Research On Flight Delay Forecast Based On Machine Learning Method

With the development of the air transportation industry,the problem of flight delays has become increasingly serious,causing huge economic losses.Therefore,achieving accurate prediction of flight delays has become a top priority.Therefore,this paper proposes a model combining Classification And Regression Trees(CART)and random forest models to predict flight delays.The aim is to mine data through machine learning algorithm models to obtain more accurate prediction results and provide references for decision-making by relevant departments.The main content of the paper is as follows:First,the flight data is introduced and preprocessed,and the data source,data volume and data information are introduced in detail,and then the data preprocessing.The data is feature capture,additional feature extraction,feature processingSecondly,build a flight delay prediction model based on CART decision tree and random forest model,introduce the mathematical principle of the model in detail,and explain the principle of finding the optimal feature of CART decision tree,and give the specific process of CART regression tree algorithm establishment.And the output of the decision tree.And finally explained the evaluation of the model,and put forward the evaluation indicators and methods.The key is to compare the model built in this article with the Logistic regression algorithm,KNN regression algorithm and single decision tree model,analyze the absolute and relative errors between the models,and use cross-validation methods to obtain each model based on the evaluation indicators.The average absolute error MAE and the goodness of fit R2.Finally,tuning and improvement are made based on the results of the model,and the parameters and tuning methods that need to be tuned are introduced,mainly on the pruning problem of the model.The optimization of the decision tree model is also mainly in the decision tree Pruning.MAE is 9.325,R~2 is 0.87.