Dynamic parking demand prediction and resource allocation method based on artificial intelligence

Abstract

The invention discloses a dynamic parking demand prediction and resource allocation method based on artificial intelligence, and the method comprises the steps: S1, constructing a data processing platform, and receiving and processing multi-dimensional data; s2, analyzing the received multi-dimensional data on the data processing platform by adopting deep learning and machine learning algorithms; s3, deploying a dynamic response system based on the identified mode and trend, and dynamically adjusting parking resources according to the changes; s4, integrating an operational research algorithm, a genetic algorithm and a machine learning model into a decision support system to optimize parking resource optimal configuration under multiple targets of parking lot income; and S5, adopting a self-adaptive learning mechanism to enable the dynamic response system to learn and adjust according to data obtained from actual operation. The method has the advantages that the parking demand prediction accuracy is improved, the parking resource allocation efficiency is optimized, and the user satisfaction and the environment change adaptation capability are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based dynamic parking demand prediction and resource allocation method. Background Art

[0002] With the acceleration of urbanization, the surge in car ownership has made parking a major problem in urban management. Traditional parking management methods mainly rely on fixed parking resource allocation and static parking demand forecasts, which are unable to cope with the growing and unpredictable parking demand. Especially in commercial centers, residential areas and areas with frequent public activities, the unreasonable allocation and inefficient use of parking resources have not only aggravated traffic congestion, but also reduced the quality of life and satisfaction of citizens.

[0003] In the existing technology, parking demand prediction mostly relies on historical data analysis, ignoring the impact of real-time changing factors such as weather changes and social activities on parking demand, resulting in a large deviation between the prediction results and actual demand. In addition, the existing parking management system lacks a flexible dynamic allocation mechanism and cannot respond to changes in parking demand in real time, resulting in uneven distribution of parking resources in time and space, further exacerbating the problem of parking difficulties. Specifically, the existing technology has obvious deficiencies in the following aspects:

[0004] 1. Static demand forecasting: Relying on historical data to predict parking demand, ignoring the role of real-time and multi-dimensional data in improving forecast accuracy.

[0005] 2. Lack of dynamic allocation mechanism: Parking resources are allocated in a fixed manner and lack the ability to make dynamic adjustments based on real-time demand, making it impossible to effectively utilize parking resources.

[0006] 3. Inefficient data processing: Faced with large and complex data sources, traditional parking management systems are unable to effectively integrate and analyze data, resulting in insufficient basis for decision-making.

[0007] 4. Delayed response time: When actual demand changes, the system cannot respond quickly, resulting in a mismatch between parking resource allocation and demand.

[0008] 5. Lack of user interaction: Traditional systems lack an effective user interaction mechanism and cannot provide personalized parking information and services, which reduces the user experience.

[0009] Therefore, how to provide a dynamic parking demand prediction and resource allocation method based on artificial intelligence is an urgent problem that technicians in this field need to solve. Summary of the invention

[0010] One purpose of the present invention is to propose a dynamic parking demand prediction and resource allocation method based on artificial intelligence. The present invention makes full use of the data processing platform, dynamic response system, intelligent optimization algorithm and adaptive learning mechanism, and describes in detail how to predict and meet the dynamically changing parking demand by integrating multi-dimensional data analysis, real-time allocation capability and comprehensive optimization algorithm, while optimizing the use of parking resources. It has the advantages of improving the accuracy of parking demand prediction, optimizing the efficiency of parking resource allocation, enhancing user satisfaction and adapting to environmental changes.

[0011] According to an embodiment of the present invention, a dynamic parking demand prediction and resource allocation method based on artificial intelligence includes the following steps:

[0012] S1. Build a data processing platform to receive and process multi-dimensional data including time, location, social media data, public event information, and weather conditions;

[0013] S2. Using deep learning and machine learning algorithms on the data processing platform to analyze the received multi-dimensional data and identify patterns and trends that affect parking demand;

[0014] S3. Based on the identified patterns and trends, deploy a dynamic response system to monitor changes in parking demand in real time and dynamically adjust parking resources according to these changes, including opening or closing certain parking areas, adjusting parking rates and recommending parking spaces to users;

[0015] S4. Integrate operations research algorithms, genetic algorithms and machine learning models into a decision support system that meets multiple objectives including minimizing user waiting time, maximizing parking lot utilization, and optimizing parking lot revenue to achieve optimal parking resource allocation;

[0016] S5. Adopt an adaptive learning mechanism to enable the dynamic response system to learn and adjust according to the data obtained from actual operation, continuously improve the prediction accuracy and efficiency of resource allocation strategies, and adapt to the evolution of urban development and user behavior.

[0017] Optionally, the S1 specifically includes:

[0018] S11, receiving multi-dimensional data input, including time data t, location data L, social media data S m , public activity information A, weather condition data W;

[0019] S12, preprocessing the received multidimensional data, including data cleaning C, data normalization NN, and missing value processing M;

[0020] S13, using data integration technology, through the weighted fusion method, the data sets t, L, S from different sources are combined m, A, W are integrated into a unified data framework D f For subsequent analysis:

[0021] F=w1T+w2L+w3S m +w4A+w5W;

[0022] Among them, W1, W2, W3, W4, and w5 are the weights of each data dimension respectively;

[0023] S14. Use time series analysis method to process time data t, and apply autoregressive moving average model (ARMA) to identify parking demand pattern P in time dimension t :

[0024] P t (t)=φ1P t (t-1)+φ2P t (t-2)+...+φ p P t (tp)+θ1ε(t-1)+θ2ε(t-2)+...+θ q ε(tq)+ε(t);

