Application of Data Mining Techniques to Identify Data Anomalies: A Case Study in the Oil and Gas Industry

摘要

This paper presents the application of the AgentMiner~(TM) tool suite to improve the efficiency of detecting data anomalies in oil well log and production data sets, which have traditionally been done by hand or through the use of database business rules. There was a need to verify the data sets, once cleansed and certified to ensure that the existing data certification process was effective. There was also a need to identify more complex relational data anomalies that cannot be addressed by simple business rules. Analysis techniques including statistical clustering, correlation and 3-D data visualization techniques were successfully utilized to identify potential complex data anomalies. A data-preprocessing tool was also applied to automatically detect simple data errors such as missing, out of range, and null values. The pre-processing tools were also used to prepare the data sets for further statistical and visualization analyses. To enhance the discovery of data anomalies two different data visualization tools for the data clusters were applied.