Parking Lot Revenue Forecasting and Calculation Method Based on Spatiotemporal Graph Learning

Abstract

The present invention belongs to the technical field of revenue prediction, and in particular, relates to a parking lot revenue prediction and accounting method based on spatiotemporal graph learning. By uniformly modeling the historical data, real-time usage and surrounding environment of the parking lot, a spatiotemporal graph convolutional network is used to capture dynamic spatiotemporal relationships, so as to provide scientific revenue prediction and accurate accounting for parking lot operation and management. In addition, multi-source data is combined to predict future revenue from historical parking lot utilization, and a dynamic relationship with the parking lot is obtained from real-time operation data. The prediction results are dynamically adjusted according to external factors such as real-time traffic and policy changes.

Description

Technical Field

[0001] The present invention belongs to the technical field of revenue prediction, and in particular to a parking lot revenue prediction and accounting method based on spatiotemporal graph learning, which is used for performing revenue evaluation before parking lot contracting and revenue accounting after contracting. Background Art

[0002] With the continuous advancement of urbanization, parking issues have become an important challenge in urban traffic management. The rapid development of cities has brought about a surge in the number of vehicles. However, the growth of parking facilities has been unable to keep up with the pace of demand, resulting in increasingly serious parking problems. For investors and operators, how to accurately evaluate the future benefits of parking lots before contracting them and to accurately calculate the benefits during the operation process are key factors in determining the success of their investment and operation. However, most of the current parking lot revenue evaluation and accounting methods mainly rely on historical data and experience, lack scientific forecasting models, and are difficult to provide accurate revenue forecasts and accounting in a changing market environment. At present, the current parking lot revenue forecasting and accounting faces the following major challenges: First, data complexity: First, parking lot revenue is affected by many factors, including geographical location, city level, regional type, number of parking lots, number of entrances and exits, and surrounding points of interest. These factors are not only diverse, but also have complex correlations with each other. For example, geographical location and city level will affect the frequency of use and charging standards of parking lots, while the number of parking lots and the number of entrances and exits will directly affect the parking lot's capacity and turnover efficiency. To accurately predict revenue, these multi-dimensional factors must be considered comprehensively. Secondly, the market environment and user needs are changing dynamically. For example, the development of surrounding commercial areas, adjustments to traffic policies, seasonal changes, etc., will all have an impact on parking lot revenue. Traditional static evaluation methods are difficult to capture these dynamic changes, resulting in inaccurate prediction results. Second, it is difficult to match similar parking lots: When making revenue predictions, it is crucial to find historical parking lot data similar to the target parking lot. However, due to the diverse characteristics of parking lots and high data dimensions, traditional methods are difficult to effectively match similar parking lots. Therefore, it is necessary to design a parking lot revenue prediction and accounting method based on spatiotemporal graph learning. Summary of the invention

[0003] In order to solve the above problems in the prior art, the present invention proposes a parking lot revenue prediction and accounting method based on spatiotemporal graph learning. By uniformly modeling the historical data, real-time usage and surrounding environment of the parking lot, and using the spatiotemporal graph convolutional network (ST-GCN) to capture dynamic spatiotemporal relationships, it provides scientific revenue prediction and accurate accounting for parking lot operation and management. In addition, it combines multi-source data to predict future revenue from historical parking lot utilization, obtains a dynamic relationship with the parking lot from real-time operation data, and dynamically adjusts the prediction results according to external factors such as real-time traffic and policy changes.

[0004] In order to achieve the above object, the present invention adopts the following technical solution:

[0005] A parking lot revenue prediction and accounting method based on spatiotemporal graph learning includes the following steps:

[0006] S1. Parking lot data collection: Collect parking lot data of target parking lots and existing parking lots as well as external real-time feature data;

[0007] S2. Preprocessing of parking lot historical data: Use the parking lot historical data to conduct statistical analysis, obtain the historical change characteristics of the existing parking lot revenue, and standardize these characteristic data, process the parking lot data, and vectorize the text data;

[0008] S3, data set division: the collected parking lot data is divided into training set, validation set and test set, which are used for training, validation and testing models;

