Differential Diagnosis of Erythematous Squamous Diseases With Feature Selection and Classification Algorithms

Differential Diagnosis of Erythematous Squamous Diseases With Feature Selection and Classification Algorithms

Aydın Çetin (Gazi University, Turkey) and Tuba Gökhan (Gazi University, Turkey)
Copyright: © 2018 |Pages: 27
DOI: 10.4018/978-1-5225-4769-3.ch005

Abstract

In this chapter, the differential diagnosis of erythematous diseases was determined using data mining and machine learning algorithms. In this chapter, data mining and its application to differential diagnosis of erythematous squamous diseases were discussed. A dermatology dataset from UCI Machine Learning Repository was used for the study. The dataset consists of 366 data items with 34 attributes. Initially, feature selection was made, and then classification was performed by using various algorithms. The number of attributes has been reduced from 34 to 19 as a result of the integration of the correlation-based filter methods and various heuristic search methods. The evaluation results show that Naive Bayes has 100% success rate in classification of psoriasis, seborrheic dermatitis, lichen planus, rose disease, chronic dermatitis, and pityriasis rubra pilaris diseases with 19 attributes selected with feature extraction algorithms.
Chapter Preview
Top

Data Mining

Data Mining is the process of obtaining previously unknown, valid and applicable information from large data sources. With data mining relationships between the variables can be deduced and future predictions can be made (Kaya & Özel, 2014). The process of obtaining raw data to knowledge using data mining is shown in Figure 1.

As shown in Figure 1, in the data mining process, the process begins with obtaining the data of the probing that is desired to be examined first. Noisy and inconsistent data is removed from the data set by data cleansing. After selecting and extracting the attributes on the data warehouse, the appropriate models for the data and probing are created, and the suitability and adequacy of these models are evaluated. Intelligent methods are applied to capture data patterns while models are being created. In the pattern evaluation step, interesting patterns representing information obtained according to predetermined criteria are defined. In the information presentation step, the obtained information is presented to the user (Kaya & Özel, 2014).

Figure 1.

Data mining processes

978-1-5225-4769-3.ch005.f01

Complete Chapter List

Search this Book:
Reset