[Paper Review] Evaluation of Predictive Data Mining Algorithms in Erythemato-Squamous Disease Diagnosis

Kwetishe Joro Danjuma, Adenike Osofisan|arXiv (Cornell University)|Jan 3, 2015

Artificial Intelligence in Healthcare43 references19 citations

TL;DR

This study evaluates Naive Bayes, Multilayer Perceptron, and J48 decision tree algorithms for diagnosing erythemato-squamous diseases using a dataset of 366 patient cases. Using 10-fold cross-validation, Naive Bayes achieved the highest accuracy of 97.4%, outperforming Multilayer Perceptron (96.6%) and J48 (93.5%), demonstrating its superior predictive performance in clinical data mining applications.

ABSTRACT

A lot of time is spent searching for the most performing data mining algorithms applied in clinical diagnosis. The study set out to identify the most performing predictive data mining algorithms applied in the diagnosis of Erythemato-squamous diseases. The study used Naive Bayes, Multilayer Perceptron and J48 decision tree induction to build predictive data mining models on 366 instances of Erythemato-squamous diseases datasets. Also, 10-fold cross-validation and sets of performance metrics were used to evaluate the baseline predictive performance of the classifiers. The comparative analysis shows that the Naive Bayes performed best with accuracy of 97.4%, Multilayer Perceptron came out second with accuracy of 96.6%, and J48 came out the worst with accuracy of 93.5%. The evaluation of these classifiers on clinical datasets, gave an insight into the predictive ability of different data mining algorithms applicable in clinical diagnosis especially in the diagnosis of Erythemato-squamous diseases.

Motivation & Objective

• To identify the most effective predictive data mining algorithms for diagnosing erythemato-squamous diseases.
• To evaluate the performance of Naive Bayes, Multilayer Perceptron, and J48 classifiers on a clinical dataset of 366 erythemato-squamous disease cases.
• To compare the predictive accuracy of these algorithms using standardized performance metrics and 10-fold cross-validation.
• To provide insights into the applicability of data mining techniques in improving clinical diagnosis of complex dermatological conditions.

Proposed method

• The study employed a dataset of 366 instances of erythemato-squamous disease cases for model training and evaluation.
• Three predictive data mining algorithms—Naive Bayes, Multilayer Perceptron, and J48 decision tree—were trained and tested on the dataset.
• 10-fold cross-validation was applied to ensure robust and unbiased performance estimation across all classifiers.
• Performance was evaluated using standard metrics such as accuracy, precision, recall, and F-measure to compare classifier effectiveness.
• The models were implemented using standard machine learning frameworks, with results aggregated across folds to determine overall performance.
• Comparative analysis was conducted to rank the classifiers based on their predictive accuracy and consistency.

Experimental results

Research questions

• RQ1 Which data mining algorithm demonstrates the highest diagnostic accuracy for erythemato-squamous diseases among Naive Bayes, Multilayer Perceptron, and J48?
• RQ2 How do the performance metrics of these classifiers compare when applied to a real clinical dataset of 366 cases?
• RQ3 What is the relative stability and reliability of each algorithm under 10-fold cross-validation?
• RQ4 Can data mining techniques significantly improve the diagnostic accuracy of erythemato-squamous diseases compared to traditional methods?

Key findings

• Naive Bayes achieved the highest accuracy of 97.4% in diagnosing erythemato-squamous diseases, making it the top-performing algorithm among the three tested.
• Multilayer Perceptron demonstrated strong performance with an accuracy of 96.6%, ranking second in predictive capability.
• J48 decision tree showed the lowest accuracy among the three, at 93.5%, indicating weaker generalization on the dataset.
• The comparative analysis confirmed that Naive Bayes offers the most reliable predictive performance for this clinical dataset.
• All three algorithms showed high overall accuracy, suggesting strong potential for data mining in dermatological diagnosis.
• The results indicate that algorithm choice significantly impacts diagnostic performance, with Naive Bayes being the optimal choice for this specific dataset.