[0009] S4, spatiotemporal graph learning: construct a spatiotemporal graph, and based on the constructed spatiotemporal graph, use the spatiotemporal graph convolutional network (ST-GCN) to extract the spatiotemporal features of existing parking lots and their revenues, perform spatiotemporal graph learning, and obtain a spatiotemporal graph prediction model;

[0010] S5. Most similar parking lot matching: clustering existing parking lots according to the parking lot data, calculating the distances between the target parking lot and various cluster centers of the existing parking lots, thereby matching the parking lot that is most similar to the target parking lot;

[0011] S6. Target parking lot revenue prediction: by analyzing the historical revenue data of the most similar parking lot, the spatiotemporal graph prediction model obtained in step S4 is used to predict the revenue of the target parking lot in the next day;

[0012] S7. Real-time revenue calculation: After the parking lot goes online, the revenue data is collected in real time, compared with the predicted revenue, and the calculation results are fed back to the space-time graph prediction model, and the model is retrained and optimized through machine learning methods.

[0013] As a further technical solution of the present invention, the parking lot data in step S1 include geographical location, city level, area type, number of parking lots, number of entrances and exits, surrounding points of interest, and parking lot utilization rate and income data in different time periods; the external real-time feature data includes daily weather, road congestion, and holiday data.

[0014] As a further technical solution of the present invention, the process of step S2 is: extract the historical utilization change characteristics of weekdays and weekends respectively, the daily revenue change trend is composed of the revenue data of all days that meet the conditions before this day, the data of all previous weekdays are selected for weekdays, and the data of all previous weekends are selected for weekends; the processed parking lot data set is , whose dimensions are , is the number of days in the dataset, and N is the number of parking lots.

[0015] As a further technical solution of the present invention, the specific process of step S4 is:

[0016] S41. Represent the parking lot distribution network as a graph , where V is the node set, representing the set of all parking lots; E is the edge set, representing the road connections between all parking lots, and the node features are , represents the historical revenue data of each parking lot, and constructs the adjacency matrix Represent the connection relationship between parking lots, and initialize A with the physical distance of each parking lot to obtain a space-time graph;

[0017] S42. Based on the spatiotemporal graph, the spatiotemporal graph convolutional network is used to extract spatial features and temporal features. The extracted spatial features are ,in, is the adjacency matrix Add self-join , yes The degree matrix of is the weight matrix, is the dimension of the weight matrix, is the ReLU activation function; the time feature is ,in, is the temporal convolution kernel, represents the convolution operation, is the dimension of temporal convolution;

[0018] S43, the vector composed of parking lot data With time characteristics Combined to get a comprehensive representation ;

[0019] S44. Predicting the revenue of parking lots : ,in is the learnable parameter matrix, It is the predicted value of the revenue of all parking lots in the next day. The minimum mean square error is used as the loss function based on the actual parking lot revenue. The network is continuously trained until the loss value no longer decreases. The network model with the minimum total loss value is saved as the spatiotemporal graph prediction model.

[0020] As a further technical solution of the present invention, the specific process of step S5 is:

[0021] S51, randomly initialize K cluster centers ;

[0022] S52. Calculate each parking lot Euclidean distance to all cluster centers , and assign it to the cluster center with the smallest distance. The new cluster center is the mean of all parking lots assigned to the cluster center;

[0023] S53, repeating step S52 until the cluster centers converge or the maximum number of iterations is reached;

[0024] S54. According to the distance between the target parking lot and each cluster center, the cluster center with the smallest distance is selected, and the parking lot closest to the target parking lot is found among the parking lots to which the cluster center belongs as the most similar parking lot. The target parking lot is set to , its most similar parking lot The matching optimization function is: ,in is the cluster center closest to the target parking lot .

[0025] As a further technical solution of the present invention, the specific process of step S6 is: analyzing the historical revenue data of the most similar parking lot, converting the longitude and latitude coordinates of the parking lot , POI information of the nearest building , City level , area 、Business Type , Berths , Number of exports and the number of entrances Input the spatiotemporal graph prediction model to predict the parking lot’s revenue for the next day , the cluster center closest to the target parking lot The mean revenue of all parking lots in the cluster for the next day is used as the revenue forecast value of the target parking lot. : .

