Research And Design Of Parking Lot Recommendation System Based On Parking Space Prediction Model

In real life,the use of parking spaces is affected by various conditions and cannot be accurately predicted.By studying the parking space vacancy rate and improving its prediction accuracy,it is of great significance to minimize the overall service cost of the guidance system.This thesis mainly includes:First,in order to improve the prediction accuracy of the parking space vacancy rate,a combined model(wavelet decomposition adaptive Kalman filter-neural network prediction model)is established.This method is to decompose the parking space vacancy rate sequence to obtain high-frequency sequence and low-frequency sequence.The low-frequency sequence uses an adaptive Kalman filter prediction model to conduct research and prediction,and the high-frequency sequence uses a neural network model to conduct research and prediction.Finally,the prediction results on the two frequencies are fused to obtain the final prediction sequence.Experiments show that the combined model has better prediction effect.Secondly,in order to solve the problem of parking lot tolerance,this thesis proposes a joint prediction model between parking lots based on Markov chain.First,find the collection of parking lots within the acceptable distance through the target parking lot.For this set of parking lots,a Markov chain is constructed by the Markov model,the corresponding transition matrix is generated,and the short-term predicted row vector is sent to the target parking lot agent for calculating cumulative prediction.Meanwhile,this can increase the probability of parking.Furthermore,a multi-attribute decision-making algorithm that maximizes dispersion is established to make the final recommendation.Using 6 attribute values as indicators,and dividing these indicators into different types for analysis.The value of the indicator parking feasibility comes from the value after prediction based on wavelet decomposition adaptive Kalman filter-neural network prediction model.The value of the index parking fault tolerance comes from the value based on the joint prediction between the parking spaces of the Markov chain,and finally the final recommendation result is comprehensively determined,and the recommendation result is displayed on the APP developed later.Finally,this thesis designed and developed a set of parking guidance system APP,the system contains tow parts.Server-side software includes: login registration module,data acquisition module,positioning and navigation module,prediction and decision module.The prediction module includes adaptive Kalman-filter prediction based on wavelet decomposition and joint prediction between parking lots based on Markov chain.The Android front-end includes: data request and analysis model,login registration interface module,parking list recommendation interface module,map marking interface module and path planning module.Finally,the parking guidance system APP developed in this thesis is tested.The test results show that the parking guidance system APP developed in this thesis effectively reduces the service cost of the guidance system.