[0025] Among them, P t (t) represents the parking demand at time t, φ1, φ2, ..., φ p are the autoregressive coefficients, θ1, θ2, ..., θ q is the moving average coefficient, ε(t) is the white noise sequence;

[0026] S15. Apply geographic information system (GIS) technology to perform spatial analysis on location data L and use Kriging interpolation method to determine the geographical distribution characteristics P of parking demand s :

[0027]

[0028] Among them, P s (x) represents the predicted parking demand at location x, λ i is the spatial position x i The weight at Z(x i ) is the known position x i The parking demand is n, where n is the number of known locations used for prediction;

[0029] S16. Analyze social media data m and public activity information A, and use natural language processing (NLP) technology and sentiment analysis methods to determine how these factors affect parking demand P e :

[0030]

[0031] in, For the Social Media Data Impact Index, SA i is the sentiment analysis score of a single post, ω i is the weight of the corresponding post, I A EI is the public activity impact index. j Assess the impact score of a single event, γ j is the weight of the corresponding event;

[0032] S17. According to the weather condition data W, the linear regression model is used to analyze the weather factor F. w , analyze the weather condition data W to quantify the specific impact of weather changes on parking demand:

[0033] P w =β0+β1T w +β2R w +β3S w ;

[0034] Among them, P w represents the impact of weather conditions on parking demand, T w is temperature, R w is the rainfall, S w is the amount of snow, β0, β1, β2, β3 are model coefficients.

[0035] Optionally, the S2 specifically includes:

[0036] S21. Analyze the integrated multidimensional data framework through deep learning algorithm D f , from which the key features F that affect parking demand are extracted k ;

[0037] S22, using the long short-term memory network (LSTM) model, based on time series data and key feature set F k Analyze and predict the parking demand at a specific time and location d :

[0038] P d (t+1)=LSTM([F k (t), F k (t-1),…,F k (tn)], Θ);

[0039] Among them, P d (t+1) represents the predicted parking demand at the future time point t+1, [F k (t), F k (t-1), ..., F k(tn)] represents the key feature set from the current time point t to the previous n time points, Θ represents the parameter set of the LSTM model, including weights and bias terms;

[0040] S23. Perform model evaluation and use accuracy A, recall R and F1 score F1 to verify the performance of the model:

[0041]

[0042] Among them, tP (true positive examples) represents the number of parking demand events correctly predicted by the model, TN (true negative examples) represents the number of non-parking demand events correctly predicted by the model, FP (false positive examples) and FN (false negative examples) represent the number of parking demand events incorrectly predicted and unpredicted by the model, respectively;

[0043] S24. Based on the model evaluation results, the gradient descent method is used to optimize the parameters θ of the LSTM model to reduce the prediction error:

[0044]

[0045] Among them, Θ old and θ new Respectively represent the model parameters before and after optimization, α is the learning rate, is the loss function, the gradient of the parameter Θ;

[0046] S25. Update the prediction model regularly and adjust the key feature set F according to the newly collected data k and model parameters Θ to adapt to the dynamic changes of parking demand and improve the efficiency and real-time response capability of parking resource allocation.

[0047] Optionally, the S3 specifically includes:

[0048] S31. Establish a real-time monitoring mechanism and use the sensor network to collect real-time usage data of each parking area. u , and combined with real-time traffic flow data D t , calculate the current comprehensive parking demand index I by weighted average method p :

[0049]

[0050] Among them, U i represents the utilization rate of the i-th parking area, n is the total number of parking areas, T j represents the traffic flow at the jth detection point, m is the total number of traffic flow detection points, and α and β are weight coefficients;

[0051] S32. Based on the collected usage data D u and forecast demand Pd , a dynamic adjustment strategy algorithm is used to determine the real-time allocation of parking resources, considering the current parking lot capacity C current and forecast demand P d , and the adjustment of parking fee rate R to formulate a specific parking space recommendation strategy, giving priority to recommending vacant parking spaces in areas with higher predicted demand to users:

[0052] Calculate the difference between predicted demand and currently available parking resources:

[0053] D gap =P d -C current ;

[0054] According to the difference D gap Dynamically adjust the sensitivity of parking resource control resource adjustment:

[0055] C adjust =C current +α·D gap ;

[0056] Among them, C adjust represents the adjusted parking resource capacity, α is the adjustment coefficient;

[0057] Adjust the parking rate R to cope with demand fluctuations:

[0058] R new =R current ·(1+β·sign(D gap ));

[0059] Among them, R new and R current denote the adjusted and current parking rates respectively, β is the rate adjustment coefficient, sign(D gap ) is D gap The sign function is used to indicate that D gap >0 or decrease D gap <0 parking rate;

[0060] S33. Apply the linear programming (LP) model to optimize the allocation of parking resources to maximize parking efficiency and user satisfaction while minimizing costs and waiting time:

[0061] The objective function is defined as minimizing the total cost C total , including operating costs C op 、User waiting cost C wait and resource allocation cost C alloc :

[0062] MinimizeC total =C op+λ1C wait +λ2C alloc ;

[0063] Among them, λ1 and λ2 are the weight coefficients of user waiting cost and resource allocation cost, respectively, balancing the impact of different types of costs;

[0064] The constraints include:

[0065] Resource capacity constraints ensure that the parking resources allocated to each parking area do not exceed its maximum capacity:

[0066] C i ≤C max，i foralli;

[0067] Demand satisfies constraints so that the predicted parking demand is met in each area:

[0068]

