Regional public parking space demand prediction method based on parking big data

Abstract

The invention discloses a regional public parking space demand prediction method based on parking big data. The method comprises the following steps: a vehicle in a parking space pre-occupation region sends a pre-occupation signal; constructing a real-time parking lot map model; analyzing a parking space distribution scheme set according to the real-time pre-parking condition based on a parking lot map model; calculating parking paths of all parking space distribution schemes in the parking space distribution scheme set based on a parking space distribution model to screen an optimal parking space distribution scheme, and updating a real-time pre-parking condition; and establishing an SVM parking space prediction model, bringing the real-time pre-parking condition into a training sample, optimizing parameters in the SVM parking space prediction model by adopting a CS algorithm based on the training sample, and predicting the number of non-idle parking spaces. According to the invention, the parking data in the parking lot for a period of time in the future are obtained in advance through a pre-occupation mode, and the parking condition in the future time is predicted through the combination of the current data and the future data, so that the prediction accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of parking lot information prediction, and in particular to a method for predicting regional public parking space demand based on parking big data. Background Art

[0002] As the number of cars in cities increases, the demand for parking spaces is increasing. The current parking planning schemes in various cities have many deficiencies, resulting in parking difficulties in certain periods and regardless of area in the city. In the formulation of parking planning schemes, accurate parking demand forecasts are an important reference. Reasonable parking planning strategies based on accurate parking demand forecasts are conducive to avoiding parking resource shortages or waste in the allocation of parking resources in various areas of the city.

[0003] In order to achieve effective utilization of parking resources, the current information sharing method is adopted to provide car owners with real-time parking lot information before they arrive at their destination. By deploying parking hardware equipment in the parking lot, information on changes in parking spaces in the parking lot is collected, and the information collected by the equipment is transmitted to the cloud through the network. The cloud then synchronizes the collected information to the user terminal, or predicts parking demand based on the usage of parking lot resources in different time periods.

[0004] However, in the prior art, the method of predicting parking demand for a certain period in the future only through real-time parking lot resource usage may result in a point in time showing that parking resources in the future are mostly idle, a large number of vehicles flock to the parking lot during this period, and insufficient parking spaces, causing parking lot resource usage to easily fluctuate greatly according to the prediction results, resulting in unstable parking lot resource allocation and poor timeliness. Summary of the invention

[0005] To this end, the present invention provides a method for predicting regional public parking space demand based on parking big data, which effectively solves the problem in the prior art that parking lot resource usage is prone to huge fluctuations according to prediction results, resulting in unstable parking lot resource allocation and poor timeliness.

[0006] In order to solve the above technical problems, the present invention specifically provides the following technical solutions: a method for predicting regional public parking space demand based on parking big data, comprising the following steps:

[0007] A parking space pre-occupancy area is set, and vehicles in the parking space pre-occupancy area send out a pre-occupancy signal;

[0008] Construct a real-time parking lot map model based on the parking space locations inside the parking lot according to the pre-occupancy signal;

[0009] Obtain the real-time pre-parking situation of the parking lot, and analyze the parking space allocation plan set according to the real-time pre-parking situation based on the parking lot map model;

[0010] Establish a parking space allocation model, calculate the parking paths of all parking space allocation schemes in the parking space allocation scheme set based on the parking space allocation model, screen out the optimal parking space allocation scheme, and update the real-time pre-parking situation;

[0011] Establish an SVM parking space prediction model, obtain training samples, incorporate real-time pre-parking conditions into the training samples, and use the CS algorithm based on the training samples to optimize the parameters in the SVM parking space prediction model;

[0012] The number of non-idle parking spaces is predicted based on the optimized SVM parking space prediction model.

[0013] Further,

[0014] After the optimal parking space allocation plan is screened out, the parking spaces corresponding to the optimal parking space allocation plan are pre-occupied;

[0015] Update the real-time pre-parking situation based on the occupied parking spaces.

[0016] Further,

[0017] The process of establishing the parking lot map model includes the following steps:

[0018] Based on the parking space location, size, and spacing parameters inside the parking lot, a number of corresponding path nodes and path edges formed by connecting the path nodes are constructed;

[0019] A connection relationship diagram between path nodes is drawn based on path nodes and path edges between different path nodes to obtain a parking lot map model.

