Large model training and application method and system based on urban brain

Abstract

The invention discloses a large model training and application method and system based on a city brain, and relates to the technical field of artificial intelligence and smart city management. Comprising the steps of 1, collecting and preprocessing data, 2, training a large-scale language model, 3, utilizing the trained large-scale language model to understand and predict city dynamics, and utilizing the large-scale language model to carry out semantic understanding and relation reasoning on city data, and 4, displaying a model analysis and prediction result by using a visual interface, and performing interactive question and answer by using the visual interface. And 5, continuously learning and updating the large-scale language model according to the feedback: continuously learning and improving the large-scale language model according to the newly acquired city data and the user feedback, regularly evaluating the performance of the model, and adjusting the structure and parameters of the model according to the evaluation result.

Description

Technical Field

[0001] The present invention discloses a large model training and application method and system based on city brain, which relate to the technical fields of artificial intelligence and smart city management. Background Art

[0002] With the acceleration of urbanization and the rapid development of information technology, modern cities are facing increasingly complex management challenges. Existing urban management methods are unable to cope with the processing of large amounts of real-time data and the dynamic changes of complex urban systems. In recent years, the City Brain has been widely used as an integrated urban data processing platform, providing rich data resources for urban management. At the same time, large-scale language models have made breakthrough progress in natural language processing and knowledge reasoning. However, existing technologies still have the following problems:

[0003] Inefficient data utilization: The city brain collects massive amounts of data, but this data is often not fully utilized.

[0004] Poor model generalizability: Existing AI models are often trained for general purposes and have difficulty accurately understanding the unique characteristics and dynamics of a specific city.

[0005] Lack of real-time performance: Many urban management decisions need to be made based on the latest data, but existing systems often have problems with data processing lags.

[0006] Weak cross-domain integration capabilities: Urban management involves multiple fields, but the existing system finds it difficult to effectively integrate and analyze cross-domain information. Summary of the invention

[0007] In response to the problems of the prior art, the present invention provides a large-model training and application method and system based on the city brain, which helps to improve urban operation efficiency, optimize resource allocation, and enhance the ability to predict and respond to urban problems.

[0008] The specific scheme proposed by the present invention is:

[0009] The present invention provides a large model training and application method based on the city brain, which is characterized by comprising:

[0010] Step 1: Collect and pre-process data: Continuously obtain real-time multi-dimensional urban data from the city brain and pre-process it. The urban data includes traffic flow, energy consumption, environmental indicators, and public service usage.

[0011] Step 2: Training a large-scale language model: Select a basic large-scale language model, use city data for continuous fine-tuning, and use transfer learning and incremental learning methods to make the large-scale language model adapt to the dynamic changes of city data.

[0012] Step 3: Use the trained large-scale language model to understand and predict urban dynamics: Use the large-scale language model to understand the semantics and reason about the relationship of urban data, and combine time series analysis and causal inference methods to predict the future state of the city.

[0013] Step 4: Use the visualization interface to display the model analysis and prediction results, and use the visualization interface for interactive Q&A.

[0014] Step 5: Continuously learn and update the large-scale language model based on feedback: Continuously learn and improve the large-scale language model based on newly acquired city data and user feedback, regularly evaluate model performance, and adjust model structure and parameters based on the evaluation results.

[0015] Furthermore, the data is preprocessed in step 1 of the large-scale model training and application method based on the urban brain, including: cleaning, standardizing and structuring the acquired urban data to ensure data quality and consistency.

[0016] Furthermore, the following formula is used in step 2 of the large model training and application method based on the city brain:

[0017]

[0018] Indicates that the incremental learning method is used to adapt the large-scale language model to the dynamic changes of urban data, where θ t represents the parameters of the model at time t, η is the learning rate, is the loss function, It is a city dataset.

[0019] Furthermore, the following formula is used in step 3 of the large model training and application method based on the city brain:

[0020]

[0021] Combining time series analysis and causal inference methods to predict the future state of the city, is the prediction of the future state, x t is the city state vector at the current time step, and f is the nonlinear function learned by the large-scale language model.

[0022] Furthermore, the following formula is used in step 5 of the large model training and application method based on the city brain:

[0023] θ new =θ old +α(θ online -θ old )

[0024] The large-scale language model is updated according to the feedback, where α is the update rate and θ online are parameters learned based on the latest city data.

[0025] The present invention also provides a large model training and application system based on the city brain, including a data acquisition and preprocessing module, a large-scale language model training module, an urban dynamic understanding and prediction module, a decision support analysis module and an update optimization module.

