Service area health degree evaluation system and method based on deep learning

Abstract

The invention provides a service area health degree evaluation system and method based on deep learning, and the system comprises an index system building module which is used for building a service area health degree evaluation index system; the data acquisition module is used for collecting data in the service area; the data preprocessing module is used for preprocessing the collected original data in the service area; the health degree label obtaining module is used for providing labels for training of the model; the deep learning model module is used for modeling by using the preprocessed data and training a model based on a label; and the health degree evaluation module is used for comprehensively evaluating the overall health condition of the service area by adopting the trained deep learning model in combination with the data of the current service area. According to the invention, comprehensiveness and scientificity of an evaluation result are ensured; meanwhile, by means of the health degree label obtaining module, the health degree of the service area is scored in combination with quantitative indexes and expert experience, and the accuracy and practicability of health degree labels are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of transportation systems, and in particular to a service area health assessment system and method based on deep learning. Background Art

[0002] The health assessment of highway service areas is a comprehensive analysis and evaluation of the overall condition of the service areas. The content of the health assessment usually covers multiple dimensions (such as the running status, operation status, and business status of the service area). Through the comprehensive evaluation of these dimensions, managers can fully understand the health status of the service area and promptly discover potential problems in the daily operation of the service area, which is conducive to managers to formulate effective improvement measures to ensure the safe and efficient operation of the service area. In addition, regular health assessments can also improve the overall service quality and user experience of the service area, ensuring that it maintains its advantage in the fierce market competition.

[0003] With the continuous development of artificial intelligence technology, deep learning technology has gradually been widely used in various fields and has demonstrated powerful data processing and analysis capabilities. The traditional service area health evaluation method lacks a systematic health evaluation system, resulting in one-sided evaluation results. At the same time, the health label acquisition process is not scientific enough and relies too much on subjective experience, making it difficult to ensure the accuracy and practicality of the labels. In addition, the evaluation method mainly relies on manual analysis and expert judgment, which makes it difficult to process complex multi-source data, thereby reducing the accuracy and real-time nature of the evaluation. Summary of the invention

[0004] In view of the defects in the prior art, the purpose of the present invention is to provide a service area health assessment method and system based on deep learning.

[0005] A service area health assessment system based on deep learning provided by the present invention includes:

[0006] The indicator system construction module is used to establish a service area health assessment indicator system;

[0007] The data collection module collects data within the service area based on the established service area health assessment indicator system;

[0008] A data preprocessing module preprocesses the raw data collected in the service area;

[0009] The health label acquisition module is used to provide labels for model training;

[0010] The deep learning model module uses the preprocessed data for modeling and trains the model based on the labels;

[0011] The health assessment module combines the current service area data with a trained deep learning model to conduct a comprehensive assessment of the overall health status of the service area.

[0012] Preferably, it also includes:

[0013] The early warning and decision support module is used to issue risk warnings based on the assessment results, provide advice to managers, and ensure the normal operation of the service area.

[0014] Preferably, the indicator system construction module constructs a service area health assessment indicator system based on the departmental structure, digitalization level and relative importance of various indicators of the service area; the service area health assessment indicator system includes multiple key dimensions of operation, management and operation.

[0015] Preferably, the preprocessing of the collected raw data in the service area includes:

[0016] Exploratory data analysis: understand the distribution, trend and possible anomalies of the data, and provide reference information for subsequent processing and modeling;

[0017] Data processing: missing values ​​and outliers in the data are processed; missing values ​​are filled with the median, and outliers are processed by the tail-shrinking method;

[0018] Variable selection: Use correlation coefficient to analyze the correlation between variables, and eliminate variables whose correlation exceeds the preset range;

[0019] Sample expansion: Use cubic spline interpolation to expand the data; smooth the original data through the interpolation function, generate new data points between existing data points, and expand the size of the data set.

[0020] Preferably, the health label acquisition module includes processing various indicators of the service area, mapping each indicator into a score from 0 to 100, combining the mapped indicator scores with the weights of manual experience, weighted fusion of each indicator, and obtaining a preliminary health score; adjusting the health score based on manual experience, and using the adjusted health score as the training label of the model.

[0021] Preferably, after obtaining the health score, the deep learning model module divides the data set into a training set and a test set, and normalizes the data; uses the Transformer model for training and prediction; the Transformer model uses mean square error as the loss function, selects Adam as the optimizer, and uses dropout technology to prevent overfitting; after the model training is completed, the root mean square error is used on the test set to evaluate the prediction effect of the model to obtain an evaluation result.