[0020] Further,

[0021] After establishing the parking lot map model, collect real-time parking conditions inside the parking lot;

[0022] Based on the parking situation inside the parking lot, the path edges of the path nodes corresponding to some occupied parking spaces are screened out to obtain a real-time parking lot map model.

[0023] Further,

[0024] The parking space allocation scheme set is analyzed based on the parking lot map model according to the real-time pre-parking situation, specifically including the following steps:

[0025] Obtain the real-time pre-parking status of the parking lot and obtain the pre-occupied parking spaces in the parking lot;

[0026] Based on the occupied parking spaces, path edges of path nodes corresponding to some of the occupied parking spaces are screened out to correct the parking lot map model;

[0027] Analyze the set of parking space allocation solutions based on the revised parking lot map model.

[0028] Further,

[0029] Analyze the parking space allocation scheme set, which specifically includes the following steps:

[0030] Establish a set W of vehicles to be occupied and vehicles to be parked:

[0031] W=(w1,w2,…,w i );

[0032] In the formula, w represents the vehicles to be reserved and the vehicles to be parked, i represents the number of vehicles to be reserved and the vehicles to be parked;

[0033] Create a set of free parking spaces P:

[0034] P=(p1,p2,…,p j );

[0035] j = J - (j1 + j2);

[0036] In the formula, p represents the vacant parking space, j represents the number of vacant parking spaces, J represents the total number of parking spaces, j1 represents the number of occupied parking spaces, and j2 represents the number of reserved parking spaces;

[0037] Based on the set of vehicles to be occupied, the set of vehicles to be parked W and the set of free parking spaces P, the parking space allocation solution set T is obtained:

[0038] T=(t1,t2,…,t n );

[0039] Where t represents the possible allocation scheme of the vehicles to be occupied or parked w to the vacant parking spaces p, and n represents the number of possible schemes.

[0040] Further,

[0041] Build a parking space allocation model:

[0042] F(T) = min(f(T));

[0043] Where f(T) represents the sum of the distances of the parking paths of the parking space allocation scheme, and min(f(T)) represents the minimum value of the sum of the distances of the parking paths of all parking space allocation schemes;

[0044] Calculate the sum of the parking path distances of all parking allocation schemes in the parking allocation scheme set:

[0045]

[0046] L(S)=∑e∈E w e ;

[0047] Where S represents the parking path of the vehicle to be occupied or parked w allocated to the vacant parking space p corresponding to the i-th parking space allocation scheme, and L(S) represents the distance of the parking path of the vehicle to be occupied or parked w allocated to the vacant parking space p corresponding to the i-th parking space allocation scheme;

[0048] E represents all the path edges of the path nodes corresponding to the parking path S, w e Represents the length of the e-th path edge.

[0049] Further,

[0050] Based on the parking space allocation model, the parking space allocation plan with the smallest total parking path distance among all parking space allocation plans is selected as the optimal parking space allocation plan.

[0051] Further,

[0052] The training samples include real-time parking conditions at different time periods in the parking lot;

[0053] Before obtaining the training samples, obtain the measured data of the parking lot and divide the measured data into training samples and test samples;

[0054] The real-time pre-parking situation is converted into pre-occupancy data of parking spaces and incorporated into the training samples to update the data in the training samples.

[0055] Further,

[0056] The CS algorithm is used to optimize the parameters in the SVM parking space prediction model, which specifically includes the following steps:

[0057] Preprocess the parking space data in the training samples;

[0058] Preliminarily determine the parameters of the CS algorithm and the thresholds of the penalty coefficient C and kernel function width parameter σ;

[0059] Set the evaluation index of the SVM parking space prediction model:

[0060] ;

[0061] In the formula, M is the evaluation index, i and n represent the order of samples, and x i Represents the model prediction value, y i represents the actual value of the training sample;

[0062] In the CS algorithm, the bird’s nest position is initialized and substituted into the SVM parking space prediction model to find the evaluation index value and retain the optimal bird’s nest position;

[0063] Update the bird's nest position and substitute it into the SVM parking space prediction model to find the current evaluation index value, compare the current evaluation index value with the previous evaluation index value, and retain the bird's nest position with a smaller evaluation index value;

[0064] After the global update is completed, a local search is performed;

[0065] After the local search is completed, the evaluation index values ​​corresponding to all the bird nest positions are sorted by size, the last n bird nests are eliminated, and n new bird nests are generated to calculate their evaluation index values;

[0066] Determine whether the iteration is terminated. If not, continue the iteration. If it is terminated, find the optimal solution of the penalty coefficient C and kernel function width parameter σ at this time.

