Review on: Techniques for Predicting Frequent Items

Volume 1, Issue 1 (2018)
Article No. 5
PP 1-8
1
www.viva-technology.org/New/IJRI
Review on: Techniques for Predicting Frequent Items
Himanshu A. Chaudhari1
, Darshana S. Vartak1
, Nidhi U. Tripathi1
, Sunita Naik2
1
(B.E. Computer Engineering, VIVA Institute of Technology/Mumbai University, India)
2
(Assistant Prof. Computer Engineering, VIVA Institute of Technology/Mumbai University, India)
Abstract : Electronic commerce(E- Commerce) is the trading or facilitation of trading in products or services
using computer networks, such as the Internet. It comes under a part of Data Mining which takes large amount
of data and extracts them. The paper uses the information about the techniques and methods used in the
shopping cart for prediction of product that the customer wants to buy or will buy and shows the relevant
products according to the cost of the product. The paper also summarizes the descriptive methods with
examples. For predicting the frequent pattern of itemset, many prediction algorithms, rule mining techniques
and various methods have already been designed for use of retail market. This paper examines literature
analysis on several techniques for mining frequent itemsets.The survey comprises various tree formations like
Partial tree, IT tree and algorithms with its advantages and its limitations.
Keywords – Association Rule Mining, Data Mining, Frequent Itemsets, IT tree, Market Basket Data,
Prediction.
1. INTRODUCTION
We live in a world where huge amount of data are collected each and every day. Analyzing such data is
an important need. Data mining is the practice of automatically searching large stores of data to discover
patterns and trends that go beyond simple analysis. Data mining is also known as Knowledge Discovery in Data
(KDD). There are huge amount of data generated in the various organizations. Therefore organizer has to take
number of decisions during extraction of useful data from the huge amount of data. But it is difficult to take out
each and every record, so organizer finds frequently occurring data in the database. Pattern mining is a subfield
of data mining. An interesting pattern is a pattern that appears frequently in a database. The purpose of frequent
itemsets mining is to identify all frequent itemsets, i.e., itemsets that have at least a precised minimum support,
the percentage of transactions containing the itemsets. Frequent patterns as a name suggest are patterns that
occur frequently in data.
A frequent itemsets typically refers to a set of items that often appear together in a transactional
dataset. For example, customer tends to purchase first laptop, followed by a digital camera and then a memory
card, is a frequent pattern. Mining frequent patterns leads to the discovery of interesting association and
correlation within data. Association rule mining is meant to find the frequent itemsets, correlations and
associations from various type of database such as relational database, transaction database, sequence databases,
streams, strings, spatial data, graphs, etc. Association rule mining tries to find the rules that direct how or why
such items are often bought together in a given transaction with multiple items. The main application of
association rule mining is market basket data. Association rule can be defined as XY where X, Y are itemsets
with antecedent and consequent respectively. Market Basket analysis[5] consist of support and confidence
where support is used to identify how frequently itemsets appears in dataset and confidence is used to identify
how frequently the rule has been found to be true. The support of a rule is the number of sequence containing
the rule divided by the total number of sequences. Supp(XY) = p (A  B). The confidence of a rule is the
number of sequence containing the rule divided by the number of sequences containing its
antecedent. Conf(XY) = supp (A, B)/supp (A). By using Support and confidence values, one can generates
rules on incoming queries and more precised prediction can be determined using prediction algorithm.

