SPATIAL TEMPORAL DATA MINING FOR CROP YIELD PREDICTION.

Background: Crop yield prediction can help agricultural departments to have strategies for improving agriculture. Towards this end many techniques came into existence. Spatiotemporal data mining is one such solution that can be employed to achieve crop yield prediction. We extend the method employed by Cao et al. for leveraging the power of parallel computing of MapReduce framework. Especially we used Spatial Hadoop which is based on the new programming paradigm besides having spatial extensions. Towards this end, we proposed a framework and implemented the extended algorithm that is compatible with MapReduce programming. Results: We collected five years data of Cotton and Maize crops of Karimnagar region of Telangana state, India. We implemented the proposed framework for crop yield prediction. Our crop yield prediction mechanism using MapReduce programming paradigm is tested using a prototype application. The proposed framework also secures data flow in the MapReduce framework. The results revealed that the proposed solution is encouraging and the error rate of prediction of Cotton and Maize is low. Conclusions: Proposed framework was able to achieve crop yield prediction. Cotton and maize crops were used for the purpose. Results revealed that the proposed system outperformed the existing one with reduced error rate.