Abstract: Software is a key point in most system development projects. Modern software systems are large and highly complex. As the complexity of the systems grows, it is becoming more and more complicated to understand the structure of the software system and to localize and analyse the errors . Nowadays it is an increasingly challenging task to ensure the robustness and reliability of a software system. In program analysis software faults can be classified into two categories: crashing faults, i.e. core dump or segmentation fault, and noncrashing faults, i.e. errors that do not incur crashes, logical errors. One could apply debugging techniques to locate the cause of a crashing fault. In case of noncrashing faults it may require a great deal of a human effort to find the cause of an error, since no crashing point, hence no backtrace is available. The aim of this project is to automate the process of locating the noncrashing faults using data mining techniques. Tracing data generated during program execution may disclose important patterns and outliers that may help discover software faults. Therefore, we try to use this approach and apply data mining methods to the data gathered during program runs. As a result we expect an analysis of the program execution that would point out the possible source of the error.