A logistics distribution visualization system based on big data

Abstract

The present invention belongs to the technical field of logistics management, and discloses a logistics distribution visualization system based on big data, comprising a data acquisition module, which is used for acquiring logistics distribution parameter data, logistics distribution image data and logistics distribution environment data; a data processing module, which is used for preprocessing the logistics distribution parameter data, logistics distribution image data and logistics distribution environment data to obtain a parameter feature data set, an image feature data set and an environment feature data set; performing correlation analysis and fusion on the parameter feature data set, the image feature data set and the environment feature data set to obtain a comprehensive feature data set; a distribution prediction module, which is used for constructing a logistics distribution prediction model, inputting the comprehensive feature data set into the logistics distribution prediction model to obtain a cargo deliverable coefficient; judging in advance whether cargo is suitable for distribution, avoiding unnecessary transportation delays and cost waste, and improving the flexibility and response speed of logistics operations.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics management, and more specifically, to a logistics distribution visualization system based on big data. Background Art

[0002] With the rapid development of e-commerce and logistics industries, the efficiency and visualization requirements of logistics distribution are increasing; traditional logistics management systems have difficulty processing and visualizing huge data sets in real time, so an improved system is needed to meet the needs of modern logistics.

[0003] The patent publication number CN116167596A discloses a distribution path analysis method and system based on big data. By acquiring 3D scanning data of logistics goods and obtaining the transportation load data of the current logistics transfer station, the cargo transportation plan analysis is performed based on the cargo 3D scanning data and the transportation load data to obtain transportation plan data, and the transportation plan data is visualized based on the logistics map model; logistics transmission is performed according to the transportation plan data, and the cargo information of each logistics transfer station is detected in real time. Based on the cargo 3D scanning data, it is checked whether there is any abnormal information condition in the logistics goods. If there is any abnormal condition, the corresponding goods will be marked as abnormal goods. The traceability distribution path query is performed based on the 3D scanning data of the abnormal goods to obtain the distribution path, and the corresponding cargo logistics data is analyzed from the logistics big data based on the distribution path, and the information of the abnormal goods is corrected according to the cargo logistics data.

[0004] The existing logistics distribution management system generally has the following problems:

[0005] The correlation between unquantified features and the target variable may cause the model to contain a large number of features that are irrelevant or weakly correlated with the target variable. These features not only fail to provide valuable information for prediction, but may also introduce noise and reduce the prediction accuracy of the model. Including a large number of irrelevant or redundant features will increase the complexity of the model, requiring more computing resources and time during the training process. At the same time, complex models are more prone to overfitting, resulting in poor performance on unseen data;

[0006] Failure to use a feature selection adjustment formula to ensure the balance of features from different categories in the final feature subset may result in features from certain categories dominating the model while features from other categories are ignored; this imbalance will limit the model’s ability to capture comprehensive information, thus affecting the accuracy of predictions;

[0007] Without a Gaussian process surrogate model, the optimization algorithm may require more evaluations to find a near-optimal hyperparameter combination, causing the optimization process to take longer. Without limiting the range of variation of the objective function value, the optimization process may be more susceptible to noise or extreme values, causing the optimization direction to deviate from the global optimal solution. This will increase the difficulty of finding the true optimal solution and may cause the optimization process to fall into a local optimum. Due to low optimization efficiency and slow convergence, existing technologies may require more computing resources and time to achieve the same optimization effect. This not only increases costs, but may also limit the application of the optimization process on larger data sets or more complex models.

[0008] In view of this, the present invention proposes a logistics distribution visualization system based on big data to solve the above problems. Summary of the invention

[0009] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a logistics distribution visualization system based on big data, comprising:

[0010] Data collection module, used to collect logistics distribution parameter data, logistics distribution image data and logistics distribution environment data;

[0011] The data processing module is used to pre-process the logistics distribution parameter data, logistics distribution image data and logistics distribution environment data to obtain parameter feature data sets, image feature data sets and environment feature data sets; perform correlation analysis and fusion on the parameter feature data sets, image feature data sets and environment feature data sets to obtain a comprehensive feature data set;

[0012] The distribution prediction module is used to build a logistics distribution prediction model, input the comprehensive feature data set into the logistics distribution prediction model, and obtain the cargo delivery coefficient;

[0013] The delivery evaluation module is used to compare the predicted cargo delivery coefficient with the preset cargo delivery coefficient threshold to determine whether the cargo meets the logistics delivery standard;

[0014] The path selection module generates a detection instruction through the logistics intelligent management terminal to inspect the goods if the goods do not meet the logistics distribution standards; if the goods meet the logistics distribution standards, the logistics intelligent management terminal generates a delivery instruction and collects the information data of the deliverable goods;

[0015] The distribution visualization module is used to construct a three-dimensional logistics distribution map based on the information data of deliverable goods and the image feature data set, select the best path for goods distribution through a path planning algorithm, and visualize the best path for goods distribution through the three-dimensional logistics distribution map; each module is connected by wired and / or wireless means.

[0016] Furthermore, the logistics distribution parameter data includes cargo quality, cargo volume, order number, order time, shipping location, receiving location and logistics distribution cost; the logistics distribution image data includes Logistics delivery within time Image data of cargo location at each time point;

[0017] Logistics distribution environment data include logistics distribution temperature, humidity, wind speed, road congestion rate, logistics distribution route length and logistics distribution road slope.

[0018] Furthermore, the method of preprocessing the logistics distribution parameter data, the logistics distribution image data and the logistics distribution environment data to obtain the parameter feature data set, the image feature data set and the environment feature data set includes:

[0019] The DBSCAN algorithm is used on the logistics distribution parameter data, logistics distribution image data and logistics distribution environment data to identify the outliers in the logistics distribution parameter data, logistics distribution image data and logistics distribution environment data, and the identified outliers are removed to obtain the processed parameter feature data set, image feature data set and environment feature data set;

[0020] The obtained parameter feature data set, image feature data set and environment feature data set are normalized by standard deviation and converted into standard normal distribution to obtain the parameter feature data set, image feature data set and environment feature data set.