[0069] Constraints on overall resource utilization, optimizing resource allocation to improve overall parking lot utilization:

[0070]

[0071] Among them, C i represents the number of resources allocated to the i-th parking area, C max，i is the maximum capacity of the i-th parking area, P d is the predicted total parking demand, U target is the target resource utilization;

[0072] S34. Implement the optimal parking resource allocation plan S opt , determine the opening status and parking fee rate of each parking area. For each parking area i, its opening status O i and the adjusted parking rate R i，new Determined by the following decision variables:

[0073]

[0074] R i，new =R i，base ·(1+δ·(C i -U target ));

[0075] Among them, O i Indicates whether parking area i is open (1 for open, 0 for closed), C i is the amount of resources allocated to the region, R i，base is the adjusted parking rate, R i，newis the base rate, δ is the rate adjustment factor, which is calculated based on the resource utilization of the area relative to the target utilization U target Deviation adjustment rate;

[0076] Update parking entrance signage and parking fee information to enable drivers to get real-time updates on parking resource allocation, using dynamic information boards (DIBs) to display the opening status and parking fee rates of each area, as well as recommended parking areas:

[0077] DIB info = {O i , R i，new , recommended area};

[0078] Provide online parking space reservation service, allowing users to select and reserve parking spaces based on real-time updated parking resource configuration. opt Adjust the number of parking spaces available for reservation and update the reservation status in real time:

[0079] P reserve (i, t) = min(C i -C booked (i,t), C max,i );

[0080] Among them, P reserve (i, t) represents the number of parking spaces that can be reserved in area i at time t, C booked (i, t) is the number of reserved parking spaces;

[0081] S35. Regularly evaluate the effect of parking resource allocation, collect and analyze parking resource utilization data util and user satisfaction data D sat :

[0082]

[0083] Among them, C used (i) represents the number of parking spaces actually used in the i-th parking area, C max，i is the maximum capacity of the i-th parking area, S user (n) represents the satisfaction score of the nth user, and N is the total number of rating users;

[0084] Using regression analysis to analyze parking resource allocation strategy S opt and parking resource utilization U rate and user satisfaction S avg The relationship between the two can help identify the effectiveness of the deployment strategy and the room for improvement:

[0085] U rate =α+β1S opt +ε;

[0086] S avg =α′+β2S opt +ε′;

[0087] Among them, α and α′ are intercept terms, β1 and β2 are coefficients corresponding to the deployment strategy, reflecting their impact on utilization and satisfaction, and ε and ε′ are error terms.

[0088] Optionally, the S4 specifically includes:

[0089] S41. Integrate operations research algorithms, genetic algorithms and machine learning models into decision support systems to analyze multi-objective optimization problems of parking resource allocation, including minimizing user waiting time T. wait , maximize parking lot utilization U use , and optimize parking revenue I revenue ;

[0090] S42. Define the objective function of the optimization problem as:

[0091] F opt =ω1·t wait +ω2·(1-U use )+ω3·I revenue ;

[0092] Among them, ω1, ω2, and ω3 are the weight coefficients of user waiting time, parking lot utilization rate, and operating income, respectively, which are used to balance the importance of each goal;

[0093] S43, using a genetic algorithm to perform a global search on parking resource configuration, generating an initial population, and finding an optimal parking resource configuration solution through selection, crossover and mutation operations;

[0094] S44. Use reinforcement learning models to predict user behavior and parking demand under different parking resource allocation schemes.

[0095] Optionally, the S5 specifically includes:

[0096] S51. Establish an adaptive learning mechanism, including continuously collecting actual parking lot usage data a and user feedback u As learning input;

[0097] S52. Collected data D a and user feedback u Analyze and extract the key factors K that affect parking demand and user satisfaction f ;

[0098] S53, using support vector machine SVM, according to the key factor Kf Adjust the prediction model M p parameters to improve the accuracy of parking demand prediction;

[0099]

[0100] in, Represents the adjusted prediction model, and Adjust represents the function of adjusting the model parameters;

[0101] S54, based on actual usage data D a and user feedback u Adjust parking resource allocation strategy S s , in order to better meet user needs and improve the operational efficiency of parking lots;

[0102]

[0103] in, represents the optimized parking resource allocation strategy, and Optimize represents the function of optimizing the strategy based on actual data and user feedback;

[0104] S55. Regularly evaluate the effectiveness of the learning mechanism by comparing the prediction accuracy of the model before and after adjustment. p and user satisfaction index S u ,ensuring that the adaptive learning mechanism can continuously improve the overall performance of the system.

[0105] The beneficial effects of the present invention are:

[0106] (1) The present invention provides a deep understanding and dynamic adaptability of parking demand by building a highly integrated data processing platform and using deep learning and machine learning algorithms for multi-dimensional data analysis. This method can effectively respond to the ever-changing parking demand in cities, especially considering the impact of factors such as weather changes and social activities on parking demand. At the same time, by using advanced data analysis technology, the key factors affecting parking demand can be accurately extracted from complex data, thereby improving the accuracy of parking demand prediction.

[0107] (2) The present invention realizes the real-time allocation and optimized management of parking resources by developing a dynamic response system and intelligent optimization algorithm. This not only greatly improves the utilization efficiency of parking resources, but also dynamically adjusts the allocation of parking resources according to real-time demand, reducing the time and effort of users to find parking spaces. In particular, by real-time monitoring of parking lot usage and traffic flow data, the system can respond to changes in parking demand in real time and recommend the best parking location for users, thereby improving the flexibility of parking management and user satisfaction.

