Special Issue Article
Multi Resolution Pruning Based Co-location Identification in Spatial Data
|Related article at Pubmed, Scholar Google|
In this paper we put forward a plant prediction system with advanced clustering and improved colocation mining. Spatial data differs from the other forms of data by the fact that the neighbouring objects will have noteworthy effect to the object under consideration. Thus mining the data item and its co-location pattern together becomes more vital. In our work, we suggest an advanced method of plant prediction scheme by two steps. Initially, clustering the location areas into three types according to the nature of the location by considering the GIS (Geographic Information System) attributes and clustering the plant species also according to thegeographical location which suits the existence of the plant. These clustering is done by modifying traditional k-means clustering algorithm by altering its repeated iteration process into single iteration and repeating the same clustering by considering multiple attributes associated with the location and plant species. Finally, a combinatorial spatial co-location algorithm is used to mine the co-locations and a plant prediction system is designed in which, for a given plant species, prediction of suitable colocations which has the highest supporting environment to grow and a set of plant species which has the highest probability of co-existence is determined. Experimental results shows the prediction to be more effective in computation time and accuracy, particularly while updating the database dynamically with the new entries as the computation and prediction is limited to the initially clustered dataset rather than the complete database.