[0021] Furthermore, the method of performing correlation analysis and fusing the parameter feature data set, the image feature data set and the environment feature data set to obtain a comprehensive feature data set includes:

[0022] The preset parameter feature data set is ;in, is the parameter feature data set; For the parameter characteristics; is the sample size; is the dimension of parameter features; is the set of real numbers;

[0023] The preset image feature dataset is ;in, It is an image feature dataset; For the image features; is the dimension of image features;

[0024] The preset environmental feature dataset is ;in, It is the environmental feature dataset; For the environmental characteristics; Dimensions of environmental characteristics;

[0025] Calculate the Pearson correlation coefficient between the features in each dataset and the cargo delivery coefficient , the specific formula is: ;in, is the target variable cargo deliverability coefficient; is a parameter feature, an image feature or an environment feature; is the covariance operator symbol; is the standard deviation operator symbol;

[0026] Summarize all calculated correlation coefficients to construct a correlation vector ;in, is the correlation vector; For the Correlation coefficients; is the index of the correlation coefficient;

[0027] According to the correlation vector The absolute value of the selected features is sorted, and the top features most relevant to the target variable are selected through the feature selection adjustment formula. features, construct a new feature subset;

[0028] The feature selection adjustment formula is: ;in, is the total number of features; is the number of categories of the feature; is the proportion of features selected from each category; To perform the floor operation, ensure that the number of features is an integer;

[0029] The new feature subset is a parameter feature subset , image feature subset and a subset of environmental features ;

[0030] The parameter feature subset, the image feature subset and the environment feature subset are fused through a weighted model to form a comprehensive feature data set;

[0031] The weighted model is: ;in, It is a comprehensive feature dataset; is the weight coefficient of the parameter feature subset; is the weight coefficient of the image feature subset; is the weight coefficient of the environmental feature subset.

[0032] Furthermore, the method for constructing the logistics distribution prediction model includes:

[0033] The data set is divided into a training set, a validation set and a test set; the data set includes a comprehensive feature data set and a corresponding cargo delivery coefficient; a logistics distribution prediction model is constructed, and the logistics distribution prediction model includes input data and output labels; the input data of the model is a historical comprehensive feature data set, and the output label of the model is the cargo delivery coefficient; RBF is selected as the kernel function of the model, and the hyperparameters of the model are initialized, and the hyperparameters of the model include the regularization parameter C and the γ value of the RBF kernel function; the cargo delivery coefficient is a support vector machine model;

[0034] The mean absolute error is used as the loss function to measure the error between the model's predicted value and the actual value; the mean absolute error loss function is: ;in, is the mean absolute error; is the number of data sets; For the The true value of the cargo deliverability coefficient; For the The predicted value of the cargo deliverability coefficient;

[0035] Use the training set data to train the model and minimize the loss function using the Adam optimizer. Use the validation set to evaluate the performance of the model and tune the model's hyperparameters until the model performance no longer improves significantly or reaches the preset stopping condition.

[0036] The test set is used to evaluate the performance of the model in the prediction task, and the current comprehensive data set is input into the trained logistics distribution prediction model to obtain the cargo deliverability coefficient.

[0037] Furthermore, the method for tuning the hyperparameters of the model includes:

[0038] S61. Define the mean absolute error of the loss function of the model is the objective function, and a set of hyperparameters is found through the objective function Make the loss function mean absolute error Minimize, the optimal hyperparameter combination that minimizes the objective function is: ;in, is the hyperparameter vector; is the search space of hyperparameters; is the objective function; The optimal hyperparameter combination to minimize the objective function;

[0039] S62. Approximate the distribution of the objective function by using the Gaussian process as the objective function proxy model: ;in, It is a Gaussian process proxy model; is the symbolic representation of Gaussian process; is a set of hyperparameter combinations The mean function of is the Gaussian kernel covariance function; is another set of hyperparameter combinations;

[0040] The Gaussian kernel covariance function ;in, is the range of variation of the objective function value; is the length scale of the Gaussian kernel covariance function, which controls the influence of distance on correlation in the input space; is a combination of hyperparameters and The Euclidean distance between

[0041] Limit the range of variation of the objective function value through the range limitation formula , the range limiting formula is: ;in, is the number of hyperparameter combinations; For the Hyperparameter combinations; is a combination of hyperparameters and The Euclidean distance between

[0042] S63. Select the next hyperparameter combination to be evaluated by the expected improvement formula, wherein the next hyperparameter combination to be evaluated is: ;in, is the next hyperparameter combination to be evaluated; For the acquisition function, decide the next hyperparameter combination to evaluate;

[0043] The expected improvement formula is: ;in, is a combination of hyperparameters The expected improvement relative to the current optimum; is the expected value operator; is the current optimal hyperparameter combination;

[0044] S64, Selection Evaluate the objective function and use the new results to update the Gaussian process proxy model ;

[0045] S65, repeat steps S62-S64 until the maximum number of iterations is reached or the objective function is no longer significantly improved; output the hyperparameter combination that minimizes the objective function , so that the objective function The minimum value is reached on the validation set; the hyperparameter combination at this time This is the final optimal hyperparameter combination.

[0046] Furthermore, the method of comparing the predicted cargo deliverability coefficient with a preset cargo deliverability coefficient threshold to determine whether the cargo meets the logistics delivery standard includes:

[0047] If the predicted cargo delivery coefficient is less than the preset cargo delivery coefficient threshold, it is determined that the cargo does not meet the logistics delivery standard;

[0048] If the predicted cargo deliverability coefficient is greater than or equal to the preset cargo deliverability coefficient threshold, the cargo is determined to have met the logistics delivery standards.

[0049] Furthermore, the deliverable goods information data includes the goods ID, the goods quantity, the shipping location and the receiving location.

