Urban parking demand prediction method, device, equipment, and storage medium

Abstract

The present application discloses a method, device, equipment, and storage medium for predicting urban parking demand, relating to the fields of artificial intelligence and smart transportation technology. The method includes: determining a target parking spot in a target city; determining at least two areas of interest that include the target parking spot; obtaining the attributes of at least two building types in each area of ​​interest to obtain building information; integrating the building information of at least two areas of interest to obtain target interest building information; regressing the target interest building information through a parking rate prediction model to obtain a target predicted parking rate; classifying the target interest building information through a parking period prediction model to obtain a target predicted parking period; and displaying the target predicted parking rate and target predicted parking period. The present application can predict parking rates and parking periods for parking spots, thereby improving the refinement of parking demand prediction granularity and effectively alleviating the contradiction between parking supply and demand.

Description

Technical Field

[0001] The present application relates to the fields of artificial intelligence and smart transportation technology, and in particular to a method and apparatus, device, and storage medium for predicting urban parking demand. Background Art

[0002] Currently, the number of vehicles is increasing year by year, and the irrational layout of parking resources has led to a relatively low parking space utilization rate in most cities (for example, less than 50%). Moreover, even in cities with sufficient parking spaces, parking facilities cannot fully meet parking demand.

[0003] Related technologies use multi-source data fusion and deep learning techniques to predict city-level parking demand. For example, they first collect multi-source city data, such as historical parking data, demographics, economic indicators, traffic flow, and other numerical variables, as well as natural language data such as transportation policies and urban planning. They then extract numerical features and use the BERT model to extract natural language features to predict parking demand. However, these technologies primarily focus on city-level predictions and lack predictions for different regions within a city. This means that the prediction granularity is insufficient and cannot effectively alleviate the contradiction between parking supply and demand.

[0004] Therefore, developing refined parking demand forecasts is crucial to alleviating the contradiction between parking supply and demand. Summary of the Invention

[0005] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a method, apparatus, device, and storage medium for predicting urban parking demand. These methods can predict parking rates and parking periods for parking spots, increasing the granularity of parking demand predictions and helping to improve the level and efficiency of urban parking management, thereby effectively alleviating the contradiction between parking supply and demand.

[0006] To achieve the above objectives, a first aspect of an embodiment of the present application provides a method for predicting urban parking demand, the method comprising:

[0007] Determine the target parking spot in the target city;

[0008] determining at least two regions of interest that include the target parking point, wherein the regional scales of any two regions of interest are different;

[0009] Obtaining attributes of at least two building types in each of the interest areas to obtain building information;

[0010] Integrating the building information of the at least two areas of interest to obtain target interest building information;

[0011] Regressing the target building of interest information using a pre-trained parking rate prediction model to obtain a target predicted parking rate for the target parking spot;

[0012] Classifying the target building of interest information using a pre-trained parking period prediction model to obtain a target predicted parking period for the target parking spot;

[0013] The target predicted parking rate and the target predicted parking period are displayed.

[0014] Optionally, the determining of at least two regions of interest including the target parking point includes:

[0015] Classifying the target parking spot according to a plurality of reference parking lots in the target city to obtain a parking spot category;

[0016] If the parking spot category indicates that the target parking spot does not belong to any of the reference parking lots, generating at least two regions of interest, each having at least two building types, with the target parking spot as a center point, wherein the regional scales of the at least two regions of interest are equidistantly increased;

[0017] If the parking point category indicates that the target parking point belongs to one of the reference parking lots, the at least two regions of interest are generated with the reference parking lot to which the target parking point belongs as a center point and with at least two area scales associated with the reference parking lot as radii.

[0018] Optionally, the method further includes pre-training the parking rate prediction model, specifically including:

[0019] Acquiring first sample interest building information of a first sample parking space, wherein the first sample interest building information indicates attributes of each of at least two building types in at least two first sample interest areas, and the first sample parking space has a parking rate label value, and the parking rate label value indicates a parking rate of the first sample parking space;

[0020] Regressing the first sample building of interest information through a preset first random forest model to obtain a parking rate prediction value;

[0021] The first random forest model is trained according to the gap between the parking rate prediction value and the parking rate label value to obtain the parking rate prediction model.

[0022] Optionally, obtaining the first sample interesting building information of the first sample parking space includes:

[0023] Generate an increasing number of candidate scales;

[0024] Generate a candidate region of interest with the first sample parking space as a center point and each candidate scale as a radius;

[0025] Evaluate a preset first random forest model based on the building information in the candidate area of ​​interest to obtain a mean absolute error and a fitting accuracy corresponding to each candidate scale;

[0026] Selecting a target scale from a plurality of candidate scales according to the mean absolute error and the fitting accuracy;

[0027] generating at least two sample region scales that increase arithmetically based on the target scale, wherein the largest scale among the at least two sample region scales is the target scale;

[0028] generating the at least two first sample interest regions with the first sample parking space as a center point and the sample area scale as a radius;

[0029] Obtaining sample building information according to the attributes of each of the at least two first sample interest areas in the at least two first sample interest areas;

[0030] The sample building information of the at least two first sample interest areas is integrated to obtain first sample interest building information.

[0031] Optionally, the at least two first sample regions of interest include a first-scale sample region of interest, a second-scale sample region of interest, and a third-scale sample region of interest, and the regional scales of the first-scale sample region of interest, the second-scale sample region of interest, and the third-scale sample region of interest are equidistantly increased;

[0032] After training the first random forest model according to the gap between the parking rate prediction value and the parking rate label value to obtain the parking rate prediction model, the method further includes:

[0033] If the region of interest in the first sample interest building information includes the first-scale sample region of interest and the second-scale sample region of interest, record the parking rate prediction model as a first candidate parking rate prediction model, and perform a performance evaluation on the first candidate parking rate prediction model to obtain first performance evaluation data;

[0034] If the region of interest in the first sample interest building information includes the first-scale sample region of interest and the third-scale sample region of interest, recording the parking rate prediction model as a second candidate parking rate prediction model, and performing a performance evaluation on the second candidate parking rate prediction model to obtain second performance evaluation data;

[0035] If the region of interest in the first sample building of interest information includes the first-scale sample region of interest, the second-scale sample region of interest, and the third-scale sample region of interest, recording the parking rate prediction model as a third candidate parking rate prediction model, and performing a performance evaluation on the third candidate parking rate prediction model to obtain third performance evaluation data;

[0036] According to the comparison among the first performance evaluation data, the second performance evaluation data and the third performance evaluation data, a final parking rate prediction model is screened out from the first candidate parking rate prediction model, the second candidate parking rate prediction model and the third candidate parking rate prediction model.

