Logistics center network fault automatic diagnosis method, apparatus and device, and storage medium

Abstract

The invention relates to the field of logistics, and discloses a logistics center network fault automatic diagnosis method, device and equipment and a storage medium, and the method is used for improving the accuracy and robustness of network fault diagnosis. The method comprises the following steps: preprocessing collected logistics center network data; extracting samples from the preprocessed network data set by adopting a self-service sampling method in advance to form a preset number of sub-data sets, constructing a decision tree for each sub-data set, and randomly selecting part of features from numerous dimension features for splitting at each node of the decision tree; traversing each decision tree, introducing a long short-term memory network model on each node of the decision tree, and constructing a decision tree-LSTM model cluster; each decision tree-LSTM model in the decision tree-LSTM model cluster is trained, and a trained decision tree-LSTM model cluster is obtained; and calling the trained decision tree-LSTM model cluster to diagnose and predict currently input logistics center network data, and outputting a diagnosis and prediction result.

Description

Technical Field

[0001] The present invention relates to the field of logistics technology, and in particular, to an automatic diagnosis method, device, equipment and storage medium for network faults in a logistics center. Background Art

[0002] With the rapid development of network technology, the network scale of logistics centers has been continuously expanding, the network structure has become increasingly complex, and the types and quantities of network faults have also been increasing. Traditional network fault diagnosis methods mainly rely on the manual experience and manual troubleshooting of logistics center personnel, and there are problems such as low efficiency, low accuracy, and difficulty in dealing with complex faults. For example, manual troubleshooting of network faults requires a large amount of time and manpower, and is easily affected by human factors, resulting in inaccurate positioning of network faults. In addition, in the face of a large amount of network device logs and performance data, the difficulty and complexity of manual analysis also increase significantly.

[0003] Therefore, the existing technology still needs to be improved and developed. Summary of the Invention

[0004] The present invention provides an automatic diagnosis method, device, equipment and storage medium for network faults in a logistics center, aiming to solve the problems that the existing network fault diagnosis methods mainly rely on the manual experience and manual troubleshooting of logistics center personnel, and there are problems such as low efficiency, low accuracy, and difficulty in dealing with complex faults.

[0005] In the first aspect of the present invention, an automatic diagnosis method for network faults in a logistics center is provided. The automatic diagnosis method for network faults in a logistics center includes: preprocessing the collected network data of the logistics center to obtain a preprocessed network data set, where the network data of the logistics center includes: system logs, network traffic, device status, application logs, and user behavior data; pre-using the bootstrap sampling method to extract samples from the preprocessed network data set according to the network scale and fault complexity of the logistics center to form a plurality of sub-data sets with a predetermined number, constructing a decision tree for each sub-data set, and at each node of the decision tree, randomly selecting some features from many dimensional features for splitting; traversing each decision tree, introducing a long short-term memory network model at each node of the decision tree to construct a decision tree-LSTM model cluster, where the long short-term memory network model is used to determine the best splitting point for some features randomly selected at the current node; training each decision tree-LSTM model in the decision tree-LSTM model cluster to obtain a trained decision tree-LSTM model cluster; calling the trained decision tree-LSTM model cluster to diagnose and predict the currently input network data of the logistics center, and outputting a diagnosis and prediction result.

[0006] Optionally, in the first implementation manner of the first aspect of the present invention, the preprocessing of the collected logistics center network data includes: collecting the logistics center network data; preprocessing the missing values and outliers in the collected logistics center network data to obtain a preprocessed network data set, where the preprocessed network data set includes normal logistics center network data and faulty logistics center network data.

[0007] Optionally, in the second implementation manner of the first aspect of the present invention, the preprocessing of the missing values and outliers in the collected logistics center network data includes: based on the data characteristics of the missing values, using the mean filling method or the linear interpolation method to process the missing values; using the three - standard - deviation principle method or the rule - based anomaly detection method to identify the outliers in the logistics center network data, and correcting and adjusting the identified outliers.

