Support Vector Machine Based Forecasting Model Selection

WITH COMPETITION BECOMING MORE AND MORE INTENSE IN AIR TRANSPORT MARKET, HOW TO FORMULATE FLIGHT SCHEDULING REASONABLY AND EFFECTIVELY BECOMES IMPORTANT CONTENT OF AIRLINE DECISION. FORECASTING FLIGHT SEGMENT VOLUME IS IMPORTANT STEP AND FORECASTING RESULT IS IMPORTANT BASIS IN FORMULATING FLIGHT SCHEDULING. THE FLIGHT SEGMENT VOLUME IS INFLUENCED BY MANY KINDS OF FACTORS, SUCH AS THE REGION ECONOMY, GEOGRAPHY, TOURISM AND THE SEASON AND SO ON. THE SOLE FORECASTING MODEL IS NOT SUITABLE FOR ALL FLIGHT SEGMENTS. THEREFORE, TO SELECT THE APPROPRIATE FORECASTING MODELS FOR A CERTAIN SEGMENT IS NEEDED.THIS THESIS SYSTEMATICALLY ANALYZES THE CHARACTERISTIC OF FLIGHT SEGMENT VOLUME AND EXISTING MODEL SELECTION METHOD,AND DISCUSSES THEORY AND ALGORITHM OF SUPPORT VECTOR MACHINE IN DETAILS, MAKES RESEARCH ON MODEL SELECTION IN FORECASTING OF FLIGHT SEGMENT VOLUME COMBINED WITH SUPPORT VECTOR MACHINE. FIRST, THE CHARACTERISTIC OF FLIGHT SEGMENT VOLUME IS ANALYZED, EXISTING MODEL SELECTION METHOD IS DISCUSSED, FORECASTING MODEL SELECTION METHOD BASED ON SUPPORT VECTOR MACHINE IS DESCRIBED. THEN, THE SUPPORT VECTOR MACHINE PRINCIPLE AND THE NUCLEAR FUNCTION ARE DISCUSSED, THE TRAINING AND CLASSIFICATION ALGORITHM OF SUPPORT VECTOR MACHINE IS DESCRIBED. FINALLY, THE DESIGN AND REALIZATION ABOUT MODEL SELECTING BASED ON SUPPORT VECTOR MACHINE IS INTRODUCED IN DETAILS. THREE KIND OF FORECASTING MODEL AND THEIR APPLICATION IN THE SYSTEM OF "AIR FLIGHT SCHEDULING & OPTIMIZING MANAGEMENT SIMULATOR" ARE PRESENTED AND EVALUATED. THE RESULT INDICATES THAT THE METHOD HAS THE GOOD EFFECT IN MODEL SELECTION FOR FORECASTING OF FLIGHT SEGMENT VOLUME.