[0022] Preferably, the deep learning model module adjusts the complexity of the model according to the obtained evaluation results, including the embedding dimension of the model, the number of heads of the multi-head attention mechanism, the dimension of the hidden layer of the feedforward neural network, the number of encoder layers, and the probability of randomly discarding neurons.

[0023] Preferably, the health assessment module performs a comprehensive assessment when the data of the current service area is input into the trained deep learning model to obtain a health score, and maps the health score into three levels: healthy, sub-healthy and unhealthy.

[0024] A service area health assessment method based on deep learning provided by the present invention includes:

[0025] Step S1: Establish a service area health assessment indicator system;

[0026] Step S2: Based on the established service area health assessment indicator system, collect data within the service area;

[0027] Step S3: pre-processing the collected raw data in the service area;

[0028] Step S4: Process various indicators of the service area to obtain a health score and provide labels for model training;

[0029] Step S5: Use the preprocessed data to build a model, and use the health score as a training label for the model;

[0030] Step S6: Combined with the current service area data, use the trained deep learning model to conduct a comprehensive assessment of the overall health status of the service area.

[0031] Preferably, it also includes:

[0032] Step S7: Issue risk warnings based on the assessment results and provide advice to managers to ensure the normal operation of the service area.

[0033] Preferably, step S1 includes constructing a service area health assessment index system based on the departmental structure, digitalization level and relative importance of various indicators of the service area; the service area health assessment index system includes multiple key dimensions of operation, management and operation.

[0034] Preferably, the preprocessing of the collected raw data in the service area includes:

[0035] Exploratory data analysis: understand the distribution, trend and possible anomalies of the data, and provide reference information for subsequent processing and modeling;

[0036] Data processing: missing values ​​and outliers in the data are processed; missing values ​​are filled with the median, and outliers are processed by the tail-shrinking method;

[0037] Variable selection: Use correlation coefficient to analyze the correlation between variables, and eliminate variables whose correlation exceeds the preset range;

[0038] Sample expansion: Use cubic spline interpolation to expand the data; smooth the original data through the interpolation function, generate new data points between existing data points, and expand the size of the data set.

[0039] Preferably, step S4 includes processing various indicators of the service area, mapping each indicator into a score from 0 to 100, combining the mapped indicator scores with the weights of manual experience, weighted fusion of each indicator, and obtaining a preliminary health score; adjusting the health score based on manual experience, and using the adjusted health score as a training label for the model.

[0040] Preferably, after obtaining the health score, step S5 divides the data set into a training set and a test set, and normalizes the data; uses a Transformer model for training and prediction; the Transformer model uses mean square error as the loss function, selects Adam as the optimizer, and uses dropout technology to prevent overfitting; after the model training is completed, the root mean square error is used on the test set to evaluate the prediction effect of the model to obtain an evaluation result.

[0041] Preferably, step S5 also includes adjusting the complexity of the model according to the obtained evaluation results, including the embedding dimension of the model, the number of heads of the multi-head attention mechanism, the dimension of the hidden layer of the feedforward neural network, the number of encoder layers, and the probability of randomly discarding neurons.

[0042] Preferably, step S6 includes performing a comprehensive evaluation when the data of the current service area is input into the trained deep learning model to obtain a health score, and mapping the health score into three levels: healthy, sub-healthy and unhealthy.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. The present invention proposes a service area health assessment method based on deep learning, which solves a series of problems existing in traditional assessment methods, such as imperfect assessment system, unscientific label acquisition process, low data processing efficiency, insufficient accuracy, and difficulty in processing complex multi-source data.

[0045] 2. The present invention relies on the indicator system construction module to establish a comprehensive and scientific service area health indicator system, ensuring the comprehensiveness and scientificity of the evaluation results. At the same time, with the help of the health label acquisition module, the service area health score is combined with quantitative indicators and expert experience to ensure the accuracy and practicality of the health label.

[0046] 3. The present invention utilizes a deep learning model module to efficiently process and analyze complex multi-source data, significantly improving the accuracy and real-time performance of predictions.

[0047] Other beneficial effects of the present invention will be explained in the specific implementation manner through the introduction of specific technical features and technical solutions. Through the introduction of these technical features and technical solutions, those skilled in the art should be able to understand the beneficial technical effects brought about by the technical features and technical solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0049] Figure 1 It is a system schematic diagram of the present invention.

[0050] Figure 2 The figure is a flow chart of data preprocessing in the present invention.

