Parking availability confidence interval prediction method and device, electronic equipment and medium

Abstract

The invention relates to a parking availability confidence interval prediction method and device, electronic equipment and a medium. The method comprises the steps of obtaining an initial arrival rate and an initial departure rate based on original observation data of a target parking lot; based on the initial arrival rate and the initial departure rate, utilizing a preset neural network model to respectively predict arrival rates and departure rates of a plurality of preset time intervals to obtain an arrival rate sequence and a departure rate sequence; and based on the arrival rate sequence and the departure rate sequence, utilizing a dynamic Markov queue model to predict probability distribution of the parking availability of each preset time interval, and obtaining a mean value and a confidence interval of the parking availability of each preset time interval according to the probability distribution. Therefore, the dynamic Markov queue model and the deep learning technology are combined to construct the parking availability confidence interval so as to evaluate the uncertainty of the parking availability, the problem that the user is difficult to park in the uncertainty environment is solved, and the operation efficiency of the whole traffic system is improved.

Description

Technical Field

[0001] The present application relates to the field of urban traffic prediction technology, and in particular to a prediction method, device, electronic device and medium for parking availability confidence interval. Background Art

[0002] As the number of vehicles in cities continues to grow, parking has become a common problem faced by drivers in many large cities. Studies have shown that the average time a driver spends looking for a parking space is about 3.5-14 minutes, and this process causes an increase in traffic volume by about 25% to 40%. In addition to the time wasted, the process of finding a parking space also causes traffic-related problems such as traffic congestion, as well as environmental problems such as air pollution. Parking availability prediction technology provides information on available parking spaces in parking lots around the destination, and is widely regarded as an effective strategy to solve parking problems and improve the overall parking efficiency of the transportation system.

[0003] In the related art, existing deep learning models all follow a point prediction framework and only predict the deterministic value (e.g., average value) of parking availability. In order to improve the prediction accuracy, researchers usually focus on the spatiotemporal correlation of the availability of parking spaces in different parking lots.

[0004] However, this method ignores the inherent uncertainty of the parking system. Due to the uncertainty of the parking environment, users face many difficulties when parking, which need to be solved urgently. Summary of the invention

[0005] The present application provides a prediction method, device, electronic device and medium for a parking availability confidence interval to solve the problem of difficulty in parking for users in an uncertain environment, provide users with information on the reliability of parking resources in the surrounding area of ​​the destination, and improve the operating efficiency of the entire transportation system.

[0006] To achieve the above-mentioned purpose, the first embodiment of the present application proposes a method for predicting a parking availability confidence interval, comprising the following steps:

[0007] Acquire original observation data of the target parking lot, and obtain an initial arrival rate and an initial departure rate based on the original observation data;

[0008] Based on the initial arrival rate and the initial departure rate, using a preset neural network model to respectively predict the arrival rate and departure rate of multiple preset time intervals to obtain an arrival rate sequence and a departure rate sequence;

[0009] Based on the arrival rate sequence and the departure rate sequence, a dynamic Markov queue model is used to predict the probability distribution of parking availability at each preset time interval, and a mean and a confidence interval of the parking availability at each preset time interval are obtained according to the probability distribution.

[0010] According to an embodiment of the present application, obtaining the initial arrival rate and the initial departure rate based on the original observation data includes:

[0011] Processing the original observation data by using a moving average strategy to obtain the processed original observation data;

[0012] Based on the processed original observation data, determining a first probability mass function for the process of a vehicle arriving at the target parking lot and a second probability mass function for the process of a vehicle leaving the target parking lot;

[0013] Based on the first probability mass function and the second probability mass function, obtaining the initial arrival rate and the initial departure rate.

[0014] According to an embodiment of the present application, predicting the arrival rate and the departure rate for multiple preset time intervals respectively by using a preset neural network model based on the initial arrival rate and the initial departure rate to obtain an arrival rate sequence and a departure rate sequence includes:

[0015] Based on the initial arrival rate and the initial departure rate, respectively obtaining the arrival rate for multiple preset periods and the departure rate for the multiple preset periods;

[0016] Using the dynamic spatio-temporal relationship between preset nodes and the Hadamard product strategy to determine the weight matrix of the preset neural network model;

[0017] Respectively inputting the arrival rate for the multiple preset periods and the departure rate for the multiple preset periods into the preset neural network model, and based on the weight matrix, obtaining the arrival rate output for the preset time interval and the departure rate output for the preset time interval;

[0018] Based on the arrival rate output for the preset time interval and the departure rate output for the preset time interval, respectively obtaining the time dependence of the arrival rate and the time dependence of the departure rate by using the residual connection layer and the gated recurrent unit of the preset neural network model;

[0019] According to the time dependence of the arrival rate and the time dependence of the departure rate, obtaining the arrival rate sequence and the departure rate sequence.