[0108] (3) The adaptive learning mechanism introduced in the present invention enables the system to continuously adjust and optimize the prediction model and resource allocation strategy based on historical data and real-time feedback. This mechanism ensures that the system can be continuously optimized over time and effectively respond to new challenges brought about by urban development and changes in user behavior. In particular, by continuously collecting and analyzing actual usage data and user feedback, the system can accurately identify the changing trends of parking demand and user satisfaction, achieve continuous optimization of parking resource allocation, and improve the intelligence and effectiveness of overall parking management. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0110] Figure 1 This is an overall framework diagram of a dynamic parking demand prediction and resource allocation method based on artificial intelligence proposed by the present invention;

[0111] Figure 2 This is a schematic diagram of the structure of a highly integrated data processing platform for a dynamic parking demand prediction and resource allocation method based on artificial intelligence proposed in the present invention. DETAILED DESCRIPTION

[0112] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0113] refer to Figure 1-Figure 2 , a dynamic parking demand prediction and resource allocation method based on artificial intelligence, comprising the following steps:

[0114] S1. Build a data processing platform to receive and process multi-dimensional data including time, location, social media data, public event information, and weather conditions;

[0115] In this implementation, S1 specifically includes:

[0116] S11, receiving multi-dimensional data input, including time data t, location data L, social media data S m , public activity information A, weather condition data W;

[0117] S12, preprocessing the received multidimensional data, including data cleaning C, data normalization NN, and missing value processing M;

[0118] S13, using data integration technology, through the weighted fusion method, the data sets t, L, S from different sources are combined m,A,W are integrated into a unified data framework D f For subsequent analysis:

[0119] F=w1T+w2L+w3S m +w4A+w5W;

[0120] Among them, w1, W2, W3, W4, and w5 are the weights of each data dimension respectively;

[0121] S14. Use time series analysis method to process time data t, and apply autoregressive moving average model (ARMA) to identify parking demand pattern P in time dimension t :

[0122] P t (t)=φ1P t (t-1)+φ2P t (t-2)+...+φ p P t (tp)+θ1ε(t-1)+θ2ε(t-2)+...+θ q ε(tq)+ε(t);

[0123] Among them, P t (t) represents the parking demand at time t, φ1, φ2, ..., φ p are the autoregressive coefficients, θ1, θ2, ..., θ q is the moving average coefficient, ε(t) is the white noise sequence;

[0124] S15. Apply geographic information system (GIS) technology to perform spatial analysis on location data L and use Kriging interpolation method to determine the geographical distribution characteristics P of parking demand s :

[0125]

[0126] Among them, P s (x) represents the predicted parking demand at location x, λ i is the spatial position x i The weight at Z(x i ) is the known position x i The parking demand is n, where n is the number of known locations used for prediction;

[0127] S16. Analyze social media data m and public activity information A, and use natural language processing (NLP) technology and sentiment analysis methods to determine how these factors affect parking demand P e :

[0128]

[0129] in, For the Social Media Data Impact Index, SA i is the sentiment analysis score of a single post, ω i is the weight of the corresponding post, I A EI is the public activity impact index. j Assess the impact score of a single event, γ j is the weight of the corresponding event;

[0130] S17. According to the weather condition data W, the linear regression model is used to analyze the weather factor F. w , analyze the weather condition data W to quantify the specific impact of weather changes on parking demand:

[0131] P w =β0+β1T w +β2R w +β3S w ;

[0132] Among them, P w represents the impact of weather conditions on parking demand, T w is temperature, R w is the rainfall, S w is the amount of snow, β0, β1, β2, β3 are model coefficients.

[0133] In this implementation, S2 specifically includes:

[0134] S21. Analyze the integrated multidimensional data framework through deep learning algorithm D f , from which the key features F that affect parking demand are extracted k ;

[0135] S22, using the long short-term memory network (LSTM) model, based on time series data and key feature set F k Analyze and predict the parking demand at a specific time and location d :

[0136] P d (t+1)=LSTM([F k (t), F k (t-1),…,F k (tn)], Θ);

[0137] Among them, P d (t+1) represents the predicted parking demand at the future time point t+1, [F k (t), F k (t-1), ..., F k(tn)] represents the key feature set from the current time point t to the previous n time points, Θ represents the parameter set of the LSTM model, including weights and bias terms;

[0138] S23. Perform model evaluation and use accuracy A, recall R and F1 score F1 to verify the performance of the model:

[0139]

[0140] Among them, tP (true positive examples) represents the number of parking demand events correctly predicted by the model, TN (true negative examples) represents the number of non-parking demand events correctly predicted by the model, FP (false positive examples) and FN (false negative examples) represent the number of parking demand events incorrectly predicted and unpredicted by the model, respectively;

[0141] S24. Based on the model evaluation results, the gradient descent method is used to optimize the parameters θ of the LSTM model to reduce the prediction error:

[0142]

[0143] Among them, Θ old and θ new Respectively represent the model parameters before and after optimization, α is the learning rate, is the gradient of the loss function J with respect to the parameter Θ;

[0144] S25. Update the prediction model regularly and adjust the key feature set F according to the newly collected data k and model parameters Θ to adapt to the dynamic changes of parking demand and improve the efficiency and real-time response capability of parking resource allocation.

