Logistics center cargo volume adjustment and personnel optimization based on data analysis

ACKNOWLEDGEMENTS: This work was supported by ministry of education industry-university cooperative education project (Grant No.: 231106441092432) and special research project on teaching reform (Grant No.: 30120300100-23-yb-jgkt03)

Abstract

With the booming development of global e-commerce business, the management efficiency of e-commerce logistics networks has become a focus of attention. In this context, this paper focuses on the prediction of cargo volume and personnel scheduling in sorting centers, aiming to improve the operational efficiency and cost control of logistics systems. Three main steps are taken for work one: data preprocessing, stationarity testing, and the establishment of an adaptive hybrid ARIMA-LSTM-XGBOOST weighted model. Data preprocessing includes interpolation of missing values and identification and processing of outliers, supplementing missing values with linear interpolation, and combining JB test and boxplot for outlier detection and processing. Secondly, perform stationarity testing, using the ADF unit root test method to verify the stationarity of the sequence, and use first-order difference to make the sequence stationary. For work two, a complex logistics network topology was established, revealing the structure and layout of the network. Using K-means algorithm to perform clustering analysis on the cargo volume of sorting centers, in order to explore the impact of transportation route changes on cargo volume, and optimize resource allocation to improve sorting efficiency. Using BP neural network for cargo volume prediction, the prediction results and model training state diagram were obtained. For work three, with the premise of ensuring the completion of daily cargo volume processing, an objective function is established to minimize the total number of human days, balance the actual hourly human efficiency per day, hourly human efficiency per site and day, average hourly human efficiency per site and day, and hourly human efficiency variance per site and day. The multi-objective optimization model and genetic algorithm are used to solve the decision variables, aiming to minimize the number of arranged human days as much as possible and pursue the actual hourly human efficiency balance per day. For work four, to ensure the average energy efficiency of the sorting center per hour, scheduling should comply with health and safety standards, ensuring that each shift has sufficient rest intervals to comply with labor health and safety laws and regulations. To achieve this goal, a 0-1 planning model is established to determine constraints such as the total picking quantity not being less than the predicted goods quantity, the maximum daily attendance of formal workers for one shift, the attendance rate of formal workers not exceeding 85%, the continuous attendance days of formal workers not exceeding 7 days and non negative constraints.