[0020] According to an embodiment of the present application, predicting the probability distribution of parking availability for each preset time interval by using a dynamic Markov queue model based on the arrival rate sequence and the departure rate sequence includes:

[0021] Obtaining the distribution of available parking spaces in the previous preset time interval of the current preset time interval;

[0022] Determine the transition probability of the current preset time interval by using the arrival rate sequence and the departure rate sequence;

[0023] Based on the distribution of available parking spaces in the previous preset time interval and the transition probability of the current preset time interval, use the dynamic Markov queue model to predict the probability distribution of parking availability in the current preset time interval.

[0024] According to an embodiment of the present application, the mean value of parking availability for each preset time interval is:

[0025]

[0026] where is the mean value of parking availability for each preset time interval, ψ + s is the current preset time interval, Round(·) is to force the input value, is the random variable form of the number of available parking spaces within the preset time interval, is the observed value of the number of available parking spaces within the preset time interval, is within the preset time interval The probability value of, ψ is the number of all preset time intervals, s is the s-th predicted preset time interval, v is the available parking space, i is from The index sampled from the distribution;

[0027] The confidence interval is:

[0028]

[0029] where is the confidence interval, is the lower bound of the confidence interval at the confidence level α / 2, is the upper bound of the confidence interval at the confidence level 1 - α / 2, and α is the preset significance level.

[0030] According to the prediction method of the parking availability confidence interval proposed by the embodiments of the present application, the initial arrival rate and the initial departure rate are obtained based on the original observation data of the target parking lot; based on the initial arrival rate and the initial departure rate, the arrival rate and the departure rate for multiple preset time intervals are respectively predicted by using a preset neural network model to obtain an arrival rate sequence and a departure rate sequence; based on the arrival rate sequence and the departure rate sequence, the probability distribution of the parking availability for each preset time interval is predicted by using a dynamic Markov queue model, and the mean value and the confidence interval of the parking availability for each preset time interval are obtained according to the probability distribution. Thus, by combining the dynamic Markov queue model with deep learning technology to construct the parking availability confidence interval to evaluate the uncertainty of the parking availability, the problem of difficult parking for users in an uncertain environment is solved, and reliable information about parking resources in the area around the destination is provided for users, improving the operation efficiency of the entire traffic system.

[0031] To achieve the above object, an embodiment of the second aspect of the present application proposes a prediction device for the parking availability confidence interval, including:

[0032] An obtaining module, configured to obtain the original observation data of the target parking lot, and obtain the initial arrival rate and the initial departure rate based on the original observation data;

[0033] A first prediction module, configured to respectively predict the arrival rate and the departure rate for multiple preset time intervals by using a preset neural network model based on the initial arrival rate and the initial departure rate to obtain an arrival rate sequence and a departure rate sequence;

[0034] A second prediction module, configured to predict the probability distribution of the parking availability for each preset time interval by using a dynamic Markov queue model based on the arrival rate sequence and the departure rate sequence, and obtain the mean value and the confidence interval of the parking availability for each preset time interval according to the probability distribution.

[0035] According to an embodiment of the present application, the obtaining module is specifically configured to:

[0036] Process the original observation data by using a moving average strategy to obtain the processed original observation data;

[0037] Based on the processed original observation data, determine a first probability mass function of the process of vehicles arriving at the target parking lot and a second probability mass function of the process of vehicles leaving the target parking lot;

[0038] Based on the first probability mass function and the second probability mass function, obtain the initial arrival rate and the initial departure rate.

[0039] According to an embodiment of the present application, the first prediction module is specifically configured to:

[0040] Based on the initial arrival rate and the initial departure rate, obtain the arrival rates of multiple preset periods and the departure rates of the multiple preset periods respectively;

[0041] Utilize the preset dynamic spatio-temporal relationship between nodes and the Hadamard product strategy to determine the weight matrix of the preset neural network model;

[0042] Input the arrival rates of the multiple preset periods and the departure rates of the multiple preset periods into the preset neural network model respectively, and based on the weight matrix, obtain the arrival rate output of the preset time interval and the departure rate output of the preset time interval;

[0043] Based on the arrival rate output of the preset time interval and the departure rate output of the preset time interval, utilize the residual connection layer and the gated recurrent unit of the preset neural network model to obtain the time dependence of the arrival rate and the time dependence of the departure rate respectively;

[0044] According to the time dependence of the arrival rate and the time dependence of the departure rate, obtain the arrival rate sequence and the departure rate sequence.

[0045] According to an embodiment of the present application, the second prediction module is specifically configured to:

[0046] Obtain the distribution of available parking spaces in the previous preset time interval of the current preset time interval;

[0047] Utilize the arrival rate sequence and the departure rate sequence to determine the transition probability of the current preset time interval;

[0048] Based on the distribution of available parking spaces in the previous preset time interval and the transition probability of the current preset time interval, utilize the dynamic Markov queue model to predict the probability distribution of the parking availability of the current preset time interval.