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] The present invention captures dynamic spatiotemporal relationships based on a spatiotemporal graph convolutional network (ST-GCN), combines historical parking revenue, real-time operation data, POI information and external factors (such as weather, policy changes, traffic conditions, etc.) to accurately predict and calculate revenue. On the one hand, the present invention can effectively solve the complex problem of multiple factors affecting parking revenue prediction. Through the spatiotemporal graph convolutional network (ST-GCN), the spatial characteristics and temporal characteristics of similar parking lots are comprehensively considered to dynamically capture the trend of parking revenue changes, thereby achieving more accurate revenue prediction; on the other hand, the present invention can comprehensively reflect the impact of the surrounding environment and dynamic changes of the parking lot by introducing surrounding POI data and other external factors, and improve the accuracy and reliability of the prediction model. Specifically, based on POI data and geographic location information, a more sophisticated spatiotemporal graph structure is constructed, so that the model can better understand the relationship and influence between parking lots. In addition, the present invention uses a spatiotemporal graph convolutional network (ST-GCN) instead of traditional static methods or simple time series models, which can significantly improve the computational efficiency and prediction accuracy of the model. Through the combination of graph convolution and time convolution, ST-GCN can maintain high efficiency when processing large-scale data and adapt to complex parking lot operation environments. The present invention also combines real-time data for dynamic adjustment and calculation, making revenue forecasts more reliable. During parking lot operations, real-time collected operational data and external factor data can dynamically update model parameters to ensure that the forecast results are highly consistent with actual conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0029] Figure 1 A flow chart of the parking lot revenue prediction and accounting method based on spatiotemporal graph learning provided by the present invention. DETAILED DESCRIPTION

[0030] The present invention is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that the following detailed descriptions are exemplary and are intended to provide further description of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those generally understood by those of ordinary skill in the art to which the present invention belongs. It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context otherwise clearly indicates, the singular form is also intended to include the plural form. In addition, it should also be understood that the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units need not be 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. In the absence of conflict, the embodiments in the present invention and the features in the embodiments may be combined with each other.

[0031] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0032] Example:

[0033] like Figure 1 As shown, this embodiment provides a parking lot revenue prediction and accounting method based on spatiotemporal graph learning, comprising the following steps:

[0034] S1. Parking lot data collection:

[0035] In order to calculate the parking lot revenue more accurately, all parking lots in the target parking lot are selected as candidate matching parking lots. There are N parking lots in total, and each parking lot is labeled with an id, and the name, latitude and longitude coordinates of the parking lot are recorded. , POI information of the nearest building , City level , area 、Business Type , Berths , Number of exports and the number of entrances , and collect the parking revenue for each time interval at daily intervals ; and collect external real-time feature data. The real-time feature data set includes daily weather, road congestion, and holiday data.

[0036] S2. Preprocess parking lot historical data:

[0037] This step uses the entire parking lot mobile phone data set to construct the historical revenue change trend. Since the number of parking lots and peak hours on weekdays and weekends are different, the historical utilization change characteristics of weekdays and weekends are extracted respectively. The daily revenue change trend is composed of the revenue data of all days that meet the conditions before this day. The data of all previous weekdays is selected for weekdays, and the data of all previous weekends is selected for weekends. The data set after processing the parking lot data is , whose dimensions are , is the number of days in the dataset, and N is the number of parking lots.

[0038] S3, data set division:

[0039] This step is to divide the data set for training the model, including parking lot data and external real-time feature data, and divide the data set into a training set, a validation set, and a test set. The training set is used to train the model, the validation set is used to verify the performance of the model during training, and the test set is used to test the performance of the model after training.

