Research Of The Prediction Model Of Online Agricultrual Based On Fuzzy Deep Learning

With the rapid development of economic society,people not only in pursuit of agricultural products with high quality and diversified species,but also in search for a more convenient way to buy it.The rise and development of agricultural products online trading,providing people convenient and efficient purchase experience of agricultural products,building a new interactive channels for the producers and consumers of agricultural products.Under the background of data drives development,the value of information is not only limited to the way to obtain it,but it is more important to extract information with valuable,innovative and understandable.Focus on this,it will provide scientific decision-making suggestions for the formulation of national policy,the implementation of enterprise strategy and the improvement of personal life.The main goal of online agricultural products trading is to completely solve the imbalance of information supply and demand,ultimately achieve the reasonable distribution of online demand and offline production.The realization of this goal can't be separated from the processing for the data generated during the online agricultural products transaction to grasp the characteristics and trends of agricultural products online trading.Consequently,the accurate prediction of the agricultural products online sales will become the key point of the current online agricultural trade information processing.As the rapid increase of agricultural products online trading data scale,the prediction performance of traditional shallow-layer model has been unable to meet the processing needs of the current online transaction data.Deep learning,as the latest research results in the field of pattern recognition and machine learning,has outstanding performance in large scale data processing for image and speech and so on,provided with its excellent model construction and feature representation ability.Starting from the original historical data of agricultural products online trading,this research established the evaluation index based on the nine factors impacting agricultural products online sales,and divided sales into four levels.Firstly,deep learning adopted the method of auto-encoder and proposed the Imperial Crown Model(ICM),which is the online agricultural product sales predicting model.This model extracts the sample features through two layers Auto-encoder network to generate a new feature vector.It trains the classifier by the labeled samples and classifies the unlabeled training samples.It also produces the optimal parameters for getting the minimum value of the loss function by fine-tuning the whole network parameters with BP to achieve the dynamic classification forecast of online agricultural products sales.Secondly,transaction data increases rapidly in diversification and there is a vague correspondence relationship between sales influence factors and level.Considering that,the model of a deep learning sales predicting model with fuzzy membership degree-Super Imperial Crown Model(SICM)is proposed,which uses data processing technology based on the fuzzy membership degree to optimize the deep learning algorithm and Auto-encoder method sparsity constraints was added.This algorithm extracts the sample features by using Sparse Auto-encoder networks,realizes the sales level classification prediction by using Softmax classifier,and achieves parameter optimization by using BP fine-tune.Finally,by using the R software with the collected of transaction data to simulate,and make a contrast of the comprehensive prediction performance of the two models.The result shows that the research methods in this paper can realize the real-time and accurate dynamic sales prediction of current online agricultural trade data,effectively overcome the imbalance between supply and demand which caused by the information imbalance.What's more,it can promote the research of deep learning used for the e-commerce transactions.