ANALYSIS OF HYBRID SYSTEM: NEURAL EXPERT SYSTEM

EXPERT SYSTEM

If we combine the two approaches then we can get the advantages of both and might be able to eliminate the disadvantages up to certain extent. The combinations of the two technologies are called hybrid system.

Hybrid Technology

Expert system and neural network both are intelligent technologies which imitate human brains. Neural network works on the concept of neurons and expert system works on the human expertise. Neural network offer faster and efficient learning and adapt to the rapidly changing environment. On another side, expert systems provide rule base and explanation facilities to the user. IF we combine the neural network with expert system, the resulting combination will be known as Neural Expert System.

An expert system do not have capacity to learn on its own , but it provides solution to the problems with requires human expertise. Contrary neural network has a learning capacity, but it doesn’t provide explanation facility. Neural network works as a black box. (Negnevitsky, 2008) The combination of neural network and expert system will allow us to create more powerful expert system based on neural network. This expert system is called neural expert system or connectionist expert system.

Difference between Neural Network and Expert System

Expert system relies on Modelling human reasoning while neural network focus on modelling human brain. Knowledge in expert system is presented by if-then- else rules. Knowledge in expert system is stored in synaptic weights between neurons. In expert system, once rules are created, it cannot be modified by the expert system itself. To modify the rules, human intervention is very much required. In neural network, error is generated and propagated backwards to change the synaptic weights. Hence no human intervention is required to learn in neural network.

In expert system, knowledge is divided into rules, which can be coded. In neural network, Knowledge is embedded in entire network. It cannot be divided into rules. Expert system cannot deal with noisy dataset while neural network allow approximate reasoning, so noisy data set can also be included in training data set.

The striking feature of the neural expert system is that it can extract ‘If-then-else’ rule from neural network with the help of neural inference engine. To train Neural expert system, we can use training algorithm such as back propagation and then extract the rule base from it. Neural expert system works best when rule base is based on Boolean logic.

Mapping Concept of Neural Expert System into Our Research Objective

As we have seen that Neural Expert system has two parts. ANN and expert system. From the sample data of human resources’ preferences and perception on motivational strategies, we can devise ANN model, which can learn to evaluate motivational strategies. At the same time, from the weights stored in neurons, we can devise rules for evaluating motivational strategies on human resources. The learned model can also be used to test the new data of human resources to evaluate motivational strategies.

CONCLUSIONS

The paper throws light on three main aspects. Exploring the concept of Neural expert system, mapping of concept for our research objective and analysis of literature review on hybrid systems and its findings. The above study helped us to establish a milestone in development of expert system. It helped us to finalised model, technique, method and tool for development of our expert system to evaluate motivational strategies on human resources.