Toward intelligent Web monitoring: performance of committee neural networks vs. single neural network

The development of communication technologies has created opportunities for Web monitoring and home care in medical practice. Due to voluminous data involved in Web based home care, intelligent systems are needed to identify if physician alert is needed and to identify the data that needs to be transmitted. Neural networks are generally used as intelligent tools, but the reliability of a single network may be questionable. In the presented study, the technique of committee neural networks for intelligent Web monitoring was developed, and the reliability of the committee neural networks was evaluated in relation to single networks. The committee neural networks based on a majority opinion of its member networks provided 100% correct results. On the other hand, individual networks proved to be less reliable and provided less than 100% correct results. Committee neural networks with a majority voting offer powerful potential tools for intelligent Web monitoring using home care devices