Application Research Of Decision Support System On IoT Vending Machine

Due to the gradual disappearance of China's demographic dividend,unmanned services and consumption have risen rapidly in all walks of life.As a consumption scenario for unmanned convenience stores,vending machines provide convenience services 24 hours a day,and are increasingly favored by businesses and consumers.With the increasing degree of precision of machinery and the continuous upgrading of payment methods,the hardware and software functions of vending machines have become quite complete.However,vending machines are a very fragmented and highly competitive market.Operation management is becoming the focus of major manufacturers.In order to realize the high efficiency and intelligence of the operation and management of vending machines,a lightweight decision support system is designed to provide scientific and data-based assistance in formulating optimal sales plans for management personnel through some intelligent algorithms and in decision management.Support is very meaningful.This project is based on actual scientific research projects.Firstly,it has conducted an in-depth understanding of the development status of vending machines and decision support systems,and has determined the research direction.After in-depth research on the existing online management system of vending machines,it proposes adding intelligence.The decision-making sub-module focuses on the prediction and decision-making parts of the vending machine decision support system for this module,and carries out corresponding application development.For the forecasting stage,first of all,through on-site investigations,a detailed analysis of the factors affecting the sales volume of vending machines is made,and the main factors that should be taken into consideration in the sales forecasting are clarified from the perspective of quantification.The pretreatment method of sales data was studied and selected.Several commonly used forecasting algorithms were analyzed and compared.Combining with the characteristics of vending machine sales data,a forecasting plan for the sales volume of vending machines was proposed.Because the out-of-stock situation in the historical sales records will result in data restriction and affect the accuracy of forecasting,this paper proposes a data revision plan to forecast the demand after the historical out-of-stock,and then uses the BP neural network to make predictions after the data is corrected,and the traditional The data corrections are compared and compared with a three-exponential smoothing model.The experimental results show that the accuracy of the forecast is improved,providing strong data support for the following comprehensive decision-making.In the comprehensive decision-making stage,other factors other than the sales volume,such as profit,unit product space,new product recommendation,product substitution,etc.,are taken into consideration,the ratio of the cargo space in the decision-making vending machine is optimized,and then the C4.5 decision is introduced.The tree algorithm makes a comprehensive decision and uses the above factors as the discriminating attributes of the decision tree to classify the commodities to achieve the purpose of formulating a sales plan.The improvement of the calculation method of the information gain rate in the C4.5 algorithm is improved.The feasibility of the algorithm is verified on the UCI public data set.The results show that the calculation speed is quickened and the efficiency of establishing a decision tree is improved.Finally,the various modules of the decision support system were designed,including three modules of human-computer interaction,data management,and interactive model management,and corresponding developments were made.The Java language was used to build the system,and R language was used for statistical calculation.Then the call between Java and R was completed,and the mobile terminal monitoring application was developed.