Using C4.5 Algorithm in Classification of Asthma in Children for

Suggesting Best Possible Treatment

Ashish Jain and Swati Sharma

BSSS College, Bhopal, India b BSSS College, Bhopal, India

Keywords: Data Mining, C4.5 Algorithm, Pre-Processing, Classification, Decision Tree.

Abstract: Millions of children worldwide suffer from asthma, and finding the best therapy is critical for treating the

disease and improving the quality of life for those afflicted. Data mining is critical for detecting hidden

patterns and trends in massive datasets, such as those used in healthcare. It has been used to identify and treat

disorders including asthma. The C4.5 algorithm is a common decision tree technique that is employed in the

proposed work to build a decision tree for selecting the optimal asthma medication in children. It employs

three primary data mining steps: pre-processing, categorization, and decision tree. Finally, if the dependent

variable matched the provided conditions, the results were gathered using a decision tree. Healthcare

practitioners can make educated judgements by using data mining techniques.

1 INTRODUCTION

Data mining provides a mechanism for combining all

methodologies naturally, allowing them to emphasise

their strengths while concealing their limitations. As

more data is generated in databases, classification

analysis has emerged as a hot study area in data

mining. Today, there are numerous classification

algorithms accessible, including Decision Trees,

Bayesian classification based on statistics, neural

networks, and others (Alexander 2020).

However, it is important to apply such an

algorithm which can deal with all types of symptoms,

and thus helps in selection of the beat possible

treatments for asthma based on symptoms entered

into the database, for proper management of the

asthma treatment very well in childhood by

considering age groups such as 0-4, 4-8, and 8-12.

According to the prior idea, the article provides a

method based on a classification algorithm that makes

use of a decision tree (Zhang and Wu 2018). Human

processes are classified into two types in cognitive

psychology: primary cognition and secondary

cognition. Furthermore, the cognitive process utilises

a variety of techniques. For intricate cases or objects,

the most significant cognitive process of humans is to

first classify the items and then further cognize each

category in order to simplify the complicated things.

Similarly, while building an application-specific

algorithm for classification for asthma management

based on the number of symptoms presented in the

dataset, it is critical to be aware of the technique in

order to simplify things. After categorizing asthma

based on dependent and independent variables, there

is other classification also which is intrinsic and

extrinsic asthma, which is further subdivided into

severe and general asthma (Paul and Sherrif 2022).

Input for classification is in the form of a .csv file, in

which symptoms and their best possible Line of

Treatment are saved and retrieved from the database.

The section next describes a method of classification

of data mining that can be used to find the most

effective treatment among the many medications that

are available. It also describes how to choose the

selection variable. In particular, the classifier, testing

the options and various attribute classes, and so on. In

the final phase, the C4.5 algorithm is used to build a

decision tree and choose the best asthma therapies for

children (Berikov and Litvinenko 2020, Breiman et al

2020, Yoos and McMullen 2022).

2 CLASSIFICATION OF ASTHMA

AS PER AGE GROUPS

Asthma is a worldwide disease, and its incidence is

rising. It is predominantly a lung condition that

manifests as the following symptoms:

Jain, A. and Sharma, S.

Using C4.5 Algorithm in Classiﬁcation of Asthma in Children for Suggesting Best Possible Treatment.

DOI: 10.5220/0012610400003739

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Artiﬁcial Intelligence for Internet of Things: Accelerating Innovation in Industry and Consumer Electronics (AI4IoT 2023), pages 195-198

ISBN: 978-989-758-661-3

195

• Reversible airway blockage, either naturally

or by therapy

• Inflammation of the airways

• Hypersensitivity of the airways

Asthma is a disorder in which the airways in the

lungs tighten and swell. It is frequent in children and

teenagers. When a child has an asthma attack, his or

her lungs do not receive enough oxygen to breathe,

and they may cough or wheeze.

Based on age categories of 0-4, 4-8, and 8-12

years, the planned work will first identify asthma as

intrinsic and extrinsic, then as severe and general. It

employs a designed technique in which numerous

symptoms like as coughing, wheezing, shortness of

breath, chest tightness, and so on are recorded when

a .csv file format is formed. The classification method

is then applied to the entered data, and the

classification is completed in conjunction with the

decision tree. Having a diagnosed asthma, quantify

the symptoms over a period of time as depicted in the

Table 1.