[0050] Furthermore, the method for constructing a three-dimensional logistics distribution map based on the deliverable goods information data and the image feature data set includes:

[0051] S91, preset shipping location coordinates are , the preset delivery location coordinates are ; Preset Logistics delivery within time The image data of each cargo position at a time point is ;in, is the index of the time point, ranging from 1 to ;

[0052] S92: For each cargo position image data , use the SSD target detection algorithm to output a set of bounding boxes and corresponding category confidences. The preset SSD target detection algorithm detects candidate bounding boxes, each of which consists of four coordinate values and a class confidence Composition; remove redundant bounding boxes through non-maximum suppression and retain the bounding box with the highest confidence;

[0053] S93. Extract each time point from the SSD target detection algorithm results The cargo location coordinates are calculated from the center point of the bounding box with the highest confidence:

[0054] ;in, For each time point The coordinates of the cargo location; For the The bounding box in The horizontal center coordinates of each time point; For the The bounding box in The vertical center coordinate of each time point; For the The bounding box in The leftmost time point coordinate; For the The bounding box in The rightmost time point coordinate; For the The bounding box in The top of the time point coordinate; For the The bounding box in The lowest time point coordinate;

[0055] S94. For each shipment, the Euclidean distance from the shipping location coordinates to the receiving location coordinates is calculated using the Euclidean distance calculation formula; the Euclidean distance calculation formula is: ;in, The Euclidean distance from the shipping location to the shipping location. is the horizontal coordinate of the shipping location; The vertical coordinate of the location of the shipping place; is the horizontal coordinate of the location of the delivery place; The vertical coordinate of the location of the delivery place;

[0056] S95. Constructing a three-dimensional diagram of logistics distribution: Preset the three-dimensional coordinate system Axis , Axis is the cargo position coordinate , The axis is the Euclidean distance from the shipping location coordinate to the receiving location coordinate ; Mark the location of each cargo at different time points in the three-dimensional coordinate system: ;in, The location of each cargo at different points in time; For time point Horizontal coordinates of the cargo location; For time point The vertical coordinate of the cargo location;

[0057] S96. Draw lines connecting the positions of each cargo at different time points on the three-dimensional logistics distribution diagram according to the cargo movement trajectory, and display the cargo movement path; the movement path is:

[0058] ;in, For time point The cargo location point.

[0059] Furthermore, the method of selecting the best path for cargo distribution by using a path planning algorithm and visually displaying the best path for cargo distribution by using a three-dimensional logistics distribution diagram includes:

[0060] The A* algorithm is used to calculate the best path from the place of shipment to the place of receipt, and real-time traffic data is integrated through the logistics intelligent management terminal. The real-time traffic data includes GPS positioning, traffic cameras and traffic broadcasts; the parameters in the heuristic function are dynamically adjusted to reflect the real-time road conditions; during the delivery process, if an emergency is detected on the current path, the logistics intelligent management terminal immediately triggers path re-planning, uses the A* algorithm to recalculate the best path, and notifies the delivery person in real time through the logistics intelligent management terminal.

[0061] Technical effects and advantages of a logistics distribution visualization system based on big data in the present invention:

[0062] The present invention can quantify the influence of different features on the cargo distribution coefficient by calculating the Pearson correlation coefficient; selecting features that are highly correlated with the target variable can significantly improve the accuracy of the prediction model; removing or reducing features with low correlation with the target variable can reduce noise and redundant information in the model, further improving the prediction effect; the feature selection adjustment formula ensures the balance of features of different categories in the final feature subset by setting the proportion of features selected from each category. This helps the model to comprehensively consider the impact of various factors on the cargo distribution coefficient and avoid deviations caused by too many or too few features in certain categories; the weighted fusion model can further simplify the calculation process while maintaining prediction accuracy by assigning reasonable weights to different feature subsets;

[0063] The Gaussian process surrogate model is used to approximate the distribution of the objective function, so that a near-optimal hyperparameter combination can be found within a smaller number of evaluations, thus automating hyperparameter optimization and reducing the need for manual intervention. It is possible to understand the degree of influence of different hyperparameter combinations on the objective function, thereby increasing the interpretability of the model tuning process. Limiting the range of variation of the objective function value helps reduce fluctuations and extreme values ​​in the optimization process, making the optimization process more stable and helping to avoid deviations from the global optimal solution due to individual extreme cases. It makes the algorithm more inclined to search in more promising areas, thereby accelerating convergence to the optimal solution. By limiting the range of variation of the objective function value, the overfitting of the model can be alleviated to a certain extent, making the model more generalized. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a schematic diagram of the structure of a logistics distribution visualization system based on big data of the present invention;

[0065] Figure 2 The figure is a flow chart of a logistics distribution visualization method based on big data of the present invention. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] Embodiment 1

[0068] See also Figure 1 As shown, the logistics distribution visualization system based on big data described in this embodiment includes:

[0069] Data collection module, used to collect logistics distribution parameter data, logistics distribution image data and logistics distribution environment data;

[0070] The data processing module is used to pre-process the logistics distribution parameter data, logistics distribution image data and logistics distribution environment data to obtain parameter feature data sets, image feature data sets and environment feature data sets; perform correlation analysis and fusion on the parameter feature data sets, image feature data sets and environment feature data sets to obtain a comprehensive feature data set;

[0071] The distribution prediction module is used to build a logistics distribution prediction model, input the comprehensive feature data set into the logistics distribution prediction model, and obtain the cargo delivery coefficient;

[0072] The delivery evaluation module is used to compare the predicted cargo delivery coefficient with the preset cargo delivery coefficient threshold to determine whether the cargo meets the logistics delivery standard;

[0073] The path selection module generates a detection instruction through the logistics intelligent management terminal to inspect the goods if the goods do not meet the logistics distribution standards; if the goods meet the logistics distribution standards, the logistics intelligent management terminal generates a delivery instruction and collects the information data of the deliverable goods;

[0074] The distribution visualization module is used to construct a three-dimensional logistics distribution map based on the information data of deliverable goods and the image feature data set, select the best path for goods distribution through a path planning algorithm, and visualize the best path for goods distribution through the three-dimensional logistics distribution map; each module is connected by wired and / or wireless means.

