Case Consistency in Case-Based Reasoning

Case-Based Reasoning (CBR) is a Machine Learning technique that models human reasoning. As it learns, however, it becomes slower as its case library grows. In addition, with noisy or changing data, it may learn incorrect data or need to remove old data from its memory. An algorithm is developed to reduce the size of a case library without reducing quality. Using this algorithm, whenever a CBR system learns new cases, it will compare these data with existing cases and determine whether old cases should be removed and or the new cases should be learned. The algorithm makes it\u27s decision based on the age of the old cases, their similarity to the learning cases, and the amount of cases already learned in that area. This thesis describes the algorithm and a prototype built to evaluate its effectiveness