Analysis And Prediction Of Flight Delay Based On Data Mining

Due to the growth of travel demand from passengers,flight delay has been a serious problem with the annual increase of flights in recent years.The research on the delayed flights would offer helpful advice to the decrease of flight delay when flight delay is inevitable.This dissertation employed the methods of statistics and data mining to seek the rules and knowledge hidden in the massive data collected from an airline company and OAG database.The main reasons for flight delay were analyzed from statistics of massive data of recent years.The most influencing factor which leaded to the current flight delay type was discovered by using the Decision-Tree model.Moreover,the condition for each type of flight delays and decision rules with confidence degree over 0.9 were also concluded from Decision-Tree model.The factors affecting the flight delay rates were sought out by analyzing the variance of delay rates for the first flight and non-first flights in a flight-string.The main factors,which had impact on the delays of first flight and non-first flights in a flight-string,and its influence levels were analyzed by applying the TAN Bayesian Network model.In addition,some factors were discovered to be dependent on other factors.The meanings of the changing tendency and extreme value about actual delay time in different months or hours were analyzed by calculating the Mean,Median,Skewness and other descriptive statistics of actual delay time.An improved model based on KNN for predicting actual delay time was proposed and then optimized by choosing an appropriate parameter K according to RMSE and MAE.In this dissertation some rules about flight delay reasons,flight delay rates and flight delay time were found by the methods of statistics and data mining.These research findings would provide reference for the airline company to recognize the regulations of flight delays and to reduce flight delays.