| |
| Acadêmico(a): Jefferson Cristian Gubetti |
| Título: Um ambiente amigável para a construção de sistemas especialistas baseados em regras de produção |
| |
| Abstract: |
| With the outdated environments used to present the subject of expert systems for freshmen of Artificial Intelligence, it's felt the need to create an environment with resources that make easy the construction and maintenance of the knowledge base of an expert system. This monograph presents the specification and implementation of an environment for building expert systems based on production rules. The user can build a knowledge base and carry out the process of reasoning based on them. The reasoning process begins with the knowledge base being transformed into a decision tree that contains the paths to all objectives and each node represents a decision (premise) or action (conclusion). With the decision tree created, lists of variables and objectives instantiated are initialized, besides the list of all possible paths to the objectives. This list is ordered according to two criteria: shortest paths and order of the rules in knowledge base. Then the data inference process starts with the environment by consulting the user about the values of each variable being examined until a goal is instantiated and the environment to display the screen with the results of the reasoner reasoning. In tests the environment presented satisfactory results, albeit with some limitations, mainly for always require user a value when a query is displayed, that is, not treating cases where this variable has no known value. In addition, it is necessary to supplement the environment with the inclusion of other features such as reasoning mode of back chaining and more options for conflict resolution. |
| |
|
