Data Mining, Florida, USA,
2004.
[8] Ying Liu, Wei-keng Liao, Alok Choudhary, A Fast High Utility Itemsets Mining Algorithm, UBDM
'05 , August 21, 2005, Chicago, Illinois, USA.
Authors
Mrs. Jyothi Pillai is Associate Professor in Department of Computer Applications at
Bhilai Institute of Technology, Durg (C.G.), India. She is a post-graduate from Barkatullah
University, India. She is a Life member of Indian Society for Technical Education. She has
a total teaching experience of 18 years. She has a total of 24 research papers published in
National / International Journals / Conferences into her credit. Presently, she is pursuing
Ph.D. from Pt. Ravi Shankar Shukla University, Raipur under the guidance of Dr.
O.P.Vyas, IIIT, Allahabad.
Dr.O.P.Vyas is currently working as Professor and Incharge Officer (Doctoral Research
Section) in Indian Institute of Information Technology-Allahabad (Govt. of India’s Center
of Excellence in I.T.). Dr.Vyas has done M.Tech.(Computer Science) from IIT Kharagpur
and has done Ph.D. work in joint collaboration with Technical University of Kaiserslautern
(Germany) and I.I.T.Kharagpur. With more than 25 years of academic experience Dr.Vyas
has guided Four Scholars for the successful award of Ph.D. degree and has more than 80
research publications with two books to his credit. His current research interests are Linked
Data Mining and Service Oriented Architectures.

Transaction Profitability Using HURI Algorithm [TPHURI]

More Related Content

What's hot (12)

Similar to Transaction Profitability Using HURI Algorithm [TPHURI] (20)

More from ijbiss (20)

Recently uploaded (20)

Transaction Profitability Using HURI Algorithm [TPHURI]