[0049] According to an embodiment of the present application, the mean value of the parking availability of each preset time interval is:

[0050]

[0051] Wherein, is the mean value of the parking availability of each preset time interval, ψ + s is the current preset time interval, Round(·) is to force the input value, is the random variable form of the number of available parking spaces within the preset time interval, is the observed value of the number of available parking spaces within the preset time interval, is within the preset time interval The probability value, ψ is the number of all preset time intervals, s is the s-th predicted preset time interval, v is the vacant parking space, and i is the index sampled from the distribution;

[0052] The confidence interval is:

[0053]

[0054] Wherein, is the confidence interval, is the lower bound of the confidence interval at the confidence level α / 2, is the upper bound of the confidence interval at the confidence level 1 - α / 2, and α is the preset significance level.

[0055] According to the prediction device for the parking availability confidence interval proposed in the embodiments of the present application, the initial arrival rate and the initial departure rate are obtained based on the original observation data of the target parking lot; based on the initial arrival rate and the initial departure rate, the arrival rate and the departure rate of multiple preset time intervals are respectively predicted by using a preset neural network model to obtain an arrival rate sequence and a departure rate sequence; based on the arrival rate sequence and the departure rate sequence, the probability distribution of the parking availability of each preset time interval is predicted by using a dynamic Markov queue model, and the mean value and the confidence interval of the parking availability of each preset time interval are obtained according to the probability distribution. Thus, by combining the dynamic Markov queue model with deep learning technology to construct the parking availability confidence interval to evaluate the uncertainty of the parking availability, the problem of difficult parking for users in an uncertain environment is solved, and reliable information about parking resources in the area around the destination is provided for users, improving the operation efficiency of the entire traffic system.

[0056] To achieve the above object, an electronic device is proposed in the third aspect embodiment of the present application, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the program to implement the prediction method for the parking availability confidence interval as described in the above embodiment.

[0057] To achieve the above object, a computer-readable storage medium is proposed in the fourth aspect embodiment of the present application, on which a computer program is stored, and the program is executed by a processor to be used to implement the prediction method for the parking availability confidence interval as described in the above embodiment.

[0058] To achieve the above object, a computer program product is proposed in the fifth embodiment of the present application, which includes a computer program, and when the computer program is executed by a processor, it is used to implement the prediction method for the parking availability confidence interval as described in the above embodiment.

[0059] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present application. Description of the Drawings

[0060] The above-mentioned and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of embodiments in conjunction with the drawings, in which:

[0061] Figure 1 is a flowchart of a method for predicting the parking availability confidence interval according to an embodiment of the present application;

[0062] Figure 2 is a schematic diagram of the area of test data according to an embodiment of the present application;

[0063] Figure 3 is a schematic diagram of an example for estimating the initial arrival rate and the initial departure rate according to an embodiment of the present application;

[0064] Figure 4 is a schematic diagram of the structure of a multi-task dynamic spatio-temporal graph convolutional network according to an embodiment of the present application;

[0065] Figure 5 is a schematic diagram of the prediction result of the parking availability on a certain parking lot according to an embodiment of the present application;

[0066] Figure 6 is a block schematic diagram of a device for predicting the parking availability confidence interval according to an embodiment of the present application;

[0067] Figure 7 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Detailed Embodiments

[0068] Embodiments of the present application will be described in detail below. Examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present application and should not be construed as limiting the present application.

[0069] The method, device, electronic device, and medium for predicting the parking availability confidence interval according to the embodiments of the present application will be described below with reference to the drawings. First, the method for predicting the parking availability confidence interval according to the embodiments of the present application will be described with reference to the drawings.

[0070] Figure 1 is a flowchart of a method for predicting the parking availability confidence interval according to an embodiment of the present application.

[0071] Exemplarily, as Figure 1As shown in the figure, the prediction method for the parking availability confidence interval includes the following steps:

[0072] In step S101, obtain the original observation data of the target parking lot, and obtain the initial arrival rate and the initial departure rate based on the original observation data.

[0073] It can be understood that the initial arrival rate refers to the number of vehicles entering the target parking lot per unit time; the initial departure rate refers to the number of vehicles leaving the target parking lot per unit time.

[0074] That is to say, first, it is necessary to obtain the original observation data of the target parking lot. These data include, but are not limited to, the time records of vehicles entering and leaving the target parking lot, the occupancy of parking spaces in the target parking lot, and any external factors that may affect vehicle flow. By collecting these detailed data, a preliminary understanding of the usage pattern of the target parking lot can be obtained. Next, based on these original observation data, the initial arrival rate and the initial departure rate can be calculated. These two indicators are crucial for understanding the dynamic balance of the target parking lot, because they can predict the congestion level of the target parking lot at different times, thereby providing a scientific basis for the management and optimization of the target parking lot.

[0075] For example, as Figure 2 shown, in consideration of data availability, the embodiments of the present application select two real roadside parking lot datasets in Los Angeles to conduct experimental demonstrations. The LA_DOWNTOWN and LA_HOLLYWOOD datasets in the research area respectively contain 50 roadside parking lots and 27 roadside parking lots.

[0076] For the convenience of understanding, the following will detail how to obtain the initial arrival rate and the initial departure rate based on the original observation data.