Article No. 5
PP 1-8
2
2. TECHNIQUES FOR PREDICTION OF FREQUENT ITEMSETS
Frequent patterns are itemsets, subsequence, or substructures that appear in a data set with frequency
no less than a user-specified threshold. Frequent itemsets are a form of frequent pattern. Discovery of all
frequent itemsets is a typical data mining task. The original use has been as part of association
rule discovery. By finding frequent itemsets, a retailer can learn what is commonly bought together. Especially
important are pairs or larger sets of items that occur much more frequently than would be expected were the
items bought independently. In this section, the methods for mining the simplest form of frequent patterns are
given.
2.1 Prediction of Missing Item Set In Shopping Cart [1]
Author invented IT-Tree (Itemsets Tree) technique. In this paper proposed algorithm makes use of
flagged IT-Tree. IT-Tree created from training data set. In this algorithm incoming itemsets were considered as
input and depend on that return graph that defines the association rule. In this algorithm they first identified all
high support, high confidence rules that have antecedent a subset of itemsets. Then after his it combines
consequent of all these rules and then created a set of items which are frequently bought. This method mainly
identify repeated occurrence of items and sort them accordingly. And most identical that is root items are
indicated with Flagged items. But there are two major drawbacks like time taken for execution is more and this
method requires more memory for processing.
Figure 1: Construction of IT tree from given database [1]
Overall paper gives brief idea about generating the IT tree to scan the dataset and sort into identical
itemset. It is Advantageous for computing itemset generation and can be used for generating candidate item sets.
2.2 Data Structure for Association Rule Mining: T -Tree and P Tree [2]
This paper demonstrates structure and algorithm using T tree and P tree with Advantage of storage and
execution time. The Total support tree (T tree) method is used to create an object node. After this method tree is
converted into array. The array format presents Partial support tree (P tree).This system proposed that the partial
support tree is increases the performance of storage and execution time. It also overcomes the Apriori algorithm.
In T tree and P tree structure branches are considered as independent therefore this structure can be used in
parallel or distributed Association Rule Mining.
Thus paper finally concluded with two different types of tree formation method in which it first form
tree and then convert it into array format which consumes memory and gives better performance in terms of
support calculation.
2.3 Itemset Trees For Targeted Association Querying [3]
The paper proposed querying the database is made even faster by rearranging the database using the
IT-tree data structure. This becomes handy especially in the batch mode Prediction (i.e., when you have to
predict `missing items' for several shopping carts). IT-tree is a compact and easily updatable representation of
the transactional dataset. Also, the construction of the itemset tree has O(N) space and time requirements. So,
this data structure is used to speed up the proposed predictor.

