Research And Implementation Of Imputation Method For Missing Data In The Trash Pickup Logistics Mangagement System

The management of municipal solid waste(MSW) has been an import social issue. With the development of urbanization and the expansion of the service area of trash pickup logistics management, there is a trend of multiplicity in managing tasks and managing difficulties. Thus, in this paper, in response to the challenges facing the management of MSW, we conducted a detailed design of the trash pickup logistics management system(TPLMS), of which purpose is to build an information & technology management system for urban residents and administrative staff to access the logistics real-time transaction data and the supervisory service.The designing of the TPLMS is mainly based on the consideration of device layer and the central system layer. The device layer module is designed to collect the information data of operation site and the data of garbage weight, and to transmit data to the central database system.The main functions of the device layer module includes real-time data transmission and off-line data storage. The designing of the central system module uses the B/S architecture. Users can know the real-time pickup operation situation, check the vehicle route, query and manage relevant data via the Internet. The main system is divided into three sub systems, including data communication subsystem, operation monitoring subsystem and data management subsystem. The main functions of the central system layer module includes data acquisition, data transmission,trash pickup operation monitoring, data report query, security management, network management,and database management. In addition, in consideration of that the data is not complete, the relevant data backup strategies are designed, which achieves data verification and audit function.To deal with the missing data in the TPLMS, this paper proposes a iterative KNN imputation method based on gray correlation analysis to fill in the missing data. This method associates with the methodologies of the grey relational analysis, KNN algorithm and multiple imputation algorithm, establishes a data model where the historical weight data is defined as an original data series. This method utilizes the gray correlation degree to search the nearest neighbors, and calculates the estimated values iteratively. In each iteration, the result dataset isused as the initial dataset for the next iteration. The iterative estimation stops until the mean change of the results converges to a certain minimal value.In this paper, we have established the trash pickup logistics management system and carried out many experiments on the transaction data. The experimental results show that the proposed method gets better performances than other commonly used imputation methods when deals with the datasets at different missing rates. The proposed method can impute the incomplete dataset efficiently and the accuracy rate of the results is as high as 80%. In addition, this method establishes an effective data model, which decreases the calculation of the matching process greatly, and has better stability and pertinence.