Learning from a Class Imbalanced Public Health Dataset: a Cost-based Comparison of Classifier Performance

Abstract

Public health care systems routinely collect health-related data from the population. This data can be analyzed using data mining techniques to find novel, interesting patterns, which could help formulate effective public health policies and interventions. The occurrence of chronic illness is rare in the population and the effect of this class imbalance, on the performance of various classifiers was studied. The objective of this work is to identify the best classifiers for class imbalanced health datasets through a cost-based comparison of classifier performance. The popular, opensource data mining tool WEKA, was used to build a variety of core classifiers as well as classifier ensembles, to evaluate the classifiers" performance. The unequal misclassification costs were represented in a cost matrix, and cost-benefit analysis was also performed. In another experiment, various sampling methods such as under-sampling, over-sampling, and SMOTE was performed to balance the class distribution in the dataset, and the costs were compared. The Bayesian classifiers performed well with a high recall, low number of false negatives and were not affected by the class imbalance. Results confirm that total cost of Bayesian classifiers can be further reduced using cost-sensitive learning methods. Classifiers built using the random under-sampled dataset showed a dramatic drop in costs and high classification accuracy.

Predicting Breast Cancer via Supervised Machine Learning Methods on Class Imbalanced Data

“…The disadvantage of this technique is that it may take longer training time and result in over-fitting since there is a significant increase in the size of the training set. A well-known oversampling technique known as Synthetic Minority Oversampling Technique (SMOTE), is used to oversample the minority class by creating synthetic instances to replicate the minority classes and increase their number of instances in the training set [31]. These synthetic instances are produced by considering two key parameters which are the number of instances (n) and the nearest neighbors (k).…”

“…The Bayesian Network is preferred as past academic works have shown that this classifier exhibits a strong correlation among the attributes in the patient disease diagnosis. Other than that, the classifier is robust to unrelated variables, noise and confounding factors that are not part of the classification [31]. Bayesian Network has been broadly employed in many medical diagnoses based on previous literature studies, www.ijacsa.thesai.org especially for cancer prediction and recently the use of Bayesian Network classifiers in breast cancer prediction is trending.…”

A Comparative Analysis on the Evaluation of Classification Algorithms in the Prediction of Diabetes

“…ROC graphs are bi-dimensional graphs where on the Y axis t p rate is plotted and on the X axis f p rate is plotted. A ROC graph describe relative trade-offs between benefits (true positives) and costs (false positives) [23].…”

Protecting sensitive information utilizing an efficient association representative rule concealing algorithm for imbalance dataset

“…The schemes are utilized to select the best subset of aspects. In [18] studied to detect the best classifiers for class imbalanced health datasets through a price depended comparison of classifier performance. The uneven misclassification prices were characterized in a cost matrix, and cost-benefit.…”