[0145] S3. Based on the identified patterns and trends, deploy a dynamic response system to monitor changes in parking demand in real time and dynamically adjust parking resources according to these changes, including opening or closing certain parking areas, adjusting parking rates and recommending parking spaces to users;

[0146] In this implementation, S3 specifically includes:

[0147] S31. Establish a real-time monitoring mechanism and use the sensor network to collect real-time usage data of each parking area. u , and combined with real-time traffic flow data D t , calculate the current comprehensive parking demand index I by weighted average method p :

[0148]

[0149] Among them, U i represents the utilization rate of the i-th parking area, n is the total number of parking areas, T jrepresents the traffic flow at the jth detection point, m is the total number of traffic flow detection points, and α and β are weight coefficients;

[0150] S32. Based on the collected usage data D u and forecast demand P d , a dynamic adjustment strategy algorithm is used to determine the real-time allocation of parking resources, considering the current parking lot capacity C current and forecast demand P d , and the adjustment of parking fee rate R to formulate a specific parking space recommendation strategy, giving priority to recommending vacant parking spaces in areas with higher predicted demand to users:

[0151] Calculate the difference between predicted demand and currently available parking resources:

[0152] D gap =P d -C current ;

[0153] According to the difference D gap Dynamically adjust the sensitivity of parking resource control resource adjustment:

[0154] C adjust =C current +α·D gap ;

[0155] Among them, C adjust represents the adjusted parking resource capacity, α is the adjustment coefficient;

[0156] Adjust the parking rate R to cope with demand fluctuations:

[0157] R new =R current ·(1+β·sign(D gap ));

[0158] Among them, R new and R current denote the adjusted and current parking rates respectively, β is the rate adjustment coefficient, sign(D gap ) is D gap The sign function is used to indicate that D gap >0 or decrease D gap <0 parking rate;

[0159] S33. Apply the linear programming (LP) model to optimize the allocation of parking resources to maximize parking efficiency and user satisfaction while minimizing costs and waiting time:

[0160] The objective function is defined as minimizing the total cost C total , including operating costs C op、User waiting cost C wait and resource allocation cost C alloc :

[0161] MinimizeC total =C op +λ1C wait +λ2C alloc ;

[0162] Among them, λ1 and λ2 are the weight coefficients of user waiting cost and resource allocation cost, respectively, balancing the impact of different types of costs;

[0163] The constraints include:

[0164] Resource capacity constraints ensure that the parking resources allocated to each parking area do not exceed its maximum capacity:

[0165] C i ≤C max，i foralli;

[0166] Demand satisfies constraints so that the predicted parking demand is met in each area:

[0167]

[0168] Constraints on overall resource utilization, optimizing resource allocation to improve overall parking lot utilization:

[0169]

[0170] Among them, C i represents the number of resources allocated to the i-th parking area, C max，i is the maximum capacity of the i-th parking area, P d is the predicted total parking demand, U target is the target resource utilization;

[0171] S34. Implement the optimal parking resource allocation plan S opt , determine the opening status and parking fee rate of each parking area. For each parking area i, its opening status O i and the adjusted parking rate R i,new Determined by the following decision variables:

[0172]

[0173] R i，new =R i，base ·(1+δ·(C i -U target ));

[0174] Among them, O iIndicates whether parking area i is open (1 for open, 0 for closed), C i is the amount of resources allocated to the region, R i，base is the adjusted parking rate, R i,new is the base rate, δ is the rate adjustment factor, which is calculated based on the resource utilization of the area relative to the target utilization U target Deviation adjustment rate;

[0175] Update parking entrance signage and parking fee information to enable drivers to get real-time updates on parking resource allocation, using dynamic information boards (DIBs) to display the opening status and parking fee rates of each area, as well as recommended parking areas:

[0176] DIB info = {O i , R i，new , recommended area};

[0177] Provide online parking space reservation service, allowing users to select and reserve parking spaces based on real-time updated parking resource configuration. opt Adjust the number of parking spaces available for reservation and update the reservation status in real time:

[0178] P reserve (i, t) = min(C i -C booked (i,t), C max，i );

[0179] Among them, P reserve (i, t) represents the number of parking spaces that can be reserved in area i at time t, C booked (i, t) is the number of reserved parking spaces;

[0180] S35. Regularly evaluate the effect of parking resource allocation, collect and analyze parking resource utilization data util and user satisfaction data D sat :

[0181]

[0182] Among them, C used (i) represents the number of parking spaces actually used in the i-th parking area, C max，i is the maximum capacity of the i-th parking area, S user (n) represents the satisfaction score of the nth user, and N is the total number of rating users;

[0183] Using regression analysis to analyze parking resource allocation strategy S opt and parking resource utilization U rate and user satisfaction S avgThe relationship between the two can help identify the effectiveness of the deployment strategy and the room for improvement:

[0184] U rate =α+β1S opt +ε;

[0185] S avg =α′+β2S opt +ε′;

[0186] Among them, α and α′ are intercept terms, β1 and β2 are coefficients corresponding to the deployment strategy, reflecting their impact on utilization and satisfaction, and ε and ε′ are error terms.

[0187] S4. Integrate operations research algorithms, genetic algorithms and machine learning models into a decision support system that meets multiple objectives including minimizing user waiting time, maximizing parking lot utilization, and optimizing parking lot revenue to achieve optimal parking resource allocation;

[0188] In this implementation, S4 specifically includes:

[0189] S41. Integrate operations research algorithms, genetic algorithms and machine learning models into decision support systems to analyze multi-objective optimization problems of parking resource allocation, including minimizing user waiting time T. wait , maximize parking lot utilization U use , and optimize parking revenue I revenue ;

[0190] S42. Define the objective function of the optimization problem as:

[0191] F opt =ω1·T wait +ω2·(1-U use )+ω3·I revenue ;

[0192] Among them, ω1, ω2, and ω3 are the weight coefficients of user waiting time, parking lot utilization rate, and operating income, respectively, which are used to balance the importance of each goal;

[0193] S43, using a genetic algorithm to perform a global search on parking resource configuration, generating an initial population, and finding an optimal parking resource configuration solution through selection, crossover and mutation operations;

[0194] S44. Use reinforcement learning models to predict user behavior and parking demand under different parking resource allocation schemes.