Table 1: Asthma Symptoms.

3 ATTRIBUTE SELECTION FOR

ASTHMA DETECTION

The process of selecting attributes comprises

searching through all potential attribute combinations

in the data to discover which proportion of attributes

works best for prediction. Variable selection is a

challenging and critical topic in machine learning. In

classification jobs, it can lead to greater accuracy or

decreased computing costs.

This approach was used in this article to assign a

value to each group of attributes. The C4.5 algorithm,

which produces a decision tree, is used to determine

the value of the dependent and independent variables.

To build decision trees from a training data set, this

approach employs the concept of information gain. It

chooses symptoms as the data feature that best

divides its sample set into subsets enriched in one of

two classes.

Figure 1: Classification of asthma as per Age group.

Its criteria is the normalized information gain

(entropy difference) that occurs as a result of

choosing an attribute for data splitting. The attribute

with the largest normalized information gain is

chosen to decide.

3.1 Classification Criteria Based

Decision Tree

A decision tree is a predictive model that connects

observations about an item to conclusions about its

target value. It is also known as classification or

regression trees.

The pseudocode algorithm based on the

symptoms provided to the C4.5 algorithm for correct

and accurate decision tree is as follows:

1. Check for basic symptoms for asthma

2. For each indefinite symptom of asthma as

selection attribute A, find normalized

information gain from splitting on A

3. Let a_best be the attribute with highest

normalized information gain.

4. Create a decision node that splits on a_best

5. Recurse on the sub-lists obtained by splitting

on a_best and add those nodes as A

, A

etc.

6. Those will be children of node A which has

highest information gain.

For example:

Let the calculated information gain for 5 nodes will

be, A

= 0.98, A

= 0.86, A

= 0.5, A

= 0.74, and A

= 0.6

Among above the node having highest information

gain will be decision node like A

. According to age

group as a selection attribute, the proposed work got

following results for the experiments done on

symptoms table.

AI4IoT 2023 - First International Conference on Artiﬁcial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry

and Consumer Electronics

196

Table 2: Classification Criteria.

Table 3: Grading Table.

4 CONCLUSION

In the twenty-first century, database and internet

technology abilities have evolved fast. Meanwhile,

a Management Information System and a Network

Data Centre have been extensively used. Data

access, data querying, and statistics all develop

throughout time. However, the high layer's decision

analysis and knowledge discovery are still

immature, resulting in the phenomena of the

"Information Explosion" and the "Knowledge

Explosion". Data mining looks to be helping to

solve these problems. Even though numerous

sectors have investigated the classification

algorithm problem, no algorithm can successfully

handle a vast volume of data while also creating a

decision tree.

Based on the findings of the studies, the

suggested paper includes one datasheet that

provides the best feasible treatment for better

asthma management in children, which is the

primary purpose of this study. (See Table III.) * For

Children above 5 years only,

** For Children below 5 years

*** Evidence to date dose not supporting using

a third long-term control medication added to

inhaled corticosteroids and acting inhaled β agonists

Using C4.5 Algorithm in Classiﬁcation of Asthma in Children for Suggesting Best Possible Treatment

197

in order to avoid using systemic corticosteroid

therapy.

However, asthma can be very well managed

from childhood as it is well known that asthma

cannot be completely cured but can be controlled

through such activities.

REFERENCES

Alexander I, Morton H. (2020), An Introduction to Neural

Computing, 2nd edition Neural Networks at Pacific

Northwest National Laboratory

Zhang Wx, Wu W Z (2018), Rough Set: Theory and

Technique, Beijing Science Press

Paul E. Keller, Sherif Hashem (2022), A Novel Approach

to Modelling and Diagnosing the Cardiovascular

System, Proceedings of the workshop on

environmental and energy applications of neural

networks

Berikov, A. Litvinenko (2020), Methods for Statistical

Data Analysis with Decision Trees, Novosibirsk

Sobolev Institute of Mathematics, Methods for

statistical data analysis with decision trees, 3(12-16)

Breiman, J. Friedman, R. A. Olshen, C. J. Stone (2020),

Classification and Regression trees.

Yoos, H. L., & McMullen, A. (2022), Illness narratives of

children with asthma. Pediatric Nursing, 22 (285-290)

AI4IoT 2023 - First International Conference on Artiﬁcial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry

and Consumer Electronics

198