Article No. 5
PP 1-8
3
The paper gives detail architecture of itemset tree and experiment with dataset. The experimental result
is fast for query answering and it can use for large dataset. But it required more memory.
2.4 Finding Localized Associations In Market Basket Data [4]
The author introduced about market basket analysis which contain support and confidence values. One
basket tells you about what one customer purchased at one time. It is a basically a theory that if you buy certain
group of items, you are likely to buy another groups of items. Market basket analysis is used in most of all
frequent mining concepts. Authors give clustering and indexing algorithms which are used to find significant
correlations for association mining.
The paper includes clustering algorithm makes computational process simple with variable type of
data. But it will increase its complexity if the problem size is increased.
2.5 An Approach for Predicting Missing Item from Large Transaction Database [5]
The system designed about the architecture to utilize the knowledge of incomplete constituents of a
“shopping cart” for the guessing of what more the customer is likely to purchase. Author takes synthetic data
obtained from IBM generator. Next step is taken as classification of clusters using Naïve Bayes text classifier
and hierarchical document clustering which is simple to implement and used for large database. These clusters
are then used for graph construction in form of Hash list which is combination of Hast table and Linked list.
And finally Combo Matrix is used for prediction purpose.
Overall the concept of clustering in paper is useful to reduce memory requirement as it does not
generate candidate set. But clustering has inability to recover from database corruption and it can arise problem
due to data scarcity.
2.6 Review On: Prediction of Missing Item Set in Shopping Cart [6]
The paper reviews for prediction of frequent items in shopping cart. Predicting the missing items from
dataset is indefinite area of research in data mining. In this paper some algorithm is introduced to identify the
frequently co-occurring group of items in the transaction database for prediction purpose. In this paper author
explains the existing approach which contain IT flagged tree. After getting IT tree the main root and identical
items sets are indicated by black dot. They modified the original tree building algorithm by flagging each node
that is identical to at least one transaction. This is called “Flagged IT tree”. Disadvantage of this approach is it
generates candidate itemsets which acquire memory space. It uses multiple passes over database. Author
proposed Dempster Combination Rule which is used to combine all the rules.
Paper actually gives overall idea about predicting the missing items in shopping cart. Paper is focused
on Dempster Shafer combination rule which is used to combine rules formed by rule generation. Proposed
system described in paper is more flexible than other system. For e.g. processing speed with IT tree is much
better than clustering the items.
2.7 Sequential Approach for Predicting Missing Items in Shopping Cart Using Apriori Algorithm. [7]
The author described sequential approach to predict the missing items in shopping cart using Apriori
algorithm. The main objective of this paper to maintain the limitation of excessive wastage of time to hold a
huge amount of candidate sets with much frequent itemsets .This system proposed to increase the performance
of support value. The authors defined the disadvantages of Apriori algorithm that it generates the number of
candidate items. The main disadvantage of this proposed system is wastage of memory.
The proposed system uses sequential approach for prediction using Apriori algorithm. It is basic
algorithm and can be applied on any type of dataset. The system gives 65% accuracy with respect to prediction
time. But it has disadvantages like storage capacity and I/O load so it cannot be use for long time.
2.8 Data Mining Approach For Retail Knowledge Discovery [8]
This Paper introduced the data mining techniques that are used in retail market for knowledge
discovery are describes as following: Market Basket Analysis: Data mining association rules, also called market
basket analysis, is one of the application areas of Data Mining. Consider a market with a collection of huge
customer transactions. An association rule is XY where X is called the antecedent and Y is the consequent. X
and Y are sets of items and the rule means that customers who buy X are likely to buy Y with probability %c
where c is called the confidence. The algorithms generally try to optimize the speed since the transaction

Article No. 5
PP 1-8
4
databases are huge in size. This type of information can be used in catalog design, store layout, product
placement, target marketing, etc. Basket Analysis is related to, but different from Customer Relationship
Management (CRM) systems where the aim is to find the dependencies between customer’s demographic data.
The paper is all about review of literature based on techniques used in data mining for retail market
knowledge discovery. And theoretically conclude with best approach as Apriori algorithm. But it cannot be used
for larger dataset.
2.9 Comparing Data Set Characteristics That Favour The Apriori, Eclat Or FP-Growth Frequent Itemset
Mining Algorithms. [9]
Existing system compares the frequent itemset mining techniques with respect to dataset
characteristics. Author mainly focused on 3 main algorithms that are Apriori, Eclat and FP-Growth. All three
algorithms are used to predict frequent item sets. Paper comprises survey on each algorithm with figure and
example. Author gives advantage and disadvantages of each algorithm. In this paper, accuracy detected with
respect to parameters as basket size vs. Runtime etc. By analysing these algorithms, author concludes that
Apriori is basic and simplest algorithm. But Apriori has serious scalability issues and exhausts available
memory much faster than Eclat and FP-Growth. Most frequent itemset applications should consider using either
FP-Growth or Eclat.
This paper has beneficial for next version of Eclat or FP-Growth algorithm which decreases the
complexity of both algorithms. The survey paper shows that Eclat and FP-growth algorithm is much better than
Apriori in all cases.
2.10 An Enhanced Prediction Technique for Missing Itemsets in Shopping Cart [10]
This system proposed the shopping cart prediction architecture. Based on passed transaction we can
easily construct a Graph structure from which association rules are generated in consideration of new incoming
instances in new transaction. Then based on threshold value set by the user and kept dynamic, the prediction
algorithm predicts the new item set to be considered for purchase. Threshold value is the minimum support
value that a particular pair has to be present before getting predicted.
2.11 Predicting Missing Items In Shopping Cart Using Associative Classification Mining [11]
This paper describes generation of Boolean matrix using AND operation. And also introduced new
concept BBA (Baysian Belief Argument) and rule selection which is used to select the rules from association
rule where all the rules are identified using support and confidence values. After getting all possible generated
rules, decision making algorithm that is Dempster Shafer algorithm is used for prediction.
Thus the system Combine all the rules using Dempster Shafer algorithm according to BBA and Rule
selection technique.

