See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/314282049

Market Basket Analysis: A Profit Based Approach to Apriori Algorithm

Conference Paper · September 2016

CITATIONS READS

4 4,256

3 authors:

Wishma Samaraweera Chekaprabha Waduge

General Sir John Kotelawala Defence University General Sir John Kotelawala Defence University
4 PUBLICATIONS   7 CITATIONS    2 PUBLICATIONS   4 CITATIONS   

SEE PROFILE SEE PROFILE

Uma Indeewari Meththananda

General Sir John Kotelawala Defence University
5 PUBLICATIONS   5 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Market Basket Analysis: A Profit Based Product Promotion Forecasting View project

sea level rise View project

All content following this page was uploaded by Wishma Samaraweera on 27 March 2017.

The user has requested enhancement of the downloaded file.

 Proceedings in Computing, 9th International Research Conference-KDU, Sri Lanka 2016

Market Basket Analysis: A Profit Based Approach to Apriori Algorithm

WJ Samaraweera1#, CP Waduge1, and RGUI Meththananda 2

1
Faculty of Computing, General Sir John Kotelawala Defence University, Sri Lanka
2
Faculty of Built Environment and Spatial Sciences, General Sir John Kotelawala Defence University, Sri Lanka
#[email protected]

Abstract—The field of data mining seeks to establishment based on customer self-service in

recognize the regularities, patterns and behaviours retailing. In supermarkets variety of products are
of large data collections. Association mining is showcased in shelves, according to their own
used to discover elements that occur frequently arrangement, making the customer comfortable
within a dataset consisting of multiple independent with purchasing products. The objective of this
selections of elements and to discover rules. This facility is to provide the customer an opportunity
mining approach can find rules which predicts the to explore different brands and prices offered by
occurrence of an item, based on the occurrences of different companies in accordance with their
other items in a particular transaction. Apriori requirements. Conversely, it provides easy access
algorithm is an influential algorithm designed to to go through products with less necessity before
operate on data collections enclosing transactions attaining true requirements. This empowers the
such as in market basket analysis. To address objective of making the basket of the customer
various issues Apriori algorithm has been extended outsize to enhance the profit of the vender, by
in different perspectives. In real world scenario, tempting customers to buy items which were not
one of the major objectives in performing a market intended to buy before entering the place. This
basket analysis is to maximize the profit. In Apriori illustrates the importance of high concentration in
algorithm, Support value and Confidence value are arranging items for floors and shelves. Association
the dominant factors in generating association rule mining is a branch of data mining, which is
rules which seems to be insufficient to achieve the sourced to address this issue, arranging shelves or
said objective as the algorithm does not consist a floors, by finding rules that will predict the
variable to maximize the profit gain. Moreover, occurrence of an item based on combinations of
consideration of frequent items, rather than rare products that frequently co-occur in transactions.
items, significantly impact the profit maximization. It helps the retailers in supermarkets to identify
Therefore, this research was focused to develop a the relationships among the purchased items by
new algorithm based on an extended Apriori customers in order to upgrade better customer
approach which maximize the profit of a satisfaction and retention.
transaction using frequent items as well as rare
items in a market basket analysis. The developed There are several algorithms to generate these
new algorithm and the extended Apriori algorithm association mining rules such as Apriori algorithm,
were applied to a real world data set and the FP-Growth algorithm, K-means, K-nearest
results were compared focusing the profit gain Neighbour Classification, Naïve Bayes, K-Apriori,
from each algorithm separately. Finally, the Eclat etc. This research is based on the influential
results conclude that the proposed algorithm algorithm, Apriori, to address market basket
derives association rules which significantly analysis, identifying products that go well
increase the profit gain, disregard of the number of together, to gain better rules for floor and shelf
items involving in the transaction. arrangements.