[0067] Based on the optimized SVM parking space prediction model, the optimal solution of the penalty coefficient C and the kernel function width parameter σ is substituted into the SVM parking space prediction model for training, and the final predicted value of the number of non-vacant parking spaces is output;

[0068] Among them, non-vacant parking spaces include parking spaces that are already occupied and occupied parking spaces according to the actual measured data of the parking lot.

[0069] Further,

[0070] The parking space pre-occupied area is a circular area with the parking lot as the center and the pre-occupied distance r as the radius.

[0071] Compared with the prior art, the present invention has the following beneficial effects:

[0072] In the present invention, a parking space pre-occupancy area is set, and before the vehicle reaches the destination parking lot, the parking space in the parking lot is pre-occupied. The real-time parking situation and pre-occupancy situation in the parking lot are included in the training samples, and an SVM parking space prediction model is established. The parameters in the SVM parking space prediction model are optimized based on the training samples, and the number of non-vacant parking spaces is predicted based on the optimized SVM parking space prediction model. The parking spaces in the parking lot are pre-occupied, which can not only avoid being unable to park in the parking lot when arriving at a specific time, but also obtain the parking data of the parking lot for a period of time in the future in advance through the pre-occupancy method. The parking situation in the future time is predicted by combining the current data and the future data, thereby improving the accuracy of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0074] Figure 1 A flowchart of a method for predicting regional public parking space demand based on parking big data provided by an embodiment of the present invention;

[0075] Figure 2 A schematic diagram showing the actual parking space locations, parking space conditions, and parking reservation conditions within a small parking lot;

[0076] Figure 3 A schematic diagram of a parking lot map model initially obtained based on parking space locations;

[0077] Figure 4 This is a schematic diagram of the real-time parking map model obtained after optimization in the first step;

[0078] Figure 5 This is a schematic diagram of the real-time parking map model obtained after optimization in the second step.

[0079] Figure 6 Schematic diagram of a parking lot map model under certain circumstances. DETAILED DESCRIPTION

[0080] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0081] like Figure 1 As shown, the present invention provides a method for predicting regional public parking space demand based on parking big data, comprising the following steps:

[0082] A parking space pre-occupancy area is set, and vehicles in the parking space pre-occupancy area send out a pre-occupancy signal;

[0083] Construct a real-time parking lot map model based on the parking space locations inside the parking lot according to the pre-occupancy signal;

[0084] Obtain the real-time pre-parking situation of the parking lot, and analyze the parking space allocation plan set according to the real-time pre-parking situation based on the parking lot map model;

[0085] Establish a parking space allocation model, calculate the parking paths of all parking space allocation schemes in the parking space allocation scheme set based on the parking space allocation model, screen out the optimal parking space allocation scheme, and update the real-time pre-parking situation;

[0086] Establish an SVM parking space prediction model, obtain training samples, incorporate real-time pre-parking conditions into the training samples, and use the CS algorithm based on the training samples to optimize the parameters in the SVM parking space prediction model;

[0087] The number of non-idle parking spaces is predicted based on the optimized SVM parking space prediction model.

[0088] In the present invention, a parking space pre-occupancy area is set, and before the vehicle reaches the destination parking lot, the parking space in the parking lot is pre-occupied. The real-time parking situation and pre-occupancy situation in the parking lot are included in the training samples, and an SVM parking space prediction model is established. The parameters in the SVM parking space prediction model are optimized based on the training samples, and the number of non-vacant parking spaces is predicted based on the optimized SVM parking space prediction model. The parking spaces in the parking lot are pre-occupied, which can not only avoid being unable to park in the parking lot when arriving at a specific time, but also obtain the parking data of the parking lot for a period of time in the future in advance through the pre-occupancy method. The parking situation in the future time is predicted by combining the current data and the future data, thereby improving the accuracy of the prediction.