[0026] Data collection and preprocessing module collects and preprocesses data: continuously obtains real-time multi-dimensional urban data from the city brain and preprocesses it. The urban data includes traffic flow, energy consumption, environmental indicators, and public service usage.

[0027] Large-scale language model training module trains large-scale language models: selects basic large-scale language models, uses city data for continuous fine-tuning, and adopts transfer learning and incremental learning methods to make large-scale language models adapt to the dynamic changes of city data.

[0028] The city dynamics understanding and prediction module uses a trained large-scale language model to understand and predict city dynamics: it uses a large-scale language model to perform semantic understanding and relational reasoning on city data, and combines time series analysis and causal inference methods to predict the future state of the city.

[0029] The decision support analysis module uses a visual interface to display model analysis and prediction results, and uses the visual interface for interactive question and answer.

[0030] The update optimization module enables the large-scale language model to continuously learn and update based on feedback: the large-scale language model continuously learns and improves based on newly acquired city data and user feedback, and regularly evaluates model performance, and adjusts the model structure and parameters based on the evaluation results.

[0031] Furthermore, the data collection and preprocessing module of the large model training and application system based on the urban brain preprocesses data, including: cleaning, standardizing and structuring the acquired urban data to ensure data quality and consistency.

[0032] Furthermore, the large-scale language model training module of the large model training and application system based on the city brain uses the following formula:

[0033]

[0034] Indicates that the incremental learning method is used to adapt the large-scale language model to the dynamic changes of urban data, where θ t represents the parameters of the model at time t, η is the learning rate, is the loss function, It is a city dataset.

[0035] Furthermore, the urban dynamics understanding and prediction module of the large model training and application system based on the city brain uses the following formula:

[0036]

[0037] Combining time series analysis and causal inference methods to predict the future state of the city, is the prediction of the future state, x t is the city state vector at the current time step, and f is the nonlinear function learned by the large-scale language model.

[0038] Furthermore, the update optimization module of the large model training and application system based on the city brain uses the following formula:

[0039] θ new =θ old +α(θ online -θ old )

[0040] The large-scale language model is updated according to the feedback, where α is the update rate and θ online are parameters learned based on the latest city data.

[0041] The benefits of the method of the present invention are:

[0042] Maximizing data value: By using the massive data of the city brain to train specialized large models, the utilization efficiency and value of data are significantly improved.

[0043] City-specific AI capabilities: The developed big models have a deep understanding of specific city characteristics and dynamics, providing more accurate and relevant analysis than general AI models.

[0044] Real-time decision support: Ability to process and analyze real-time data streams to provide timely insights and decision-making recommendations to city managers.

[0045] Cross-domain knowledge integration: The powerful semantic understanding and reasoning capabilities of the large model enable the system to effectively integrate and analyze information from different urban subsystems.

[0046] Enhanced predictive capabilities: A deeper understanding of urban dynamics enables more accurate predictions of future trends and potential issues.

[0047] Improved scientific nature of decision-making: Decision support based on data-driven and AI analysis improves the scientific nature and effectiveness of urban management decisions.

[0048] Optimal resource allocation: Forecasting and analysis help better plan and allocate urban resources and improve resource utilization efficiency.

[0049] Strengthening emergency response capabilities: Through real-time data analysis and prediction, potential risks can be identified early and the city’s emergency response capabilities can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic diagram of the application flow of the system of the present invention.

[0051] Figure 2 The present invention relates to a visualization interface schematic diagram. DETAILED DESCRIPTION

[0052] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0053] Example 1

[0054] The present invention provides a large model training and application method based on the city brain, which is characterized by comprising:

[0055] Step 1: Collect and preprocess data: Continuously obtain real-time multi-dimensional urban data from the city brain and preprocess it. The urban data includes traffic flow, energy consumption, environmental indicators, public service usage, etc.

[0056] The data preprocessing may include: cleaning, standardizing and structuring the acquired urban data to ensure data quality and consistency.

[0057] Step 2: Train a large-scale language model: Select a basic large-scale language model, such as the GPT series, and use urban data for continuous fine-tuning. Use transfer learning and incremental learning methods to make the large-scale language model adapt to the dynamic changes of urban data.

[0058] The following formula can be further used:

[0059]

[0060] Indicates that the incremental learning method is used to adapt the large-scale language model to the dynamic changes of urban data, where θ t represents the parameters of the model at time t, η is the learning rate, is the loss function, It is a city dataset.