Article No. 5
PP 1-8
5
Figure 2: Shopping cart prediction architecture [11]
2.12 Missing Item Prediction And its Recommendation Based On Users Approach In E-Commerce [12]
This system proposed the algorithm in this spectrum use fast and effective technique. The system uses
association rule mining techniques. This method produces high support and high confidence rules. This
technique proves to appear better than the traditional techniques in association rule mining. But the cons of this
technique are complexity increases with the increase in average length of items. The alternative method to
predict missing items uses Boolean vector and the relational AND operation to discover frequent itemsets
without generating candidate items. It directly generates the association rules.
By this proposed system, one can gain the information of predict the missing items uses Boolean
matrix and AND relation operation.
2.13 A Survey on Approaches for Mining Frequent Itemsets [13]
Paper described algorithms for mining from Horizontal Layout Database for non-frequently bought
items. Direct Hashing and Pruning (DHP) Algorithm: DHP can be derived from Apriori by introducing
additional control. To this purposes DHP makes use of an additional hash table that aims at limiting the
generation of candidates in set as much as possible. DHP also progressively trims the database by discarding
attributes in transaction or even by discarding entire transactions when they appear to be subsequently useless.
This method, support is counted by mapping the items from the candidate list into the buckets which is divided
according to support known as Hash table structure. As the new itemset is encountered if item exist earlier then
increase the bucket count else insert into new bucket. Thus in the end the bucket whose support count is less the
minimum support is removed from the candidate set.
2.14 Association Rule Mining Using Improved Apriori Algorithm [14]
Author explained that Apriori algorithm generates interesting frequent or infrequent candidate item sets
with respect to support count. Apriori algorithm can require to produce vast number of candidate sets. To
generate the candidate sets, it needs several scans over the database. Apriori acquires more memory space for
candidate generation process. While it takes multiple scans, it must require a lot of I/O load. The approach to
overcome the difficulties is to get better Apriori algorithm by making some improvements in it. Also will
develop pruning strategy as it will decrease the scans required to generate candidate item sets and accordingly
find a valence or weightage to strong association rule. So that, memory and time needed to generate candidate
item sets in Apriori will reduce. And the Apriori algorithm will get more effective and sufficient.
This Paper gives advantages of Improved Apriori algorithm is that it has less complex structure and
less number of transaction as it scans the dataset less number of times than Apriori. But then also it has
limitation of multiple scan with limited memory capacity.