[0075] Logistics distribution parameter data includes cargo quality, cargo volume, order number, order time, shipping location, receiving location and logistics distribution cost; logistics distribution image data includes Logistics delivery within time Image data of cargo location at each time point;

[0076] Logistics distribution environment data include logistics distribution temperature, humidity, wind speed, road congestion rate, logistics distribution route length and logistics distribution road slope.

[0077] The quality of goods is acquired through weight sensors, the volume of goods is acquired through laser scanners, the order number, order time and logistics distribution cost are acquired through the intelligent logistics management terminal, and the location of the shipping place and the location of the receiving place are acquired through GPS positioning technology;

[0078] Logistics delivery within time The image data of the cargo location at each time point is obtained through the camera installed on the delivery truck; the temperature of the logistics distribution is obtained through the temperature sensor, the humidity is obtained through the humidity sensor, the wind speed is obtained by installing a wind speed sensor on the outside of the vehicle, the road congestion rate is obtained by GPS tracking the vehicle location and speed calculation, the logistics distribution path length is obtained through GPS, and the logistics distribution road slope is obtained through the vehicle-mounted tilt sensor;

[0079] The method of preprocessing the logistics distribution parameter data, the logistics distribution image data and the logistics distribution environment data to obtain the parameter feature data set, the image feature data set and the environment feature data set includes:

[0080] The DBSCAN algorithm is used on the logistics distribution parameter data, logistics distribution image data and logistics distribution environment data to identify the outliers in the logistics distribution parameter data, logistics distribution image data and logistics distribution environment data, and the identified outliers are removed to obtain the processed parameter feature data set, image feature data set and environment feature data set;

[0081] The obtained parameter feature data set, image feature data set and environment feature data set are normalized by standard deviation and converted into standard normal distribution to obtain the parameter feature data set, image feature data set and environment feature data set.

[0082] The method of performing correlation analysis and fusing the parameter feature dataset, the image feature dataset and the environment feature dataset to obtain the comprehensive feature dataset includes:

[0083] The preset parameter feature data set is ;in, is the parameter feature data set; For the parameter characteristics; is the sample size; is the dimension of parameter features; is the set of real numbers;

[0084] The preset image feature dataset is ;in, It is an image feature dataset; For the image features; is the dimension of image features;

[0085] The preset environmental feature dataset is ;in, It is the environmental feature dataset; For the environmental characteristics; Dimensions of environmental characteristics;

[0086] Calculate the Pearson correlation coefficient between the features in each dataset and the cargo delivery coefficient , the specific formula is: ;in, is the target variable cargo deliverability coefficient; is a parameter feature, an image feature or an environment feature; is the covariance operator symbol; is the standard deviation operator symbol;

[0087] Summarize all calculated correlation coefficients to construct a correlation vector ;in, is the correlation vector; For the Correlation coefficients; is the index of the correlation coefficient;

[0088] According to the correlation vector The absolute value of the selected features is sorted, and the top features most relevant to the target variable are selected through the feature selection adjustment formula. features, construct a new feature subset;

[0089] The feature selection adjustment formula is: ;in, is the total number of features; is the number of categories of the feature; is the proportion of features selected from each category; To perform the floor operation, ensure that the number of features is an integer;

[0090] For example: Assume that the parameter feature data set The number of samples is 20, the parameter feature dimension is 4; image feature dataset The number of samples is 15, the image feature dimension is 3; the environmental feature dataset The sample size is 10, the environmental feature dimension is 5; the cargo distribution coefficient is a vector of length 5; calculate the Pearson coefficient between each feature and the cargo distribution coefficient, assuming that the Pearson coefficient between the parameter feature and the cargo distribution coefficient is [-0.8, 0.5, 0.9, -0.3]; the Pearson coefficient between the image feature and the cargo distribution coefficient is [0.6, -0.4, 0.2]; the Pearson coefficient between the environmental feature and the cargo distribution coefficient is [-0.1, 0.7]; summarize all the correlation coefficients to form a correlation vector R, and arrange them according to the absolute value of the correlation coefficient to obtain R=[0.9, -0.8, 0.7, 0.6, 0.5, -0.4, 0.2, -0.3, -0.1];

[0091] Assuming that 50% of the features are selected from each class, the amount of selection adjustment is calculated using the feature adjustment formula: ; From the feature data set of each category, select the features with the largest absolute value of the correlation coefficient to construct a new feature subset;

[0092] The new feature subset is a parameter feature subset , image feature subset and a subset of environmental features ;

[0093] The parameter feature subset, the image feature subset and the environment feature subset are fused through a weighted model to form a comprehensive feature data set;

[0094] The weighted model is: ;in, It is a comprehensive feature dataset; is the weight coefficient of the parameter feature subset; is the weight coefficient of the image feature subset; is the weight coefficient of the environmental feature subset.

[0095] The construction method of logistics distribution prediction model includes:

[0096] The data set is divided into a training set, a validation set and a test set; the data set includes a comprehensive feature data set and a corresponding cargo delivery coefficient; a logistics distribution prediction model is constructed, and the logistics distribution prediction model includes input data and output labels; the input data of the model is a historical comprehensive feature data set, and the output label of the model is the cargo delivery coefficient; RBF is selected as the kernel function of the model, and the hyperparameters of the model are initialized, and the hyperparameters of the model include the regularization parameter C and the γ value of the RBF kernel function; the cargo delivery coefficient is a support vector machine model;

[0097] The mean absolute error is used as the loss function to measure the error between the model's predicted value and the actual value; the mean absolute error loss function is: ;in, is the mean absolute error; is the number of data sets; For the The true value of the cargo deliverability coefficient; For the The predicted value of the cargo deliverability coefficient;

[0098] Use the training set data to train the model and minimize the loss function using the Adam optimizer. Use the validation set to evaluate the performance of the model and tune the model's hyperparameters until the model performance no longer improves significantly or reaches the preset stopping condition.

