The invention discloses a mass data quality verification method based on Hadoop. The method comprises the following steps: formulating a data quality standard; for the DDL instruction, writing metadata information of the creation table into Hive; for the DQL statement, converting the SQL character string into an abstract syntax tree, performing syntax analysis on the abstract syntax tree, analyzing whether latest generated SQL semantics are wrong or not according to a data quality standard, and adding extension information; compiling the abstract syntax tree to generate a corresponding logic execution plan, optimizing the logic execution plan, converting the optimized logic execution plan into a physical plan, generating a MapReduce job, submitting the MapReduce job to the Yarn for execution, and finally returning an execution result; storing a returned execution result into the HDFS, and carrying out data visualization and abnormal data exporting, tracking and tracing. Therefore, thedata quality verification effects of abnormal data display, traceability, easy configuration and easy classification are achieved.