Differential Diagnosis of Erythemato-Squamous Diseases using Ensemble of Decision Trees

Book Chapter
, Mohamed El Bachir Menai and Nuha Altayash, . 2014
Publication Work Type: 
Publishing City: 
Kaohsiung, Taiwan
Publisher Name: 
Book Title: 
Lecture Notes in Artificial Intelligence series (LNAI)
Publication Abstract: 

Abstract. The differential diagnosis of erythemato-squamous diseases
(ESD) in dermatology is a difficult task because of the overlapping of
their signs and symptoms. Automatic detection of ESD can be useful to
support physicians in making decisions if the model gives comprehensible
explanations and conclusions. Several approaches have been proposed to
automatically diagnosis ESD, including artificial neural networks (ANN)
and support vector machines (SVM). Although, these methods achieve
high performance accuracy, they are not attractive for dermatologists
because their models are not directly usable. Decision trees can be converted
into a set of if-then rules, which makes them particularly suitable
for rule-based systems. They have been already used for the diagnosis of
ESD. In this paper, we investigate the performance of boosting decision
trees as an ensemble strategy for the diagnosis of ESD. We consider two
decision tree models, namely unpruned decision tree and pruned decision
tree. The experimental results obtained on UCI dermatology data
set show that boosting decision trees leads to a relative increase in accuracy
that attains 5.35%. Comparison results with other related methods
demonstrate the competitiveness of the ensemble of unpruned decision
trees. It performs 96.72% accuracy, which is better than those of some
methods, such as genetic algorithms and K-means clustering.