Extraction Of Features (Pattern Recognition) From Web Data Using Neural Expert System (NES)

Expert system and neural network both are intelligent technologies and share common goals. Both are used to deduct some logical inferences from a given data. While expert systems rely on logical inferences and focus on modeling human reasoning, neural network rely on parallel data processing and focus on modeling a human brain. Expert system treat the brain as a black box, since we don’t know what’s going inside the system, whereas neural network look at its structure and functions, particularly at its ability to learn. These difference can be seen in knowledge representation and data processing techniques used in expert systems and neural networks. Knowledge acquisition from a rule based (If-Then) expert system is done by a human experts. Once rules are stored in knowledge base, they cannot be modified by system itself. Expert system cannot learn from experiences or adapt to new environments but a neural network can. Only a human being can manually modify the knowledge base by adding deleting or changing some rules. Knowledge in neural network is stored as synaptic weights between neurons. When a training set of data is presented to the network during learning phase, knowledge is obtained. Network propagates input data from layer to layer until the output data is generated. Neural network can learn without human intervention like hebbian’s unsupervised learning [1]. Learning generalization, robustness and parallel information processing make neural network a right component for building a new breed of expert systems. A neural expert system can extract If-Then rules from neural network which enables it to justify and explain its conclusion. Neural expert systems use a trained neural network in place of the knowledge base. The neural network is capable of generalization. In other words the new input data does not need precisely match the data that was used in network training. We can apply new data set also. This allows neural expert system to deal with noisy and incomplete data. This ability is called approximate reasoning. So explanation facilities and approximate reasoning are two most important part & heart of this approach as we can see in fig. 1. Now suppose we want to cluster websites according to their features like content, construction, accessibility, navigation, contact,