[0089] In practice, some vehicles have the pre-occupancy function, while some do not. The present invention mainly predicts the parking lot usage of vehicles that are estimated to have an overall parking intention through future parking data of pre-occupied vehicles.

[0090] For some vehicles that can be reserved, a parking space reservation area is set. This setting is also for obtaining the pre-parking situation for a certain period of time in the future. The parking space reservation area is a circular area with the parking lot as the center and the reservation distance r as the radius.

[0091] The larger the parking pre-occupied area, the more accurate the pre-parking situation obtained. However, in order to avoid excessive complexity of the data, the present invention only collects data within a certain parking space pre-occupied area. This design is mainly to obtain future data, which can play a more accurate role in predicting the demand for regional public parking spaces.

[0092] In actual application, it is also possible to reserve a specific parking space in the parking lot by reserving parking time. However, in reality, travel is generally random and sudden, and a reservation made some time ago may be canceled at a specific time. The method of setting a parking reservation area can accurately obtain the pre-parking situation for a period of time in the future, because the probability of a vehicle that has already traveled suddenly canceling the reservation is generally much smaller than the probability of the user canceling the reservation in advance. In addition, only collecting the reservation signals of vehicles in the parking space reservation area can greatly reduce the data that needs to be collected. This part of the data can be included in the SVM parking space prediction model as future data. The data is small and the prediction efficiency is high. Moreover, it belongs to future data in a short period of time with high accuracy, which can optimize the SVM parking space prediction model to a large extent and improve the prediction accuracy.

[0093] In the present invention, a parking lot map model is established in advance, and the process of establishing the parking lot map model includes the following steps:

[0094] Based on the parking space location, size, and spacing parameters inside the parking lot, a number of corresponding path nodes and path edges formed by connecting the path nodes are constructed;

[0095] A connection relationship diagram between path nodes is drawn based on path nodes and path edges between different path nodes to obtain a parking lot map model.

[0096] Assume that Figure 2 As shown, the actual parking space location, parking space situation and parking reservation situation in a small parking lot. The large boxes in the figure represent different parking spaces, the circles represent occupied parking spaces in the parking lot, and the small boxes represent reserved parking spaces in the parking lot.

[0097] according to Figure 2 For each parking space, a corresponding path node is established for each parking space-related path, such as Figure 3 As shown, 1, 2, 3, ..., 8 are all path nodes corresponding to parking spaces, 35, 36, ..., 42 represent path nodes corresponding to passages in the parking lot, and adjacent path nodes between feasible paths can be connected to form path edges, for example, 33-34. Assuming that the vehicle needs to park at path node 1, it needs to pass through path edges 33-34, 34-35, and 35-1 to reach path node 1.

[0098] like Figure 3 As shown, the parking lot map model is obtained by drawing path nodes and path edges.

[0099] The initial parking lot map model is a map model that reflects the locations of parking spaces in the entire parking lot. In order to obtain a further real-time parking lot map model based on the actual parking situation, after the parking lot map model is established, the real-time parking situation inside the parking lot is collected. Based on the parking situation inside the parking lot, the path edges of the path nodes corresponding to some occupied parking spaces are screened out to obtain a real-time parking lot map model.

[0100] like Figure 2 As shown, path nodes 9, 12, and 14 are obviously occupied by vehicles. Therefore, path edges 9-35, 12-38, and 14-40 corresponding to path nodes of occupied parking spaces are screened out to obtain a real-time parking lot map model, such as Figure 4 The figure shows the real-time parking lot map model obtained after the first step of optimization.

[0101] After obtaining the real-time parking lot map model, the parking space allocation scheme set is analyzed based on the real-time parking lot map model according to the real-time pre-parking situation, specifically including the following steps:

[0102] Obtain the real-time pre-parking status of the parking lot and obtain the pre-occupied parking spaces in the parking lot;

[0103] Based on the occupied parking spaces, path edges of path nodes corresponding to some of the occupied parking spaces are screened out to correct the parking lot map model;

[0104] Analyze the set of parking space allocation solutions based on the revised parking lot map model.

