Learning from Ensembles: Using Artificial Neural Network Ensemble for Medical Outcomes Prediction

Predicting the outcome of a medical procedure or event with high level of accuracy can be a challenging task. To answer the challenge, data mining can play a significant role. The main objective of this study is to examine the performances of an artificially intelligent (Al)-based data mining technique namely artificial neural network ensemble (ANNE) in prediction of medical outcomes. It also describes a novel approach, namely "RIDC-ANNE". This approach tries to improve data quality by configuring an ensemble of bagged networks as a filter and identifying the regions in the data space that have high impact on the system performance. Furthermore, it can also be used to extract explanations and knowledge from several combined neural network classifiers. The methodology employed utilizes a series of clinical datasets. The datasets embody a number of important properties, which make them a good starting point for the purpose of this research. This study reveals that the RIDC-ANNE approach can be used to successfully extract the regions in the data space that have high impact on the system performance and enhance the overall utility of current neural network models