[0077] As a possible implementation manner, in some embodiments, obtaining the initial arrival rate and the initial departure rate based on the original observation data includes: processing the original observation data by using a moving average strategy to obtain the processed original observation data; determining a first probability mass function for the process of vehicles arriving at the target parking lot and a second probability mass function for the process of vehicles leaving the target parking lot based on the processed original observation data; and obtaining the initial arrival rate and the initial departure rate based on the first probability mass function and the second probability mass function.

[0078] It can be understood that the moving average strategy is a technique for smoothing data, which can help eliminate short-term fluctuations in the data so as to more clearly see the long-term trend. The probability mass function (PMF) is a function that describes the probability of a discrete random variable taking a specific value.

[0079] Specifically, after obtaining the original observation data, the embodiments of the present application can use a moving average strategy to smooth the original observation data to obtain a more accurate and reliable observation data sequence (i.e., the processed original observation data). Based on the processed original observation data, the embodiments of the present application can represent multiple target parking lots in the urban road network as a weighted directed graph This directed graph consists of a vertex set V (i.e., the set of target parking lots), an edge set E (i.e., the set of edges between nodes), and a weight matrix set A (i.e., the adjacency matrix). Each vertex in the graph represents each target parking lot, and the edges and weights represent the road connections and road capacities between the target parking lots respectively. The number of idle parking spaces within a preset time interval t the number of vehicles arriving at the target parking lot and the number of vehicles leaving the target parking lot are represented as random variables. Assume that the number of arriving vehicles and the number of leaving vehicles within the preset time interval t both follow the Poisson distribution (the Poisson distribution is a probability distribution that describes the number of occurrences of random events within a fixed time or space), that is, it indicates that in the scenario of a parking lot, although the number of vehicles arriving at and leaving the target parking lot is random, within a certain time interval, the number of their occurrences has an expected average value. and The probability distributions of can be estimated from the corresponding observed values (i.e., the number of arriving vehicles and the number of leaving vehicles in the original observation data). The first probability mass function of the vehicle arrival process at the target parking lot can be expressed as:

[0080]

[0081] where, is the number of vehicles arriving at the target parking lot in the original observation data, is the number of vehicles arriving at the target parking lot in the t th th preset time interval, λ t is the initial arrival rate, and t is the preset time interval.

[0082] Similarly, the second probability mass function of the vehicle departure process from the target parking lot can be expressed as:

[0083]

[0084] where, is the number of vehicles leaving the target parking lot in the original observation data, is the number of vehicles leaving the target parking lot in the t th th preset time interval, pt is the initial departure rate, and t is the preset time interval.

[0085] Without loss of generality, assume that the vehicle arrival and departure patterns are stable within the time interval (t - δ, t + δ). Then, through the average observations within this time interval, the initial arrival rate λ t and the initial departure rate p t can be estimated, that is:

[0086]

[0087] where is the number of arriving vehicles within the preset time interval 2δ, is the number of departing vehicles within the preset time interval 2δ, and δ is half of the sliding window size.

[0088] Furthermore, the time series of the initial arrival rate is:

[0089] λ = {λ 1 , …, λ ψ} T ;

[0090] The time series of the initial departure rate is:

[0091] p = {p 1 , …, p ψ} T ;

[0092] where ψ is the number of preset time intervals.

[0093] As Figure 3 shown, it presents an example graph of the estimated initial arrival rate and initial departure rate obtained by performing a moving average on the processed original observation data within a sliding window of size 2δ.

[0094] In step S102, based on the initial arrival rate and the initial departure rate, the arrival rates and departure rates for multiple preset time intervals are predicted respectively using a preset neural network model, obtaining an arrival rate sequence and a departure rate sequence.

[0095] Among them, the preset neural network model refers to a multi - task dynamic spatio - temporal graph convolutional network (Dynamic Graph Convolutional Network, abbreviated as DGCN).

[0096] That is to say, based on the initially estimated arrival rate λ and departure rate p obtained in step S101, the embodiments of the present application can use DGCN to predict the arrival rate sequence and departure rate sequence for the next n preset time intervals. Among them, DGCN includes two interconnected branches, namely, the arrival rate prediction branch and the departure rate prediction branch. These two branches use the same structure and can be used to simultaneously predict the arrival rate sequence and departure rate sequence for the next n preset time intervals.

[0097] Next, a detailed description will be given on how to respectively predict the arrival rate and departure rate for multiple preset time intervals based on the initial arrival rate and initial departure rate, using a preset neural network model, to obtain the arrival rate sequence and departure rate sequence.