The Apriori Algorithm was introduced by Aggarwal

Key Words: Apriori Algorithm, Support Value,
and Srikant (1994) which delivers a way to find
Confidence Value, Market Basket Analysis
frequent itemsets in a market basket analysis.
I. INTRODUCTION Predefined Minimum Support Value, a factor
based on industrial exposure, Minimum Support,
Retailing is one of the leading businesses in the how frequent the item appears in the transaction
world and supermarket is a commercial set calculated with respect to the data set, and

127
 Proceedings in Computing, 9th International Research Conference-KDU, Sri Lanka 2016

Confidence Value, the probability of purchasing an

item when an another item has already been
purchased, are the basic filters, help to generate
association rules in this algorithm which absorbs A. Useful Concepts in Apriori Algorithm.
patterns associates with frequent items only.
 Itemset - A collection of one or more
Although the ultimate objective of the vender is to items (that represents together a single
maximize profit, it is identified that the constraints entity)
confederate with frequency of an item are not Eg: - {Milk, Bread, Diaper}
sufficient to encounter this objective.  Minimum Support – A user defined value
Alternatively, exponentially proportional growth of which helps to eliminate non-frequent
association rules with the expansion of the dataset items from a database.
is another major observation in utilizing the Apriori  Frequent Itemset – An itemset that occurs
algorithm. The research is conducted to discover a in at least a user specific percentage of
new approach to incapacitate the above the database (the sets of item which has
mentioned observations. The proposed minimum support).
modification to the Apriori algorithm is an  Support – The support of a rule, 𝑋 → 𝑌,
introduction of a new constraint associates with is the percentage of transactions in T that
profit. The new constraint, running parallel to the contains 𝑋 ∪ 𝑌, and can be seen as an
minimum support and the support, enhances the estimate of the probability, 𝑃(𝑋 ∪ 𝑌).
total profit gain by generating rules, considering Support determines how frequent the
both rare and frequent items, to arrange the rule is applicable in the transaction set T.
shelves of a supermarket. Further it prunes the The support of rule 𝑋 → 𝑌 is computed
unnecessary rules generates with the expansion of as follows:
the data set intensifying the efficiency of the
process. 𝑆𝑢𝑝𝑝𝑜𝑟𝑡 = 𝑃(𝑋 ∪ 𝑌)
𝑐𝑜𝑢𝑛𝑡(𝑋 ∪ 𝑌)
=
𝑛
II. LITERATURE REVIEW  Confidence - The confidence of a rule,
X → Y, is the percentage of transactions in
T that contains X also contains Y. It is the
A large number of association rule mining conditional probability, 𝑃(𝑋|𝑌). The
algorithms have been developed with different confidence of the rule X → Y is computed
mining efficiencies. Apriori (Agrawal and Srikant, as follows:
1994), FP Growth (Han et al., 2000), Eclat (Han and
Kamber, 2001), K-Apriori (Annie and Kumar, 2011), 𝑐𝑜𝑢𝑛𝑡(𝑋 ∪ 𝑌)
𝐶𝑜𝑛𝑓𝑖𝑑𝑒𝑛𝑐𝑒 =
K-Means (Liu et al., 2014), K-Nearest Neighbor 𝑐𝑜𝑢𝑛𝑡(𝑋)
(Larose and Larose, 2005) and Naïve Bayes
(Kamruzzaman and Rahman, 2010) are some of
the association rule mining algorithms. These B. Apriori Algorithm.
algorithms can be categorized into two types
called candidate generation or pattern growth. The Apriori algorithm works in two steps:
Apriori Algorithm is one of the most popular and 1. Generate all frequent itemsets – A
influential algorithms in association rule mining frequent itemset is an itemset that has
categorized under candidate generation. It is an transaction support above minimum
algorithm for frequent itemset support.
mining and association rule learning over 2. Generate all confident association rules
transactional databases. The Apriori Algorithm was from frequent itemsets – A confident
first introduced by Agarwal and Srikant (1994) association rule is a rule with confidence
which generates frequent itemsets based on a above minimum confidence.
threshold called ‘Minimum Support’.

