In some e-commerce system, users do not know performance of commodities described by numeric values, but they can provide conceptual descriptions, such as “high”, “medium” and “low” on its price. The query based on concepts as inputs is named conceptual query. The basic method of conceptual query is to convert user's fuzzy concept into standard SQL statement based on the domain knowledge. However, if a condition is composed of the high and multilevel concepts, then its conversion of the concepts is very complex. Its efficiency will be poor. A new knowledge-based query method on multilevel generalized data is proposed in this paper. The knowledge of fuzzy concepts needs to be established, and multilevel generalized data is generated, a knowledge-based multilevel query is achieved on generalized data. Experiments on the synthetic data sets show that on a higher level and multiple levels the query efficiency of the method proposed in the paper is better than the basic method. The knowledge-based conceptual query on multilevel generalized data will provide efficient methods for the query about housing products in rural town and can meet the needs of customers with different knowledge, and has higher practical value.