[0051] Figure 3 It is a box plot after data normalization in the present invention.

[0052] Figure 4 This is a schematic diagram of using the Pearson correlation coefficient to screen health indicators in the present invention.

[0053] Figure 5 This is a flow chart for obtaining health status in the present invention.

[0054] Figure 6 This is a workflow diagram of the deep learning model module in the present invention.

[0055] Figure 7 This is a schematic diagram of the model in the present invention.

[0056] Figure 8 This is a comparison chart of the loss functions of the training set and the test set of the model in the present invention.

[0057] Fig. 9 It is a comparison chart of the predicted value and the true value of the model in the present invention.

[0058] Fig.10 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0059] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0060] The present invention provides more accurate technical support for the assessment of the health of service areas by collecting, processing and modeling multi-source data. It can automatically extract key features from the data and conduct a comprehensive assessment of the health of service areas through deep learning models, thereby providing managers with more scientific decision-making support and effectively improving the operation and management efficiency of service areas.

[0061] Reference Figure 1 As shown, a service area health assessment system based on deep learning includes:

[0062] Indicator system construction module, data collection module, data preprocessing module, health label acquisition module, deep learning model module, health assessment module and early warning and decision support module.

[0063] Among them, the indicator system construction module is used to establish a scientific service area health assessment indicator system, covering key indicators such as operation, operation, and management, providing a specific basis for data collection.

[0064] Data collection module: responsible for collecting data on operation, management and operation from various data sources within the service area, providing a comprehensive data basis for subsequent evaluation.

[0065] Data preprocessing module: cleans, filters and extracts features of the collected raw data to ensure data quality and consistency, and provide reliable data input for model training and evaluation.

[0066] Health label acquisition module: Generates health scores through data-driven or rule-based calculation, weighted fusion and expert experience to ensure that the scores can effectively reflect the actual health status of the service area and provide high-quality labels for model training.

[0067] Deep learning model module: Based on the Transformer architecture, it models the preprocessed multi-dimensional data, generates service area health assessment results, and improves the accuracy and comprehensiveness of the assessment.

[0068] Health assessment module: Combined with the current service area data, the trained deep learning model is used to conduct a comprehensive assessment of the overall health status of the service area.

[0069] Early warning and decision support module: Based on the assessment results, risk warnings are issued and suggestions are provided to managers to ensure the safety and efficient operation of the service area.

[0070] The above are basic embodiments of the present invention. The technical solution of the present invention is further described below through two preferred embodiments.

[0071] Example 1

[0072] Reference Figure 1 As shown, a service area health assessment system based on deep learning includes:

[0073] 1. Index system construction module

[0074] Based on the department structure, digitalization level and relative importance of various indicators of the service area, an evaluation index system for the health of the service area is constructed. This system covers multiple key dimensions such as operation, management and operation, and aims to comprehensively and accurately evaluate the overall health of the service area. The specific indicator system is as follows:

[0075]

[0076] 2. Data acquisition module

[0077] After building the indicator system, it is necessary to systematically collect various types of data, including qualitative data (such as supervision and inspection scores, effective complaints, etc.) and quantitative data (such as traffic flow monitoring, energy and resource monitoring, etc.). Qualitative data is usually collected by manual scoring to ensure that subjective feedback and evaluations in service area operations can be captured. Quantitative data is collected through automated systems, such as obtaining operating income data through the cash register system and using cameras to monitor traffic flow and passenger flow.

[0078] 3. Data preprocessing module

[0079] After data collection, data preprocessing is performed. Figure 2 As shown, specifically including:

[0080] Exploratory data analysis: Conduct exploratory data analysis to understand the distribution, trends, and possible anomalies of the data, and provide reference information for subsequent processing and modeling.

[0081] like Figure 3 As shown, after normalizing the data and drawing a box plot, it can be found that except for the consumption of hazardous chemicals vehicles and electricity consumption, other variables contain outliers. The outliers of each variable need to be processed later.

[0082] Data processing: Missing values ​​are filled with the median to reduce the adverse effects of missing data on model training. The median filling method can reduce the impact of the filling process on data distribution while retaining the overall trend of the data. Outliers are processed by tail shrinking (e.g., adjusting the extreme values ​​of the data to between the 1% and 99% quantiles) to reduce the impact of outliers on model training and avoid adverse effects of extreme data on model training results.

