Research On Association Rules Mining For Marine Environmental Data Using MapReduce

Marine environmental data variations play an important role in determining ocean conditions and influence global climates, so marine environmental data can be used to forecast global climate variations. In this paper, ocean temperature and salinity data have been selected from marine environmental data to reveal the association between temperature variations and salinity changes. Different from market basket analysis, mining temperature and salinity patterns is a difficult task due to the temporal-spatial nature of the ocean data.The focus of this study is extracting previously unknown patterns of abnormal ocean temperature and salinity variations from Argo data that can be further applied to predict ocean current variations. Temperature and salinity patterns in the ocean data are mined using MapReduce framework. The main works are as followed:First, Argo data are converted to market-basket type data that are used to find temporal-spatial association rules. The discovered rules reveal the associations of abnormal ocean salinity and temperature variations.Second, combined with the methodology of Inter-transactions affairs, a method based on the concept of neighborhood has been proposed to analysis the ocean changes of salinity and temperature in different marine areas.Third, after having analyzed the characteristics of Inter-transaction, we have proposed a theory and method to reduce the dimension attribute of ocean temperature and salinity variation inter-transactions.Last, we redefined the extented item order in inter-transactions which is used for sorting the extended items in descending order. We use the Parallel FP-Growth algorithm that based on MapReduce programing paradigm to discover the temporal and spatial variation patterns which were used to predict future ocean salinity and temperature variations using specific space models in designated area.