[0061] Step 3: Use the trained large-scale language model to understand and predict urban dynamics: Use the large-scale language model to perform semantic understanding and relational reasoning on urban data, and combine time series analysis and causal inference methods to predict the future state of the city.

[0062] The following formula can be further used:

[0063]

[0064] Combining time series analysis and causal inference methods to predict the future state of the city, is the prediction of the future state, x t is the city state vector at the current time step, and f is the nonlinear function learned by the large-scale language model.

[0065] Step 4: Use the visualization interface to display the model analysis and prediction results, and use the visualization interface for interactive Q&A, allowing city managers to ask natural language questions and get AI-assisted answers. Implement multi-scenario analysis functions to help evaluate the potential impact of different decisions. The interface can be referenced Figure 2 .

[0066] Step 5: Continuously learn and update the large-scale language model based on feedback: Continuously learn and improve the large-scale language model based on newly acquired city data and user feedback, regularly evaluate model performance, and adjust model structure and parameters based on the evaluation results.

[0067] The following formula can be further used:

[0068] θ new =θ old +α(θ online -θ old )

[0069] The large-scale language model is updated according to the feedback, where α is the update rate and θ online are parameters learned based on the latest city data.

[0070] Example 2

[0071] The present invention also provides a large model training and application system based on the city brain, including a data acquisition and preprocessing module, a large-scale language model training module, an urban dynamic understanding and prediction module, a decision support analysis module and an update optimization module.

[0072] Data collection and preprocessing module collects and preprocesses data: continuously obtains real-time multi-dimensional urban data from the city brain and preprocesses it. The urban data includes traffic flow, energy consumption, environmental indicators, and public service usage.

[0073] Large-scale language model training module trains large-scale language models: selects basic large-scale language models, uses city data for continuous fine-tuning, and adopts transfer learning and incremental learning methods to make large-scale language models adapt to the dynamic changes of city data.

[0074] The city dynamics understanding and prediction module uses a trained large-scale language model to understand and predict city dynamics: it uses a large-scale language model to perform semantic understanding and relational reasoning on city data, and combines time series analysis and causal inference methods to predict the future state of the city.

[0075] The decision support analysis module uses a visual interface to display model analysis and prediction results, and uses the visual interface for interactive question and answer.

[0076] The update optimization module enables the large-scale language model to continuously learn and update based on feedback: the large-scale language model continuously learns and improves based on newly acquired city data and user feedback, and regularly evaluates model performance, and adjusts the model structure and parameters based on the evaluation results.

[0077] As the information interaction and execution process between the modules in the above-mentioned system are based on the same concept as the embodiment of the method of the present invention, the specific contents can be found in the description of the embodiment of the method of the present invention and will not be repeated here.

[0078] Likewise, the system of the present invention has the following beneficial effects:

[0079] Maximizing data value: By using the massive data of the city brain to train specialized large models, the utilization efficiency and value of data are significantly improved.

[0080] City-specific AI capabilities: The developed big models have a deep understanding of specific city characteristics and dynamics, providing more accurate and relevant analysis than general AI models.

[0081] Real-time decision support: The system can process and analyze real-time data streams to provide city managers with timely insights and decision-making recommendations.

[0082] Cross-domain knowledge integration: The powerful semantic understanding and reasoning capabilities of the large model enable the system to effectively integrate and analyze information from different urban subsystems.

[0083] Enhanced predictive capabilities: With a deeper understanding of urban dynamics, the system can more accurately predict future trends and potential problems.

[0084] Improved scientific nature of decision-making: Decision support based on data-driven and AI analysis improves the scientific nature and effectiveness of urban management decisions.

[0085] Optimal resource allocation: The system's forecasting and analysis capabilities help to better plan and allocate urban resources and improve resource utilization efficiency.

[0086] Enhanced emergency response capabilities: Through real-time data analysis and prediction, the system can identify potential risks early and improve the city’s emergency response capabilities.

[0087] It should be noted that not all steps and modules in the above-mentioned processes and system structures are necessary, and some steps or modules can be ignored according to actual needs. The execution order of each step is not fixed and can be adjusted as needed. The system structure described in the above-mentioned embodiments can be a physical structure or a logical structure, that is, some modules may be implemented by the same physical entity, or some modules may be implemented by multiple physical entities, or some components in multiple independent devices may be implemented together.