[0083] Variable selection: In order to avoid the problem of collinearity between feature variables, the correlation coefficient (such as Pearson correlation coefficient) is used to analyze the correlation between variables. Variables with high correlation (such as correlation coefficient greater than 0.9) are eliminated to reduce redundant features.

[0084] like Figure 4 As shown, by using the Pearson correlation coefficient to screen the health indicators, it can be found that the correlation coefficient between traffic flow and passenger flow, sales volume, and sales order number exceeds 0.9, and the correlation is too high. Consider retaining one of the variables (such as traffic flow) and eliminating the remaining variables.

[0085] Sample expansion: Since the label data of the health score needs to rely on manual experience, which is costly and insufficient, the present invention introduces the cubic spline interpolation method to expand the data. The cubic spline interpolation method is a mathematical method that smoothes the original data through an interpolation function. It can generate new data points between existing data points, thereby expanding the size of the data set. This method can not only maintain the original trend of the data, but also increase the diversity of the data, thereby reducing the problem of insufficient model generalization ability caused by insufficient sample size.

[0086] 4. Health label acquisition module

[0087] Reference Figure 5 As shown, the indicators of the service area are processed through rules (such as thresholds, rankings) or data-driven methods (such as interval statistics), and each indicator is mapped to a score from 0 to 100. The mapped indicator scores are combined with the weights of manual experience to weightedly fuse the indicators to obtain a preliminary health score. In order to ensure that the score can more accurately reflect the actual health status of the service area, the health score is further adjusted based on manual experience (such as adding or deducting points). Finally, the adjusted health score is used as the training label of the model. The following is the relevant formula for obtaining the health label:

[0088] Indicator score mapping: Map each service area indicator X i Mapped to a score S from 0 to 100 i .

[0089] Threshold mapping: According to the predetermined threshold Tlow and T high , the index X i Mapped to score S i :

[0090]

[0091] Ranking mapping: Based on the ranking of the indicator values, convert them into corresponding percentile scores S i :

[0092]

[0093] Among them, Rank(X i ) is the indicator X i The rank in the sample, N is the total number of samples.

[0094] Interval statistic mapping: Use statistics within the interval (such as mean or median) to normalize the data:

[0095]

[0096] Among them, min(X i ) and max(X i ) are indicators X i The minimum and maximum values ​​of

[0097] Weighted fusion: The scores of each indicator S i According to the weight w of manual experience i Perform weighted fusion to obtain a preliminary health score H:

[0098]

[0099] Among them, n is the number of indicators, w i Satisfy the conditions

[0100] Health score adjustment: adjust the initial health score H based on manual experience, such as adding or deducting points, to obtain the final health score H adj :

[0101] H adj =H+ΔH

[0102] Among them, ΔH is the adjustment score based on the specific situation.

[0103] Model label: The adjusted health score H adj Will be used as the training label of the model for supervised learning model training.

[0104] 5. Deep Learning Model Module

[0105] After obtaining the health label, the deep learning model is trained, tested, and evaluated. Figure 6 As shown, the process is as follows:

[0106] Dataset splitting: The dataset is divided into a training set and a test set, where the training set occupies the first 80% of the data and the test set occupies the last 20% of the data. This division ensures that the model can learn the laws of the data through the training set and verify the generalization ability of the model through the test set.

[0107] Data normalization: Normalize the data so that the values ​​of all feature data are scaled to the same range (such as 0 to 1) to eliminate the impact of different dimensions between features, thereby improving the training effect and stability of the model.

[0108] Model construction: reference Figure 7 As shown in the figure, the Transformer model is used for training and prediction. Since this is a regression task, the input is a feature vector of fixed length and the output is a scalar value, so there is no sequence-to-sequence conversion involved. Based on this feature, only the encoder part of the Transformer is used in the model without the decoder. The encoder can effectively capture the complex relationship between input features and generate accurate predictions of the target variable through a multi-layer self-attention mechanism.

[0109] Model training: During the model training process, the mean square error (MSE) is used as the loss function, the optimizer selects Adam, and the dropout technology is used to prevent overfitting. During the training process, the training strategy is adjusted by observing the loss changes of the training set and the test set to obtain the best model performance.

[0110] Reference Figure 8 As shown in the figure, the loss function of the model's training set and test set drops rapidly in the first few epochs and gradually stabilizes. The loss curves of the training set and the test set are similar, and there is no overfitting phenomenon, indicating that the model performs well on both the training and test sets and has good generalization ability.