[0105] like Figure 2 As shown in FIG, path nodes 1, 2, 3, and 4 have been pre-occupied by vehicles. Therefore, path edges 1-35, 2-36, 3-37, and 4-38 corresponding to path nodes of the pre-occupied parking spaces are screened out to obtain a further optimized parking lot map model, as shown in FIG. Figure 5 The figure shows the real-time parking lot map model obtained after the second step of optimization.

[0106] After obtaining the final parking lot map model, it is necessary to analyze the parking space allocation plan set. The present invention takes into account the situation where vehicles without the pre-occupancy function and vehicles with the pre-occupancy function occupy parking spaces simultaneously. At this time, the parking space allocation plans for the vehicles to be pre-occupied and the vehicles to be parked are integrated at the same time.

[0107] Analyze the parking space allocation scheme set, which specifically includes the following steps:

[0108] Establish a set W of vehicles to be occupied and vehicles to be parked:

[0109] W=(w1,w2,…,w i );

[0110] In the formula, w represents the vehicles to be reserved and the vehicles to be parked, i represents the number of vehicles to be reserved and the vehicles to be parked;

[0111] Create a set of free parking spaces P:

[0112] P=(p1,p2,…,p j );

[0113] j = J - (j1 + j2);

[0114] Where p represents an empty parking space, j represents the number of empty parking spaces, J represents the total number of parking spaces, j1 represents the number of occupied parking spaces, j2 represents the number of pre-occupied parking spaces, and (j1+j2) represents the total number of occupied and pre-occupied parking spaces in the parking lot.

[0115] Based on the set of vehicles to be occupied, the set of vehicles to be parked W and the set of free parking spaces P, the parking space allocation solution set T is obtained:

[0116] T=(t1,t2,…,t n );

[0117] Where t represents the possible allocation scheme of the vehicles to be occupied or parked w to the vacant parking spaces p, and n represents the number of possible schemes.

[0118] In the above embodiment, T represents the set of all possible parking space allocation schemes for allocating the set of vehicles to be occupied and parked W to the set of free parking spaces P. There are multiple allocation methods in one parking space allocation scheme. For example, one parking space allocation scheme may include allocating w1 to p1, w2 to p2, and w i Assign to p i distribution method.

[0119] After obtaining the set of all possible parking space allocation plans, it is necessary to establish a parking space allocation model to calculate the sum of the paths corresponding to each parking space allocation plan in the set of all possible parking space allocation plans, so as to screen out the parking space allocation plan with the smallest total distance of the parking path, and occupy and pre-occupy the parking space based on the screened parking space allocation plan.

[0120] In the present invention, a parking space allocation model is established:

[0121] F(T) = min(f(T));

[0122] Where f(T) represents the sum of the distances of the parking paths of the parking space allocation scheme, and min(f(T)) represents the minimum value of the sum of the distances of the parking paths of all parking space allocation schemes;

[0123] Calculate the sum of the parking path distances of all parking allocation schemes in the parking allocation scheme set:

[0124]

[0125] L(S)=∑ e∈E w e ;

[0126] Where S represents the parking path of the vehicle to be occupied or parked w allocated to the vacant parking space p corresponding to the i-th parking space allocation scheme, and L(S) represents the distance of the parking path of the vehicle to be occupied or parked w allocated to the vacant parking space p corresponding to the i-th parking space allocation scheme;

[0127] E represents all the path edges of the path nodes corresponding to the parking path S, w e Represents the length of the e-th path edge.

[0128] In the above embodiment, it is necessary to calculate the sum of the paths corresponding to each parking space allocation scheme in the set of all possible parking space allocation schemes. In the process of calculating the sum of the paths corresponding to one parking space allocation scheme, all parking paths S in the parking space allocation scheme are first obtained, and the path edges corresponding to all parking paths are obtained. That is to say, for a vehicle in one of the parking space allocation schemes, the sum of the lengths of the path edges of its parking path is calculated to obtain the distance L(S) of the parking path of the vehicle to be occupied or the vehicle to be parked allocated to the vacant parking space p corresponding to the i-th parking space allocation scheme. Then, the L(S) corresponding to all vehicle allocation schemes in the parking space allocation scheme are calculated and added to obtain the sum of the distances of the parking paths corresponding to each parking space allocation scheme.

