Performance Analysis: Apriori Algorithm with Regression in Weka Tool

This report presents an experimental analysis of the Apriori algorithm and regression techniques implemented within the Weka tool. The study focuses on optimizing execution time when generating frequent patterns, strong rules, and maximal rules. The implementation uses synthetic and real datasets, including a supermarket dataset, to evaluate the performance of the algorithms under varying support and confidence levels. The report includes implementation snapshots, result tables, and graphs comparing the execution times of Apriori and improved Apriori algorithms, along with Apriori with regression. Key findings highlight the impact of different support and confidence values on the execution time, with a specific emphasis on how linear regression can reduce the execution time for item sets with a predicted confidence of zero. Additionally, the report touches upon big data concepts like clustering and association rule mining, with a focus on minimizing costs associated with big data processing.

CHAPTER 1
EXPERIMENTAL RESULTS
5.1 IMPLEMENTATION OF PROPOSED WORK
In our thesis a weka tool of apriori and apriori algorithm with regression technique has been
actualized in weka tool to analyze the execution time of generating the frequent pattern,
strong rules, close rules and maximal. With the help of this application, we can easily find
the time which has taken to discover the frequent designs through apriori algorithm as well
as apriori with regression technique, in this weka tool we have found the frequent patterns
on dynamic value of support and confidence. By this work we can easily find out which
algorithm takes less execution time.
5.1.1 WEKA TOOL
Weka tool 4.5 was discharged on 15 August 2012; an arrangement of new or enhanced
features were included into this version. The Weka tool 4.5 will also run on Windows
Vista or later. The Weka tool 4.5 uses Common Language Runtime 4.0, with some
additional runtime features.
. Weka tool 4.5 is supported on Windows Vista, Server 2008, 7, Server 2008 R2, 8, Server
2012, 8.1 and Server 2012 R2. Applications utilizing Weka tool 4.5 will likewise keep
running on computers with Weka tool 4.6 installed, which supports additional operating
systems.
.NET Weka tool is available for building Metro-style apps using C# or Visual Basic.
CORE FEATURES
 Ability to restrict to what extent the regular expression engine will endeavor to
determine a regular articulation before it times out.
 Ability to characterize the culture for an application domain.
1

 Console bolster for Unicode (UTF-16) encoding.
 Support for forming of social string requesting and comparison data. Better
execution while recovering resources.
 Native help for Zip compression (previous versions supported the compression
algorithm, but not the archive format).
 Ability to customize a reflection context to override default reflection behavior through
the Custom Reflection Context class.
 New asynchronous elements were added to the C# and Visual Basic languages. These
components include a task-based model for performing offbeat
operations, implementing futures and promises.
5.1.2 IMPLEMENTATION SNAPSHOT
 We will create the weka tool frame work with the help of regression and apriori
calculation to design the basic generating frequent use of data set values.
 In this frame work we will define some selected item set for implementation of
synthesis data set. In this, we compare the values of apriori and regression on different
confidence and support values.
Fig.5.1 Regression implemented for optimization weka snap shot
2

Fig 5.1.1 snap shot of regression implemented in weka tool for time consuming.
Fig. 5.2 Snap Shot of Time Optimization through Apriori on support and confidence
Weka snap shot
3

Fig. 5.3 Snap Shot of Time Optimization through improve Apriori support and
confidence using weka snap shot
.
4

5.5 Snap Shot of Time Optimization through Apriori with regression on same support
and confidence
5

Fig. 5.2 Snap Shot of Time Optimization through Apriori on help and certainty
Weka depiction with data set upload
Super market data set upload in weka tool for analyzing
5.2 DATASET OF THE PROPOSED APPROACH
6

Remembering the true objective to assess the three proposed approaches, only single class of
dataset is utilized. They are,
 Synthetic dataset
 Supermarket dataset
 Real dataset
 Synthetic Dataset
Synthetic databases were created utilizing the MS-EXCEL. The data impersonates the
exchanges in a retailing domain. The execution of the algorithms is shown for manufactured
datasets. The accompanying datasets were produced with the true goal of assessing the three
proposed approaches,
 super market data set
 We have taken the data set of Super-market, in which there are 106 attributes
and 4627 instances.
 We can convert the real database into synthetic database on the basis of apriori
algorithm format given by a, b………..z values.
 Example:
a=bread
b= butter and so on.
 REAL DATASET
Association Rule mining likewise assumes significant part in finding the frequent example
from Real databases, overview data from Real research, data about the items and frequent
patterns, data containing, data concerning. This can confidence basic leadership concerning
the choice of items to be developed in a specific way in less time consuming.
5.2.1 ASSESSMENT PARAMETERS
7