[0195] S5. Adopt an adaptive learning mechanism to enable the dynamic response system to learn and adjust according to the data obtained from actual operation, continuously improve the prediction accuracy and efficiency of resource allocation strategies, and adapt to the evolution of urban development and user behavior.

[0196] In this implementation, S5 specifically includes:

[0197] S51. Establish an adaptive learning mechanism, including continuously collecting actual parking lot usage data a and user feedback u As learning input;

[0198] S52. Collected data D a and user feedback u Analyze and extract the key factors K that affect parking demand and user satisfaction f ;

[0199] S53, using support vector machine SVM, according to the key factor K f Adjust the prediction model M p parameters to improve the accuracy of parking demand prediction;

[0200]

[0201] in, Represents the adjusted prediction model, and Adjust represents the function of adjusting the model parameters;

[0202] S54, based on actual usage data D a and user feedback u Adjust parking resource allocation strategy S s , in order to better meet user needs and improve the operational efficiency of parking lots;

[0203]

[0204] in, represents the optimized parking resource allocation strategy, and Optimize represents the function of optimizing the strategy based on actual data and user feedback;

[0205] S55. Regularly evaluate the effectiveness of the learning mechanism by comparing the prediction accuracy of the model before and after adjustment. p and user satisfaction index S u ,ensuring that the adaptive learning mechanism can continuously improve the overall performance of the system.

[0206] Embodiment 1:

[0207] In a parking lot scenario in a metropolitan commercial district, with the frequent commercial activities and the substantial increase in traffic, parking demand has increasingly become a major problem faced by managers and car owners. Especially during peak hours on weekdays and weekends, parking lots are often overloaded, causing car owners to spend a long time looking for parking spaces, which seriously affects car owners' satisfaction and the operational efficiency of the commercial district. In order to effectively solve this problem, the management of a commercial district decided to deploy the dynamic parking demand prediction and resource allocation method based on artificial intelligence proposed in this invention.

[0208] The implementation of this method is based on a sensor network installed in the parking lot and its surrounding roads, which collects various data in real time, including parking space occupancy, vehicle entry and exit frequency, and traffic flow on nearby roads. In addition, the system also integrates multi-dimensional data such as commercial district event schedules, social media activity information, and weather forecasts to build a comprehensive data processing and analysis platform. Through deep learning and machine learning algorithms, the system analyzes this data to predict parking demand in different time periods.

[0209] In the embodiment, the system predicts that on the eve of a holiday, due to a large-scale promotion and good weather conditions, the parking demand is expected to increase by 40% compared to usual. Based on this prediction result, the dynamic response system is immediately activated and takes measures including adding temporary parking areas, adjusting the flow guidance of parking lot entrances and exits, and recommending the best parking time and area to customers through the mobile application of the commercial area. In addition, the system also appropriately adjusts the parking fee rate according to the predicted demand peak to guide the parking demand to be dispersed to non-peak hours.

[0210] Table 1: Comparison of parking lot operation data in commercial areas

[0211] Data indicators Before implementing the method After implementing the method Average parking lot utilization rate 98％ 92％ Average time it takes a car owner to find a parking space More than 30 minutes Within 15 minutes Number of parking-related customer complaints Average 50 cases per event No more than 10 events per event Parking fee rate adjustment No adjustment Adjust according to demand peaks

[0212] Specific data show that before the implementation of the method of the present invention, the average utilization rate of commercial parking lots on the eve of holidays was as high as 98%, the average time for car owners to find parking spaces was more than 30 minutes, and the number of parking-related customer complaints per activity reached an average of 50. After the implementation of the method of the present invention, even in the case of a surge in passenger flow, the average utilization rate of parking lots dropped to below 92%, the average time for car owners to find parking spaces was shortened to less than 15 minutes, and the number of parking-related customer complaints was reduced to no more than 10 per activity.

[0213] Through the method of the present invention, the commercial district successfully coped with the surge in parking demand, improved the utilization efficiency of parking resources, shortened the time for car owners to find parking spaces, reduced management costs, and improved car owner satisfaction and the overall operational efficiency of the commercial district. This embodiment demonstrates the effectiveness and beneficial effects of the present invention in practical applications, especially in complex scenarios that require dynamic response to changes in parking demand.

[0214] In summary, the present invention provides an innovative solution that can dynamically predict parking demand and allocate parking resources in real time by combining advanced data processing technology and artificial intelligence algorithms. This not only solves the problems of unreasonable parking resource allocation and low prediction accuracy in the prior art, but also greatly improves the efficiency of parking management and the satisfaction of car owners, and provides an effective solution to urban parking problems. The above is only a preferred specific implementation of the present invention, but the protection scope of the present invention is not limited to this. Any technician familiar with the technical field within the technical scope disclosed by the present invention, according to the technical solution and its inventive concept of the present invention, shall be covered within the protection scope of the present invention.

