Data Mining and Visualization for Business Intelligence

This article discusses the practical work on classifier algorithms like Decision Tree, Naïve Bayes, K-Nearest Neighbour and their consequences on the vote.arff dataset using Weka tool. It also explains the confusion matrix and the advantages of each algorithm. The article is relevant for students studying Data Mining and Visualization for Business Intelligence.

Data Mining and Visualization for Business Intelligence

Table of Contents
1. Part – B Practical Part...........................................................................................................2
1.1 Classifier Algorithms......................................................................................................3
1.2 Practical Work on Decision Tree Algorithm................................................................4
1.3 Practical Work on K-Nearest Neighbour Algorithm...................................................7
1.4 Practical Work on Naïve Bayes Algorithm...................................................................9
1.5 Discussion and Consequences......................................................................................11
References.....................................................................................................................................13
1

1. Part – B Practical Part
This project is to analyse the vote.arff dataset done by using weka software. So, user needs
download the vore.arff file (Brownlee, 2018).
After, Open Weka Tool
Then, load the vote.arff file on weka tool. It is shown below.
2

Once dataset is loaded successfully, after applying the classifier algorithms. Basically,
the Weka tools is used to classify the vote data by using the following algorithms such as the
naïve Bayes, decision tree and the K-Nearest neighbour algorithm, for further investigation more
classifier algorithms will be used to compare the results.
1.1 Classifier Algorithms
The classifier algorithm is a precise way to deal with make classification models, starting
with the input data set. For instance, the neural systems, rule based classifiers, naïve Bayes
classifiers, decision tree classifiers and the support vector machines all represent the distinctive
method for taking care of the issue like classification. Each method gets learning algorithm for
distinguishing the model which is best for expressing the connection among the trait set along
with the class name of data. Hence, the primary target of learning algorithms refers to fabricating
the predictive model which foresees the class marks of the unknown records. Here, we are using
the following classifier algorithms like,
3

 Decision Tree
 Naïve Bayes
 K-Nearest neighbour
1.2 Practical Work on Decision Tree Algorithm
Decision Tree Classifier are the direct as well as generally utilized for the techniques like
classification, which uses straightforward idea for challenging the classifier related problem.
This algorithm presents the progression of effectively made investigations in terms of test
record’s characteristics. Every time it gets the answer, it receives class mark data from the
subsequent investigation in the record. As the rule, may decision trees can be built from a given
arrangement of properties. While tress’s portion are more exact than others, identifying ideal tree
is technically not feasible, because of the exponential size of search space. But, different
algorithms are created for building precise, but problematic, decision tree in a sensible measure
of time. Such algorithms usually makes use of insatiable system which grows the decision tree
with local ideal decision series related to ascribe and to utilize for dividing the data ("Chapter 3 :
Decision Tree Classifier — Theory – Machine Learning 101 – Medium", 2018).
4

5

In Decision tree classifier, the observed result can see in confusion matrix correctly classified
instance and incorrectly classified instances. The training set contains total 423instances
Correctly Classified Instances 284 67.1395 %
Incorrectly Classified Instances
139 32.8605
%
If we observe the above instances, out of 423 instances 284 instances are correctly classified
which is a good sign of the result and 139 instances are incorrectly classified.
Confusion Matrix
A B
147 89
50 137
1.3 Practical Work on K-Nearest Neighbour Algorithm
The K-Nearest Neighbours refers to a standout among the most common and basic
classification algorithm in the Machine Learning. This algorithm has managed learning space
6

and identifies the extreme application in the design acknowledgment, data interruption
recognition and in data mining. It is broadly expendable, all things considered, situations as it is
non-parametric, and this means that, it does not make any basic assumptions about the data
exchange. We are given some earlier data, which characterizes organizes into bunches
distinguished by an attributes ("K-Nearest Neighbours - GeeksforGeeks", 2018).
7

In K-nearest neighbour classifier, the observed result can see in confusion matrix
correctly classified instance and incorrectly classified instances. The training set contains total
423instances
Instances Classified Correctly 337
83.4158
percentage
Instances Classified Incorrectly
67 16.5842
percentage
If we observe the above instances, out of 404 instances 337 instances are correctly
classified which is a good sign of the result and 67 instances are incorrectly classified.
Confusion Matrix
A B
203 30
8