[0111] Model evaluation: After model training is completed, the root mean square error (RMSE) is used on the test set to evaluate the prediction effect of the model. RMSE can quantify the deviation between the model prediction value and the true value, and the evaluation result is used to judge the accuracy and practicality of the model.

[0112] Reference Fig. 9 As shown, a line graph is drawn to visually observe the matching effect between the true value and the predicted value.

[0113] It can be seen that the predicted value of the model (red line) generally follows the trend of the true value (blue line) well, reflecting the effectiveness of the model in capturing the main change patterns.

[0114] Model tuning: Based on the evaluation results, further adjust the complexity of the model, including the embedding dimension of the model, the number of heads of the multi-head attention mechanism, the dimension of the hidden layer of the feedforward neural network, the number of encoder layers, and the probability of randomly discarding neurons (dropout rate). Through these adjustments, the structure of the model is optimized to better adapt to task requirements and improve prediction accuracy.

[0115] 6. Health Assessment Module

[0116] The data of the current service area is input into the trained deep learning model for comprehensive evaluation to obtain the health score, and the health score is mapped into three levels: healthy, sub-healthy, and unhealthy. Through the level division, managers can quickly understand the health status of the service area.

[0117] 7. Early warning and decision support module

[0118] According to the health assessment results (for example, when the service area is rated as sub-healthy or unhealthy), the system will issue a warning to the manager to indicate the possible operational risks. At the same time, based on historical data and assessment results, it will put forward corresponding optimization suggestions and countermeasures to help managers take timely actions to prevent further deterioration of potential problems, thereby ensuring the safety and efficient operation of the service area.

[0119] Example 2

[0120] The traditional service area health evaluation method lacks a systematic health evaluation system, resulting in one-sided evaluation results. At the same time, the health label acquisition process is not scientific enough and relies too much on subjective experience, making it difficult to ensure the accuracy and practicality of the labels. In addition, the evaluation method mainly relies on manual analysis and expert judgment, which makes it difficult to handle complex multi-source data, thereby reducing the accuracy and real-time performance of the evaluation.

[0121] The method mentioned in this embodiment has the following effects:

[0122] (1) Design a health indicator system for service areas, from the operation, management and operation of service areas

[0123] Evaluate the health of the service area from multiple dimensions to ensure the comprehensiveness and scientificity of the evaluation;

[0124] (2) Design a process for obtaining service area health labels, combining quantitative indicators with expert experience

[0125] Conduct service area health scoring to ensure the accuracy and practicality of labels;

[0126] (3) Design a service area health assessment method based on deep learning. By introducing a deep learning model, the multi-source data of the service area can be automatically processed and analyzed to achieve accurate prediction and assessment of the health of the service area.

[0127] Reference Fig.10 As shown, a service area health assessment method based on deep learning includes:

[0128] Step S1: Construct a service area health assessment index system based on the departmental structure, digital intelligence level and relative importance of various indicators of the service area; the service area health assessment index system includes multiple key dimensions of operation, management and operation.

[0129] Step S2: Based on the established service area health assessment indicator system, collect data within the service area;

[0130] Step S3: pre-processing the collected raw data in the service area, specifically including the following steps:

[0131] Exploratory data analysis: understand the distribution, trend and possible anomalies of the data, and provide reference information for subsequent processing and modeling;

[0132] Data processing: missing values ​​and outliers in the data are processed; missing values ​​are filled with the median, and outliers are processed by the tail-shrinking method;

[0133] Variable selection: Use correlation coefficient to analyze the correlation between variables, and eliminate variables whose correlation exceeds the preset range;

[0134] Sample expansion: Use cubic spline interpolation to expand the data; smooth the original data through the interpolation function, generate new data points between existing data points, and expand the size of the data set.

[0135] Step S4: Process various indicators of the service area to obtain a health score and provide labels for model training;

[0136] Specifically, each indicator is mapped into a score from 0 to 100, and the mapped indicator score is combined with the weight of manual experience to perform weighted fusion on each indicator to obtain a preliminary health score; the health score is adjusted based on manual experience, and the adjusted health score is used as the training label of the model.

[0137] Step S5: Use the preprocessed data to build a model, and use the health score as a training label for the model;

[0138] After obtaining the health score, the data set is divided into a training set and a test set, and the data is normalized; the Transformer model is used for training and prediction; the Transformer model uses mean square error as the loss function, the optimizer selects Adam, and the dropout technology is used to prevent overfitting; after the model training is completed, the root mean square error is used on the test set to evaluate the prediction effect of the model to obtain the evaluation result. According to the obtained evaluation results, the complexity of the model is adjusted, including the embedding dimension of the model, the number of heads of the multi-head attention mechanism, the dimension of the hidden layer of the feedforward neural network, the number of encoder layers, and the probability of randomly discarding neurons.

