| Summary: | Currently there are many methods of collecting geoscience data, such as station observations, satellite images, sensor networks, etc. All of these data sources from different regions and time intervals are combined in geoscience research activities today. Using a mixture of several different data sources may have benefits but may also lead to severe data quality problems, such as inconsistent data and missing values. There have been efforts to produce more consistent data sets from multiple data sources. However, because of the huge gaps in data quality among the different sources, data quality inequality among different regions and time intervals has still occurred in the resultant data sets. As the construction methods of these data sets are quite complicated, it would be difficult for users to know the data quality of a dataset not to mention the data quality for a specified location or a given time interval. In this paper, the authors address the problem by generating a data quality measure for all regions and time intervals of a dataset. The data quality measure is computed by comparing the constructed datasets and their sources or other relevant data, using data mining techniques. This paper also demonstrates how to handle major quality problems, such as outliers and missing values, by using data mining techniques in the geoscience data, especially in global climate data.
|
|---|