[0037] Optionally, the method further includes pre-training the parking period prediction model, specifically including:

[0038] Obtaining the number of parking times of each of at least two of the first sample parking spaces, and taking each of the at least two parking times in turn as an initial parking times threshold;

[0039] The first sample parking space whose parking number is greater than the initial parking number threshold is used as a reference sample parking space, and the number of the reference sample parking spaces is counted to obtain the total number of remaining parking spaces corresponding to each initial parking number threshold;

[0040] Evaluating a preset second random forest model based on the building information associated with each reference sample parking space to obtain a model accuracy corresponding to each initial parking number threshold;

[0041] Filtering a target parking number threshold from the multiple initial parking number thresholds according to the total number of remaining parking spaces and the model accuracy;

[0042] The first sample parking space whose parking times is greater than the target parking times threshold is reserved to obtain a second sample parking space;

[0043] Regressing the second sample building of interest information of the second sample parking space using the second random forest model to obtain a parking period prediction value; wherein the second sample parking space has a parking period label value, and the parking period label value indicates the parking period of the second sample parking space;

[0044] The second random forest model is trained according to the gap between the parking period prediction value and the parking period label value to obtain the parking period prediction model.

[0045] Optionally, after determining the target parking point in the target city, the method further includes:

[0046] Obtaining the number of residents in the largest area of ​​interest associated with the target parking point;

[0047] Obtaining the estimated number of parking spaces set for the target parking spot;

[0048] The step of regressing the target building of interest information using a pre-trained parking rate prediction model to obtain a target predicted parking rate of the target parking spot includes:

[0049] The target building information of interest, the number of residents and the estimated number of parking spaces are regressed using a pre-trained parking rate prediction model to obtain a target predicted parking rate based on the estimated number of parking spaces.

[0050] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a device for predicting urban parking demand, the device comprising:

[0051] A parking point determination module is used to determine a target parking point in a target city;

[0052] an area of ​​interest determination module, configured to determine at least two areas of interest including the target parking point, wherein the area scales of any two areas of interest are different;

[0053] An information acquisition module, configured to acquire attributes of at least two building types in each of the interest areas to obtain building information;

[0054] an information integration module, configured to integrate the building information of the at least two areas of interest to obtain target building information of interest;

[0055] a parking rate prediction module, configured to regress the target building of interest information using a pre-trained parking rate prediction model to obtain a target predicted parking rate for the target parking spot;

[0056] A parking period prediction module is used to classify the target building of interest information using a pre-trained parking period prediction model to obtain a target predicted parking period for the target parking spot;

[0057] A data display module is used to display the target predicted parking rate and the target predicted parking period.

[0058] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an urban parking demand prediction device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the first aspect above when executing the computer program.

[0059] To achieve the above-mentioned purpose, the fourth aspect of the embodiments of the present application proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method described in the first aspect.

[0060] The urban parking demand prediction method and apparatus, urban parking demand prediction device, and storage medium proposed in this application address the problem that existing methods can only predict parking demand for entire cities or regions, but not for specific parking spots. By incorporating building information surrounding parking spots into the basic information used for parking demand prediction, this method first identifies at least two regions of interest (ROIs) of varying scales (e.g., one 500-meter region and another 1000-meter region). Next, the attributes of different building types (e.g., schools, shopping malls, businesses, etc.) within the respective RIOs are obtained to obtain building information (e.g., a school has an area of ​​1000 square meters within a 500-meter RIO and an area of ​​25,000 square meters within a 1000-meter RIO). This information is then integrated to obtain target building information. Target parking rates and target parking time periods are then predicted and displayed. In summary, this application selects areas of interest at different regional scales to observe and quantify the building information around parking spots, capturing spatial changes and environmental characteristics at different regional scales. It can predict parking rates and parking periods for parking spots with high accuracy, thereby improving the granularity of parking demand prediction, helping to improve the level and efficiency of urban parking management, and effectively alleviating the contradiction between parking supply and demand.

[0061] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become obvious from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a flow chart of the urban parking demand prediction method provided by an embodiment of the present application;

[0063] Figure 2 is a schematic diagram of the region of interest provided in an embodiment of the present application;

[0064] Figure 3 This is a parking demand prediction interface diagram provided by an embodiment of the present application;

[0065] Figure 4 This is a flow chart of the model training phase provided in an embodiment of the present application;

[0066] Figure 5 Schematic diagram of the parking rate fitting provided in the embodiment of the present application;

[0067] Figure 6is the total number of remaining parking spaces provided in the embodiment of this application, R 2 Schematic diagram of the relationship between the initial parking number threshold;

[0068] Figure 7 This is a flow chart of the model application phase provided by the embodiment of the present application;

[0069] Figure 8 It is a structural diagram of the urban parking demand prediction device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0071] It should be noted that although the device schematics illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the device or the sequence in the flowcharts. The terms "first," "second," and so on, in the specification, claims, and drawings, are used to distinguish similar items and are not necessarily used to describe a specific sequence or precedence.

[0072] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0073] The urban parking demand prediction method provided in the embodiments of the present application can be applied to any terminal or server, and can also be software running on the server or terminal. The server can be configured as an independent physical server, or as a server cluster or distributed system consisting of multiple physical servers. It can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The software can be an application or computer program that implements the urban parking demand prediction method, but is not limited to the above forms.

[0074] The urban parking demand prediction method, urban parking demand prediction device, urban parking demand prediction equipment, and computer-readable storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the urban parking demand prediction method in the embodiments of the present application is described.

[0075] Please refer to Figure 1 , Figure 1An optional flow chart of a method for predicting urban parking demand is disclosed. Figure 1 The method may include but is not limited to steps 101 to 107.

[0076] Step 101, determining a target parking spot in a target city;

[0077] Step 102: determining at least two regions of interest that include the target parking point, wherein the scales of any two regions of interest are different;

[0078] Step 103, obtaining attributes of at least two building types in each area of ​​interest to obtain building information;

[0079] Step 104 , integrating the building information of at least two areas of interest to obtain target building information of interest;

[0080] Step 105 , regressing the target building of interest information using a pre-trained parking rate prediction model to obtain a target predicted parking rate for the target parking spot;

[0081] Step 106 , classifying the target building of interest information using a pre-trained parking period prediction model to obtain a target predicted parking period for the target parking spot;

[0082] Step 107: Display the target predicted parking rate and the target predicted parking period.

[0083] Steps 101 to 107, as shown in this embodiment of the present application, address the problem of existing parking demand predictions that only cover entire cities or regions, but not specific parking spots. By incorporating building information surrounding parking spots into the basic information used for parking demand prediction, the method first identifies at least two regions of interest (ROIs) of varying scales (e.g., one 500-meter region and another 1000-meter region). Next, the attributes of different building types (e.g., schools, shopping malls, businesses, etc.) within the respective ROIs are obtained to obtain building information (e.g., a school has an area of ​​1000 square meters within a 500-meter region and an area of ​​25,000 square meters within a 1000-meter region). This information is then integrated to obtain target building information. Target parking rates and target parking hours are then predicted and displayed. In summary, this application selects areas of interest at different regional scales to observe and quantify the building information around parking spots, capturing spatial changes and environmental characteristics at different regional scales. It can predict parking rates and parking periods for parking spots with high accuracy, thereby improving the granularity of parking demand prediction, helping to improve the level and efficiency of urban parking management, and effectively alleviating the contradiction between parking supply and demand.