Article No. 5
PP 1-8
6
2.15 An Efficient Prediction of Missing Itemset In Shopping Cart. [15]
The system proposed the shopping cart prediction architecture. Based on passed transaction we can
easily construct a Graph structure from which association rules are generated in consideration of new incoming
instances in new transaction. Then based on threshold value set by the user and kept dynamic, the Prediction
algorithm predicts the new item set to be considered for purchase. Threshold Value is the minimum support
value that a particular pair has to be present before getting predicted.
3. ANALYSIS
The papers are analyzed by techniques which are studied in module 2. The table analyzes according to
techniques with respect to the parameters like support values, prediction time, transaction length, execution time
etc.
Table 1: Analysis Table
Sr.
No.
Title Technique/Methods Parameter Accuracy
1
Prediction Of Missing
Item Set In Shopping
Cart [1]
Specific IT flagged tree
, BBA
Transaction length
vs. Prediction time
and support threshold
vs. Execution time
Minimum support,
execution time =
56*103
s, Threshold
=30%, if prediction
time is 40s then
average transaction
length is 15.
2
Data Structure For
Association Rule Mining
:T -Tree And P Tree [2]
T tree and P tree
formation in Apriori
algorithm
support vs. Time,
support vs. Storage,
time vs. No. of
records
If support is 4% then
time required is 1 s. If
time require is 30s
then no. of records are
300*103
3
Itemset Trees For
Targeted Association
Querying [3]
IT tree formation,
association rule using
Market Basket
Basket size vs. Time
For 10,000 distinct
items, if there are
4000 baskets then
time required to
prediction is 10s
4
Finding Localized
Associations In Market
Basket Data [4]
Clustering algorithm,
merging operation
No. of cluster vs.
runtime and
N.A.
5
An Approach For
Predicting Missing Item
From Large Transaction
Database [5]
Association rule
(market basket analysis)
Length of transaction
vs. Avg. Size of
transaction
N.A.
6
Review On: Prediction
Of Missing Item Set In
Shopping Cart [6]
Flagged IT tree,
Dempster combination
rule (DCR)
N.A. N.A.
7
Sequential Approach For
Predicting Missing Items
In Shopping Cart Using
Apriori Algorithm. [7]
Apriori algorithm N.A. N.A.
8
Data Mining Approach
For Retail Knowledge
Discovery [8]
Market basket analysis
and Apriori algorithm
N.A. N.A.

Article No. 5
PP 1-8
7
4. CONC
LUSI
ON
The goal of data
mining is to predict
the future or to
understand the past.
The paper includes
analysis of various
techniques used for
predicting the
frequent item set in
shopping cart. The
paper is all about
review of literature
based on techniques
used in data mining
for retail market
knowledge
discovery. Paper
defines methods to
find association rules
with calculation of
support and
confidence values to
get the rules. New
algorithms like
Improved Apriori as
well as modifications
of existing
algorithms are often
introduced
thoroughly. From the
above literature review on different techniques of frequent itemset, paper concludes as Improved Apriori is
better for generating candidate items. To combine each rule DS-ARM is used based on threshold value to get
predicted item. The limitations found in literature are unnecessary generation of candidate itemsets which takes
more utilization of memory. Besides the technical limitations of any decision making (DS-ARM) its usability
and popularity among practitioners should be a matter of concern Also found that the algorithms like Apriori
make multiple scans in database. The drawbacks can be overcome by using less utilization of memory and less
number of scan can decrease the execution time which will be better for performance of prediction of items.
REFERENCES
[1] K. Wickramaratna and M. Kubat, “Predicting Missing Item In Shopping Cart”, IEEE Transactions On Knowledge And Data
Engineering, Volume 21 Issue 7, July 2009.
[2] F. Coenen, P. Leng, and S. Ahmed, “Data Structure for Association Rule Mining: T-Trees and P-Trees”, IEEE Transactions on
Knowledge and Data Engineering, Vol. 16, No. 6, June 2004.
[3] M. Kubat, A. Hafez, V. V. Raghavan, J. Lekkala, And W. K. Chen, “Itemset Trees For Targeted Association Querying”, IEEE
Transactions On Knowledge And Data Engineering, Vol. 15, No. 6, November/December 2003.
9
Comparing Data Set
Characteristics That
Favour The Apriori, Eclat
Or FP-Growth Frequent
Itemset Mining
Algorithms. [9]
Eclat , Apriori, FP-
growth, naive brute
method
density of frequent
item vs. Runtime and
size of basket vs.
Runtime
N.A.
10
An Enhanced Prediction
Technique For Missing
Itemset In Shopping Cart
[10]
Prediction accuracy
measure to find
prediction and recall
transaction length vs.
Prediction time and
execution time vs.
Minimum support
N.A.
11
Predicting Missing Items
In Shopping Cart Using
Associative Classification
Mining [11]
Association rule and
Dempster Shafer
theory
N.A. N.A.
12
Missing Item Prediction
And its Recommendation
Based On Users
Approach In E-
Commerce. [12]
Association rule and
Boolean matrix
N.A. N.A.
13
A Survey On Approaches
For Mining Frequent
Itemsets [13]
Association rule N.A. N.A.
14
Association Rule Mining
Using Improved Apriori
Algorithm [14]
Improved Apriori
Algorithm
Number of scan the
dataset and time
No of scan to Ap
=272 while no. of
to improved Aprio
15
An Efficient Prediction
Of Missing Itemset In
Shopping Cart. [15]
Association rule mining
Precision, recall, F-
value and prediction
time
Time required to
predict item is les
than existing syste