[0215] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A dynamic parking demand prediction and resource allocation method based on artificial intelligence, characterized in that: The steps include: S1. Build a data processing platform to receive and process multi-dimensional data including time, location, social media data, public event information, and weather conditions; S2. Using deep learning and machine learning algorithms on the data processing platform to analyze the received multi-dimensional data and identify patterns and trends that affect parking demand; S3. Based on the identified patterns and trends, deploy a dynamic response system to monitor changes in parking demand in real time and dynamically adjust parking resources according to these changes, including opening or closing certain parking areas, adjusting parking rates and recommending parking spaces to users; S4. Integrate operations research algorithms, genetic algorithms and machine learning models into a decision support system that meets multiple objectives including minimizing user waiting time, maximizing parking lot utilization, and optimizing parking lot revenue to achieve optimal parking resource allocation; S5. Adopt an adaptive learning mechanism to enable the dynamic response system to learn and adjust according to the data obtained from actual operation, continuously improve the prediction accuracy and efficiency of resource allocation strategies, and adapt to the evolution of urban development and user behavior.

2. The method for dynamic parking demand prediction and resource allocation based on artificial intelligence according to claim 1 is characterized in that: The S1 specifically includes: S11, receiving multi-dimensional data input, including time data t, location data L, social media data S m , public activity information A, weather condition data W; S12, preprocessing the received multidimensional data, including data cleaning C, data normalization NN, and missing value processing M; S13, using data integration technology, through the weighted fusion method, the data sets T, L, S from different sources are combined m , A, W are integrated into a unified data framework D f For subsequent analysis: F=w1T+w2L+w3S m +w4A+w5W; Among them, w1, W2, W3, W4, and w5 are the weights of each data dimension respectively; S14. Use time series analysis method to process time data t, and apply autoregressive moving average model (ARMA) to identify parking demand pattern P in time dimension t : P t (t)=φ1P t (t-1)+φ2P t (t-2)+…+φ p P t (tp)+θ1ε(t-1)+θ2ε(t-2)+…+θ q ε(tq)+ε(t); Among them, P t (t) represents the parking demand at time t, φ1, φ2, ..., φ p are the autoregressive coefficients, θ1, θ2, ..., θ q is the moving average coefficient, ε(t) is the white noise sequence; S15. Apply geographic information system (GIS) technology to perform spatial analysis on location data L and use Kriging interpolation method to determine the geographical distribution characteristics of parking demand P s : Among them, P s (x) represents the predicted parking demand at location x, λ i is the spatial position x i The weight at Z(x i ) is the known position x i The parking demand is n, where n is the number of known locations used for prediction; S16. Analyze social media data m and public activity information A, and use natural language processing (NLP) technology and sentiment analysis methods to determine how these factors affect parking demand P e : in, For the Social Media Data Impact Index, SA i is the sentiment analysis score of a single post, ω i is the weight of the corresponding post, I A EI is the public activity impact index. j Assess the impact score of a single event, γ j is the weight of the corresponding event; S17. According to the weather condition data W, the linear regression model is used to analyze the weather factor F. w , analyze the weather condition data W to quantify the specific impact of weather changes on parking demand: P w =β0+β1T w +β2R w +β3S w ; Among them, P w represents the impact of weather conditions on parking demand, T w is temperature, R w is the rainfall, S w is the amount of snow, β0, β1, β2, β3 are model coefficients.

3. The method for dynamic parking demand prediction and resource allocation based on artificial intelligence according to claim 1 is characterized in that: The S2 specifically includes: S21. Analyze the integrated multidimensional data framework through deep learning algorithm D f , from which the key features F that affect parking demand are extracted k ; S22, using the long short-term memory network (LSTM) model, based on time series data and key feature set F k Analyze and predict the parking demand at a specific time and location d : P d (t+1)=LSTM([F k (t),F k (t-1),…,F k (t-n)],Θ); Among them, P d (t+1) represents the predicted parking demand at the future time point t+1, [F k (t), F k (t-1), ..., F k (tn)] represents the key feature set from the current time point t to the previous n time points, Θ represents the parameter set of the LSTM model, including weights and bias terms; S23. Perform model evaluation and use accuracy A, recall R and F1 score F1 to verify the performance of the model: Among them, tP (true positive examples) represents the number of parking demand events correctly predicted by the model, TN (true negative examples) represents the number of non-parking demand events correctly predicted by the model, FP (false positive examples) and FN (false negative examples) represent the number of parking demand events incorrectly predicted and unpredicted by the model, respectively; S24. Based on the model evaluation results, the gradient descent method is used to optimize the parameters θ of the LSTM model to reduce the prediction error: Among them, Θ old and θ new Respectively represent the model parameters before and after optimization, α is the learning rate, is the gradient of the loss function J with respect to the parameter Θ; S25. Update the prediction model regularly and adjust the key feature set F according to the newly collected data k and model parameters Θ to adapt to the dynamic changes of parking demand and improve the efficiency and real-time response capability of parking resource allocation.