[0084] In step 101, the target city refers to the city for which parking demand forecasting is required. The target city can be user-specified. In one example, when planning a parking lot, an administrator of a parking management company selects a target city from multiple candidate cities on a map. For example, if the candidate cities include City A, City B, and City C, the administrator may select City B as the target city.

[0085] A target parking spot is a designated parking spot in a target city. The target parking spot may be a parking space in the target city, such as a parking space in a parking lot. Alternatively, the target parking spot may be a point in any area of ​​the target city. Any area may include a street, a commercial district, an administrative district, and the like. In one example, a parking spot selection interface including a map may be displayed, and the user may select a point on the map to obtain the target parking spot.

[0086] It should be noted that the target parking spots determined in this application are used for parking demand prediction, that is, it is assumed that the target parking spots are available for parking, without considering the impact of some policies.

[0087] In step 102, the area of ​​interest (AOI) is the area surrounding the target parking point, and the AOI includes the target parking point. The scales of any two AOIs are different. For example, at least two AOIs include areas with scales of 500 meters, 1000 meters, and 1500 meters, respectively.

[0088] In one embodiment, step 102 may include:

[0089] Step 201, classifying a target parking spot according to multiple reference parking lots in a target city to obtain a parking spot category;

[0090] Step 202: If the parking spot category indicates that the target parking spot does not belong to any reference parking lot, generating at least two regions of interest (ROIs) with the target parking spot as the center point, each having at least two building types, wherein the regional scales of the at least two regions of interest are equidistantly increased;

[0091] In step 203 , if the parking spot category indicates that the target parking spot belongs to one of the reference parking lots, at least two regions of interest are generated with the reference parking lot to which the target parking spot belongs as the center point and with at least two area scales associated with the reference parking lot as the radius.

[0092] In step 201, a reference parking lot is a parking lot in a target city used to assist in parking demand prediction. For example, if the target city includes three parking lots, T1, T2, and T3, at least one of T1, T2, and T3 can be set as a reference parking lot. The reference parking lot contains multiple reference parking spaces; for example, T1 contains 200 reference parking spaces. Each reference parking space corresponds to a location, and the reference parking lot corresponds to a reference location sequence, which includes the locations of multiple reference parking spaces.

[0093] Step 201 may specifically include comparing the target parking spot's location with the reference location sequence of each reference parking lot; if the target parking spot's location does not exist in the reference location sequence of any reference parking lot, generating a parking spot category indicating that the target parking spot does not belong to any reference parking lot; and if the target parking spot's location exists in the reference location sequence of one of the reference parking lots, generating a parking spot category indicating that the target parking spot belongs to one of the reference parking lots. In this way, the parking spot category can be quickly determined.

[0094] In step 202, if the parking spot category indicates that the target parking spot does not belong to any reference parking lot, it indicates that the regional scale cannot be determined based on the relevant information of the reference parking lots. In this case, the reliability of the regional scale is improved by taking advantage of the number of building types. Specifically, the area surrounding the target parking spot contains multiple building types (e.g., schools, shopping malls, businesses, medical facilities, real estate, educational training facilities, tourist attractions, etc.). Indirect regional scale determination is performed based on the presence of at least two building types. This ensures that the generated region of interest contains relevant information of at least two building types, which helps improve the accuracy of subsequent parking rate and parking time prediction based on the region of interest.

[0095] Step 202 may specifically include: generating a candidate ROI with the target parking point as the center point and a preset scale value as the radius; if the number of building types in the candidate ROI is less than two, increasing the scale value to update the candidate ROI; if the number of building types in the candidate ROI is greater than or equal to two, setting the candidate ROI as the smallest ROI; and generating the next ROI with the target parking point as the center point and a multiple of the scale of the smallest ROI as the radius. In this way, while ensuring that the generated ROI includes information on different building types, the adaptability of the generated ROI is improved, resulting in a high degree of universality.

[0096] It should be noted that when the at least two interest regions are specifically three or more interest regions, the regional scales of the three or more interest regions are increased arithmetic progression. Figure 2 , Figure 2The regions of interest with regional scales of 500 meters, 1000 meters, and 1500 meters are shown, where point p represents the target parking point.

[0097] In step 203, if the parking spot category indicates that the target parking spot belongs to one of the reference parking lots, this indicates that the regional scale can be determined based on the relevant information of the reference parking lot. Specifically, different reference parking lots vary in size, with some having thousands of parking spaces while others may have only dozens. For these reference parking lots of varying sizes, the at least two regional scales associated with them are generally different. For example, larger reference parking lots have larger regional scales, while smaller ones have smaller regional scales. In addition to considering the impact of parking lot size on regional scale, the presence of facilities such as subways, bus stops, and bus stations near the reference parking lot also affects regional scale. It is necessary to pre-assign at least two regional scales to the reference parking lot. In practical applications, after determining the reference parking lot to which the target parking spot belongs, at least two regions of interest are generated, with the reference parking lot as the center point and the at least two regional scales associated with the reference parking lot as the radius.

[0098] It should be noted that the reference parking lot as the center point can specifically be the exit of the reference parking lot, the entrance of the reference parking lot, or other points in the reference parking lot. This embodiment does not make specific limitations on this.

[0099] The benefit of the embodiment of steps 201 to 203 is that different ROI generation methods are dynamically selected according to whether the target parking spot belongs to a reference parking lot, which can improve the flexibility, accuracy and universality of generating ROIs.

[0100] In step 103, the attributes of the building type in the area of ​​interest include the total area occupied, the number of buildings, and the like.

[0101] In one example, the building types include shopping and tourist attractions, and the area scales of the interest areas are 500m (meters), 1000m and 1500m respectively. Referring to Table 1, the total area occupied by shopping in the area scale of 500m is 1481m 2 (square meters) and the number is 2, the total area of ​​the interest area with a regional scale of 1000m is 4801m 2 There are 4 of them, and the total area of ​​the area of ​​interest with a regional scale of 1500m is 9606m 2 The total area of ​​tourist attractions in the area of ​​interest with a regional scale of 500m is 10314m 2The total area of ​​the interest area with a regional scale of 1000m is 121302m 2 There are three of them, and the total area of ​​the area of ​​interest is 492,867 m2 at a regional scale of 1,500 m2. 2 And the number is 6.

[0102]

[0103] Table 1

[0104] In step 104, for example, referring to Table 1, for the target parking spot, the statistically calculated building information includes the total area of ​​shopping within 500 meters, the area of ​​tourist attractions within 500 meters, the total area of ​​shopping within 1000 meters, the area of ​​tourist attractions within 1000 meters, the total area of ​​shopping within 1500 meters, and the area of ​​tourist attractions within 1500 meters. If the two area scales of 500 meters and 1000 meters are selected, the target building information is [1481, 10314, 121302, 9606]. If the three area scales of 500 meters, 1000 meters, and 1500 meters are selected, the target building information is [1481, 10314, 4801, 121302, 9606, 492867]. Alternatively, the two area scales of 500 meters and 1500 meters can be selected, or the two area scales of 1000 meters and 1500 meters can be selected, and this is not specifically limited in this embodiment.

