| With the vigorous development of smart city,smart parking,as an important part of smart city,is gaining attention from experts and scholars.At present,regional traffic congestion tends to occur during major events or bad weather,due to the low level of parking lot management and the lack of good forecasting and early warning technology for parking demand.In view of the above problems,this paper uses historical data to predict parking berth demand in semi enclosed areas.The original data used in this paper are the parking records of three months in a university campus in Hangzhou.The main work and results are as follows:(1)The changing pattern of parking demand in the campus is analyzed,and the Kalman filter is used to forecast the parking demand every other day,with the method improved by adding the factors behind the parking berth demand.The method produces higher accuracy.(2)In order to further improve the accuracy of parking berth demand forecasting,neural network algorithm is introduced.BP neural network and LSTM cyclic neural network are successively used to forecast short-term parking berth demand,and correct the results of next-day forecast.The experimental results show that LSTM recurrent neural network achieves obviously significantly better accuracy than BP neural network.(3)Firstly,we use Java to compile Kalman filter prediction model.Then we use TensorFlow to implement two kinds of neural network prediction models and store the results in MySQL database.Finally,we use JFrame to develop a visual interface which provides managers and users with intuitive results of parking berth demand prediction.In order to improve the utilization rate of parking lots and alleviate regional traffic congestion,this paper uses real-time data and historical data to forecast the demand of parking berths in the region and develops a parking berth demand prediction system.The research results are of great value to the traffic and parking management in the region. |
PDF Full Text Request