37 134
1.4 Practical Work on Naïve Bayes Algorithm
Naive Bayes is a classification algorithms. Customarily it expect that the data regards are
supposed, despite the fact that it numerical sources of information are upheld by accepting a
distribution ("How the Naive Bayes Classifier works in Machine Learning", 2018).
9

In naïve Bayes classifier, the observed result can see in confusion matrix correctly
classified instance and incorrectly classified instances ("How the Naive Bayes Classifier works
in Machine Learning", 2018). The training set contains total 418 instances
Instances Classified Correctly
356 85.1675
percentage
Instances Classified Incorrectly
62 14.8325
percentage
If we observe the above instances, out of 404 instances 337 instances are correctly
classified which is a good sign of the result and 67 instances are incorrectly classified.
Confusion Matrix
A B
157 13
49 199
10

1.5 Discussion and Consequences
Look at three classifier, Naïve Bayes utilizes a straightforward usage of the theorem
called Bayes Theorem where the earlier likelihood for each class is computed from the
preparation data and thought to be autonomous of one another. This is an unlikely assumption
since we anticipate that the factors will associate and be needful, despite the fact that this
suspicion will make the chances quick as well as simple for computing it. Indeed, also under the
unlikely suspicion, the Naive Bayes has appeared as an extremely compelling classification
algorithms. Naive Bayes figures the back likelihood for each class and makes an expectation for
class which has most astounding likelihood. In that capacity, it bolsters both double classification
and multi-class classification issues.
The decision trees will support the classification and regression issues. Decision trees are
all the more as of late alluded to as Classification and Regression Trees (CART). It works by
making a tree to assess the data example, begin at the foundation of the tree and moves down to
leaves (roots) till the point that an expectation could be done. The way toward making the
decision tree works by covetously choosing the split point which is best keeping in mind the end
goal to make forecasts and rehashing the procedure until the point when the tree is an established
("Decision Tree Classifier", 2018).
The k-nearest neighbour’s algorithms underpins both classification and regression. It is
additionally called kNN for short. It works by putting away the whole preparing dataset and
questioning it to find the k most comparable preparing designs when making an expectation. All
things considered, there is no model other than the simple preparing dataset and the main
algorithms performed is the questioning of the preparation dataset when a forecast is asked
(Srivastava, 2018). It is a straightforward algorithms, yet one that does not expect particularly
about the issue other than that the separation between data examples is significant in making
forecasts. Accordingly, it frequently accomplishes great execution.
At long last, the guileless Bayes classifier is utilized to give the powerful data
investigation contrasted with other classifier. It has simple and quick data examination to give
the productive framework to sharing the proof are set up and it is can acquire and successfully
utilize the best of accessible confirmation including lodging deals data assessment. In this Bayes
Classification, the common sense learning algorithms, watched data and earlier learning will be
joined. So we can get a valuable viewpoint data mining to comprehend the all the more learning
11

algorithms and effortlessly assess them. This algorithms additionally decides the probabilities for
theory expressly (Ray, 2018).
References
12

Brownlee, J. (2018). How To Compare the Performance of Machine Learning Algorithms in
Weka. Retrieved from https://machinelearningmastery.com/compare-performance-machine-
learning-algorithms-weka/
Chapter 3 : Decision Tree Classifier — Theory – Machine Learning 101 – Medium. (2018).
Retrieved from https://medium.com/machine-learning-101/chapter-3-decision-trees-theory-
e7398adac567
Decision Tree Classifier. (2018). Retrieved from
http://mines.humanoriented.com/classes/2010/fall/csci568/portfolio_exports/lguo/
decisionTree.html
How the Naive Bayes Classifier works in Machine Learning. (2018). Retrieved from
http://dataaspirant.com/2017/02/06/naive-bayes-classifier-machine-learning/
K-Nearest Neighbours - GeeksforGeeks. (2018). Retrieved from
https://www.geeksforgeeks.org/k-nearest-neighbours/
Ray, S. (2018). 6 Easy Steps to Learn Naive Bayes Algorithm (with code in Python). Retrieved
from https://www.analyticsvidhya.com/blog/2017/09/naive-bayes-explained/
Srivastava, T. (2018). Introduction to KNN, K-Nearest Neighbors : Simplified. Retrieved from
https://www.analyticsvidhya.com/blog/2018/03/introduction-k-neighbours-algorithm-
clustering/
13