Related Project Tycho Datasets
In this dissertation, I develop a novel inconsistency detection and data fusion method for data integration systems. Inconsistent data may lead to incorrect query results and induce unexplainable outcomes. I propose an inconsistency detection method to find out which data items (e.g., temporal or spatial report) have the higher potential to cause data conflicts as well as to estimate a reasonable consistent reported value. My approach is based on representing overlapping data reports as a characteristic linear system. The characteristic linear system can be used to estimate consistent reported values within overlapping time and space intervals. I explore applicability of the proposed approach in different domains. In particular, I perform temporal data fusion with time-overlapping reports using a historical database. I also experiment with spatial data fusion involving space-overlapping reports using simulation of sensor data sets of robots performing search and rescue task. Finally, I apply the proposed approach to combine temporal and spatial fusion and demonstrate that such multidimensional fusion improves inconsistency detection and target value estimation.
Read the full article