[0088] The above-described embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Equivalent substitutions or changes made by those skilled in the art based on the present invention are within the protection scope of the present invention. The protection scope of the present invention shall be subject to the claims.

Claims

1. A large model training and application method based on the city brain, characterized by include: Step 1: Collect and pre-process data: Continuously obtain real-time multi-dimensional urban data from the city brain and pre-process it. The urban data includes traffic flow, energy consumption, environmental indicators, and public service usage. Step 2: Training a large-scale language model: Select a basic large-scale language model, use city data for continuous fine-tuning, and use transfer learning and incremental learning methods to make the large-scale language model adapt to the dynamic changes of city data. Step 3: Use the trained large-scale language model to understand and predict urban dynamics: Use the large-scale language model to understand the semantics and reason about the relationship of urban data, and combine time series analysis and causal inference methods to predict the future state of the city. Step 4: Use the visualization interface to display the model analysis and prediction results, and use the visualization interface for interactive Q&A. Step 5: Continuously learn and update the large-scale language model based on feedback: Continuously learn and improve the large-scale language model based on newly acquired city data and user feedback, regularly evaluate model performance, and adjust model structure and parameters based on the evaluation results.

2. According to claim 1, a large model training and application method based on the city brain is characterized by: The data is preprocessed in step 1, including: cleaning, standardizing and structuring the acquired urban data to ensure data quality and consistency.

3. A large model training and application method based on the city brain according to claim 1 or 2, characterized in that In step 2, use the following formula: Indicates that the incremental learning method is used to adapt the large-scale language model to the dynamic changes of urban data, where θ t represents the parameters of the model at time t, η is the learning rate, is the loss function, It is a city dataset.

4. According to the large model training and application method based on the city brain according to claim 3, it is characterized in that the following formula is used in step 3: Combining time series analysis and causal inference methods to predict the future state of the city, is the prediction of the future state, x t is the city state vector at the current time step, and f is the nonlinear function learned by the large-scale language model.

5. The large model training and application method based on the city brain according to claim 1 is characterized by: In step 5, use the following formula: i new =θ old +α(θ online -θ old ) The large-scale language model is updated according to the feedback, where α is the update rate and θ online are parameters learned based on the latest city data.

6. A large model training and application system based on the city brain, characterized by It includes data collection and preprocessing module, large-scale language model training module, urban dynamic understanding and prediction module, decision support analysis module and update optimization module. Data collection and preprocessing module collects and preprocesses data: continuously obtains real-time multi-dimensional urban data from the city brain and preprocesses it. The urban data includes traffic flow, energy consumption, environmental indicators, and public service usage. Large-scale language model training module trains large-scale language models: selects basic large-scale language models, uses city data for continuous fine-tuning, and adopts transfer learning and incremental learning methods to make large-scale language models adapt to the dynamic changes of city data. The city dynamics understanding and prediction module uses a trained large-scale language model to understand and predict city dynamics: it uses a large-scale language model to perform semantic understanding and relational reasoning on city data, and combines time series analysis and causal inference methods to predict the future state of the city. The decision support analysis module uses a visual interface to display model analysis and prediction results, and uses the visual interface for interactive question and answer. The update optimization module enables the large-scale language model to continuously learn and update based on feedback: the large-scale language model continuously learns and improves based on newly acquired city data and user feedback, and regularly evaluates model performance, and adjusts the model structure and parameters based on the evaluation results.

7. The large model training and application system based on the city brain according to claim 6 is characterized by: The data collection and preprocessing module preprocesses the data, including: cleaning, standardizing and structuring the acquired urban data to ensure data quality and consistency.

8. A large model training and application system based on the city brain according to claim 6 or 7, characterized in that The large-scale language model training module uses the following formula: Indicates that the incremental learning method is used to adapt the large-scale language model to the dynamic changes of urban data, where θ t represents the parameters of the model at time t, η is the learning rate, is the loss function, It is a city dataset.

9. The large model training and application system based on the city brain according to claim 8 is characterized by: The urban dynamics understanding and prediction module uses the following formula: Combining time series analysis and causal inference methods to predict the future state of the city, is the prediction of the future state, x t is the city state vector at the current time step, and f is the nonlinear function learned by the large-scale language model.

10. The large model training and application system based on the city brain according to claim 6 is characterized by: The update optimization module uses the following formula: i new =θ old +α(θ online -θ old ) The large-scale language model is updated according to the feedback, where α is the update rate and θ online are parameters learned based on the latest city data.

Images