[0129] Based on the parking space allocation model, the parking space allocation plan with the smallest total parking path distance among all parking space allocation plans is selected as the optimal parking space allocation plan.

[0130] by Figure 6 For example, assuming that the number of vehicles to be occupied and parked is 1, and the currently vacant parking spaces include 10, 11, and 13, the parking paths corresponding to the parking space allocation plan for this vehicle include 33-34-35-36-10, 33-34-35-36-37-11, and 33-34-35-36-37-38-39-13. It is obvious that the first parking path has the smallest total distance. Therefore, 10 as the target parking space is the optimal parking space allocation plan.

[0131] After the optimal parking space allocation plan is screened out, the parking space corresponding to the optimal parking space allocation plan is allocated to the corresponding vehicle number for pre-occupancy. If there is a vehicle to be parked simultaneously, the corresponding plan in the optimal parking space allocation plan is obtained for the vehicle, and parking is performed directly according to the path of the plan.

[0132] After completing a series of parking and pre-occupancy actions, the real-time pre-parking situation is updated based on the pre-occupied parking spaces, wherein the real-time pre-parking situation includes the location, size, spacing, and number of the pre-occupied parking spaces.

[0133] In the present invention, each time a prediction process is performed, it is necessary to obtain the real-time pre-parking situation, and each time a vehicle sends a pre-occupancy signal, a pre-occupancy is performed, and the real-time pre-parking situation is synchronously updated.

[0134] In the present invention, the training samples include the real-time parking conditions of the parking lot at different time periods;

[0135] Before obtaining the training samples, the measured data of the parking lot is obtained, and the measured data is divided into training samples and test samples, wherein the training samples are substituted into the parameters of the SVM parking space prediction model for optimization.

[0136] The real-time pre-parking situation is converted into pre-occupancy data of parking spaces and incorporated into the training samples to update the data in the training samples.

[0137] In the present invention, the CS algorithm is used to optimize the parameters in the SVM parking space prediction model, which specifically includes the following steps:

[0138] Step 100, preprocessing the parking space data in the training sample;

[0139] Step 200, preliminarily determine the parameters of the CS algorithm and the thresholds of the penalty coefficient C and the kernel function width parameter σ;

[0140] Among them, the parameters of the CS algorithm preliminarily determined include the number of bird nests, the probability of a bird nest being discovered, the maximum number of iterations, the number of abandoned bird nests, and the elimination factor.

[0141] Step 300, setting the evaluation index of the SVM parking space prediction model:

[0142] ;

[0143] In the formula, M is the evaluation index (mean relative error), i and n represent the order of samples, and x i Represents the model prediction value, y i represents the actual value of the training sample;

[0144] Step 400, in the CS algorithm, initialize the bird's nest position (C, σ) and substitute it into the SVM parking space prediction model, calculate the evaluation index value through the above formula and retain the optimal bird's nest position;

[0145] Step 500, update the bird's nest position (C, σ), and substitute the updated bird's nest position into the SVM parking space prediction model, calculate the current evaluation index value according to the above formula, compare the current evaluation index value with the previous evaluation index value, and retain the bird's nest position with a smaller evaluation index value;

[0146] If the current evaluation index value is smaller, the current bird's nest position will replace the previous bird's nest position as the optimal bird's nest position;

[0147] If the previous evaluation index is smaller, the previous nest position is kept as the optimal nest position.

[0148] Step 600, after the global update is completed, a local search is performed;

[0149] Each bird's nest randomly generates a number between [0 and 1]. If the number is greater than the probability of the nest being discovered, the current nest position is eliminated and a new nest position is randomly generated.

[0150] If this number is less than the probability of the nest being found, the location of the nest remains unchanged.

[0151] Step 700, after the local search is completed, the evaluation index values ​​corresponding to all the bird nest positions are sorted in order of size, the n last bird nests are eliminated, and n new bird nests are generated, and their evaluation index values ​​are calculated;

[0152] Among them, n is generally selected as 3.

[0153] Step 800, determine whether the iteration is terminated, if not, continue the iteration, if it is terminated, find the optimal solution of the penalty coefficient C and the kernel function width parameter σ at this time.

