Database system is a very important information infrastructure of modern society. The consistent sharing of data is the key to the realization of reliable application of database system in different fields of society. With respect to the problems of data inconsistency that are widely existed in database system applications, we focus on the entity-relation modeling, the origin of data generation, combine with practical application cases and propose active and effective prior precautions which are extracted from our many years of teachings and scientific researches. First, we use the objective uniqueness of recognizing entity and relation to overcome the subjective biases in data modeling fundamentally. Next, some empirical rules for optimizing the entity-relation model are employed to further rectify the inappropriateness in the model. The effectiveness and feasibility of the proposed principles of the entity-relation modeling optimization for database system are sufficiently approved by the demon-stration of practical application cases.