128
 Proceedings in Computing, 9th International Research Conference-KDU, Sri Lanka 2016

C. Pseudocode for Apriori Algorithm. It is profit oriented that Peanut Butter and Bread
or Peanut Butter and Jelly are arranged in side by
𝐶𝑘 - Candidate itemset of size 𝑘
side in shelves of the grocery store. Such
𝐿𝑘 - Frequent itemset of size 𝑘 information will help the grocery store to decide
which items can be put together in order to tempt
𝐿1 = {frequent items}; the customer to buy more things in a logical
manner.
For (𝑘 = 1; 𝐿𝑘 != ∅; 𝑘++) do begin
But Apriori Algorithm suffers from some main
𝐶𝑘+1 = candidates generated from 𝐿𝑘 ;
limitations such as unnecessary memory utilization
for each transaction 𝑡 in database do by generating a vast number of candidate sets with
increment the count of all candidates in higher frequent itemsets, low minimum support or
𝐶𝑘+1 that are contained in t. large itemsets. (Rao and Gupta, 2012) Furthermore
Apriori Algorithm has a high scanning time since it
𝐿𝑘+1 = candidates in 𝐶𝑘+1 with needs to check for many more itemsets and they
min_support (minimum support) have to be scanned repeatedly in consequent
end steps.

return 𝑈𝑘 𝐿𝑘 ; Several aspects of Apriori Algorithm have been

studied in academic literature in order to generate
Generation of Association Rules is one of the association rules while declining limitations of
major tasks in Data Mining. Association Rule Apriori Algorithm. One of such aspect is mining
Mining is all about finding rules whose support and association rules with multiple minimum supports
confidence exceed the threshold and minimum (Liu et al., 1999). The Extended Model (MSApriori)
support values. These association rules can be allows the user to specify multiple minimum
used in numerous real world tasks such as Market supports to reflect the items and their frequencies
basket analysis, Customer segmentation, Fraud in the database. It generates all large itemsets by
detection, Detection of patterns in text and making multiple passes over the data. This model
Medical diagnosis. emphasizes that having a single minimum support
value is insufficient. If it is set too high, necessary
Apriori Algorithm can be mainly utilized to
rules may not be generated and on the other hand
generate the association rules in Market Basket
if it is set too low, combinatorial explosion will
Analysis. Market Basket Analysis (Raorane et al.,
occur. It is proved here that using multiple
2012; Annie and Kumar, 2011) is one of the most
minimum supports instead of single minimum
frequently used data mining technique used to
support value will provide two conclusions; rare
generate association rules. The purpose of Market
items will not be ignored and number of generated
Basket Analysis is to discover purchasing patterns
rules will be less compared to initial Apriori
of products from a supermarket’s transactional
algorithm. Another approach of Apriori Algorithm
database. Typically in supermarkets very large and
is introducing new parameters to maximize profit
constantly growing databases are maintained.
(Trikha and Singh, 2014). This algorithm enhance
From these large collection of data, it is really
the efficiency of generating association mining
difficult to extract the data related to the pattern
rules by making a model which will be beneficial in
of buying products of customers. Association rules
eliminating the shortcomings of Apriori Algorithm.
in Market Basket Analysis are frequently used by
Two new parameters called, Q-factor using profit
retail stores to assist in marketing, advertising,
ratio and Profit Weighting factor (PW factor) were
floor placement and inventory control. Direct
introduced in order to identify interesting patterns
Marketers could use this technique to determine
from transactional databases and to maximize
the layout of their catalogue and order forms also.
profit.
Eg: A grocery store noticed that 100% of the time
A different approach called Improvised Apriori
that Peanut Butter is purchased, so is Bread.
Algorithm using frequent pattern tree was
Furthermore, 33.3% of the time Peanut Butter is
suggested for real time applications. This algorithm
purchased, Jelly has also been purchased.
focuses on reducing time spent to scan large