[0154] If the number of iterations exceeds the maximum number of iterations or the set minimum error, it indicates that the iteration has terminated;

[0155] If the number of iterations does not exceed the maximum number of iterations and does not exceed the set minimum error, it indicates that the iteration has not terminated. If the iteration has not terminated, return to step 500 to continue iterating.

[0156] The above steps can calculate the optimal solution of the penalty coefficient C and the kernel function width parameter σ, which is the optimized parameters in the SVM parking space prediction model. Based on the optimized SVM parking space prediction model, the optimal solution of the penalty coefficient C and the kernel function width parameter σ is substituted into the SVM parking space prediction model for training, and the final predicted value of the number of non-idle parking spaces is output.

[0157] In the above embodiment, the non-vacant parking spaces include parking spaces where vehicles have parked and reserved parking spaces in the actual measured data of the parking lot.

[0158] In the present invention, after each real-time pre-parking situation is updated, a prediction process for the number of non-free parking spaces in the parking lot can be triggered to predict the number of free parking spaces in the parking lot within a certain period of time in the future, and the predicted data can be synchronized to the client in real time.

[0159] The number of non-vacant parking spaces in a parking lot in the future can just reflect the demand for public parking spaces, thereby realizing the prediction of the demand for public parking spaces in the present invention.

[0160] The present invention does not require that all vehicles have the pre-occupancy function. The parking conditions of some vehicles with the pre-occupancy function in the future are collected to predict the future parking conditions. The future data trend is predicted using some "future" data, which can improve the accuracy of data prediction.

[0161] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A method for predicting regional public parking space demand based on parking big data, characterized in that: The following steps are involved: A parking space pre-occupancy area is set, and vehicles in the parking space pre-occupancy area send out a pre-occupancy signal; Construct a real-time parking lot map model based on the parking space locations inside the parking lot according to the pre-occupancy signal; Obtain the real-time pre-parking situation of the parking lot, and analyze the parking space allocation plan set according to the real-time pre-parking situation based on the parking lot map model; Establish a parking space allocation model, calculate the parking paths of all parking space allocation schemes in the parking space allocation scheme set based on the parking space allocation model, screen out the optimal parking space allocation scheme, and update the real-time pre-parking situation; Establish an SVM parking space prediction model, obtain training samples, incorporate real-time pre-parking conditions into the training samples, and use the CS algorithm based on the training samples to optimize the parameters in the SVM parking space prediction model; The number of non-idle parking spaces is predicted based on the optimized SVM parking space prediction model.

2. The regional public parking space demand prediction method based on parking big data according to claim 1 is characterized in that: After the optimal parking space allocation plan is screened out, the parking spaces corresponding to the optimal parking space allocation plan are pre-occupied; Update the real-time pre-parking situation based on the occupied parking spaces.

3. The regional public parking space demand prediction method based on parking big data according to claim 1 is characterized in that: The process of establishing the parking lot map model includes the following steps: Based on the parking space location, size, and spacing parameters inside the parking lot, a number of corresponding path nodes and path edges formed by connecting the path nodes are constructed; A connection relationship diagram between path nodes is drawn based on path nodes and path edges between different path nodes to obtain a parking lot map model.

4. The method for predicting regional public parking space demand based on parking big data according to claim 3 is characterized in that: After establishing the parking lot map model, collect real-time parking conditions inside the parking lot; Based on the parking situation inside the parking lot, the path edges of the path nodes corresponding to some occupied parking spaces are screened out to obtain a real-time parking lot map model.

5. The method for predicting regional public parking space demand based on parking big data according to claim 4 is characterized in that: Based on the parking lot map model, the parking space allocation scheme set is analyzed according to the real-time pre-parking situation, which specifically includes the following steps: Obtain the real-time pre-parking status of the parking lot and obtain the pre-occupied parking spaces in the parking lot; Based on the occupied parking spaces, path edges of path nodes corresponding to some of the occupied parking spaces are screened out to correct the parking lot map model; Analyze the set of parking space allocation solutions based on the revised parking lot map model.

