Research And Application Of Cigarette Market Status Monitoring Based On Bayesian Network

With the rapid development of modern society and economy,under the background of "cigarette structure upgrading,terminal situation fission and consumer market transformation",the cigarette market environment accurately collects cigarette market data,and realizes the analysis of market state information and application information,which is directly related to the control and response ability of tobacco industry to cigarette retail market,and directly determines the demand forecast and supply Procurement,commodity delivery,terminal construction and brand cultivation also directly affect the development of cigarette marketing,accurately predict the current cigarette market status and implement cigarette regulation,which has great research and guidance significance in other fields such as market competitiveness,new product delivery strategy and cigarette marketing in the tobacco industry.In order to study and accurately predict the cigarette market status and construct the cigarette market status prediction model,This thesis makes correlation coefficient analysis,data discretization,K-means clustering analysis on the index data of 2017-2019 in Jiangxi Province(including the index data of cigarette sales price,social inventory sales ratio,delivery volume,ordering volume,etc.),and provides the real-time increased and changed market status data In this thesis,a learning method of Bayesian network structure based on incremental learning of BDeu score is proposed.The method uses the function of BDeu score to obtain the criteria used to evaluate whether incremental learning evaluation should be needed in the construction of network structure,and uses Hill pruning search algorithm to retrieve the optimal Bayesian network structure in the current state,so as to improve the prediction of Bayesian network prediction model Performance,and the relevant cigarette data index items are also input into the SVM algorithm prediction model,AdaBoost algorithm prediction model,GBDT algorithm prediction model and other machine learning algorithm prediction models for parameter adjustment and optimization,and the methods proposed in this thesis are compared and analyzed.Through the current cigarette market state prediction and classification research and the comparison experiment of the evaluation index of each machine learning algorithm model,the experiment is carried out The results show that the Bayesian network intelligent model based on the incremental learning of BDeu score can accurately predict the market status of cigarettes and accurately reflect the trend of market status.Finally,the article designs and realizes the monitoring platform of cigarette market status in Jiangxi Province,including two parts:the front-end platform and the back-end management system.The back-end management system mainly realizes the query and display of the original cigarette index data and the new index data,makes rules and establishes models.The front-end platform realizes the monitoring and display,prediction and analysis of the daily sales data of each cigarette brand Current market status;query the market status probability of historical forecast,display the progress of cigarette market regulation,and help suppliers monitor the whole cigarette market status and cigarette regulation progress.