129
 Proceedings in Computing, 9th International Research Conference-KDU, Sri Lanka 2016

number of candidate itemsets and saving space III. METHODOLOGY

utilized by unnecessary association rules (Bhandari
et al., 2015). The improvised algorithm will scan Proposed research work is based on an
only some transactions by a formula which improvement of MS Apriori algorithm that
partitions the set of transactions into sections and enhances the effectiveness of the process by
select one particular section among them. In new constructing a model which is beneficial in
model it has been observed that the time overcoming the shortcomings of Apriori algorithm.
consumed in group of transaction is less than the The frequent itemset which gives even 100%
confidence with the classical Apriori algorithm may
classical Apriori Algorithm and the difference
not provide maximum profit gain to the vender.
increases more when the number of transactions
The proposed algorithm calculates a profit factor
increases. Though this approach reduces which supports to maximize profit, associates with
consumed time than the original Apriori Algorithm, various frequent itemsets generated. The
it only reduces the time consuming by 67.87%. proposed improvement of the algorithm is
implemented using Matlab.
There are several other contemporary approaches
Different data sets from different market
to Apriori Algorithm such as a secure mining of
environments has been used to check the internal
Association Rules which is based on the Fast consistency of the proposed algorithm.
Distributed Mining Algorithm (Tassa, Open, and
Road, 2014). Furthermore an Adaptive The workflow of the proposed work is shown
Implementation of Apriori Algorithm was below in Figure 1.
proposed in order to reduce the response time
significantly by using the approach of mining the
frequent itemsets (Balaji et al., 2013). An
association classification based on compactness of
rules is proposed but it suffers from a difficulty of
over fitting (Qiang et al., 2009). How to maximize
the efficiency of the parallel Apriori Algorithm is
discussed and it is suggested that the efficiency
can be improved effective load balancing (Shah
and Mahajan, 2009).

A new approach primarily based on Apriori

Algorithm is proposed in this paper, which
considers profit as a variable when generating
frequent itemsets. Though under theoretical
framework the main variables that have been
considered in Apriori Algorithm are Minimum
Support and Confidence, when considering the Figure 1: Work flow of the proposed algorithm
real world scenarios, profit is the main variable
that should be considered. This paper focuses on a
new variable called profit of each product which Following are the steps involved in the proposed
calculates the profit margin with respect to the methodology. It explains how the proposed work
number of transactions other than the mean has been done.
constraint of minimum supports (Samaraweera et
al., 2014). Furthermore, proposed algorithm INPUT
controls the exponential growth of association A set of 𝑛 transaction data, each item, 𝑖 =
1, … , 𝑚 with support (𝑠𝑢𝑝𝑖 ), user defined
rules quantity as the size of the dataset increases.
minimum support (𝑚𝑠𝑖 ), profit (𝑝𝑟𝑜𝑓𝑖 ) and user
In addition to these reasons, Rare Item Problem is
defined minimum profit (𝑚𝑝𝑖 ), and a minimum
also addressed through this new approach. Since confidence value 𝜆 where,
rare items generate more profit than frequent 𝑐𝑜𝑢𝑛𝑡(𝑖) 𝑐𝑜𝑢𝑛𝑡(𝑖)∗𝑢𝑛𝑖𝑡 𝑃𝑟𝑜𝑓𝑖𝑡𝑖
𝑠𝑢𝑝𝑖 = and 𝑝𝑟𝑜𝑓𝑖 =
items, it is necessary to consider rare items as well. 𝑛 𝑛

