Error locating method based on program failure clustering analysis

The invention provides an error locating method based on program failure clustering analysis. A failure caused by the same error has similarity trigger. The method comprises the following steps of firstly, obtaining execution slices for each failure test and computing a distance between the execution slices according to a Jaccard distance formula; secondly, carrying out clustering on program failure execution slides according to a K-Means clustering algorithm to obtain a failure execution slice cluster; thirdly, obtaining a program status sequence chart of the failure execution slices according to a runtime program status division method of a jump instruction; fourthly, respectively establishing execution path coverage vectors in basic blocks/rows and a coverage matrix according to the function nest call frequency in the program status sequence chart; and lastly, respectively computing suspiciousness degree of each basic block/row and sorting in a descending order, solving a frequent set in basic blocks/rows