[0098] As a possible implementation manner, in some embodiments, based on the initial arrival rate and initial departure rate, using a preset neural network model to respectively predict the arrival rate and departure rate for multiple preset time intervals, to obtain the arrival rate sequence and departure rate sequence, includes: respectively obtaining the arrival rate for multiple preset cycles and the departure rate for multiple preset cycles based on the initial arrival rate and initial departure rate; determining the weight matrix of the preset neural network model using the dynamic spatio-temporal relationship between preset nodes and the Hadamard product strategy; respectively inputting the arrival rate for multiple preset cycles and the departure rate for multiple preset cycles into the preset neural network model, and based on the weight matrix, obtaining the arrival rate output for the preset time interval and the departure rate output for the preset time interval; based on the arrival rate output for the preset time interval and the departure rate output for the preset time interval, using the residual connection layer and gated recurrent unit of the preset neural network model to respectively obtain the time dependence of the arrival rate and the time dependence of the departure rate; and obtaining the arrival rate sequence and departure rate sequence according to the time dependence of the arrival rate and the time dependence of the departure rate.

[0099] Specifically, taking the arrival rate prediction branch as an example, considering the multi-periodicity of the estimated arrival rate input sequence, including recent (i.e., ), daily-period (i.e., ), and weekly-period (i.e., ) segments, the arrival rate for multiple preset cycles is input into DGCN to learn the spatial dependence of the vehicle arrival patterns between all target parking lots in the parking system. By considering the dynamic characteristics of the spatial relationship in different time periods, DGCN realizes the extension of the classical static GCN (Graph Convolutional Network). DGCN introduces a learnable dynamic node embedding NE, and through the training of the neural network model, the dynamic spatial relationship between nodes can be generated, expressed as:

[0100]

[0101] Among them, A d is the dynamic weight matrix, and NE·NE T is the similarity between node embeddings, which is used to calculate the dynamic weight matrix; ReLU(x) = Max(0, x) is used to ensure non-linearity; is used to normalize the dynamic weight matrix A d and is the degree matrix, is the normalized adjacency matrix.

[0102] To maintain the interpretability of the weight matrix, the final weight matrix A f (i.e., the weight matrix of the preset neural network model) is achieved by using the Hadamard product between the dynamic matrix A d and the static 0-1 adjacency matrix A s , that is:

[0103] A f = A d ⊙ A s .

[0104] Therefore, the DGCN operation can be expressed as:

[0105]

[0106] Among them, is the output of DGCN during the preset time interval t (i.e., the arrival rate output of the preset time interval), X t is the input of the preset neural network model, and W t is the parameter of the preset neural network model.

[0107] Furthermore, the input and output of DGCN are passed into the residual connection layer to stabilize the training process, and the output can be obtained and further input into the Gated Recurrent Unit (GRU) to capture the time dependence of the arrival rate, thereby obtaining the output:

[0108]

[0109] Among them, is the hidden output of the GRU unit in the arrival rate prediction branch, and L is the window size of the input sequence.

[0110] Similarly, the output of the departure rate prediction branch passing through GRU is:

[0111]

[0112] Among them, is the hidden output of the GRU unit in the departure rate prediction branch.

[0113] As Figure 4 shown, under the multi-task learning framework, the soft parameter sharing unit is used to connect the corresponding GRU networks in the two prediction branches, so as to realize the sharing of hidden layer parameters. Among them, the fusion mode in the soft parameter sharing unit is "weighted fusion", which is expressed as:

[0114]

[0115] Among them, ψ s is the adjustment factor.

[0116] Subsequently, through the fully connected layer (Fully Connected Layer, abbreviated as FC), the final output of the arrival rate prediction branch (i.e., the arrival rate sequence) is Similarly, the final output of the departure rate prediction branch (i.e., the departure rate sequence) is

[0117] In step S103, based on the arrival rate sequence and the departure rate sequence, use the dynamic Markov queue model to predict the probability distribution of parking availability for each preset time interval, and obtain the mean value and confidence interval of parking availability for each preset time interval according to the probability distribution.

[0118] It can be understood that the dynamic Markov queue model is a mathematical model that can be used to predict random processes that change over time, such as the dynamic changes of vehicle arrivals and departures from the parking lot. Parking availability refers to the number of parking spaces available in the parking lot at a specific time point or time period. The probability distribution is used to represent the likelihood of different parking availability situations occurring. The confidence interval refers to the numerical range within which the parking availability may fall under a certain probability guarantee.

[0119] That is to say, based on the predicted arrival rate sequence and the departure rate sequence the dynamic Markov queue model can be used to predict the probability distribution of parking availability for the next n preset time intervals, and then infer the mean value and confidence interval of parking availability for the next n preset time intervals according to the probability distribution.

[0120] Next, the process of how to predict the probability distribution of parking availability for each preset time interval using the dynamic Markov queue model based on the arrival rate sequence and the departure rate sequence will be elaborated in detail.

[0121] As a possible implementation method, in some embodiments, based on the arrival rate sequence and the departure rate sequence, a dynamic Markov queue model is used to predict the probability distribution of parking availability for each preset time interval, including: obtaining the distribution of free parking spaces in the previous preset time interval of the current preset time interval; using the arrival rate sequence and the departure rate sequence to determine the transition probability of the current preset time interval; based on the distribution of free parking spaces in the previous preset time interval and the transition probability of the current preset time interval, using the dynamic Markov queue model to predict the probability distribution of parking availability for the current preset time interval.

