Improving the classification of chronic diseases using the naive Bayes algorithm

Abstract EN

Manual classification of disease into different classes based on residential patient areas is a tedious task and may vary depending on the scenario studied.

Therefore, the standard Naïve Bayesian classification algorithm was used to classify the disease based on several characteristics that represent their residential areas and their numbers, and for quick and easy prediction also in the process of classifying the areas most susceptible to chronic disease, and the importance of the study lies in reducing the normal manual work.

A methodology focused on the problem of extending the traditional naive Bayesian model was used to classify uncertain data.

The main problem of the naive Bayes model is estimating the conditional probability of the class, and estimating the kernel density is a common method so the kernel density estimation method has been extended to deal with uncertain data.

This reduces the problem to considering double integrals.

For finite kernel functions and probability distributions, the double integral can be evaluated analytically to give a closed formula, allowing an efficient formula-dependent algorithm in general, however, double integral cannot be simplified in closed forms.

In this case, a sample-based approach is proposed.

The remarkable experimental results also indicate that the proposed classification method can be promising and can be applied elsewhere and help in the diagnosis process by patient area.

The Naïve base algorithm was used to validate the proposed method experimentally to be 90% accurate, which proves its efficiency.