[0105] In step 105, the parking rate prediction model is a machine learning model used to predict parking rates, such as a random forest model. The target building information is input into the parking rate prediction model for regression analysis to obtain a target predicted parking rate for the target parking spot. The target predicted parking rate is the parking probability of the target parking spot as predicted by the parking rate prediction model. For example, inputting the target building information [1481, 10314, 4801, 121302, 9606, 492867] into the parking rate prediction model for regression analysis yields a target predicted parking rate of 40%.

[0106] In another embodiment, the urban parking demand prediction method may further include: obtaining the number of residents in the maximum interest area associated with the target parking point; obtaining the estimated number of parking spaces set for the target parking point; then step 105 may include: regressing the target interest building information, the number of residents and the estimated number of parking spaces through a pre-trained parking rate prediction model to obtain a target predicted parking rate under the estimated number of parking spaces.

[0107] For example, assuming that the number of residents is 17832 and the expected number of parking spaces is 5, the fused information based on the target interest building information, the number of residents and the expected number of parking spaces can be obtained as [1481, 10314, 4801, 121302, 9606, 492867, 17832, 5]. This fused information is input into the parking rate prediction model for regression analysis, and the target predicted parking rate is 50% when the expected number of parking spaces is 5.

[0108] It should be noted that by dynamically adjusting the estimated number of parking spaces, multiple target predicted parking rates can be obtained, which is beneficial for parking management companies and other businesses or users to manage parking around target parking spots, such as whether to develop parking lots around target parking spots and parking fee pricing.

[0109] The benefit of the above embodiment of considering the estimated number of parking spaces is that the prediction granularity of parking demand prediction is further improved, and the predicted target predicted parking rate can assist in tasks such as parking management.

[0110] In step 106, the parking period prediction model is a machine learning model used to predict parking periods, such as a random forest model. The target building information is input into the parking period prediction model for classification analysis to obtain the target predicted parking period for the target parking spot. The target predicted parking period is the parking period predicted by the parking period prediction model for the target parking spot. For example, inputting the target building information [1481, 10314, 4801, 121302, 9606, 492867] into the parking period prediction model for classification analysis yields the target predicted parking period as daytime. Daytime corresponds to the period from 9:00 AM to 6:00 PM.

[0111] In step 107, the target predicted parking rate and the target predicted parking period are displayed. Figure 3 , the parking demand forecast interface can be displayed to the user. The interface includes an area for displaying the parking demand forecast results, and the specific display content of the area is "Parking rate forecast: 40%", 24-hour mode: daytime, latitude and longitude: 114, 22". The interface can also include a building information display area, and the specific display content of the area is "Companies and enterprises, a total of 2 items, with a total area of ​​10677m 2 : XX Industrial Park, XX Industrial Zone", and "Medical, a total of 1 project, with a total area of ​​30404m 2 :XX Hospital" etc.

[0112] In one embodiment, before step 105, the urban parking demand prediction method further includes pre-training a parking rate prediction model and a pre-training parking period prediction model. The training data required for pre-training will now be introduced.

[0113] In one example, refer to Figure 4 , the urban parking demand prediction method may include the following steps:

[0114] (1) Data collection:

[0115] Cameras are used to sense the entry and exit time of each parking space.

[0116] (2) Data preprocessing:

[0117] The parking duration of each parking space is calculated based on the difference between the departure time and the entry time, and then abnormal data with a time interval of less than 10 seconds or more than one week are removed.

[0118] (3) Count the building information associated with each parking space:

[0119] Building types include tourist attractions, government agencies, sports and fitness, real estate, education and training, healthcare, companies, administrative divisions, transportation facilities, administrative landmarks, lifestyle services, finance, cultural media, shopping, hotels, food, leisure and entertainment, entrances and exits, car services, natural features, gates, and roads. It can be used to identify and analyze key urban areas, understand the city's spatial structure and functional layout, and help decision makers formulate reasonable development strategies, resource allocation, and policy measures.

[0120] This application uses a multi-scale AOI spatial analysis method. Multi-scale AOI refers to the consideration and analysis of areas of interest (AOIs) at different scales around a specific area. This method plays an important role in spatial analysis and is a key measure for improving model reliability. By selecting AOIs of different scales to observe and quantify building information around parking spaces, it is possible to capture spatial variations and environmental characteristics at different scales.

[0121] (4) Parking rate statistics:

[0122] The total opening time of each parking space is the last departure time minus the initial use time, which is recorded as t total The sum of the parking time of each parking space is taken as the actual occupancy time, recorded as t use The ratio of the occupied time to the total open time is the parking rate of the car, which is recorded as This indicator can measure the utilization rate of the parking space. The higher the parking rate, the more cars are parked in the parking space at all times, and the utilization rate is high. If the parking rate is low, it means that there are no cars parked in the parking space most of the time, and the utilization rate is very low.

[0123] (5) Parking period statistics:

[0124] To accurately determine whether parking spaces are more frequently parked during the day or at night, parking statistics for each parking space are collected over a 24-hour period. If parking occurs between 9:00 AM and 6:00 PM, the space is classified as daytime; otherwise, it is classified as nighttime. By distinguishing between daytime and nighttime parking spaces, parking management companies can better formulate pricing strategies to maximize pricing profits.

[0125] For example, a parking space has two parking records: the first record is "parked at 09:01 and left at 09:02 the next day," and the second record is "parked at 09:30 and left at 10:00." Initialize the value of each hourly time period in the 24-hour period to 0. For the first record, add 60 minutes to each hourly time period, including 61 minutes from 9:00 to 10:00. For the second record, add 30 minutes from 9:00 to 10:00, leaving all other hours untouched. Finally, except for the 91-minute period from 9:00 to 10:00, all other hourly time periods are 60 minutes. Finally, use softmax to normalize the 24 time periods. Calculate the percentage of parking time in each hourly time period. Then compare the percentage of parking time during the day and at night.

[0126] In one embodiment, a clustering algorithm can be used to determine daytime and nighttime time periods. For example, consider the parking time percentages of {1, [0.1, 0.1, 0.8]; 2, [0.2, 0.2, 0.6]; 3, [0.6, 0.2, 0.2]; 4, [0.8, 0.1, 0.1]}. After clustering, the eigenvectors of 1 and 2 are very close, and they are clustered into one category. The eigenvectors of 3 and 4 are also relatively close, and they are clustered into another category. In this embodiment, the specific 24-hour parking duration percentage of each parking space is statistically analyzed. K-means clustering is then used to cluster these into two categories. Clustering reveals that parking rates are very high after 9:00 AM, while they decrease significantly after 6:00 PM. Therefore, parking between 9:00 AM and 6:00 PM is classified as daytime, and the remaining time periods are classified as nighttime. In subsequent data classification, this rule achieves a high accuracy rate, indicating a certain degree of parameter reliability. For optimization, different clustering analyses can be performed in different cities to obtain different time classification rules.