[0122] Specifically, using the dynamic Markov queue model, the free parking space distribution It can be obtained by iterative calculation through the previous preset time interval ψ+s-1. For parking lot The distribution of free parking spaces within the preset time interval ψ+s (i.e., the current preset time interval) is a probability distribution, which includes C l The discrete states {0,…,j,…,C l}, where C l For parking lot Capacity. Forecast This is equivalent to predicting its probability mass function, namely:

[0123]

[0124] Assume that the distribution within the time interval ψ+s-1 (i.e. the previous preset time interval) It has been determined that its probability mass function can be expressed as:

[0125]

[0126] According to the Markov property, we can determine:

[0127]

[0128] in, is the transition probability It can be further expressed as:

[0129]

[0130] Where k is the number of vehicles leaving within the preset time interval ψ+s, k+ij is the number of vehicles arriving within the same preset time interval, is the first probability mass function mentioned above, is the second probability mass function mentioned above, is the predicted arrival rate, is the predicted departure rate.

[0131] By calculating the transition probabilities of all possible states the number of available parking spaces in the parking lot at the preset time interval ψ + s can be obtained Therefore, by calculating for all target parking lots within the preset time interval ψ + s the probability distribution of parking availability can be obtained

[0132] After obtaining the probability distribution of parking availability the confidence interval at a certain confidence level (1 - α)% can be obtained to quantify the uncertainty of available parking spaces, expressed as:

[0133]

[0134] where is the confidence interval, is the lower bound of the confidence interval at the confidence level α / 2, is the upper bound of the confidence interval at the confidence level 1 - α / 2, and α is the preset significance level.

[0135] where

[0136]

[0137] where is the inverse cumulative distribution function, and Round(·) forces the input value to be an integer.

[0138] The mean of parking availability can be expressed as:

[0139]

[0140] where is the mean of parking availability for each preset time interval, ψ + s is the current preset time interval, Round(·) is the forced input value, is the random variable form of the number of available parking spaces within the preset time interval, is the observed value of the number of available parking spaces within the preset time interval, is within the preset time interval the probability value, ψ is the number of all preset time intervals, s is the s-th predicted preset time interval, v is the available parking space, and i is the index sampled from the distribution.

[0141] Such as Figure 5As shown, an example of the prediction results of parking availability in the FLOWER ST parking lot is presented. It can be seen that the prediction results of parking availability not only have excellent point prediction performance; at the same time, the predicted confidence intervals (gray areas) can well cover the observed values (red dots) of the number of available parking spaces, and it achieves a confidence interval coverage rate far exceeding the confidence level (10%) of 6.2% with a reasonable interval width.

[0142] According to the prediction method of the parking availability confidence interval proposed in the embodiments of the present application, the initial arrival rate and the initial departure rate are obtained based on the original observation data of the target parking lot; based on the initial arrival rate and the initial departure rate, the arrival rate and the departure rate for multiple preset time intervals are respectively predicted by using a preset neural network model to obtain an arrival rate sequence and a departure rate sequence; based on the arrival rate sequence and the departure rate sequence, the probability distribution of the parking availability for each preset time interval is predicted by using a dynamic Markov queue model, and the mean value and the confidence interval of the parking availability for each preset time interval are obtained according to the probability distribution. Thus, by combining the dynamic Markov queue model with deep learning technology to construct the parking availability confidence interval to evaluate the uncertainty of parking availability, the problem of difficult parking for users in an uncertain environment is solved, providing users with reliable information on parking resources in the surrounding area of the destination, and improving the operation efficiency of the entire transportation system.

[0143] Next, the prediction device for the parking availability confidence interval proposed in the embodiments of the present application will be described with reference to the accompanying drawings.

[0144] Figure 6 It is a block diagram of the prediction device for the parking availability confidence interval according to an embodiment of the present application.

[0145] As shown in Figure 6 the prediction device 10 for the parking availability confidence interval includes: an acquisition module 100, a first prediction module 200, and a second prediction module 300.

[0146] Among them, the acquisition module 100 is used to obtain the original observation data of the target parking lot and obtain the initial arrival rate and the initial departure rate based on the original observation data;

[0147] The first prediction module 200 is used to respectively predict the arrival rate and the departure rate for multiple preset time intervals by using a preset neural network model based on the initial arrival rate and the initial departure rate to obtain an arrival rate sequence and a departure rate sequence;

[0148] The second prediction module 300 is used to predict the probability distribution of the parking availability for each preset time interval by using a dynamic Markov queue model based on the arrival rate sequence and the departure rate sequence, and obtain the mean value and the confidence interval of the parking availability for each preset time interval according to the probability distribution.

[0149] Further, in some embodiments, the obtaining module 100 is specifically configured to:

[0150] Process the original observation data by using a moving average strategy to obtain the processed original observation data;

[0151] Based on the processed original observation data, determine a first probability mass function for the process of the vehicle arriving at the target parking lot and a second probability mass function for the process of the vehicle leaving the target parking lot;

[0152] Based on the first probability mass function and the second probability mass function, obtain an initial arrival rate and an initial departure rate.

