Study On Knowledge Discovery And Prediction Model Of Passenger Ticket Revenue Of The Trains Based On Rough Set

In order to increase the railway passenger ticket revenue and allocate transportation resources more reasonably,the management department forms an income decision table based on the existing train passenger ticket revenue,operating conditions,and the number of passengers traveling along the train station.Forecasting the ticket revenue of trains can not only provide the railway station with a more reasonable amount of the next-phase passenger income index,but also provide guarantees for the management department to effectively carry out cost control and passenger transportation organization.However,with the rapid development of China's high-speed railway network construction,trains` operating conditions are becoming more complex and diverse.Faced with new situation,relevant managers still rely on experience or manual intervention to deal with factors that affect passenger ticket revenue.The result of this forecast is ticket revenue.It has been proved that the results can no longer meet the needs of the work.At present,the knowledge discovery model based on rough set theory method has been applied in many fields and has achieved remarkable results.However,the knowledge discovery and forecast of passenger ticket revenue of train is still in its preliminary way through the model above.Because there are many factors affecting the railway ticket revenue,the data type is complex,and it is nonlinear,high noise and so on.Therefore,this paper proposes a passenger-ticket revenue knowledge discovery and forecast model based on rough set,which can be used to mine the potential rules behind ticket revenue knowledge and form a set of knowledge discovery model systems that can effectively forecast passenger ticket revenue are of great significance in both theoretical and practical applications.In this paper,the total ticket revenue of the train and the ticket revenue of each stop station along the way are taken as the research objects.The data pre-processing,knowledge discovery based on rough set and income forecasting are constructed in the model.In the first module,a K-means-CACC algorithm is used to discretize the target variables according to the characteristics of the data found in the ticket revenue knowledge.This algorithm avoids the unsupervised discretization method to ignore the data distribution information and the boundary boundaries are determined to be unrepresentative and other shortcomings.In the second module,based on the rough set theory method,the dependency between the condition attribute and the decision attribute is calculated.The heuristic reduction algorithm based on the core attribute is used to reduce the redundant attributes in the initial decision table.The random forest algorithm is used to construct passenger ticket revenue knowledge discovery rules,this method avoids the association analysis method can not calculate the rules and contradictory rules.Finally,depend on the above knowledge-rule base,this paper proposes a rough set-ensemble learning model to predict passenger ticket revenue,and uses LSTM(Long short-term Memory),XGBoost(Extreme Gradient Boosting)and the target selection algorithm based on error interval intersection are used as individual learners to predict separately and integrate according to Stacking algorithm.At the end of the paper,based on the methods proposed and used in each module,validation was conducted based on actual ticket sales data.The verification results show that the above method can effectively form a concise and easy-to-understand decision table and can more accurately predict the ticket revenue than the current method.