KNOWLEDGE DISCOVERY FROM LAND RECORD SYSTEMS

SUMMARY Decision support, planning, and policy making are core processes in land management. Performing these tasks may require the examination and analysis of large numbers of land records. In current land management systems, the analysis of data is done by people, which may be slow, inaccurate and require substantial resources. This paper explores data mining as a technique for automating data processing and analysis of land records. Automated data mining may be useful in two ways. Firstly, it may assist in the discovery of important knowledge and patterns of behaviour from large digital datasets, which otherwise would be extremely difficult to do manually. Secondly, automation may reduce the level of skills required to run a system. Scenarios are explored for applying data mining techniques to detect manipulations of land records, where post conflict is the situational context. Of particular interest are fraudulent changes in the land records by the state or powerful factions to grab land from unsuspecting land holders on the ground. The results of preliminary tests on simulated data are reported. Specifically, outlier detection techniques are used to potentially identify patterns such as an unusual number of transactions in short periods of time, and periods with no or very few transactions.