[0099] The test set is used to evaluate the performance of the model in the prediction task, and the current comprehensive data set is input into the trained logistics distribution prediction model to obtain the cargo deliverability coefficient.

[0100] Methods for tuning model hyperparameters include:

[0101] S61. Define the mean absolute error of the loss function of the model is the objective function, and a set of hyperparameters is found through the objective function Make the loss function mean absolute error Minimize, the optimal hyperparameter combination that minimizes the objective function is: ;in, is a hyperparameter vector, such as the regularization parameter C and the γ value of the RBF kernel function; is the search space of hyperparameters; is the objective function; The optimal hyperparameter combination to minimize the objective function;

[0102] S62. Approximate the distribution of the objective function by using the Gaussian process as the objective function proxy model: ;in, It is a Gaussian process proxy model; is the symbolic representation of Gaussian process; is a set of hyperparameter combinations The mean function of is the Gaussian kernel covariance function; is another set of hyperparameter combinations;

[0103] Gaussian kernel covariance function ;in, is the range of variation of the objective function value; is the length scale of the Gaussian kernel covariance function, which controls the influence of distance on correlation in the input space; larger The value indicates that the function changes slowly in the input space. The value indicates that the function changes rapidly in the input space; is a combination of hyperparameters and The Euclidean distance between

[0104] Limit the range of variation of the objective function value through the range limitation formula , the range restriction formula is: ;in, is the number of hyperparameter combinations; For the Hyperparameter combinations; is a combination of hyperparameters and The Euclidean distance between

[0105] S63. Select the next hyperparameter combination to be evaluated by the expected improvement formula, wherein the next hyperparameter combination to be evaluated is: ;in, is the next hyperparameter combination to be evaluated; For the acquisition function, decide the next hyperparameter combination to evaluate;

[0106] The expected improvement formula is: ;in, is a combination of hyperparameters The expected improvement relative to the current optimum; is the expected value operator; is the current optimal hyperparameter combination;

[0107] S64, Selection Evaluate the objective function and use the new results to update the Gaussian process proxy model ;

[0108] S65, repeat steps S62-S64 until the maximum number of iterations is reached or the objective function is no longer significantly improved; output the hyperparameter combination that minimizes the objective function , so that the objective function The minimum value is reached on the validation set; the hyperparameter combination at this time This is the final optimal hyperparameter combination.

[0109] The method of comparing the predicted cargo delivery coefficient with the preset cargo delivery coefficient threshold to determine whether the cargo meets the logistics delivery standard includes:

[0110] If the predicted cargo delivery coefficient is less than the preset cargo delivery coefficient threshold, it is determined that the cargo does not meet the logistics delivery standard;

[0111] If the predicted cargo deliverability coefficient is greater than or equal to the preset cargo deliverability coefficient threshold, the cargo is determined to have met the logistics delivery standards.

[0112] The deliverable goods information data includes the goods ID, the goods quantity, the shipping location and the receiving location.

[0113] The method for constructing a three-dimensional logistics distribution map based on the deliverable goods information data and the image feature data set includes:

[0114] S91, preset shipping location coordinates are , the preset delivery location coordinates are ; Preset Logistics delivery within time The image data of each cargo position at a time point is ;in, is the index of the time point, ranging from 1 to ;

[0115] S92: For each cargo position image data , use the SSD target detection algorithm to output a set of bounding boxes and corresponding category confidences. The preset SSD target detection algorithm detects candidate bounding boxes, each of which consists of four coordinate values and a class confidence Composition; remove redundant bounding boxes through non-maximum suppression and retain the bounding box with the highest confidence;

[0116] S93. Extract each time point from the SSD target detection algorithm results The cargo location coordinates are calculated from the center point of the bounding box with the highest confidence:

[0117] ;in, For each time point The coordinates of the cargo location; For the The bounding box in The horizontal center coordinates of each time point; For the The bounding box in The vertical center coordinate of each time point; For the The bounding box in The leftmost time point coordinate; For the The bounding box in The rightmost time point coordinate; For the The bounding box in The top of the time point coordinate; For the The bounding box in The lowest time point coordinate;

[0118] S94. For each shipment, the Euclidean distance from the shipping location coordinates to the receiving location coordinates is calculated using the Euclidean distance calculation formula; the Euclidean distance calculation formula is: ;in, The Euclidean distance from the shipping location to the shipping location. is the horizontal coordinate of the shipping location; The vertical coordinate of the location of the shipping place; is the horizontal coordinate of the location of the delivery place; The vertical coordinate of the location of the delivery place;

[0119] S95. Constructing a three-dimensional diagram of logistics distribution: Preset the three-dimensional coordinate system Axis , Axis is the cargo position coordinate , The axis is the Euclidean distance from the shipping location coordinate to the receiving location coordinate ; Mark the location of each cargo at different time points in the three-dimensional coordinate system: ;in, The location of each cargo at different points in time; For time point Horizontal coordinates of the cargo location; For time point The vertical coordinate of the cargo location;

[0120] S96. Draw lines connecting the positions of each cargo at different time points on the three-dimensional logistics distribution diagram according to the cargo movement trajectory, and display the cargo movement path; the movement path is:

[0121] ;in, For time point The cargo location point.

