| As one of the world’s top ten ports with the largest number of ports and the largest cargo throughput,China’s port logistics plays an important role in the development of the national economy.In recent years,the rapid development of the port economy also continues to put forward new requirements for port enterprises,how to improve the efficiency of port transport logistics services,reduce port operating losses and risks have become urgent problems to be solved.Many scholars have focused their research around these issues and have achieved results.But these studies are mainly from the perspective of port layout optimization,resource allocation management,enterprise process management and reengineering to improve port logistics efficiency,using port operation data and building predictive algorithm models to solve port production delays is still less research.As the port production operation has the characteristics of a large number of constituent operations,flexible activity arrangements,interrelated effects,and in the production process by mechanical equipment,weather,personnel and other factors,these characteristics and factors are very easy to cause delays in the actual production process of port operations,seriously affecting the efficiency of the port.At present,the port’s production management is still mainly based on experience and systems,and it is difficult to meet the need for early detection,early adjustment and change,accurate positioning and risk reduction.This paper uses Bayesian network as the basic algorithm model for delay prediction,improves the network structure and parameter learning for port production characteristics,combines the workflow mining algorithm to obtain the workflow model,and finally conducts instance verification analysis of the prediction model and applies it to the construction of the operational delay prediction system.The main work of this study has three points.First,we analyze the current situation of the production operation in the break-bulk port and summarize the production operation activities and the factors that affect the operation activities;we select the operation data from the port production system,obtain the real operation process model through the process mining algorithm,and propose a parallel inspired mining algorithm to improve the efficiency in view of the large amount of port data and complex process model.Second,the production delay prediction problem is transformed into a Bayesian network probability model solving problem.The network learning phase transforms the production activity process and activity influencing factors into two separate structural networks,introduces the process model into the activity process structural learning,and finally merges it;the parameter learning intermediate phase revises the parameter learning formula considering the cyclical nature of the process activity.Validate the validity,accuracy and interpretation of the improved predictive model using real data from the port.Finally,based on the existing port systems and data sources,the framework of the port production delay prediction application system that integrates the prediction model is designed and developed using modern programming language techniques to achieve further application of the model.In this paper,the process model is combined with Bayesian network to improve the learning speed and interpretation of the prediction model structure,and the influence of the job cycle structure is considered in the parameter learning,which can make the prediction model more adaptable to the complex port operation situation,improve the accuracy of job delay prediction and business interpretation.It effectively helps to optimize the port operation plan,adjust the operation schedule in a timely manner,reduce the risk of blockage delays and improve port efficiency.Provides ideas and guidance for other businesses and scenarios in the port that can be learned from. |