130
 Proceedings in Computing, 9th International Research Conference-KDU, Sri Lanka 2016

STEP 01 Item A B C D E F G
In order to determine frequent items (𝐿1 ) which 𝒎𝒔𝒊 0.4 0.7 0.3 0.7 0.6 0.2 0.4
are highly consumable and profitable in vender’s 𝒎𝒑𝒊 1.0 2.2 2.0 1.9 2.5 1.4 2.0
perspective, should satisfy the following condition. Table 1: Minimum Support (𝑚𝑠𝑖 )and Minimum profit
𝑠𝑢𝑝𝑖 ≥ 𝑚𝑠𝑖 and 𝑝𝑟𝑜𝑓𝑖 ≥ 𝑚𝑝𝑖 (𝑚𝑝𝑖 )
The items are filtered based on the minimum
Consider following set of transaction, profit margin
support and minimum profit.
and support value in Table 2, Table 3 and Table 4
STEP 02
Generate the candidate set of k-itemsets (𝐶𝑘+1 ) respectively.
by pairing the items in 𝐿𝑘 , 𝑘 = 1,2,3, …. Then
compute the average minimum support of 𝑖 𝑡ℎ and
𝑗𝑡ℎ items (𝑎𝑚𝑠𝑖𝑗 ) and average minimum profit of ID Transaction
𝑖 𝑡ℎ and 𝑗𝑡ℎ items (𝑎𝑚𝑝𝑖𝑗 ) of each candidate item.
1 ABDG
So as to sort the highly consumable and profitable
k-itemsets (𝑅𝑘 ), individual support and profit of 2 BDE
items should be greater than or equal to average
minimum support and profit respectively. 3 ABCEF
𝑠𝑢𝑝𝑖 ≥ 𝑎𝑚𝑠𝑖𝑗 and 𝑠𝑢𝑝𝑗 ≥ 𝑎𝑚𝑠𝑖𝑗 with
4 BDEG
𝑝𝑟𝑜𝑓𝑖 ≥ 𝑎𝑚𝑝𝑖𝑗 and 𝑝𝑟𝑜𝑓𝑗 ≥ 𝑎𝑚𝑝𝑖𝑗 where
(𝑚𝑠𝑖 +𝑚𝑠𝑗 ) (𝑚𝑝𝑖 +𝑚𝑝𝑗 )
𝑎𝑚𝑠𝑖𝑗 = and 𝑎𝑚𝑝𝑖𝑗 = 5 ABCEF
2 2
6 BEG
STEP 03
The sorted k-itemsets(𝑅𝑘 ) is pruned to obtain 7 ACDE
𝐿𝑘+1, by comparing the support and profit of 𝑖 𝑡ℎ
8 BE
and 𝑗𝑡ℎ items together, with average minimum
support and profit respectively as follows; 9 ABEF
𝑠𝑢𝑝𝑖∪𝑗 ≥ 𝑎𝑚𝑠𝑖𝑗 and 𝑝𝑟𝑜𝑓𝑖∪𝑗 ≥ 𝑎𝑚𝑝𝑖𝑗
where 𝑖, 𝑗 are items. 10 ACDE

Table 2: Transaction Data

STEP 04
Repeat STEP 02 and STEP 03 until 𝐿𝑘+1 = ∅. Item A B C D E F G
prof(i) 1.2 2.4 2.4 2.0 2.7 1.2 2.1
STEP 05 Table 3: Profit margin of each item with respect to the
Construct the association rules for each k-itemset transactions in Table 2
in 𝐿𝑘 . Compute the confidence values of all
association rules and compare it with the user Item A B C D E F G
defined confidence value 𝜆. sup(i) 0.6 0.8 0.4 0.5 0.9 0.3 0.3
Table 4: support of candidate items
OUTPUT
STEP 01
Association rules of frequent itemsets which are
giving maximum profit to the business. Utilizing the Table 1 to Table 4, the frequent 1-
itemset 𝐿1 is generated, according to the first step
of the algorithm. The result of this step is in Table
IV. RESULTS AND DISCUSSION 5.