Article No. 5
PP 1-8
8
[4] C. Aggarwal, C. Procopiuc and P. Yu, “Finding Localized Associations In Market Basket Data”, IEEE Transactions On Knowledge And
Data Engineering, Vol. 14, No. 1, January/February 2002.
[5] P. Meshram, D. Gupta, P. Dahiwale, “An Approach For Predicting The Missing Items From Large Transaction Database”, IEEE
Sponsored 2nd International Conference On Innovations In Information Embedded And Communication Systems Iciiecs’15.
[6] S. Yende, P. Shirbhate, “Review On: Prediction Of Missing Item Set In Shopping Cart”, International Journal Of Research In Science &
Engineering, Volume 1, Issue 1, April 2017.
[7] R. Bodakhe, P. Gotarkar, A. Dahiwade, P. Gosavi, J.Syed, “A Sequential Approach For Predicting Missing Items In Shopping Cart
Using Apriori Algorithm”, Imperial Journal Of Interdisciplinary Research (IJIR) Volume 3, Issue4, 2017.
[8] J. Vohra, “Data Mining Approach For Retail Knowledge Discovery”, International Journal Of Advanced Research In Computer Science
And Software Engineering, Volume 6, Issue 3, March 2016.
[9] J. Heaton, “Comparing Dataset Characteristics That Favour the Apriori, Eclat or FP-Growth Frequent Itemset Mining Algorithms”, 30
Jan 2017
[10] M. Nirmala, V. Palanisamy, “An Enhanced Prediction Technique For Missing Itemset In Shopping Cart”, International Journal Of
Emerging Technology And Advanced Engineering, Volume 3, Issue 7, July 2013 .
[11] K. Kumar, S. Sairam, “Predicting Missing Items In Shopping Cart Using Associative Classification Mining”, International Journal Of
Computer Science And Mobile Computing, Volume 2, Issue 11, November 2013.
[12] H. Deulkar, R. Shelke, “Implementation of Users Approach for Item Prediction and Its Recommendation In Ecommerce”, International
Journal Of Innovative Research In Computer And Communication Engineering, Volume 5, Issue 4, April 2017.
[13] S. Neelima, N. Satyanarayana and P. Krishna Murthy, “A Survey On Approaches For Mining Frequent Itemsets”, IOSR Journal Of
Computer Engineering (IOSR-JCE), Volume 16, Issue 4, Ver. Vii, (Jul – Aug. 2014), Pp 31-34.
[14] M. Ingle, N. Suryavanshi, “Association Rule Mining Using Improved Apriori Algorithm”, International Journal Of Computer
Applications, Volume 112,Issue 4, February 2015.
[15] M. Nirmala. and V. Palanisamy, “An Efficient Prediction Of Missing Itemset In Shopping Cart”, Journal Of Computer Science, Volume
9 (1), 2013, pp 55-62.

Review on: Techniques for Predicting Frequent Items

More Related Content

What's hot (18)

Similar to Review on: Techniques for Predicting Frequent Items (20)

More from vivatechijri (20)

Recently uploaded (20)

Review on: Techniques for Predicting Frequent Items