The execution of the proposed Market Basket methodology is assessed utilizing the
parameter like,
 Execution time
Execution time
Execution time demonstrates the time occupied by the standard Apriori procedure and
regression strategies to execute the procedure in all datasets. The technique which takes less
time in execution is best. This parameter is critical in Market Basket Investigation as
through this one can arrive at a decision about the relative execution of the proposed
methods. These performance values cause changes in the advancement of Super Market.
Assessment Parameters
The execution of the proposed Market Basket Investigation methodologies are assessed
utilizing the parameters like,
 Accuracy,
 Area Under the Curve :
 Sensitivity and
 Specificity
 Execution time
5.3 RESULT ANALYSIS
In this section, we have examined about the results like tables, graphs on different types of
support and confidence estimations of proposed work in different ways.
5.3.1 SIMULATION PARAMETERS
8

1. In this table, there are distinctive time of execution (in sec) of both algorithm i.e., apriori
calculation and apriori with regression technique on different support count and different
confidence.
Results snap shots
Table 5.1 Comparison of Time Execution in Apriori and Improve Apriori with
supportand different Confidence using weka tool
Algo Sec
Apriori Algo 115
Improve apriori 76
Support Count
Min_sup=0.15
Confidence
Min_conf=0.9
9
Size of
large
item set
L(1) L(2) L(3) L(4) L(5)
Apriori 41 347 855 617 105
Improved
apriori
34 173 226 96 10

Min_sup=0.15 Min_conf=0.9
Fig. 5.6 Time execution with different support and different confidence.
In this result analysis we will create the graph based on different support and different
confidence rules. As well as we have used the dot net frame work for implementation of this
time execution result. We analysis the linear regression technique reduced the time of
execution of those item set values whose confidence to be predict should be zero.
Time of execution in sec Sec
Aprioiri algorithm 115
Improve aprioiri 76
10

Fig.5.7 Time optimization of Apriori with Improve Apriori in sec
Table 5.2 Comparison of Time Execution in Apriori and improve apriori with different
support and same Confidence
Algo Sec
Apriori Algo 102
Apriori Imrove
55
Support Count
Min_sup=0.16
Confidence
Min_conf
=0.9
Size of large item-
sets
L(1) L(2) L(3) L
Apriori 38 322 707 4
11

Improved Apriori 31 148 78 5
Fig. 5.8 Time execution with diverse support and same confidence.
In this result analysis we will create the graph in light of different support and same
confidence rules. And in addition we have used the dot net frame work for implementation
of this time execution result. We analysis the linear regression technique reduced the time of
execution of individuals item set values whose confidence to be predict should be zero.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Fig.5.9 Time optimization of Apriori with Improve on different support and same
confidence
12

Time execution of different support and same confidence
5.3 Comparison of Time Execution in improve Apriori and Apriori with Regression on same
support and Confidence
Algo Sec
Apriori Algo 65
Apriori with Regression 41
Support Count 0.16
Confidence 0.9
13

Fig. 5.10 Time execution with same support and different confidence.
14
Time of Execution (In msec)
REGRESSION 45
Improved_Apriori 76

shown improve apriori time of execution values with regression approach
Results of improve apriori with regression on different support and same confidence
In this outcome analysis we will create the graph based on same support and different
confidence rules. As well as we have used the dot net frame work for implementation of this
time execution result. We analysis the linear regression technique reduced the time of
execution of those thing set values whose confidence to be predict should be zero.
15
Algorithm parameters
Improve apriori 60
regression 39
support 0.16
confidence 0.9

Fig.5.11 Time optimization of Apriori with Regression on same support and different
confidence
Table 5.4 Comparison of Time Execution in Apriori and Apriori with Regression on same support and
same Confidence
Algo Sec
Apriori Algo 281
Apriori with Regression 126
Support Count 2-6
Confidence 2-6
16