Suppose a supermarket tracks sales data for seven 𝑳𝟏 {A , B , C , E }

items denoted by ‘A’, ‘B’, ‘C’, ‘D’, ‘E’, ‘F’, ‘G’. The Table 5: The frequent 1-itemset
obtained results by the implementation of the
The items obtained in this result are relatively
proposed algorithm are discussed below.
frequent and relatively profitable as it has been
filtered from both profit and support (frequency)
INPUT
constraints.
The predefined minimum support and minimum
profit values are given in Table 1.

131
 Proceedings in Computing, 9th International Research Conference-KDU, Sri Lanka 2016

STEP 02 This association rule illustrates the reliability of the

profit gain and the pruning capacity of the
Candidate 2-itemsets (𝐶2 ) are generated from 𝐿1 .
proposed algorithm.
Table 6 consist of arithmetic mean on support and
profit of 𝐶2 calculated according to the formulas V. CONCLUSION
given in the algorithm.

𝑪𝟐 AB AC AE BC BE CE The extended Apriori algorithm generates rules to

𝒂𝒎𝒔𝒊𝒋 0.55 0.35 0.5 0.5 0.65 0.45 arrange floors and shelves of a supermarket based
𝒂𝒎𝒑𝒊𝒋 1.6 1.5 1.75 2.1 2.35 2.25 on the frequent items in the transaction database
Table 6: ams and amp of 𝐶2
which is insufficient to accomplish the
As supports of the two items in each item set in 𝐶2 requirement of the venders, maximization of profit
must be larger than or equal to the 𝑎𝑚𝑠𝑖𝑗 AND gain. The proposed algorithm in this research
profits of the two items in each item set in 𝐶2 must consists of a profit constraint with effect from the
be larger than or equal to the 𝑎𝑚𝑝𝑖𝑗 , 𝑅2 = {𝐵𝐸}. commencement of the process so as to generate
rules based on both frequent and rare items and
This step clearly depicts the prohibition of profit of itemsets. This newly acquainted
generating unnecessary rules as it has been constraint enhances the profit gain in a
filtered from arithmetic mean of profit and transaction. Simultaneously this profit constraint
support. facilitates the rare items without disturbing
frequent items. It has been inspected with
STEP 03
credible data sets and the outcomes conclude that
Since 𝑠𝑢𝑝𝑖∪𝑗 > 𝑎𝑚𝑠𝑖𝑗 AND 𝑝𝑟𝑜𝑓𝑖∪𝑗 > 𝑎𝑚𝑝𝑖𝑗 of BE the rules generated by the proposed algorithm
2-itemset 𝐿2 = {𝐵𝐸}. heightens the profit gain while pruning
unnecessarily generated rules. When negotiating
𝑹𝟐 BE with outsized real world data sets the results
𝒔𝒖𝒑𝒊∪𝒋 0.7 might vary depending on the predefined values.
𝒑𝒓𝒐𝒇𝒊∪𝒋 2.55
Therefore, the algorithm is subjected to further
𝒂𝒎𝒔𝒊𝒋 0.65
perfections to optimize the circumstances.
𝒂𝒎𝒑𝒊𝒋 2.35
Table 7:R 2 2-itemsets VI. REFERENCES
The outcome of this step illustrates the consistency of
the algorithm as it provides an itemset comprises with a
Agrawal, R., Srikant, R., 1994. Fast algorithms for
maximum profit and support.
mining association rules, in: Proceedings of the
20th VLDB conference, pp 487–499.
STEP 04
Annie M.C.L.C., Kumar D.A., 2011. Frequent Item set
Since 𝐿3 is null, the process terminates.
mining for Market Basket Data using K-Apriori
algorithm, in: International Journal of
STEP 05
Computational Intelligence and Informatics,
Association rules are formed for 𝐿2 . Volume 1, No. 1, pp.14-18.