[0040] S4. Spatiotemporal graph learning:

[0041] Construct a spatiotemporal graph, and based on the constructed spatiotemporal graph, use the spatiotemporal graph convolutional network (ST-GCN) to extract the spatiotemporal features of existing parking lots and their revenues, perform spatiotemporal graph learning, and obtain a spatiotemporal graph prediction model; the specific process is as follows:

[0042] S41. Represent the parking lot distribution network as a graph , where V is the node set, representing the set of all parking lots; E is the edge set, representing the road connections between all parking lots, and the node features are , represents the historical revenue data of each parking lot, and constructs the adjacency matrix Represent the connection relationship between parking lots, and initialize A with the physical distance of each parking lot to obtain a space-time graph;

[0043] S42. Based on the spatiotemporal graph, the spatiotemporal graph convolutional network is used to extract spatial features and temporal features. The extracted spatial features are ,in, is the adjacency matrix Add self-join , yes The degree matrix of is the weight matrix, is the dimension of the weight matrix, is the ReLU activation function; the time feature is ,in, is the temporal convolution kernel, represents the convolution operation, is the dimension of temporal convolution;

[0044] S43, the vector composed of parking lot data With time characteristics Combined to get a comprehensive representation ;

[0045] S44. Predicting the revenue of parking lots : ,in is the learnable parameter matrix, It is the predicted value of the revenue of all parking lots in the next day. The minimum mean square error is used as the loss function based on the actual parking lot revenue. The network is continuously trained until the loss value no longer decreases. The network model with the minimum total loss value is saved as the spatiotemporal graph prediction model.

[0046] S5, most similar parking lot matching:

[0047] This step is based on the properties of the parking lot. Clustering is performed and the distances between the target parking lot and the various cluster centers of the existing parking lots are calculated, so as to match the parking lot that is most similar to the target parking lot. The specific process is as follows:

[0048] S51, randomly initialize K cluster centers ;

[0049] S52. Calculate each parking lot Euclidean distance to all cluster centers , and assign it to the cluster center with the smallest distance. The new cluster center is the mean of all parking lots assigned to the cluster center;

[0050] S53, repeating step S52 until the cluster centers converge or the maximum number of iterations is reached;

[0051] S54. According to the distance between the target parking lot and each cluster center, the cluster center with the smallest distance is selected, and the parking lot closest to the target parking lot is found among the parking lots to which the cluster center belongs as the most similar parking lot. The target parking lot is set to , its most similar parking lot The matching optimization function is: ,in is the cluster center closest to the target parking lot .

[0052] S6. Target parking lot revenue prediction: Analyze the historical revenue data of the most similar parking lot and convert the latitude and longitude coordinates of the parking lot , POI information of the nearest building , City level , area 、Business Type , Berths , Number of exports and the number of entrances Input the spatiotemporal graph prediction model to predict the parking lot’s revenue for the next day , the cluster center closest to the target parking lot The mean revenue of all parking lots in the cluster for the next day is used as the revenue forecast value of the target parking lot. : .

[0053] S7. Real-time revenue calculation:

[0054] Used to collect actual revenue in real time after the parking lot goes online and forecasted returns , if the difference between the actual return and the predicted return If it is not 0, the actual income And the corresponding parking lot historical revenue data Re-input into the space-time graph for feedback learning to continuously improve the accuracy of the revenue prediction model.

[0055] The method of this embodiment is used to conduct simulation experiments with existing Arima, RNN, and Transformer models. The simulation experiments are measured using root mean square error and mean absolute error. The results are shown in Table 1. It can be seen from Table 1 that the method of this embodiment greatly improves the accuracy of parking lot revenue prediction.

[0056] Table 1 Comparison of the accuracy of this embodiment with other algorithms

[0057] .