[0153] Further, in some embodiments, the first prediction module 200 is specifically configured to:

[0154] Based on the initial arrival rate and the initial departure rate, respectively obtain the arrival rates for multiple preset periods and the departure rates for multiple preset periods;

[0155] Use the preset dynamic spatio-temporal relationship between nodes and the Hadamard product strategy to determine the weight matrix of the preset neural network model;

[0156] Respectively input the arrival rates for multiple preset periods and the departure rates for multiple preset periods into the preset neural network model, and based on the weight matrix, obtain the arrival rate output for a preset time interval and the departure rate output for a preset time interval;

[0157] Based on the arrival rate output for a preset time interval and the departure rate output for a preset time interval, use the residual connection layer and the gated recurrent unit of the preset neural network model to respectively obtain the time dependence of the arrival rate and the time dependence of the departure rate;

[0158] According to the time dependence of the arrival rate and the time dependence of the departure rate, obtain an arrival rate sequence and a departure rate sequence.

[0159] Further, in some embodiments, the second prediction module 300 is specifically configured to:

[0160] Obtain the distribution of available parking spaces in the previous preset time interval of the current preset time interval;

[0161] Use the arrival rate sequence and the departure rate sequence to determine the transition probability of the current preset time interval;

[0162] Based on the distribution of available parking spaces in the previous preset time interval and the transition probability of the current preset time interval, use the dynamic Markov queue model to predict the probability distribution of parking availability in the current preset time interval.

[0163] According to an embodiment of the present application, the mean of the parking availability for each preset time interval is:

[0164]

[0165] Wherein, is the mean of the parking availability for each preset time interval, ψ + s is the current preset time interval, Round(·) is to force the input value, is the random variable form of the number of available parking spaces within the preset time interval, is the observed value of the number of available parking spaces within the preset time interval, within the preset time interval is the probability value, ψ is the number of all preset time intervals, s is the s-th predicted preset time interval, v is the available parking space, i is from the index sampled from the

[0166] The confidence interval is:

[0167]

[0168] Wherein, is the confidence interval, is the lower bound of the confidence interval at the confidence level ɑ / 2, is the upper bound of the confidence interval at the confidence level 1 - ɑ / 2, and ɑ is the preset significance level.

[0169] It should be noted that the foregoing explanation of the embodiment of the prediction method for the parking availability confidence interval also applies to the prediction device for the parking availability confidence interval of this embodiment, and will not be elaborated here.

[0170] The prediction device for the parking availability confidence interval proposed according to the embodiment of the present application obtains the initial arrival rate and the initial departure rate based on the original observation data of the target parking lot; based on the initial arrival rate and the initial departure rate, uses a preset neural network model to predict the arrival rate and the departure rate for multiple preset time intervals respectively to obtain an arrival rate sequence and a departure rate sequence; based on the arrival rate sequence and the departure rate sequence, uses a dynamic Markov queue model to predict the probability distribution of the parking availability for each preset time interval, and obtains the mean and the confidence interval of the parking availability for each preset time interval according to the probability distribution. Thus, by combining the dynamic Markov queue model with deep learning technology to construct the parking availability confidence interval to evaluate the uncertainty of the parking availability, the problem of difficult parking for users in an uncertain environment is solved, reliable information about parking resources in the surrounding area of the destination is provided for users, and the operation efficiency of the entire traffic system is improved.

[0171] Figure 7Schematic diagram of the structure of the electronic device provided by the embodiments of the present application. The electronic device may include:

[0172] A memory 701, a processor 702, and a computer program stored on the memory 701 and executable on the processor 702.

[0173] When the processor 702 executes the program, it implements the prediction method for the parking availability confidence interval provided in the above embodiments.

[0174] Furthermore, the electronic device further includes:

[0175] A communication interface 703 for communication between the memory 701 and the processor 702.

[0176] The memory 701 is used to store a computer program executable on the processor 702.

[0177] The memory 701 may include a high-speed RAM (Random Access Memory) memory, and may also include a non-volatile memory, such as at least one disk memory.

[0178] If the memory 701, the processor 702, and the communication interface 703 are implemented independently, the communication interface 703, the memory 701, and the processor 702 may be interconnected through a bus and communicate with each other. The bus may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, Figure 7 only a thick line is shown in the figure, but it does not mean that there is only one bus or one type of bus.

[0179] Optionally, in a specific implementation, if the memory 701, the processor 702, and the communication interface 703 are integrated on a chip, the memory 701, the processor 702, and the communication interface 703 may communicate with each other through an internal interface.

[0180] The processor 702 may be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0181] The embodiments of the present application further provide a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the prediction method of the parking availability confidence interval as described above is implemented.

[0182] The embodiments of the present application further provide a computer program product, which includes a computer program, and when the computer program is executed by a processor, the prediction method of the parking availability confidence interval as described above is implemented.