[0122] For example, the number of goods: 10 boxes, the location coordinates of the shipping place: (10, 20), the location coordinates of the receiving place: (50, 80); at three time points within 10 minutes, the image data of the goods location is obtained:

[0123]

[0124] For images , the SSD algorithm detects 3 candidate bounding boxes with coordinates:

[0125] Bounding Box 1: Top Left Corner ( =12, = 22), lower right corner ( =18, =28), confidence level: 0.8;

[0126] Bounding Box 2: Top Left Corner ( =15, = 25), lower right corner( = 21, = 31), confidence level: 0.6;

[0127] Bounding Box 3: Top Left Corner ( =10, = 20), lower right corner ( =16, = 26), confidence level: 0.4;

[0128] Redundant bounding boxes are removed through non-maximum suppression, and the bounding box with the highest confidence is retained as 1;

[0129] Calculate the center coordinates of the cargo location based on the coordinates of the optimal bounding box 1: ;

[0130] Use the Euclidean distance formula to calculate the distance from the shipping location to the receiving location: ;

[0131] Mark the location of the goods at different time points in the three-dimensional coordinate system, time point 1: ( , 15, 25); Time point 2: ( , 30, 40); Time point 3: ( , 45, 65); Draw the movement path according to the location of the goods at different points in time ;

[0132] Methods for selecting the best path for cargo distribution by using a path planning algorithm and visually displaying the best path for cargo distribution by using a three-dimensional logistics distribution diagram include:

[0133] The A* algorithm is used to calculate the best path from the place of shipment to the place of receipt, and real-time traffic data is integrated through the logistics intelligent management terminal. The real-time traffic data includes GPS positioning, traffic cameras and traffic broadcasts; the parameters in the heuristic function are dynamically adjusted to reflect the real-time road conditions; during the delivery process, if an emergency is detected on the current path, the logistics intelligent management terminal immediately triggers path re-planning, uses the A* algorithm to recalculate the best path, and notifies the delivery person in real time through the logistics intelligent management terminal.

[0134] The preset cargo deliverable coefficient threshold is set by the staff, and different cargo deliverable coefficients are collected through the logistics intelligent management terminal, and the average value of multiple cargo deliverable coefficients is taken as the preset cargo deliverable coefficient threshold.

[0135] In this embodiment, by calculating the Pearson correlation coefficient, it is possible to quantify the degree of influence of different features on the cargo distribution coefficient; selecting features that are highly correlated with the target variable can significantly improve the accuracy of the prediction model; removing or reducing features that are less correlated with the target variable can reduce noise and redundant information in the model, further improving the prediction effect; the feature selection adjustment formula ensures the balance of features of different categories in the final feature subset by setting the proportion of features selected from each category. This helps the model to comprehensively consider the impact of various factors on the cargo distribution coefficient and avoid deviations caused by too many or too few features in certain categories; the weighted fusion model can further simplify the calculation process while maintaining prediction accuracy by assigning reasonable weights to different feature subsets;

[0136] The Gaussian process surrogate model is used to approximate the distribution of the objective function, so that a near-optimal hyperparameter combination can be found within a smaller number of evaluations, thus automating hyperparameter optimization and reducing the need for manual intervention. It is possible to understand the degree of influence of different hyperparameter combinations on the objective function, thereby increasing the interpretability of the model tuning process. Limiting the range of variation of the objective function value helps reduce fluctuations and extreme values ​​in the optimization process, making the optimization process more stable and helping to avoid deviations from the global optimal solution due to individual extreme cases. It makes the algorithm more inclined to search in more promising areas, thereby accelerating convergence to the optimal solution. By limiting the range of variation of the objective function value, the overfitting of the model can be alleviated to a certain extent, making the model more generalized.

[0137] Embodiment 2

[0138] See also Figure 2As shown, the part not described in detail in this embodiment is described in Example 1, which provides a logistics distribution visualization method based on big data, including:

[0139] S1, collecting logistics distribution parameter data, logistics distribution image data and logistics distribution environment data;

[0140] S2. Preprocess the logistics distribution parameter data, logistics distribution image data and logistics distribution environment data to obtain parameter feature data set, image feature data set and environment feature data set; perform correlation analysis and fusion on the parameter feature data set, image feature data set and environment feature data set to obtain a comprehensive feature data set;

[0141] S3, construct a logistics distribution prediction model, input the comprehensive feature data set into the logistics distribution prediction model, and obtain the cargo delivery coefficient;

[0142] S4, comparing the predicted cargo delivery coefficient with the preset cargo delivery coefficient threshold to determine whether the cargo meets the logistics delivery standard;

[0143] S5. If the goods do not meet the logistics distribution standards, a detection instruction is generated through the logistics intelligent management terminal to test the goods; if the goods meet the logistics distribution standards, a delivery instruction is generated through the logistics intelligent management terminal, and information data on deliverable goods is collected;

[0144] S6. Construct a three-dimensional logistics distribution map based on the deliverable goods information data and the image feature data set, select the best path for goods delivery through a path planning algorithm, and visualize the best path for goods delivery through the three-dimensional logistics distribution map.

[0145] Since the electronic device introduced in this embodiment is an electronic device used to implement a logistics distribution visualization system based on big data in the embodiment of this application, based on the logistics distribution visualization system based on big data introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not described in detail here. As long as the technical personnel of this field implement the electronic device used in the logistics distribution visualization system based on big data in the embodiment of this application, it belongs to the scope of protection of this application.

