ABSTRACT
ABSTRACT: In agricultural risk assessment, an important issue is to measure the weatherrelated risk in crop planting. In this article, methods of data mining were used to recognize the relationship between yield and meteorology factors (such as precipitation, temperature and so on), with paddy as a representative crop planted in Hunan Province. Based on the county-level meteorological data and paddy yield data in Hunan Province, the relationship between yield risk and meteorological factors were analyzed in the view of big data. Firstly, with a trend removing, the non-meteorological part of paddy yield was removed from the data of yield. Then the abnormity of meteorological factors (precipitation and temperature) and the declines of paddy yield were transformed into binary information flows within different identification standards. Then rules analysis in the field of machine self-learning would be taken into association. Finally, with an iteration of all the probable outcomes from the combinations and a marking of the results of binary information flow matching, the maximum possible loss of paddy’s yield per unit area in different meteorological conditions could be worked out in a certain degree of confidence. Using the result, the evaluation could cover the real risk of yield declines in 97% confidence with only 3%∼5% risk appendix, which means conditional expected loss. At the same time, the risk appendix of traditional method in the same confidence could be 8% to 10%.