[0058] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0059] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A parking lot revenue prediction and accounting method based on spatiotemporal graph learning, characterized in that: The following steps are involved: S1. Parking lot data collection: Collect parking lot data of target parking lots and existing parking lots as well as external real-time feature data; S2. Preprocessing of parking lot historical data: Use the parking lot historical data to conduct statistical analysis, obtain the historical change characteristics of the existing parking lot revenue, and standardize these characteristic data, process the parking lot data, and vectorize the text data; S3, data set division: the collected parking lot data is divided into training set, validation set and test set, which are used for training, validation and testing models; S4, spatiotemporal graph learning: construct a spatiotemporal graph, and based on the constructed spatiotemporal graph, use the spatiotemporal graph convolutional network to extract the spatiotemporal features of existing parking lots and their revenues, perform spatiotemporal graph learning, and obtain a spatiotemporal graph prediction model; S5. Most similar parking lot matching: clustering existing parking lots according to the parking lot data, calculating the distances between the target parking lot and various cluster centers of the existing parking lots, thereby matching the parking lot that is most similar to the target parking lot; S6. Target parking lot revenue prediction: Analyze the historical revenue data of the most similar parking lot and convert the latitude and longitude coordinates of the parking lot , POI information of the nearest building , City level , area 、Business Type , Berths , Number of exports and the number of entrances Input the spatiotemporal graph prediction model to predict the parking lot's revenue for the next day , the cluster center closest to the target parking lot The mean revenue of all parking lots in the cluster for the next day is used as the revenue forecast value of the target parking lot. : ; S7. Real-time revenue calculation: After the parking lot goes online, the revenue data is collected in real time, compared with the predicted revenue, and the calculation results are fed back to the space-time graph prediction model, and the model is retrained and optimized through machine learning methods.

2. The parking lot revenue prediction and accounting method based on spatiotemporal graph learning according to claim 1 is characterized in that: The parking lot data in step S1 include geographical location, city level, area type, number of parking lots, number of entrances and exits, surrounding points of interest, and parking lot utilization rate and income data at different time periods; external real-time feature data include daily weather, road congestion, and holiday data.

3. The parking lot revenue prediction and accounting method based on spatiotemporal graph learning according to claim 2 is characterized in that: The process of step S2 is: extract the historical utilization change characteristics of weekdays and weekends respectively. The daily revenue change trend is composed of the revenue data of all days that meet the conditions before this day. The data of all previous weekdays are selected for weekdays, and the data of all previous weekends are selected for weekends. The processed parking lot data set is , whose dimensions are , is the number of days in the dataset, and N is the number of parking lots.

4. The parking lot revenue prediction and accounting method based on spatiotemporal graph learning according to claim 3 is characterized in that: The specific process of step S4 is: S41. Represent the parking lot distribution network as a graph , where V is the node set, representing the set of all parking lots; E is the edge set, representing the road connections between all parking lots, and the node features are , represents the historical revenue data of each parking lot, and constructs the adjacency matrix Represent the connection relationship between parking lots, and initialize A with the physical distance of each parking lot to obtain a space-time graph; S42. Based on the spatiotemporal graph, the spatiotemporal graph convolutional network is used to extract spatial features and temporal features. The extracted spatial features are ,in, is the adjacency matrix Add self-join , yes The degree matrix of is the weight matrix, is the characteristic dimension of the weight matrix, is the ReLU activation function; the time feature is ,in, is the temporal convolution kernel, represents the convolution operation, is the feature dimension of temporal convolution; S43, the vector composed of parking lot data With time characteristics Combined to get a comprehensive representation ; S44. Predicting the revenue of parking lots : ,in is the learnable parameter matrix, It is the predicted value of the revenue of all parking lots in the next day. The minimum mean square error is used as the loss function based on the actual parking lot revenue. The network is continuously trained until the loss value no longer decreases. The network model with the minimum total loss value is saved as the spatiotemporal graph prediction model.

5. The parking lot revenue prediction and accounting method based on spatiotemporal graph learning according to claim 4 is characterized in that: The specific process of step S5 is: S51, randomly initialize K cluster centers ; S52, calculate each parking lot Euclidean distance to all cluster centers , and assign it to the cluster center with the smallest distance. The new cluster center is the mean of all parking lots assigned to the cluster center; S53, repeating step S52 until the cluster centers converge or the maximum number of iterations is reached; S54. According to the distance between the target parking lot and each cluster center, the cluster center with the smallest distance is selected, and the parking lot closest to the target parking lot is found among the parking lots to which the cluster center belongs as the most similar parking lot. The target parking lot is set to , its most similar parking lot The matching optimization function is: ,in is the cluster center closest to the target parking lot .

Images