[0146] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0147] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A logistics distribution visualization system based on big data, characterized in that: include: Data collection module, used to collect logistics distribution parameter data, logistics distribution image data and logistics distribution environment data; The data processing module is used to pre-process the logistics distribution parameter data, logistics distribution image data and logistics distribution environment data to obtain parameter feature data sets, image feature data sets and environment feature data sets; perform correlation analysis and fusion on the parameter feature data sets, image feature data sets and environment feature data sets to obtain a comprehensive feature data set; The distribution prediction module is used to build a logistics distribution prediction model, input the comprehensive feature data set into the logistics distribution prediction model, and obtain the cargo delivery coefficient; The delivery evaluation module is used to compare the predicted cargo delivery coefficient with the preset cargo delivery coefficient threshold to determine whether the cargo meets the logistics delivery standard; The path selection module generates a detection instruction through the logistics intelligent management terminal to inspect the goods if the goods do not meet the logistics distribution standards; if the goods meet the logistics distribution standards, the logistics intelligent management terminal generates a delivery instruction and collects the information data of the deliverable goods; The distribution visualization module is used to construct a three-dimensional logistics distribution map based on the information data of deliverable goods and the image feature data set, select the best path for goods distribution through the path planning algorithm, and visualize the best path for goods distribution through the three-dimensional logistics distribution map; each module is connected by wire and / or wireless means; The method of performing correlation analysis and fusing the parameter feature data set, the image feature data set and the environment feature data set to obtain a comprehensive feature data set includes: The preset parameter feature data set is X = [X1, X2, ..., X q ]∈R P×q ; Where X is the parameter feature data set; X q is the qth parameter feature; P is the number of samples; q is the dimension of the parameter feature; R is a real number set; The preset image feature data set is Y = [Y1, Y2, ..., Y q′ ]∈R P×q′ ; Where Y is the image feature dataset; Y q′ is the q′th image feature; q′ is the dimension of the image feature; The preset environmental feature data set is Z = [Z1, Z2, ..., Z q″ ]∈R P×q″ ; Where Z is the environmental feature dataset; Z q″ is the q″th environmental feature; q″ is the dimension of the environmental feature; Calculate the Pearson correlation coefficient ρ between the features in each data set and the cargo distribution coefficient. The specific formula is: Among them, I is the target variable cargo deliverability coefficient; J is the parameter feature, image feature or environmental feature; Cov is the covariance operator; σ is the standard deviation operator; Summarize all the calculated correlation coefficients and construct the correlation vector R = [ρ1, ρ2, ..., ρ α ]; where R is the correlation vector; ρ α is the αth correlation coefficient; α is the index of the correlation coefficient; Sort by the absolute value of the correlation vector R, and select the top K features most relevant to the target variable through the feature selection adjustment formula to construct a new feature subset; The feature selection adjustment formula is: Among them, O is the total number of features; U is the number of feature categories; fot is the proportion of features selected from each category; To perform the floor operation, ensure that the number of features is an integer; The new feature subsets are parameter feature subset X′, image feature subset Y′ and environment feature subset Z′; The parameter feature subset, the image feature subset and the environment feature subset are fused through a weighted model to form a comprehensive feature data set; The weighted model is: PA = β1·X′+β2·Y′+β3·Z′; wherein PA is a comprehensive feature data set; β1 is a weight coefficient of a parameter feature subset; β2 is a weight coefficient of an image feature subset; β3 is a weight coefficient of an environment feature subset; The method for constructing the logistics distribution prediction model includes: The data set is divided into a training set, a validation set and a test set; the data set includes a comprehensive feature data set and a corresponding cargo delivery coefficient; a logistics distribution prediction model is constructed, and the logistics distribution prediction model includes input data and output labels; the input data of the model is a historical comprehensive feature data set, and the output label of the model is the cargo delivery coefficient; RBF is selected as the kernel function of the model, and the hyperparameters of the model are initialized, and the hyperparameters of the model include a regularization parameter C and a γ value of the RBF kernel function; the logistics distribution prediction model is a support vector machine model; The mean absolute error is used as the loss function to measure the error between the model's predicted value and the actual value; the mean absolute error loss function is: Among them, MAE is the mean absolute error; n is the number of data sets; y i is the true value of the i-th cargo deliverability coefficient; is the predicted value of the deliverability coefficient of the i-th cargo; Use the training set data to train the model and minimize the loss function using the Adam optimizer. Use the validation set to evaluate the performance of the model and tune the model's hyperparameters until the model performance no longer improves significantly or reaches the preset stopping condition. The test set is used to evaluate the performance of the model in the prediction task, and the current comprehensive data set is input into the trained logistics distribution prediction model to obtain the cargo deliverability coefficient.

2. The logistics distribution visualization system based on big data according to claim 1 is characterized in that: The logistics distribution parameter data includes cargo quality, cargo volume, order number, order time, shipping location, receiving location and logistics distribution cost; the logistics distribution image data includes cargo location image data at m time points of logistics distribution within a time period of t; Logistics distribution environment data include logistics distribution temperature, humidity, wind speed, road congestion rate, logistics distribution route length and logistics distribution road slope.

3. The logistics distribution visualization system based on big data according to claim 2 is characterized in that: The method of preprocessing the logistics distribution parameter data, the logistics distribution image data and the logistics distribution environment data to obtain the parameter feature data set, the image feature data set and the environment feature data set comprises: The DBSCAN algorithm is used on the logistics distribution parameter data, logistics distribution image data and logistics distribution environment data to identify the outliers in the logistics distribution parameter data, logistics distribution image data and logistics distribution environment data, and the identified outliers are removed to obtain the processed parameter feature data set, image feature data set and environment feature data set; The obtained parameter feature data set, image feature data set and environment feature data set are normalized by standard deviation and converted into standard normal distribution to obtain the parameter feature data set, image feature data set and environment feature data set.