[0183] In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0184] In the description of this specification, the descriptions with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc., mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms are not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples.

[0185] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for predicting a parking availability confidence interval, characterized in that: The following steps are involved: Acquire original observation data of the target parking lot, and obtain an initial arrival rate and an initial departure rate based on the original observation data; Based on the initial arrival rate and the initial departure rate, using a preset neural network model to respectively predict the arrival rate and departure rate of multiple preset time intervals to obtain an arrival rate sequence and a departure rate sequence; Based on the arrival rate sequence and the departure rate sequence, a dynamic Markov queue model is used to predict the probability distribution of parking availability at each preset time interval, and a mean and a confidence interval of the parking availability at each preset time interval are obtained according to the probability distribution.

2. The method according to claim 1, characterized in that: The obtaining of the initial arrival rate and the initial departure rate based on the original observation data comprises: Processing the original observation data using a moving average strategy to obtain processed original observation data; Based on the processed original observation data, determining a first probability mass function of a process in which a vehicle arrives at the target parking lot and a second probability mass function of a process in which a vehicle leaves the target parking lot; The initial arrival rate and the initial departure rate are obtained based on the first probability mass function and the second probability mass function.

3. The method according to claim 1, characterized in that The method of predicting the arrival rate and the departure rate of a plurality of preset time intervals based on the initial arrival rate and the initial departure rate using a preset neural network model to obtain an arrival rate sequence and a departure rate sequence includes: Based on the initial arrival rate and the initial departure rate, respectively obtaining arrival rates of a plurality of preset periods and departure rates of the plurality of preset periods; Determine the weight matrix of the preset neural network model by using the dynamic spatiotemporal relationship between preset nodes and the Hadamard product strategy; Inputting the arrival rates of the plurality of preset periods and the departure rates of the plurality of preset periods into the preset neural network model respectively, and obtaining the arrival rate output of the preset time interval and the departure rate output of the preset time interval based on the weight matrix; Based on the arrival rate output at the preset time interval and the departure rate output at the preset time interval, the time dependency of the arrival rate and the time dependency of the departure rate are respectively obtained using the residual connection layer and the gated recurrent unit of the preset neural network model; The arrival rate sequence and the departure rate sequence are obtained according to the time dependency of the arrival rate and the time dependency of the departure rate.

4. The method according to claim 1, characterized in that: The method of predicting the probability distribution of parking availability at each preset time interval by using a dynamic Markov queue model based on the arrival rate sequence and the departure rate sequence includes: Obtaining the distribution of free parking spaces in the previous preset time interval before the current preset time interval; Determine the transition probability of the current preset time interval by using the arrival rate sequence and the departure rate sequence; Based on the distribution of free parking spaces in the last preset time interval and the transition probability in the current preset time interval, the dynamic Markov queue model is used to predict the probability distribution of parking availability in the current preset time interval.

5. The method according to claim 1, characterized in that The average parking availability for each preset time interval is: in, is the mean of the parking availability in each preset time interval, ψ+s is the current preset time interval, Round(·) is a mandatory input value, is a random variable of the number of free parking spaces within the preset time interval, is the observed value of the number of free parking spaces within the preset time interval, For the preset time interval The probability value of ψ is the number of all preset time intervals, s is the sth predicted preset time interval, v is the vacant parking space, and i is the number of The index of the sample in the distribution; The confidence interval is: in, is the confidence interval, is the lower bound of the confidence interval at confidence level α / 2, is the upper bound of the confidence interval at the confidence level 1-α / 2, where α is the preset significance level.

6. A parking availability confidence interval prediction device, characterized in that: include: An acquisition module, used to acquire original observation data of the target parking lot, and obtain an initial arrival rate and an initial departure rate based on the original observation data; A first prediction module is used to predict the arrival rate and the departure rate of multiple preset time intervals respectively based on the initial arrival rate and the initial departure rate by using a preset neural network model to obtain an arrival rate sequence and a departure rate sequence; The second prediction module is used to predict the probability distribution of parking availability at each preset time interval based on the arrival rate sequence and the departure rate sequence using a dynamic Markov queue model, and obtain the mean and confidence interval of the parking availability at each preset time interval according to the probability distribution.

7. The device according to claim 6, characterized in that The obtaining module is specifically used for: Processing the original observation data using a moving average strategy to obtain processed original observation data; Based on the processed original observation data, determining a first probability mass function of a process in which a vehicle arrives at the target parking lot and a second probability mass function of a process in which a vehicle leaves the target parking lot; The initial arrival rate and the initial departure rate are obtained based on the first probability mass function and the second probability mass function.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for predicting a parking availability confidence interval according to any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the method for predicting a parking availability confidence interval as described in any one of claims 1 to 5.

10. A computer program product, characterized in that The method comprises a computer program, which, when executed by a processor, is used to implement the method for predicting a parking availability confidence interval according to any one of claims 1 to 5.

Images