4. The method for dynamic parking demand prediction and resource allocation based on artificial intelligence according to claim 1 is characterized in that: The S3 specifically includes: S31. Establish a real-time monitoring mechanism and use the sensor network to collect real-time usage data of each parking area. u , and combined with real-time traffic flow data D t , calculate the current comprehensive parking demand index I by weighted average method p : Among them, U i represents the utilization rate of the i-th parking area, n is the total number of parking areas, T j represents the traffic flow at the jth detection point, m is the total number of traffic flow detection points, and α and β are weight coefficients; S32. Based on the collected usage data D u and forecast demand P d , a dynamic adjustment strategy algorithm is used to determine the real-time allocation of parking resources, considering the current parking lot capacity C current and forecast demand P d , and the adjustment of parking fee rate R to formulate a specific parking space recommendation strategy, giving priority to recommending vacant parking spaces in areas with higher predicted demand to users: Calculate the difference between predicted demand and currently available parking resources: D gap =P d -C current ; According to the difference D gap Dynamically adjust the sensitivity of parking resource control resource adjustment: C adjust =C current +α·D gap ; Among them, C adjust represents the adjusted parking resource capacity, α is the adjustment coefficient; Adjust the parking rate R to cope with demand fluctuations: R new =R current ·(1+β·sign(D gap )); Among them, R new and R current denote the adjusted and current parking rates respectively, β is the rate adjustment coefficient, sign(D gap ) is D gap The sign function is used to indicate that D gap >0 or decrease D gap <0 parking rate; S33. Apply the linear programming (LP) model to optimize the allocation of parking resources to maximize parking efficiency and user satisfaction while minimizing costs and waiting time: The objective function is defined as minimizing the total cost C total , including operating costs C op 、User waiting cost C wait and resource allocation cost C alloc : MinimizeC total =C op +λ1C wait +λ2C alloc ; Among them, λ1 and λ2 are the weight coefficients of user waiting cost and resource allocation cost, respectively, balancing the impact of different types of costs; The constraints include: Resource capacity constraints ensure that the parking resources allocated to each parking area do not exceed its maximum capacity: C i ≤C max，f foralli; Demand satisfies constraints so that the predicted parking demand is met in each area: Constraints on overall resource utilization, optimizing resource allocation to improve overall parking lot utilization: Among them, C i represents the number of resources allocated to the i-th parking area, C max，i is the maximum capacity of the i-th parking area, P d is the predicted total parking demand, U target is the target resource utilization; S34. Implement the optimal parking resource allocation plan S opt , determine the opening status and parking fee rate of each parking area. For each parking area i, its opening status o i and the adjusted parking rate R i,new Determined by the following decision variables: R i，new =R i,base ·(1+δ·(C i -U target )); Among them, i Indicates whether parking area i is open (1 for open, 0 for closed), C i is the amount of resources allocated to the region, R i，base is the adjusted parking rate, R i,new is the base rate, δ is the rate adjustment factor, which is calculated based on the resource utilization of the area relative to the target utilization U target Deviation adjustment rate; Update parking entrance signage and parking fee information to enable drivers to get real-time updates on parking resource allocation, using dynamic information boards (DIBs) to display the opening status and parking fee rates of each area, as well as recommended parking areas: DIB info = {O i , R i，new , recommended area}; Provide online parking space reservation service, allowing users to select and reserve parking spaces based on real-time updated parking resource configuration. opt Adjust the number of parking spaces available for reservation and update the reservation status in real time: P reserve (i,t)=min(C i -C booked (i,t),C max,i ); Among them, P reserve (i, t) represents the number of parking spaces that can be reserved in area i at time t, C booked (i, t) is the number of reserved parking spaces; S35. Regularly evaluate the effect of parking resource allocation, collect and analyze parking resource utilization data util and user satisfaction data D sat : Among them, C used (i) represents the number of parking spaces actually used in the i-th parking area, C max，i is the maximum capacity of the i-th parking area, S user (n) represents the satisfaction score of the nth user, and N is the total number of rating users; Using regression analysis to analyze parking resource allocation strategy S opt and parking resource utilization U rate and user satisfaction S avg The relationship between the two can help identify the effectiveness of the deployment strategy and the room for improvement: U rate =α+β1S opt +e; S avg =α′+β2S opt +e′; Among them, α and α′ are intercept terms, β1 and β2 are coefficients corresponding to the deployment strategy, reflecting their impact on utilization and satisfaction, and ε and ε are error terms.

5. The method for dynamic parking demand prediction and resource allocation based on artificial intelligence according to claim 1 is characterized in that: The S4 specifically includes: S41. Integrate operations research algorithms, genetic algorithms and machine learning models into decision support systems to analyze multi-objective optimization problems in parking resource allocation, including minimizing user waiting time t wait , maximize parking lot utilization U use , and optimize parking revenue I revenue ; S42. Define the objective function of the optimization problem as: F opt =ω1·t wait +ω2·(1-U use )+ω3·I revenue ; Among them, ω1, ω2, and ω3 are the weight coefficients of user waiting time, parking lot utilization rate, and operating income, respectively, which are used to balance the importance of each goal; S43, using a genetic algorithm to perform a global search on parking resource configuration, generating an initial population, and finding an optimal parking resource configuration solution through selection, crossover and mutation operations; S44. Use reinforcement learning models to predict user behavior and parking demand under different parking resource allocation schemes.

6. The method for dynamic parking demand prediction and resource allocation based on artificial intelligence according to claim 1 is characterized in that: The S5 specifically includes: S51. Establish an adaptive learning mechanism, including continuously collecting actual parking lot usage data a and user feedback u As learning input; S52. Collected data D a and user feedback u Analyze and extract the key factors K that affect parking demand and user satisfaction f ; S53, using support vector machine SVM, according to the key factor K f Adjust the prediction model M p parameters to improve the accuracy of parking demand prediction; in, Represents the adjusted prediction model, and Adjust represents the function of adjusting the model parameters; S54, based on actual usage data D a and user feedback u Adjust parking resource allocation strategy S s , in order to better meet user needs and improve the operational efficiency of parking lots; in, represents the optimized parking resource allocation strategy, and optimize represents the function of optimizing the strategy based on actual data and user feedback; S55. Regularly evaluate the effectiveness of the learning mechanism by comparing the prediction accuracy of the model before and after adjustment. p and user satisfaction index S u ,ensuring that the adaptive learning mechanism can continuously improve the overall performance of the system.

Images