4. The logistics distribution visualization system based on big data according to claim 3 is characterized in that: The method for tuning the hyperparameters of the model includes: S61. Define the mean absolute error (MAE) of the loss function of the model as the objective function. Use the objective function to find a set of hyperparameters x that minimizes the mean absolute error (MAE) of the loss function. The optimal hyperparameter combination that minimizes the objective function is: Where x is the hyperparameter vector; χ is the search space of hyperparameters; MAE is the objective function; x * The optimal hyperparameter combination to minimize the objective function; S62. Approximate the distribution of the objective function by using a Gaussian process as a proxy model of the objective function: g(x)~GP(μ(x), k(x, x′)); where g(x) is a Gaussian process proxy model; GP is the symbolic representation of a Gaussian process; μ(x) is the mean function of a set of hyperparameter combinations x; k(x, x′) is the Gaussian kernel covariance function; x′ is another set of hyperparameter combinations; The Gaussian kernel covariance function Among them, δ 2 is the range of variation of the objective function value; l is the length scale of the Gaussian kernel covariance function, which controls the degree of influence of distance on correlation in the input space; ||xx′|| is the Euclidean distance between the hyperparameter combinations x and x′; The range of variation of the objective function value is limited by the range limiting formula δ 2 , the range limiting formula is: Where N is the number of hyperparameter combinations; x b is the bth hyperparameter combination; ||x b -x c || is the hyperparameter combination x b and x c The Euclidean distance between S63. Select the next hyperparameter combination to be evaluated by the expected improvement formula, wherein the next hyperparameter combination to be evaluated is: Among them, x next is the next hyperparameter combination to be evaluated; a(x, g(x)) is the acquisition function that determines the next hyperparameter combination to be evaluated; The expected improvement formula is: EI(x)=E[max(0,MAE(x best )-g(x))]; where EI(x) is the expected improvement of the hyperparameter combination x relative to the current optimal value; E is the expected value operator; x best is the current optimal hyperparameter combination; S64, select x next Evaluate the objective function and use the new result to update the Gaussian process proxy model g(x); S65, repeat steps S62-S64 until the maximum number of iterations is reached or the objective function is no longer significantly improved; output the hyperparameter combination x that minimizes the objective function * , so that the objective function MAE reaches the minimum value on the validation set; at this time, the hyperparameter combination x * This is the final optimal hyperparameter combination.

5. The logistics distribution visualization system based on big data according to claim 4 is characterized in that: The method of comparing the predicted cargo delivery coefficient with a preset cargo delivery coefficient threshold to determine whether the cargo meets the logistics delivery standard includes: If the predicted cargo delivery coefficient is less than the preset cargo delivery coefficient threshold, it is determined that the cargo does not meet the logistics delivery standard; If the predicted cargo deliverability coefficient is greater than or equal to the preset cargo deliverability coefficient threshold, the cargo is determined to have met the logistics delivery standards.

6. The logistics distribution visualization system based on big data according to claim 5 is characterized in that: The deliverable goods information data includes the goods ID, the goods quantity, the shipping location and the receiving location.

7. The logistics distribution visualization system based on big data according to claim 6 is characterized in that: The method for constructing a three-dimensional logistics distribution graph based on the deliverable goods information data and the image feature data set includes: S91. The preset coordinates of the shipping place are (Xs″tart,j,Ys′t′art,j), and the preset coordinates of the receiving place are (X″ end,j ,Y″ end,j );The image data of each cargo location at m time points in the logistics distribution within the preset time t is Where w is the index of the time point, ranging from 1 to m; S92: For each cargo position image data The SSD target detection algorithm is used to output a set of bounding boxes and corresponding category confidences. The preset SSD target detection algorithm detects M candidate bounding boxes, each of which consists of four coordinate values ​​(X″ min ,Y″ min ,X″ max ,Y″ max ) and a category confidence G; redundant bounding boxes are removed by non-maximum suppression, and the bounding box with the highest confidence is retained; S93, extract each time point t from the SSD target detection algorithm results w The cargo location coordinates are calculated from the center point of the bounding box with the highest confidence: ; Among them, (X″ j,w ,Y″ j,w ) for each time point t w The cargo location coordinates; X″ j,w Y″ is the horizontal center coordinate of the jth bounding box at the wth time point; j,w is the vertical center coordinate of the jth bounding box at the wth time point; X″ min,j,w is the leftmost X″ coordinate of the jth bounding box at the wth time point; X″ max,j,w is the rightmost X″ coordinate of the jth bounding box at the wth time point; Y″ min,j,w is the Y″ coordinate of the top edge of the jth bounding box at the wth time point; Y″ max,j,w is the Y″ coordinate of the bottom edge of the j-th bounding box at the w-th time point; S94. For each shipment, the Euclidean distance from the shipping location coordinates to the receiving location coordinates is calculated using the Euclidean distance calculation formula; the Euclidean distance calculation formula is: Among them, d j X″ is the Euclidean distance from the shipping location to the receiving location; start,j Y″ is the horizontal coordinate of the shipping location; start,j X″ is the vertical coordinate of the shipping location; end,j Y″ is the horizontal coordinate of the destination location; end,j The vertical coordinate of the location of the delivery place; S95. Construct a three-dimensional diagram of logistics distribution: The X″ axis of the preset three-dimensional coordinate system is t w ,Y″ axis is the cargo location coordinate (X″ j,w ,Y″ j,w ), the Z″ axis is the Euclidean distance d from the shipping location coordinate to the receiving location coordinate j ; Mark the position of each cargo at different time points in the three-dimensional coordinate system: Apj = (t w ,X″ j,w ,Y″ j,w ); where Apj is the location of each cargo at different time points; X″ j,w Y″ is the horizontal coordinate of the cargo position at time point w; j,w is the vertical coordinate of the cargo position at time point w; S96. Draw lines connecting the positions of each cargo at different time points on the three-dimensional logistics distribution diagram according to the cargo movement trajectory, and display the cargo movement path; the movement path is: in, For time point The cargo location point.

8. The logistics distribution visualization system based on big data according to claim 7 is characterized in that: The method of selecting the best path for cargo distribution by using a path planning algorithm and visually displaying the best path for cargo distribution by using a three-dimensional logistics distribution diagram includes: The A* algorithm is used to calculate the best path from the place of shipment to the place of receipt, and real-time traffic data is integrated through the logistics intelligent management terminal. The real-time traffic data includes GPS positioning, traffic cameras and traffic broadcasts; the parameters in the heuristic function are dynamically adjusted to reflect the real-time road conditions; during the delivery process, if an emergency is detected on the current path, the logistics intelligent management terminal immediately triggers path re-planning, uses the A* algorithm to recalculate the best path, and notifies the delivery person in real time through the logistics intelligent management terminal.

Images