Fig. 5.12 Time execution with same support and same confidence.
In this outcome analysis we will create the graph based on same support and same
confidence rules. As well as we have used the dot net frame work for implementation of this
time execution result. We analysis the linear regression technique reduced the time of
execution of those thing set values whose confidence to be predict should be zero.
Fig.5.13 Time optimization of Apriori with Regression on same support and same
confidence
The above graph shows the difference of time on same set support and same confidence
value in between the Apriori algorithm and implemented algorithm i.e. Apriori with
regression algorithm in view of Table 5.4.
17

CHAPTER 2
Report
Report on features , classes and algorithm in big data
Big data preparing is an appeal range which forces a substantial weight on calculation,
correspondence, capacity in data focuses, which causes significant operational cost to data
focus supplier. So limiting expense has turn into an issue for the up and coming big data.
Unique in relation to traditional cloud benefit one of the fundamental component of the big
data benefit is the tight coupling amongst data and calculation, as calculation errand can be
directed just when the relating data are accessible. Thus, three factors that is
communicational cost, computational cost, operational cost impacts the use cost of data
focuses. So keeping in mind the end goal to limit the cost clustering is utilized. Clustering
bunches a chosen objects into classes of comparable items. Highlight Selection Removes
Irrelevant Features-it happens in the cluster preparing (planning calculation) Redundant
Features its happens in the bunch arrangement (data-driven calculation) joint optimization–
18

2 stages Features isolated into clusters(subsets) MST Cluster delegates are chosen Efficient,
Effective, Free. In view of these criteria, an element clustering in light of determination
calculation is proposed and tentatively assessed for a specimen disease dataset. This work
finds the successful qualities utilized and expels repetition.
Introduction
Big data usually includes data set with sizes beyond the ability of commonly used software
tools to capture, manage and process data within a tolerable elapsed time. Its size is
constantly moving target as of 2012 ranging from a few Dozen of terabyte to many
petabytes of data "massively parallel software running on tens, hundreds, or even thousands
of servers". Advantages: Big data is timely, accessible, trustworthy, relevant and secure.
 Reduced maintenance cost
 Big data tools allow us to identify the threats that we face internally
. It keeps data safe.
Association rule mining Prune step procedure
Prune Step: Any (k-1)-itemset that is not frequent cannot be a subset of a frequent k-itemset
1.Ck: Candidate itemset of size k
2.Lk: frequent itemset of size k
3.L1= {frequent items};
4.for(k=1; Lk!=0; k++) do begin
5.Ck+1= candidates generated from Lk;
6. for each transaction T in database do increment count of all candidates in Ck+1 that are
contained in T
Lk+1= candidates in Ck+1 with min_support
7. end .
proposed problem
. PROBLEM JUSTIFICATION
19

Frequent Pattern Mining is an important knowledge discovery task and it has been a focus
theme in knowledge discovery research. Frequent Pattern Mining over large database is
fundamental to many knowledge discovery applications. One of the main issues in Frequent
Pattern Mining is Sequential Pattern Mining retrieved the relationships among objects in
sequential dataset[2]. Apri ori All is a typical algorithm to solve the problem in Sequential
Pattern Mining but its complexity is so high and it is difficult to apply in large datasets.
Recently, to overcome the technical difficulty, there are a lot of researches on new
approaches as follow
1. Custom built Apriori algorithm
2. Modified Apriori algorithm
3. Frequent Pattern-tree and its development
4. Integrating Genetic algorithms
5. Rough set Theory/ Dynamic Functions
BIG DATA ANALYTICS TOOLS
There are varieties of applications and tools developed by various organizations to process
and analyze Big Data. The Big Data analysis applications support parallelism with the help
of computing clusters. These computing clusters are collection of hardware connected by
Ethernet cables. The following are major applications in the area of Big Data analytics.
Map Reduce method
The framework is divided as following:
Map: A function that parcels out work to different nodes in the distributed cluster
Reduce: A function that collects the work and resolves the results into a single value.
Frequent item set using common diviser method: big data set
20

