In: Computer Science
The specific issues which a developer may not have been able to deal with, even though the new system functions normally liked the old system are:
1. The developer may not have been able to deal with dynamic data. In an expert system, facts do not remain static and they may become invalid over time, or in case the state of an expert system gets changed due to course of events. The developer may not have taken this into consideration.
2. The responses of the system may be normal, but might not be within a deadline. An expert system's response should fall within a constraint or time bound.
3.The developer may not have updated the knowledge representation schemes in the new system. Such rules are important so as to do reasoning about the course of events in the past and present. The sequence of events are important to anticipate the decisions for the future.
4.A given sequence of rules is used to deduce a particular conclusion in an expert system. The new system may not be justifiable in terms of the conclusions reached, though it may be functioning normally. A developer should check if the decisions reached are explanatory. He might overrule a decision in case this is not so.
5. The developer should determine the response of expert system to external events. In case a decision making process is interrupted, the system should eb capable of resuming the process after dealing with higher priority tasks
6. The new system might function normally but may not be compatible with real time systems which perform tasks like feature extraction or data processing.
7. The knowledge base of the expert system may have structural deficiencies. A weakly designed knowledge base may not capture important business rules and may not be able to represent deep knowledge with many layers, where one input becomes important for the next layer.
8. The developer may have chosen a lower end development for environment, which may be simple in terms of quality and cost.
9. The knowledge base may not be complete. This means the size may be small (in terms of the number of rules) or there might be holes in the base (the expert system may not have any answers for a particular condition though it may be functioning normally.
10. The expert system may not be running in a continuous manner. There may not be a proper mechanism for garbage collection or it may not be prioritizing the goals.