6. The method for predicting regional public parking space demand based on parking big data according to claim 5 is characterized in that: Analyze the parking space allocation scheme set, which specifically includes the following steps: Establish a set W of vehicles to be occupied and vehicles to be parked: W=(w1,w2,…,w i ); In the formula, w represents the vehicles to be reserved and the vehicles to be parked, i represents the number of vehicles to be reserved and the vehicles to be parked; Create a set of free parking spaces P: P=(p1,p2,…,p j ); j = J - (j1 + j2); In the formula, p represents the vacant parking space, j represents the number of vacant parking spaces, J represents the total number of parking spaces, j1 represents the number of occupied parking spaces, and j2 represents the number of reserved parking spaces; Based on the set of vehicles to be occupied, the set of vehicles to be parked W and the set of free parking spaces P, the parking space allocation solution set T is obtained: T=(t1,t2,…,t n ); Where t represents the possible allocation scheme of the vehicles to be occupied or parked w to the vacant parking spaces p, and n represents the number of possible schemes.

7. The method for predicting regional public parking space demand based on parking big data according to claim 6, characterized in that: Build a parking space allocation model: F(T) = min(f(T)); Where f(T) represents the sum of the distances of the parking paths of the parking space allocation scheme, and min(f(T)) represents the minimum value of the sum of the distances of the parking paths of all parking space allocation schemes; Calculate the sum of the parking path distances of all parking allocation schemes in the parking allocation scheme set: L(S)=∑ e∈E w e ; Where S represents the parking path of the vehicle to be occupied or parked w allocated to the vacant parking space p corresponding to the i-th parking space allocation scheme, and L(S) represents the distance of the parking path of the vehicle to be occupied or parked w allocated to the vacant parking space p corresponding to the i-th parking space allocation scheme; E represents all the path edges of the path nodes corresponding to the parking path S, w e Represents the length of the e-th path edge.

8. The method for predicting regional public parking space demand based on parking big data according to claim 7 is characterized in that: Based on the parking space allocation model, the parking space allocation plan with the smallest total parking path distance among all parking space allocation plans is selected as the optimal parking space allocation plan.

9. The method for predicting regional public parking space demand based on parking big data according to claim 1, characterized in that: The training samples include real-time parking conditions at different time periods in the parking lot; Before obtaining the training samples, obtain the measured data of the parking lot and divide the measured data into training samples and test samples; The real-time pre-parking situation is converted into pre-occupancy data of parking spaces and incorporated into the training samples to update the data in the training samples.

10. The method for predicting regional public parking space demand based on parking big data according to claim 9, characterized in that: The CS algorithm is used to optimize the parameters in the SVM parking space prediction model, which specifically includes the following steps: Preprocess the parking space data in the training samples; Preliminarily determine the parameters of the CS algorithm and the thresholds of the penalty coefficient C and kernel function width parameter σ; Set the evaluation index of the SVM parking space prediction model: ； In the formula, M is the evaluation index, i and n represent the order of samples, and x i Represents the model prediction value, y i Represents the actual value of the training sample; In the CS algorithm, the bird’s nest position is initialized and substituted into the SVM parking space prediction model to find the evaluation index value and retain the optimal bird’s nest position; Update the bird's nest position and substitute it into the SVM parking space prediction model to find the current evaluation index value, compare the current evaluation index value with the previous evaluation index value, and retain the bird's nest position with a smaller evaluation index value; After the global update is completed, a local search is performed; After the local search is completed, the evaluation index values ​​corresponding to all the bird nest positions are sorted by size, the last n bird nests are eliminated, and n new bird nests are generated to calculate their evaluation index values; Determine whether the iteration is terminated. If not, continue the iteration. If it is terminated, find the optimal solution of the penalty coefficient C and kernel function width parameter σ at this time. Based on the optimized SVM parking space prediction model, the optimal solution of the penalty coefficient C and the kernel function width parameter σ is substituted into the SVM parking space prediction model for training, and the final predicted value of the number of non-vacant parking spaces is output; Among them, non-vacant parking spaces include parking spaces that are already occupied and occupied parking spaces according to the actual measured data of the parking lot.

11. The method for predicting regional public parking space demand based on parking big data according to claim 1, characterized in that: The parking space pre-occupied area is a circular area with the parking lot as the center and the pre-occupied distance r as the radius.

Images