[0139] Step S6: Combined with the current service area data, the trained deep learning model is used to comprehensively evaluate the overall health status of the service area. When the data of the current service area is input into the trained deep learning model, a comprehensive evaluation is performed to obtain a health score, and the health score is mapped into three levels: healthy, sub-healthy, and unhealthy.

[0140] Step S7: Issue risk warnings based on the assessment results and provide advice to managers to ensure the normal operation of the service area.

[0141] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0142] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A service area health assessment system based on deep learning, characterized in that: include: The indicator system construction module is used to establish a service area health assessment indicator system; The data collection module collects data within the service area based on the established service area health assessment indicator system; A data preprocessing module preprocesses the raw data collected in the service area; The health label acquisition module is used to provide labels for model training; The deep learning model module uses the preprocessed data for modeling and trains the model based on the labels; The health assessment module combines the current service area data with a trained deep learning model to conduct a comprehensive assessment of the overall health status of the service area.

2. The service area health assessment system based on deep learning according to claim 1 is characterized in that: Also includes: The early warning and decision support module is used to issue risk warnings based on the assessment results, provide advice to managers, and ensure the normal operation of the service area.

3. The service area health assessment system based on deep learning according to claim 1, characterized in that: The indicator system construction module constructs a service area health assessment indicator system based on the departmental structure, digitalization level and relative importance of various indicators of the service area; the service area health assessment indicator system includes multiple key dimensions of operation, management and operation.

4. The service area health assessment system based on deep learning according to claim 1, characterized in that: The preprocessing of the collected raw data in the service area includes: Exploratory data analysis: understand the distribution, trend and possible anomalies of the data, and provide reference information for subsequent processing and modeling; Data processing: missing values ​​and outliers in the data are processed; missing values ​​are filled with the median, and outliers are processed by the tail-shrinking method; Variable selection: Use correlation coefficient to analyze the correlation between variables, and eliminate variables whose correlation exceeds the preset range; Sample expansion: Use cubic spline interpolation to expand the data; smooth the original data through the interpolation function, generate new data points between existing data points, and expand the size of the data set.

5. The service area health assessment system based on deep learning according to claim 1, characterized in that: The health label acquisition module includes processing various indicators of the service area, mapping each indicator into a score from 0 to 100, combining the mapped indicator scores with the weights of manual experience, and weighted fusion of each indicator to obtain a preliminary health score; The health score is adjusted based on manual experience, and the adjusted health score is used as the training label of the model.

6. The service area health assessment system based on deep learning according to claim 5, characterized in that: After obtaining the health score, the deep learning model module divides the data set into a training set and a test set, and normalizes the data; the Transformer model is used for training and prediction; the Transformer model uses mean square error as the loss function, the optimizer selects Adam, and uses dropout technology to prevent overfitting; after the model training is completed, the root mean square error is used on the test set to evaluate the prediction effect of the model to obtain an evaluation result.

7. The service area health assessment system based on deep learning according to claim 6, characterized in that: The deep learning model module adjusts the complexity of the model according to the obtained evaluation results, including the embedding dimension of the model, the number of heads of the multi-head attention mechanism, the dimension of the hidden layer of the feedforward neural network, the number of encoder layers, and the probability of randomly discarding neurons.

8. The service area health assessment system based on deep learning according to claim 5, characterized in that: The health assessment module performs a comprehensive assessment when the data of the current service area is input into the trained deep learning model to obtain a health score, and maps the health score into three levels: healthy, sub-healthy and unhealthy.

9. A service area health assessment method based on deep learning, characterized in that: include: Step S1: Establish a service area health assessment indicator system; Step S2: Based on the established service area health assessment indicator system, collect data within the service area; Step S3: pre-processing the collected raw data in the service area; Step S4: Process various indicators of the service area to obtain a health score and provide labels for model training; Step S5: Use the preprocessed data to build a model, and use the health score as a training label for the model; Step S6: Combined with the current service area data, use the trained deep learning model to conduct a comprehensive assessment of the overall health status of the service area.

10. The service area health assessment method based on deep learning according to claim 9 is characterized in that: Also includes: Step S7: Issue risk warnings based on the assessment results and provide advice to managers to ensure the normal operation of the service area.

Images