| The paper starts from the research background of public data predict ion,and elaborates the current research statu s of deep learning methods and public data analysis in related fields on the basis o f exist ing machine learning met hods.For public big data predict ion in the field of music and meteorology,the fo llowing are mainly carried out jobs:(1)In this paper,based on the data set of Aliyun music in a specific p eriod,the maximum mutual informat ion coefficient method is adopted to select the combinat io n of "daily playback amount" and "average playback amount for 3 consecut ive days" as the co mbined index for the predict ion o f the playback amount of future singers.The predict ion model co mbining the short and long time memory network and attention mechanism was used to predict the amount of songs played by relevant singers in a certain period o f t ime.Co mpared wit h the single model and the support vector machine method,the method adopted in this paper achieves better results on the given evaluat ion index.(2)For the selected historical meteorological data set of Qinghai Lake,"1.5m average air temperature","1.5m a verage water pressure" and historical precipitat ion were selected as the co mbined indexes for target predict ion.The short-term precipitat ion predict ion model of Qinghai Lake was established by co mbining the bidirect ional lo ng-term memory network and at tention mechanism.The short-term precipitat ion of Qinghai Lake in the next 48 hours is predicted.Compared with Bi LSTM and LSTM-ATTENTION,the model adopted in this paper is obviously superior to the other two models in terms o f average abso lute error and root mean square error.This research mainly analyzes and experimentally simulates the predictio n of time series public data in related fields.The experimental result s show that the combined model based on deep learning shows excellent results,which provides a certain future and ideas for the predict ion research of public data in other fields. |
PDF Full Text Request