[0127] (6) Pre-training parking rate prediction model. The specific implementation process is described below.

[0128] (7) Pre-training the parking period prediction model. The specific implementation process is described below.

[0129] (8) Platform deployment: The platform deploys the parking rate prediction model and parking period prediction model to the prediction system based on technologies such as Flask and Layui. The specific process of predicting parking rate and parking period in the prediction system will be described below.

[0130] In one embodiment, the process of pre-training the parking rate prediction model may specifically include:

[0131] Step 301, obtaining information of a first sample building of interest of a first sample parking space;

[0132] Step 302: regressing the first sample building of interest information using a preset first random forest model to obtain a parking rate prediction value;

[0133] Step 303 : Training the first random forest model based on the gap between the parking rate prediction value and the parking rate label value to obtain a parking rate prediction model.

[0134] In step 301, the first sample parking space refers to a parking space used for model training. For example, if a parking lot has 2,000 parking spaces, all 2,000 parking spaces can be used as first sample parking spaces. The first sample building information of interest refers to the building information associated with the first sample parking space, specifically indicating the attributes of each of at least two building types within at least two first sample areas of interest. The first sample parking space has a parking rate label value, which indicates the parking rate of the first sample parking space.

[0135] It should be noted that the definition of the first sample building information of interest is essentially the same as the building information in step 103 above, except that the former is used in the model training phase, while the latter is used in the model application phase. Therefore, the interpretation of the first sample building information of interest can refer to the building information in step 103 above.

[0136] In another embodiment, the process of pre-training the parking rate prediction model may specifically include:

[0137] Obtaining information of a first sample building of interest of a first sample parking space;

[0138] Obtaining the number of first sample residents in the maximum interest area associated with the first sample parking space;

[0139] Obtain the number of first sample parking spaces in the parking lot where the first sample parking space is located;

[0140] A first random forest model is used to regress the first sample building of interest information, the first sample number of residents, and the first sample number of parking spaces to obtain a predicted parking rate value;

[0141] According to the gap between the parking rate prediction value and the parking rate label value, the first random forest model is trained to obtain a parking rate prediction model.

[0142] The benefit of this embodiment is that by comprehensively considering the number of surrounding residents and the number of parking spaces and combining AOI to train the model, the model performance can be further improved, thereby improving the accuracy and applicability of parking rate prediction in actual applications.

[0143] In one embodiment, step 301 may include:

[0144] Step 401, generating a plurality of candidate scales in increasing order;

[0145] Step 402 , generating a candidate region of interest with the first sample parking space as the center point and each candidate scale as the radius;

[0146] Step 403: Evaluate the preset first random forest model based on the building information in the candidate ROI to obtain the mean absolute error and fitting accuracy corresponding to each candidate scale;

[0147] Step 404: Filtering a target scale from multiple candidate scales based on the mean absolute error and the fitting accuracy;

[0148] Step 405: generating at least two sample region scales that are arithmetically increasing based on the target scale, and the largest scale among the at least two sample region scales is the target scale;

[0149] Step 406 , generating at least two first sample interest regions with the first sample parking space as the center point and the sample area scale as the radius;

[0150] Step 407 , obtaining sample building information according to the attributes of each of the at least two building types in the at least two first sample interest regions;

[0151] Step 408: Integrate the sample building information of at least two first sample interest areas to obtain first sample interest building information.

[0152] For steps 401 to 405, in one example, refer to Figure 5 , the candidate scale can be set to 0, 250, 500, 750, 1000, 1250, 1500, 1750. It can be observed that when the candidate scale increases from 0m to 1500m, the mean absolute error MAE gradually decreases and the fitting accuracy R 2 Gradually increases, and after exceeding 1500m, the mean absolute error MAE increases, and the fitting accuracy R 2 Therefore, 1500m can be selected as the target scale, and then the sample area scales generated based on 1500m can include 500m, 1000m, and 1500m. The sample area scales generated based on 1500m can also include 700m, 1100m, and 1500m.2 The details are described below.

[0153] In one embodiment, step 404 may include determining, for the current candidate scale, if the mean absolute error is smaller than the mean absolute error of the next candidate scale, and if the fitting accuracy is greater than the fitting accuracy of the next candidate scale, then determining the current candidate scale as the target scale. This can improve the accuracy of determining the target scale.

[0154] It should be noted that the specific implementation details of steps 406 to 408 can be found in the above description of steps 102 to 104 and will not be repeated here.

[0155] The benefit of the embodiment of steps 401 to 408 is that it can dynamically determine the sample area scale for the first sample parking space and reduce the impact on model performance, thereby improving the flexibility and accuracy of determining the first sample building of interest information and having significant applicability.

[0156] In steps 302 and 303, the first random forest model is based on the Random Forest algorithm. The basic concept of Random Forest is to use a voting mechanism among multiple decision trees to make predictions. Each decision tree is generated from a random subset of the training data, thus avoiding overfitting.

[0157] In regression tasks, the Mean Absolute Error (MAE) and the Coefficient of Determination (R²) are commonly used to evaluate model performance. MAE is a commonly used regression model evaluation metric that measures the average difference between predicted and observed values. The steps for calculating MAE are as follows: for each data point, the absolute error between the predicted value and the true value is calculated. The absolute errors for all data points are then summed. Finally, the total absolute error is divided by the number of data points to obtain the mean absolute error. A notable feature of MAE is that it is robust to outliers, meaning it is not affected by them. Compared to other regression model evaluation metrics, MAE is easier to interpret and understand because it directly represents the absolute magnitude of the average error. However, it is important to note that MAE cannot determine the direction of the deviation between the predicted and actual values. It only focuses on the absolute magnitude of the error and does not provide information on whether the predicted values ​​are biased upward or downward. Therefore, if the directional bias of the predicted values ​​is important, other metrics such as the Mean Squared Error (MSE) or the Mean Percentage Error (MAPE) may be more appropriate. R² (R-squared) is a commonly used regression model evaluation metric used to measure how well a regression model fits the observed data. Its value ranges from 0 to 1, with values ​​closer to 1 indicating a better fit for the data. Specifically, R² measures the proportion of the variability in the dependent variable that is explained by the model. When R² is 0, the model fails to explain any variability in the dependent variable. When R² is between 0 and 1, the model partially explains the variability in the dependent variable, with values ​​closer to 1 indicating a higher degree of explanation. When R² is 1, the model fully explains the variability in the dependent variable, meaning that the model fits the observed values ​​perfectly.

[0158] The benefit of the embodiment of the above steps 301 to 303 is that it can improve the recognition capability of the trained parking rate prediction model for multi-scale AOIs, thereby improving the accuracy of parking rate prediction.

