Classification on multi label dataset using rule mining technique

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
_______________________________________________________________________________________
Volume: 03 Issue: 06 | Jun-2014, Available @ http://www.ijret.org 154
CLASSIFICATION ON MULTI-LABEL DATASET USING RULE
MINING TECHNIQUE
Ravi Patel1
, Jay Vala2
, Kanu Patel3
1
Information Technology, GCET
2
Information Technology, GCET
3
Information Technology, BVM
Abstract
Most recent work has been focused on associative classification technique. Most research work of classification has been done on
single label data. But it is not appropriate for some real world application like scene classification, bioinformatics, and text
categorization. So that here we proposed multi label classification to solve the issues arise in single label classification. That is
very useful in decision making process. Multi-label classification is an extension of single-label classification, and its generality
makes it more difficult to solve compare to single label. Also we proposed classification based on association rule mining so that
we can accumulate the advantages of both techniques. We can get the benefit of discovering interesting rules from data using rule
mining technique and using rule ranking and rule pruning technique we can classified that rules so that redundant rules can be
reduced. So that Here proposed work is done on multi label dataset that is classified using rule mining algorithm. Here proposed
approach is an accurate and effective multi label classification technique, highly competitive and scalable than other traditional
and associative classification approaches.
--------------------------------------------------------------------***-----------------------------------------------------------------
1. INTRODUCTION
Data mining functionalities are used to specify the kind of
patterns to be found in data mining tasks.[1,2] In general,
data mining tasks can be classified into two categories:
descriptive and predictive. Descriptive mining tasks
characterize the general properties of the data in the
database. Predictive mining tasks perform inference on the
current data in order to make predictions. Various types of
task of data mining are: Classification, Clustering,
Association Rule Discovery, Regression etc.
2. ASSOCIATION RULE MINING
Association rule mining is the discovery of association rules.
Association rule mining, one of the most important and well
researched techniques of data mining, was first introduced in
[3]. It studies the frequency of items occurring together in
transactional databases, and according to support count,
detects the frequent item sets. Another threshold,
confidence, which is the conditional probability than an item
appears in a transaction when another item appears, is used
to pinpoint association rules. Association rule mining is used
for market basket analysis. For example, it could be useful
for the OurVideoStore manager to know what movies are
often rented together or if there is a relationship between
renting a movies and buying chips or popcorn. The
discovered association rules are of the form: A B [s,c],
where A and B are conjunctions of attribute value-pairs, and
s (for support) is the probability that A and B appear
together in a transaction and c (for confidence) is the
provisional probability that B appears in a transaction when
A is in transaction. For example, the association rules:
Rent Type(X, “game”) ^Age(X, “13-19”) Buys(X,
“chips”) [s=2% ,c=55%] would indicate that 2% of the
transactions considered are of customers aged between 13
and 19 who are renting a game and buying a chips, and that
there is a certainty of 55% that teenage customers who rent a
game also buy chips.
Two main algorithms: apriori algorithm and FP-Growth
algorithm
Here we are using FP-Growth algorithm instead of apriori
for associative classification. some advantages of FP-
Growth :Uses compact data structure ,Eliminates repeated
database scan[12]
3. CLASSIFICATION
In classification [4], by the help of the analysis of training
data we develop a model which used to predict the class of
objects whose class label is unknown. The model is trained
so that it can distinguish different data classes. The training
data is having data objects whose class label is known in
advance.
Classification analysis is the Also known as supervised
classification, uses given class labels to order the objects in
the data collection. Classification approaches normally use a
training set where all objects are already associated with
known class labels. Whilst single-label classification, which
assigns each rule in the classifier the most obvious class
label, has been widely studied [9] little work has been
conducted on multi-label classification. The classification
algorithm learns from the training set and builds a model.
This model is used to classify new unclassified data objects.
For example, after starting a credit policy, the
OurVideoStore manage rs could analyze the customers’

