Implementation of the Q-learning Algorithm for Optimising Judicial Advisory Expert System (jaes)

The work presented in this paper is an implementation of the Q-learning algorithm to optimise the rule execution process of Judicial Advisory Expert System (JAES). JAES covers the Sale of Goods Act (SGA) sections 16 to 20, UK. Addition of further sections to JAES led to its performance degradation; hence adopting an algorithm which improves the rule execution process was necessary. The existing approach of depth-first search was computationally expensive; therefore we had to adopt the Q-learning algorithm to learn the paths of the execution process to optimise its performance. Prior to implementing the Qlearning algorithm, we evaluated JAES with real world legal case hearings, to determine its effectiveness to reach legal conclusions. For a successful implementation of the Q-learning algorithm, we had to modify the existing modular decomposition structure to accommodate the changes in the design. Furthermore, after the implementation, we evaluated the effectiveness and the performance of JAES and compared the results using scenario-based legal cases. The results were impressive, as we were able to optimise the JAES rule execution process by 33%.