[0159] In one embodiment, the at least two first sample regions of interest include a first-scale sample region of interest, a second-scale sample region of interest, and a third-scale sample region of interest, and the regional scales of the first-scale sample region of interest, the second-scale sample region of interest, and the third-scale sample region of interest increase in arithmetic progression. For example, the first-scale sample region of interest is a 500-meter region of interest, the second-scale sample region of interest is a 1000-meter region of interest, and the third-scale sample region of interest is a 1500-meter region of interest.

[0160] After step 303, the urban parking demand prediction method may further include:

[0161] Step 501: If the region of interest in the first sample building of interest information includes the first-scale sample region of interest and the second-scale sample region of interest, the parking rate prediction model is recorded as a first candidate parking rate prediction model, and a performance evaluation is performed on the first candidate parking rate prediction model to obtain first performance evaluation data.

[0162] Step 502: If the region of interest in the first sample building of interest information includes the first-scale sample region of interest and the third-scale sample region of interest, the parking rate prediction model is recorded as a second candidate parking rate prediction model, and a performance evaluation is performed on the second candidate parking rate prediction model to obtain second performance evaluation data.

[0163] Step 503: If the region of interest in the first sample building of interest information includes the first-scale sample region of interest, the second-scale sample region of interest, and the third-scale sample region of interest, the parking rate prediction model is recorded as a third candidate parking rate prediction model, and a performance evaluation is performed on the third candidate parking rate prediction model to obtain third performance evaluation data.

[0164] Step 504 : Based on the comparison among the first performance evaluation data, the second performance evaluation data, and the third performance evaluation data, a final parking rate prediction model is selected from the first candidate parking rate prediction model, the second candidate parking rate prediction model, and the third candidate parking rate prediction model.

[0165] In steps 501 to 503, the performance evaluation can specifically select the mean absolute error (MAE) or the coefficient of determination (R²) to evaluate the performance of the model. The smaller the mean absolute error, the higher the first model performance evaluation data. 2 The higher the value, the higher the first model performance evaluation data. The second model performance evaluation data and the third model performance evaluation data are similar to the first model performance evaluation data and will not be described in detail here.

[0166] In step 504, it specifically includes: obtaining a comparison result based on the comparison among the first performance evaluation data, the second performance evaluation data and the third performance evaluation data; if the comparison result indicates that the first performance evaluation data is optimal, determining the first candidate parking rate prediction model as the final parking rate prediction model; if the comparison result indicates that the second performance evaluation data is optimal, determining the second candidate parking rate prediction model as the final parking rate prediction model; if the comparison result indicates that the third performance evaluation data is optimal, determining the third candidate parking rate prediction model as the final parking rate prediction model.

[0167] The benefit of the embodiment of steps 501 to 504 is that model performance evaluation can be performed based on different combinations of multiple regional scales to determine a parking rate prediction model with better performance, thereby improving the accuracy of parking rate prediction.

[0168] In one example, multiple AOI scales (500, 1000, and 1500 meters) were selected to cover land use within different ranges. Based on the provided data, parking rates were statistically analyzed, and various regression models were fitted using the total area occupied by the multi-scale AOIs as features, as shown in Table 2 below. Analysis of the results revealed that the random forest combined with multi-scale AOIs approach achieved the best fit, with a fitting efficiency of 0.911. This demonstrates that in the parking rate prediction scenario, the random forest combined with multi-scale AOIs has the strongest explanatory power and can more accurately explain parking rate fluctuations.

[0169]

[0170] Table 2

[0171] In one embodiment, the process of pre-training the parking period prediction model may specifically include:

[0172] Step 601: Obtain the first sample interesting building information and parking period label value of the first sample parking space.

[0173] Step 602: regress the first sample building of interest information using a preset second random forest model to obtain a parking period prediction value;

[0174] Step 603: Based on the gap between the parking period prediction value and the parking rate label value, a second random forest model is trained to obtain a parking period prediction model.

[0175] In classification, precision, recall, and F1 are commonly used for evaluation. True Positives (TP) and True Negatives (TN): indicate the number of examples correctly predicted as positive and negative by the classifier. Higher TP and TN values ​​indicate a high classifier accuracy for these categories. False Positives (FP) and False Negatives (FN): indicate the number of examples incorrectly predicted as positive and negative by the classifier. Higher FP and FN values ​​may indicate misclassification or missed examples. Accuracy: calculated as (TP+TN) / (TP+TN+FP+FN), indicates the proportion of examples correctly predicted by the classifier. A high accuracy indicates a strong overall classifier's predictive power, but this may be affected by the ratio of positive and negative examples in imbalanced datasets. Precision: calculated as TP / (TP+FP), indicates the proportion of true positives among all examples predicted as positive. A high precision indicates a low false positive rate among examples predicted as positive. Recall (also known as sensitivity, true positive rate): The calculation formula is TP / (TP+FN), which indicates the proportion of samples that are correctly predicted to be positive examples among all samples that are actually positive examples. A higher recall rate means that the classifier can capture more positive examples. F1 value: It takes into account the precision and recall rates, and is the harmonic mean of the precision and recall rates. The calculation formula is 2*(Precision*Recall) / (Precision+Recall). A higher F1 value means that the classifier performs better in maintaining a balance between precision and recall. This embodiment found that the method of using random forest combined with multi-scale AOI can achieve the highest accuracy. Therefore, this algorithm was chosen to classify parking periods.

[0176] In another embodiment, the process of pre-training the parking period prediction model may specifically include:

[0177] Obtaining information of a first sample building of interest of a first sample parking space;

[0178] Obtaining the number of first sample residents in the maximum interest area associated with the first sample parking space;

[0179] Obtain the number of first sample parking spaces in the parking lot where the first sample parking space is located;

[0180] The preset second random forest model is used to regress the first sample building information of interest, the first sample number of residents, and the first sample number of parking spaces to obtain a predicted parking rate value;

[0181] According to the gap between the parking rate prediction value and the parking rate label value, the second random forest model is trained to obtain the parking rate prediction model.

[0182] The benefit of this embodiment is that by comprehensively considering the number of surrounding residents and the number of parking spaces and combining AOI to train the model, the model performance can be further improved, thereby improving the accuracy and applicability of parking period prediction in actual applications.

[0183] Considering that too few parking times in some parking spaces will affect the accuracy of parking period prediction, it is necessary to delete some abnormal parking spaces during the pre-training of the parking period prediction model. Based on this, in one embodiment, the process of pre-training the parking period prediction model may specifically include:

[0184] Step 701: Obtain the parking times of at least two first sample parking spaces, and use each of the at least two parking times in turn as an initial parking times threshold;

[0185] Step 702: The first sample parking space with a parking count greater than the initial parking count threshold is used as a reference sample parking space, and the number of reference sample parking spaces is counted to obtain the total number of remaining parking spaces corresponding to each initial parking count threshold.

