The invention discloses a memory prediction method and device suitable for a dependency analysis tool, and the method specifically comprises the steps of extracting a source code file from a Java program package file; analyzing the extracted source code file, generating an abstract syntax tree, and obtaining the number of the instance objects of each type of node class in the abstract syntax treegenerated by each source code file; calculating the size of a memory occupied by each type of node type instance objects; calculating the memory occupied by the abstract syntax tree generated by eachsource code file; and calculating the memory size required by the whole Java program package. According to the method, the memory is predicted when the dependency analysis tool runs, so that the dependency analysis tool container can be allocated to the container running nodes with enough memory resources through the predicted memory every time, and the running stability of the dependency analysistool and the running stability of a distributed code detection system are ensured.