This study proposes the RO-RA-RV structure for rule-based knowledge and integrates methods of conditional probability, vector matrices, and artificial intelligence to establish a conditional probability knowledge similarity algorithm and calculation system. This calculation system can quickly and accurately calculate rule-based knowledge similarity matrices and determine the relationship among knowledge items, this relationship can function as the knowledge source for treatments that increase the addedvalue of the knowledge, through the inference of value-added treatment such as merging, integration, deletion, innovation and additions, the accuracy of the knowledge itself can be securely ensured and wrong decisions be avoided. According to the knowledge similarity matrices, the knowledge case most similar to the testing case can be quickly retrieved, and used for all types of with Knowledge-Based Reasoning or Case-Based Reasoning to help decision making and prediction.