[0186] Step 703: Evaluate the preset second random forest model based on the building information associated with each reference sample parking space to obtain the model accuracy corresponding to each initial parking number threshold;

[0187] Step 704: Filtering a target parking number threshold from multiple initial parking number thresholds based on the total number of remaining parking spaces and model accuracy;

[0188] Step 705: retain the first sample parking spaces whose parking times are greater than the target parking times threshold to obtain second sample parking spaces;

[0189] Step 706: Regress the second sample building of interest information of the second sample parking space using a second random forest model to obtain a parking period prediction value; wherein the second sample parking space has a parking period label value, and the parking period label value indicates the parking period of the second sample parking space;

[0190] Step 707 : Based on the difference between the parking period prediction value and the parking period label value, the second random forest model is trained to obtain a parking period prediction model.

[0191] It should be noted that the second sample interest buildings refer to the building information associated with the second sample parking spaces, specifically indicating the attributes of each of the at least two second sample interest areas of at least two building types.

[0192] In one example, the total number of parking spaces is 11,070. If the number of parking times for a parking space is too small, it means that the parking space may have just been used or has not been used much. In order to ensure the accuracy of the parking rate of parking spaces, this embodiment needs to use data with a large number of parking times to train the model. Figure 6 As shown in the figure, the horizontal axis represents the initial parking threshold, the left vertical axis represents the total number of remaining parking spaces, and the right vertical axis represents model accuracy. Experiments have found that when the total number of remaining parking spaces is approximately 1300, the total number of remaining parking spaces is not too small, and the model accuracy is also at a high level. Therefore, 1300 is used as the minimum parking threshold for model training.

[0193] Before training, the data was divided into training and test sets. The training set accounted for 70% of the total data, and the remaining 30% served as the test set. After screening, 3,629 parking spaces that met the requirements were identified. For parking space classification, a multi-scale AOI (area of ​​interest) algorithm combined with multiple classifiers was used. Parking time periods were classified and the classification performance of different methods was evaluated. The classification report is shown in Table 3. "Support" refers to the number of parking spaces in the first sample.

[0194]

[0195] Table 3

[0196] The benefit of the embodiment of steps 501 to 507 is that the parking spaces are first filtered based on the number of parking times and then used for model training, thereby improving the accuracy of the trained parking period prediction model in predicting parking periods.

[0197] In one embodiment, step 504 may include:

[0198] Screening multiple initial parking number thresholds according to the model accuracy and a preset accuracy threshold to obtain at least two candidate parking number thresholds;

[0199] At least two candidate parking number thresholds are sorted according to the total number of remaining parking spaces, so as to determine the largest candidate parking number threshold as the target parking number threshold.

[0200] The benefit of the above embodiment is that by first screening based on model accuracy and then screening based on the total number of remaining parking spaces, the rationality of the target parking number threshold can be effectively improved, and both model accuracy and sample number can be taken into account.

[0201] After pre-training the parking rate prediction model and parking period prediction model, the platform deploys technologies such as Flask and Layui to embed these models into the prediction system. If the user clicks on a parking spot that is a parking space, information such as the parking rate and parking period can be obtained. If there are no parking spaces, the backend automatically calculates the area of ​​interest (AOI) of the target parking spot and then uses the trained model for regression and classification, resulting in predictions of parking rates and parking periods for unknown areas.

[0202] In practical applications, refer to Figure 7 , the urban parking demand prediction method may include the following steps:

[0203] (1) Start the prediction system to display the parking demand prediction interface;

[0204] (2) Click on the parking demand prediction interface to determine the target parking spot;

[0205] (3) Determine whether the target parking spot is a parking space; if not, proceed to step (4); if yes, proceed to step (5);

[0206] (4) Display parking space statistics associated with the parking space, including parking rate and parking period, and may also include parking space ID, road section name, number of parking times, total parking time, earliest parking time, latest parking time, usage time, and parking situation at each hour.

[0207] (5) Obtaining the target building of interest information at the target parking spot;

[0208] (6) Model prediction, including: regressing the target interest building information through the parking rate prediction model to obtain the target predicted parking rate, and classifying the target interest building information through the parking period prediction model to obtain the target predicted parking period;

[0209] (7) Display the prediction results, including: displaying the target predicted parking rate and the target predicted parking period.

[0210] Based on the above embodiments, this application has the following innovations:

[0211] 1. Multi-scale AOI spatial analysis method: This paper adopts a multi-scale AOI spatial analysis method. By selecting AOIs with different buffer sizes to observe and quantify the building information around parking spaces, it captures spatial changes and environmental characteristics at different scales and improves the reliability of the model.

[0212] 2. Application of artificial intelligence algorithms: The random forest algorithm combined with multi-scale AOI data was used to perform regression analysis of parking rates and classification analysis of parking periods, achieving excellent fitting results and classification accuracy of 0.911 and 98%, respectively.

[0213] 3. Refined prediction: Compared with related technologies, this application not only makes predictions at the city level, but also at the meso- and micro-levels, including parking demand predictions for specific streets and specific parking spaces, meeting the refined needs of urban traffic management and optimal allocation of parking resources.

[0214] Based on these innovations, this application offers significant advantages in the accuracy and practicality of parking demand forecasting, providing scientific decision-making support for urban parking management. For example, this application can effectively estimate parking rates and parking time periods at user-selected coordinates (target parking spots). This application utilizes a multi-scale AOI algorithm combined with a random forest algorithm to ensure accurate calculations, enabling parking management companies to optimize resource allocation, improve operational efficiency, and develop dynamic pricing strategies.

[0215] The various technical features in the above embodiments can be combined arbitrarily as long as there is no conflict or contradiction between the combinations of features. However, due to space limitations, they are not described one by one. Therefore, the arbitrary combination of the various technical features in the above embodiments also falls within the scope of disclosure of this specification.

[0216] The present application also discloses a device for predicting urban parking demand, referring to Figure 8 The urban parking demand prediction device includes: a parking point determination module 801, used to determine a target parking point in a target city; an interest area determination module 802, used to determine at least two interest areas containing the target parking point, where the regional scales of any two interest areas are different; an information acquisition module 803, used to obtain the attributes of at least two building types in each interest area to obtain building information; an information integration module 804, used to integrate the building information of at least two interest areas to obtain target interest building information; a parking rate prediction module 805, used to regress the target interest building information through a pre-trained parking rate prediction model to obtain a target predicted parking rate of the target parking point; a parking period prediction module 806, used to classify the target interest building information through a pre-trained parking period prediction model to obtain a target predicted parking period of the target parking point; and a data display module 807, used to display the target predicted parking rate and the target predicted parking period.

[0217] In one embodiment, the urban parking demand prediction device further includes a first training module for pre-training a parking rate prediction model.

[0218] In one embodiment, the urban parking demand prediction device further includes a second training module for pre-training a parking period prediction model.

[0219] In one embodiment, the urban parking demand prediction device further includes: a resident number acquisition module for acquiring the number of residents in the maximum interest area associated with the target parking point; and a parking number acquisition module for acquiring the estimated number of parking spaces set for the target parking point.

[0220] It should be noted that the specific implementation of the urban parking demand prediction device is basically the same as the specific embodiment of the above-mentioned urban parking demand prediction method, and will not be repeated here.