_______________________________________________________________________________________
behaviors of their credit, and label accordingly the
customers who received credits with three possible labels
“safe”, “risky” and “very risky”. The classification analysis
would generate a model that could be used for acceptance or
rejection of credit requests in the future. Many techniques
for classification are neural network [8], rule based
classifier[5], Bayesian network (BN)[6], Decision tree[7].
3.1 Associative Classification
Recent studies propose the extraction of a set of high quality
association rules from the training data set which satisfies
certain user-specified frequency and confidence thresholds.
Effective and efficient classifiers have been built by careful
selection of rules, e.g., CBA [14], CAEP [15], and ADT
[16]. Such a method takes the most effective rule(s) from
among all the rules mined for classification. Since
association rules explore highly confident associations
among multiple variables, it may overcome some constraints
introduced by a decision-tree induction method which
examines one variable at a time. Extensive performance
studies [ 14, 15, 16] show that association based
classification may have better accuracy in general.
3.2 Multi-Label Classification [13]
For the growing quantity of data in fields such as
bioinformatics, scene classification and text categorization,
multi-label classification is more appropriate In multi-label
classification is a variant of the classification problem where
multiple target labels must be assigned to each instance.
Multi-label classification should not be confused with
multiclass classification, which is the problem is
categorizing instances into more than two classes.
There are two main methods for tackling the multi-label
classification problem. Problem transformation methods and
algorithm adaptation methods. Problem transformation
methods transform the multi-label problem into a set of
binary classification problems. Algorithm adaptation
methods adapt the algorithms to directly perform multi-label
classification.
3. PROPOSED WORK
Here proposed work consists of following steps:
1. Discover the set of frequent items.
2. Rule generation using threshold value of
confidence and support.
3. Re-rank the execution order of rule using rule
ranking technique.
4. Prune the redundant rule using some pruning
technique.
5. Generate prune rule.
6. Predict test case using the prune rule.
In Step 3 Two Techniques is used to improve the existing
algorithm. There are two techniques is used in
implementation of the proposed algorithm which are
described below.
1. Rule Pruning Technique
2. Rule Ranking Technique
Before pruning starts the rules must be sorted in according
to descending order in confidence, support, and number of
items in the rule antecedent.
Rule Ranking Technique of CaR is conducted according to
confidence in descending order. Then, CaR with the same
confidence are ranked according to the support value in
descending order. if the CaR with the same confidence level
as well as same support level they rank according to rule
length with long rule to short rule.
Rule Pruning Technique Here find the rule which satisfy
the if rule on right hand side (consequent) is the class label,
and the rule on left hand side (antecedent) is attribute
values[10,11].
Data Set format: Suppose One Dataset Weather has
following nominal attributes with values.
Data Set Weather:
(a). Data set with nominal attributes (b). Converted into numeric data set
Data Set Format
Here we have to take unique numeric value for every value
of attributes.
 Outlook=Sunny=1, Outlook=Overcast=2,
Outlook=Rainy=3
 Temperature=hot=4, temperature=mild=5,
temperature=cool=6
 Humidity=high=7, humidity=normal=8
 Windy=TRUE=9, windy=FALSE=10
 Play=yes=11, play=no=12
3.1 Implementation
For that implementation we are not using weka tool because
there is no provision in weka to set more than one attribute
as a class-label. Means weka tool is not supporting to
multilabel classification. That’s why here we are using net
beans for generating association rule from FP-Growth
algorithm.

_______________________________________________________________________________________
3.2 Function of Proposed Algorithm
In Existing Algorithm there are the functions which are used
to generate association rules those are given below.
 Outputsrules() is used for calling Outputsrule().
 Outputsrule() is used for displaying the rules.
 OutputsItemSet() is used for rule format how will
be display from itemset.
 InserrulesintoRulelist() is used for inserting rules as
per Descending order of confidence.
Changes made in the functions of proposed algorithm:
The Function Outputsrules() which is used to generate
Association rules based on threshold of Confidence value
and Support value Parameter of the algorithm.
So, in this proposed pruning method we have implemented
one new function Outputsrules1() for generating association
rules which satisfy our condition :
Condition: Right side of the rule:-Class Attribute
Left Side of the rule:-Simple Attribute
In this Outputsrule1() function First of all we are checking
all the attribute at the right side of the rule if right side of the
rule is class label then this rule will be consider as a
classified rule. If the rule satisfy above condition then we
will check left hand side of the rule if left hands side of rule
have Simple attribute then this rule will be generate after
this pruning step Using InserrulesintoRulelist(), Rules are
sorted according to Descending order of Confidence value.
if Confidence is same then it will check Support of the rule,
if support of the rules are also same then it will check for the
no of items in antecedent of the rule. Means according to
length rules are sorted and pruned.
4. RESULTS AND ANALYSIS
Here we have taken two data sets for associative
classification. We have shown graphs for no. of rules
reduction and accuracy before and after applying our
proposed Classification approach.
Number of Rules with Different Confidence Level
Table 1 Accuracy with different confidence level
Confidence
60 70 80 90 100
Before After Before After Before After Before After Before After
DataSet1 203 75 159 58 154 58 153 57 153 57
DataSet2 1087 342 1052 342 1052 342 818 171 818 171
Table 2 Here graphs shows no. Of rules reduction and accuracy
Confidence
60 70 80 90 100
Before After Before After Before After Before After Before After
DataSet1 51.72 72.42 51.72 72.42 51.72 72.42 51.72 72.42 51.72 72.42
DataSet2 20.83 77.08 20.83 77.08 20.83 77.08 16 20.83 16 20.83