𝐵 → 𝐸 and 𝐸 → 𝐵 are the generated association Annie M.C.L.C., Kumar D.A., 2012, Market Basket
Analysis For A Supermarket Based On
rules.
Frequent Itemset Mining, in: International
Calculate the confidence values of the above Journal of Computer Science 9.5.
association rules. Balaji Mahesh, Rao, V.R.k., Subrahmanya, G., 2013. An
Adaptive Implementation Case Study of
1. Confidence of 𝐵 → 𝐸 = 0.875
Apriori Algorithm for a Retail Scenario, in: a
2. Confidence of 𝐸 → 𝐵 = 0.777 Cloud Environment, ccgrid, pp.625629, 2013
Predefined confidence value 𝜆 = 0.8 13th IEEE/ACM International Symposium on
Cluster, Cloud, and Grid Computing, 2013.
∴ Final result generated by the proposed
algorithm is; 𝐵 → 𝐸 Bhandari, Akshita, Ashutosh Gupta, Debasis Das, 2015.
Improvised Apriori Algorithm Using Frequent

132
 Proceedings in Computing, 9th International Research Conference-KDU, Sri Lanka 2016

Pattern Tree For Real Time Applications In International Conference on Advances in

Data Mining, in: Procedia Computer Science 46 Recent Technologies in Communication and
(2015): 644-651. Computing, doi: 10.1109/artcom.2009.73.

Han, J., Kamber, M., 2001. Data Mining: Concepts and Samaraweera, W.J., Vasanthapriyan, S. and Oza, K.S.
Techniques, in: Morgan Kaufmann Publishers, (2011) Designing a multi-level support based
San Francisco, CA. association mining algorithm. Available at:
http://www.ijsrp.org/research-paper-
Han, J., Pei, H., Yin, Y., 2000. Mining Frequent Patterns 0414.php?rp=P282520 (Accessed: 2016).
without Candidate Generation, in: Proc. Conf.
on the Management of Data SIGMOD’00, ACM
Press, New York, NY, USA.

Kamruzzaman, S. M., Rahman, C. M., 2010. Text

Categorization Using Association Rule And
Naïve Bayes Classifier.

Larose, D.T., Larose, P.D.T., 2005. Discovering

knowledge in data: An introduction to data
mining. New York: Wiley-Interscience.

Liu, Bing, Wynne, H.s.u., Yiming Ma, 1999. Mining

Association Rules With Multiple Minimum
Supports. Knowledge Discovery & Data Mining
(KDD-99). San Diego, in: ACM SIGKDD
International Conference, 1999.

Liu, G., Huang, S., Lu, C. and Du, Y., 2014. An improved
k-means algorithm based on association rules,
in: International Journal of Computer Theory
and Engineering, 6(2), pp. 146–149. doi:
10.7763/ijcte.2014.v6.853.

Qiang Niu, Shi-Xiong Xia, Lei, Zhang, 2009. Association

Classification Based on Compactness of
Rules,in: WKDD 2009, Second International
Workshop on Knowledge Discovery and Data
Mining, pp. 245-247.

Rao, S., Gupta, R.,2012. Implementing Improved

Algorithm Over APRIORI Data Mining
Association Rule Algorithm, in: International
Journal of Computer Science And Technology
Mar. 2012, pp. 489-493.

Raorane A. A, Kulkarni R. V., Jitkar B.D., 2012.

Association Rule - Extracting Knowledge Using
Market Basket Analysis, in: Research Journal of
Recent Sciences, Vo11 (2)19.

Trikha, R. and Singh, J. (2014) ‘Improvement in Apriori

Algorithm with New Parameters’, International
Journal of Science and Research, 3(9).

Tassa, T., Open, T. and Road (2014) ‘Secure mining of

association rules inHorizontally distributed
databases’, IEEE Transactions on Knowledge &
Data Engineering, (4), pp. 970–983. doi:
10.1109/TKDE.2013.41.

Shah, K. and Mahajan, S. (2009) ‘Maximizing the

efficiency of parallel Apriori algorithm’, 2009

133

View publication stats