[0221] The present application also discloses an urban parking demand prediction device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the urban parking demand prediction method described above is implemented. The urban parking demand prediction device can be, for example, a mobile phone or a vehicle.

[0222] An embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned urban parking demand prediction method.

[0223] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0224] An embodiment of the present application also provides a computer program product, which includes a computer program. The computer program is read and executed by a processor, so that when the processor executes the computer program, the urban parking demand prediction method as described above is implemented.

[0225] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0226] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0227] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0228] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0229] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0230] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0231] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0232] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0233] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0234] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: various media that can store programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0235] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A method for predicting urban parking demand, characterized in that: The method comprises: Determine the target parking spot in the target city; determining at least two regions of interest that include the target parking point, wherein the regional scales of any two regions of interest are different; Obtaining attributes of at least two building types in each of the interest areas to obtain building information; Integrating the building information of the at least two areas of interest to obtain target interest building information; Regressing the target building of interest information using a pre-trained parking rate prediction model to obtain a target predicted parking rate for the target parking spot; Classifying the target building of interest information using a pre-trained parking period prediction model to obtain a target predicted parking period for the target parking spot; displaying the target predicted parking rate and the target predicted parking period; The method further includes pre-training the parking rate prediction model, specifically including: Obtaining first sample building information of a first sample parking space, wherein the first sample building information indicates attributes of each of at least two building types in at least two first sample regions of interest, the first sample parking space having a parking rate label value, the parking rate label value indicating a parking rate of the first sample parking space, and the at least two first sample regions of interest including a first-scale sample region of interest, a second-scale sample region of interest, and a third-scale sample region of interest, with equidistantly increasing regional scales; Regressing the first sample building of interest information through a preset first random forest model to obtain a parking rate prediction value; Training the first random forest model according to the gap between the parking rate prediction value and the parking rate label value to obtain a parking rate prediction model; If the region of interest in the first sample interest building information includes the first-scale sample region of interest and the second-scale sample region of interest, record the parking rate prediction model as a first candidate parking rate prediction model, and perform a performance evaluation on the first candidate parking rate prediction model to obtain first performance evaluation data; If the region of interest in the first sample interest building information includes the first-scale sample region of interest and the third-scale sample region of interest, recording the parking rate prediction model as a second candidate parking rate prediction model, and performing a performance evaluation on the second candidate parking rate prediction model to obtain second performance evaluation data; If the region of interest in the first sample building of interest information includes the first-scale sample region of interest, the second-scale sample region of interest, and the third-scale sample region of interest, recording the parking rate prediction model as a third candidate parking rate prediction model, and performing a performance evaluation on the third candidate parking rate prediction model to obtain third performance evaluation data; According to the comparison among the first performance evaluation data, the second performance evaluation data and the third performance evaluation data, a final parking rate prediction model is screened out from the first candidate parking rate prediction model, the second candidate parking rate prediction model and the third candidate parking rate prediction model.

2. The method according to claim 1, characterized in that The determining of at least two regions of interest including the target parking point includes: Classifying the target parking spot according to a plurality of reference parking lots in the target city to obtain a parking spot category; If the parking spot category indicates that the target parking spot does not belong to any of the reference parking lots, generating at least two regions of interest, each having at least two building types, with the target parking spot as a center point, wherein the regional scales of the at least two regions of interest are equidistantly increased; If the parking point category indicates that the target parking point belongs to one of the reference parking lots, the at least two regions of interest are generated with the reference parking lot to which the target parking point belongs as a center point and with at least two area scales associated with the reference parking lot as radii.

3. The method according to claim 1, characterized in that The obtaining of the first sample interesting building information of the first sample parking space includes: Generate an increasing number of candidate scales; Generate a candidate region of interest with the first sample parking space as a center point and each candidate scale as a radius; Evaluate a preset first random forest model based on the building information in the candidate area of ​​interest to obtain a mean absolute error and a fitting accuracy corresponding to each candidate scale; Selecting a target scale from a plurality of candidate scales according to the mean absolute error and the fitting accuracy; generating at least two sample region scales that increase arithmetically based on the target scale, wherein the largest scale among the at least two sample region scales is the target scale; generating the at least two first sample interest regions with the first sample parking space as a center point and the sample area scale as a radius; Obtaining sample building information according to the attributes of each of the at least two first sample interest areas in the at least two first sample interest areas; The sample building information of the at least two first sample interest areas is integrated to obtain first sample interest building information.

4. The method according to claim 1, wherein The method further includes pre-training the parking period prediction model, specifically comprising: Obtaining the number of parking times of each of at least two of the first sample parking spaces, and taking each of the at least two parking times in turn as an initial parking times threshold; The first sample parking space whose parking number is greater than the initial parking number threshold is used as a reference sample parking space, and the number of the reference sample parking spaces is counted to obtain the total number of remaining parking spaces corresponding to each initial parking number threshold; Evaluating a preset second random forest model based on the building information associated with each reference sample parking space to obtain a model accuracy corresponding to each initial parking number threshold; Filtering a target parking number threshold from the multiple initial parking number thresholds according to the total number of remaining parking spaces and the model accuracy; The first sample parking space whose parking times is greater than the target parking times threshold is reserved to obtain a second sample parking space; Regressing the second sample building of interest information of the second sample parking space using the second random forest model to obtain a parking period prediction value; wherein the second sample parking space has a parking period label value, and the parking period label value indicates the parking period of the second sample parking space; The second random forest model is trained according to the gap between the parking period prediction value and the parking period label value to obtain the parking period prediction model.

5. The method according to any one of claims 1 to 4, characterized in that After determining the target parking point in the target city, the method further includes: Obtaining the number of residents in the largest area of ​​interest associated with the target parking point; Obtaining the estimated number of parking spaces set for the target parking spot; The step of regressing the target building of interest information using a pre-trained parking rate prediction model to obtain a target predicted parking rate of the target parking spot includes: The target building information of interest, the number of residents and the estimated number of parking spaces are regressed using a pre-trained parking rate prediction model to obtain a target predicted parking rate based on the estimated number of parking spaces.

6. A device for predicting urban parking demand, characterized in that: For executing the method according to any one of claims 1 to 5, the apparatus comprises: A parking point determination module is used to determine a target parking point in a target city; an area of ​​interest determination module, configured to determine at least two areas of interest including the target parking point, wherein the area scales of any two areas of interest are different; An information acquisition module, configured to acquire attributes of at least two building types in each of the interest areas to obtain building information; an information integration module, configured to integrate the building information of the at least two areas of interest to obtain target building information of interest; a parking rate prediction module, configured to regress the target building of interest information using a pre-trained parking rate prediction model to obtain a target predicted parking rate for the target parking spot; A parking period prediction module is used to classify the target building of interest information using a pre-trained parking period prediction model to obtain a target predicted parking period for the target parking spot; A data display module is used to display the target predicted parking rate and the target predicted parking period.

7. An urban parking demand prediction device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which implements the method according to any one of claims 1 to 5 when executed.

Images