The proposed method depends on calculating GCD between the prime multiplications of
each subset resulting from the partitioning and calculating cross GCD among subsets.
Before starting the computation of itemset frequencies, a pool of subsets is created- as
illustrated in the previous section- to give the distributed nodes, ALUs or processor cores the
ability to handle subsets in parallel.
For each subset we start calculating GCD using Euclidean method. The Euclidean algorithm,
is an efficient method for computing the greatest common divisor (GCD) of two numbers. It
is based on the principle that the greatest common divisor of two numbers does not change if
the larger number is replaced by the difference between it and the smaller number.
In this section, the standard Apriori Algorithm (AA) and the proposed three methodologies
are assessed. The exhibitions of the algorithms have been assessed utilizing the different
parameters.
The proposed methodologies are,
 Apriori Algorithm ,
 Regression modeling technique.
Frame work of dot net is utilized for the calculation of the numerical examination and is
considered as a fourth-era programming dialect Frame work of dot net is business Matrix
Laboratory bundle which works as an intelligent programming environment. It is an
establishment of the Mathematics Department software lineup and is likewise accessible for
PC's and Macintoshes and might be found . Consequently for this examination Frame work
of dot net has been thought about and all the three strategies have been actualized utilizing
Frame work of dot net
C denotes the set of candidate k-itemsets and k F denotes the set of frequent k-itemsets:
 The algorithm initially makes one pass over the data set to determine the support of
every item. Upon completion of this step, the set of all frequent one-itemsets one F
will be celebrated (steps 1 and 2).
 Next, the algorithm can iteratively generate new candidate k-itemsets exploitation
the frequent (k − 1) - itemsets found within the previous iteration (step 5). Candidate
generation is enforced employing a operate known as apriorigen.
 To count the support of the candidates, the algorithm has to build a further leave out
the info set (steps 6–10). The set operate is employed to see all the candidate itemsets
in k C that are contained in every dealings t.
21

 After tally their supports, the algorithm eliminates all candidate itemsets whose
support counts are but min_sup (step 12).
 The algorithm terminates once there aren't any new frequent itemsets generated, i.e.,
K F (step 13). The frequent itemset generation a part of the Apriori algorithm has 2
necessary characteristics.
In principle, there are many ways to come up with candidate itemsets. The subsequent may
be a list of needs for a good candidate generation procedure:
1) It ought to avoid generating too several needless candidates. A candidate itemset makes
no sense if a minimum of one in every of its subsets is sporadic. Such a candidate is bound
to be sporadic in keeping with the antimonot one property of support.
2) It should make sure that the candidate set is complete, i.e., no frequent itemsets are
unnoticed by the candidate generation procedure. to make sure completeness, the set of
candidate itemsets should subsume the set of all frequent itemsets.
3) It shouldn't generate identical candidate itemset over once. Generation of duplicate
candidates ends up in wasted computations and therefore ought to be avoided for potency
reasons.
Flowchart
YES
22
Generate k- itemset
candidates (join step)
Check support (pruning)
Frequent Itemsets
Regression
Set of k-1
frequent
itemsets
is Φ?

NO
Apriori algorithm with Regression
Conclusion
we explore pattern mining and its algorithm which are used to find patterns by integrating
structured and unstructured data together using Map Reduce framework. Sequential pattern
mining methods have been used to analyze this data and identify patterns. Such patterns
have been used to implement efficient systems that can recommend based on earlier
pragmatic patterns, help in making predictions, to improve usability of system, detect events
and in general help in making strategic product decisions. We have also seen applications of
pattern mining in variety of domain. Apart from this, new sequential pattern mining methods
may also be developed to handle special scenarios of colossal patterns, approximate
sequential patterns and other kinds of sequential patterns specific to the applications.
References:
1. Quan Fang, Jitao Sang, Changsheng Xu, Fellow, IEEE, and Yong Rui, Fellow,
IEEE,” Topic-Sensitive Influencer Mining in Interest-Based Social Media Networks
via Hypergraph Learning”, IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 16,
NO. 3, APRIL 2014.
2. Rahul Ramachandran, John Rushing, Amy Lin, Helen Conover, Xiang Li, Sara
Graves, U. S. Nair, Kwo-Sen Kuo, Member, IEEE, and Deborah K. Smith,” Data
Prospecting–A Step Towards Data Intensive Science”, IEEE JOURNAL OF
SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE
SENSING, VOL. 6, NO. 3, JUNE 2013
3. Tao Gu, Member, IEEE, Liang Wang, Student Member, IEEE, Zhanqing Wu,
Xianping Tao, Member, IEEE, and Jian Lu,“ A Pattern Mining Approach to Sensor-
23
Stop