_______________________________________________________________________________________
These results show that if value of confidence is increase
then the numbers of rules are decreased but also the
accuracy of the algorithm is decrease. After 80% confidence
in dataset2 the accuracy of the algorithm is decreased
drastically and in dataset 1 no more reduction in the number
of rules after confidence level 80%. So the better value of
confidence for the datasets is 80%.
5. CONCUSIONS AND FUTURE WORK
Our Proposed work of Multi-label Associative Classification
algorithm uses two methods called rule ranking and rule
Pruning, for ranking and removing unnecessary rules which
are generated by the FP-Growth association rule mining
algorithm. So, we can able to increase the accuracy of FP-
Growth algorithm. Using this technique we are able to
overcome the problems of single label classification and
redundant rules generation.
Our proposed multi label classification algorithm could be
improve for executing dataset which have character or
Nominal attribute. And also improve the accuracy of this
algorithm by applying heuristic search methods of artificial
intelligence to discover more informative rules Or generated
rule could be passed in the genetic algorithm as an initial
population to achieve more accuracy.
REFERENCES
[1] Shelly Gupta , Dharminder kumar ,Anand Sharma
“data mining classification techniques applied for
breast cancer diagnosis and prognosis”, ijcse vol. 2
no. 2 ,2011
[2] Tipawan Silwattananusarn and Dr.
KulthidaTuamsuk “Data Mining and Its Applications
for Knowledge Management : A Literature Review
from 2007 to 2012” IJDKP Vol.2, No.5, September
2012
[3] Agrawal, R., Srikant, R.: Fast algorithms for mining
association rule. In: Proceedings of the 20th
International Conference on Very Large Data Bases,
pp. 487–499 (1995)
[4] R. Duda and P. Hart. Pattern Classification and
Scene Analysis. JohnWiley & Sons, 1973
[5] G.L. Pappa and A.A. Freitas, Automating the Design
of Data Mining Algorithms. An Evolutionary
Computation Approach, Natural Computing Series,
Springer, 2010
[6] G.F. Cooper, P. Hennings-Yeomans, S.Visweswaran
and M. Barmada, “An Efficient Bayesian Method for
Predicting Clinical Outcomes from Genome-Wide
Data”, AMIA 2010 Symposium Proceedings, 2010,
pp. 127-131
[7] M. Garofalakis, D. Hyun, R. Rastogi and K. Shim,
“Building Decision Trees with Constraints”, Data
0
50
100
150
200
250
60 70 80 90 100
Noofrules
Confidence
Dataset1
Before
After
0
200
400
600
800
1000
1200
60 70 80 90 100
Noofrules
confidence
Dataset2
Before
After

_______________________________________________________________________________________
Mining and Knowledge Discovery, vol. 7, no. 2,
2003, pp. 187 – 214
[8] Y. Singh Y, A.S. Chauhan, “Neural Networks in
Data Mining”, Journal of Theoretical and Applied
Information echnology, 2005, pp. 37-42
[9] B.Santhosh Kumar, K.V.Rukmani .,“Implementation
of web usage mining using APRIORI and FP growth
algorithms” IJANA'2010
[10] Fadi Thabtah,Wa'el Hadi,Hussein Abu-Mansour,L.
McCluskey "A New Rule Pruning Text
Categorisation Method ,7th International Multi-
Conference on Systems, Signals and Devices,2010.
[11] Yuhanis Yusof, Mohammed Hayel Refai "MMCAR:
Modified Multi-class Classification based on
Association Rule"978-1-4673-1090-1/12/2012 IEEE
[12] B.Santhosh Kumar, K.V.Rukmani .,“Implementation
of web usage mining using APRIORI and FP growth
algorithms” IJANA'2010
[13] Grigorios Tsoumakas, Ioannis Katakis:"Multi-Label
Classification: An Overview “International Journal of
Data Warehousing and Mining, 3(3), 1-13, July-
September 2007
[14] B. Liu, W. Hsu, and Y. Ma. Integrating classification
and association rule mining. In KDD’98, New York,
NY, Aug.1998.
[15] G. Dong, X. Zhang, L. Wong, and J. Li. Caep:
Classification by aggregating emerging patterns. In
DS’99 (LNCS1721), Japan, Dec. 1999.
[16] K. Wang, S. Zhou, and Y. He. Growing decision tree
on support-less association rules. In KDD’00,
Boston, MA,Aug. 2000.

Classification on multi label dataset using rule mining technique

More Related Content

What's hot (18)

Viewers also liked (12)

Similar to Classification on multi label dataset using rule mining technique (20)

More from eSAT Publishing House (20)

Recently uploaded (20)

Classification on multi label dataset using rule mining technique