[0008] Optionally, in the third implementation manner of the first aspect of the present invention, the training of each decision tree - LSTM model in the decision tree - LSTM model cluster to obtain the trained decision tree - LSTM model cluster includes: dividing the preprocessed network data set into a training data set and a test data set, and training each decision tree - LSTM model based on the training set data; using the test data set to evaluate each trained decision tree - LSTM model, and adjusting the decision tree - LSTM model based on the evaluation results.

[0009] Optionally, in the fourth implementation manner of the first aspect of the present invention, the training of each decision tree - LSTM model based on the training set data includes: initializing the weight parameters of each decision tree - LSTM model; inputting the training set data into the decision tree - LSTM model, and performing forward propagation calculation according to the initialized weight parameters of the current decision tree - LSTM model to obtain a predicted value; calculating the difference between the predicted value and the true value using a loss function or the mean square error according to the prediction target to obtain a loss gradient; backpropagating the loss gradient, and optimizing the weight parameters of the current decision tree - LSTM model based on the loss gradient and using an optimizer.

[0010] Optionally, in the fifth implementation manner of the first aspect of the present invention, the using the test data set to evaluate each trained decision tree-LSTM model and adjusting the decision tree-LSTM model based on the evaluation results includes: inputting each data record in the test set data into the trained decision tree-LSTM model to obtain a predicted value; comparing the predicted value with the true value to determine whether the prediction is accurate, and counting the total number of accurately predicted values; calculating the accuracy rate based on the total number of accurately predicted values and the total number of data records in the test set data. If the accuracy rate is less than the preset threshold, increase the number of hidden layers of the LSTM long short-term memory network and adjust the learning rate of the decision tree-LSTM model.

[0011] Optionally, in the sixth implementation manner of the first aspect of the present invention, the calling the trained decision tree-LSTM model cluster to perform diagnostic prediction on the currently input logistics center network data and outputting the diagnostic prediction result includes: constructing a network fault diagnosis system based on the front-end and back-end separation architecture, and integrating the trained decision tree-LSTM model cluster into the network fault diagnosis system; using the network fault diagnosis system to receive user front-end requests and receive network data in real time; the network fault diagnosis system calling the trained decision tree-LSTM model cluster to analyze and diagnose the network data received in real time and output the diagnostic prediction result.

[0012] Optionally, a logistics center network fault automatic diagnosis device is provided in the second aspect of the present invention, including: a preprocessing module for preprocessing the collected logistics center network data to obtain a preprocessed network data set, where the logistics center network data includes: system logs, network traffic, device status, application logs, and user behavior data; an extraction module for pre-using the bootstrap sampling method to extract samples from the preprocessed network data set according to the logistics center network scale and fault complexity to form a plurality of sub-data sets with a predetermined number, constructing a decision tree for each sub-data set, and randomly selecting some features from many dimensional features for splitting at each node of the decision tree; a construction module for traversing each decision tree and introducing a long short-term memory network model at each node of the decision tree to construct a decision tree-LSTM model cluster, where the long short-term memory network model is used to determine the best splitting point for some features randomly selected at the current node; a training module for training each decision tree-LSTM model in the decision tree-LSTM model cluster to obtain a trained decision tree-LSTM model cluster; a calling module for calling the trained decision tree-LSTM model cluster to perform diagnostic prediction on the currently input logistics center network data and outputting the diagnostic prediction result.

[0013] Optionally, in the first implementation manner of the second aspect of the present invention, the preprocessing module includes: a collection unit for collecting logistics center network data; a preprocessing unit for preprocessing missing values and outliers in the collected logistics center network data to obtain a preprocessed network data set, where the preprocessed network data set includes normal logistics center network data and faulty logistics center network data.

[0014] Optionally, in the second implementation manner of the second aspect of the present invention, the preprocessing unit includes: an inspection subunit for inspecting the collected logistics center network data through a data analysis tool to check whether there are missing values in each field of the logistics center network data; a processing subunit for processing the missing values by using the mean filling method or the linear interpolation method based on the data characteristics of the missing values; an identification subunit for identifying outliers in the logistics center network data by using the three-standard-deviation principle method or a rule-based anomaly detection method, and correcting and adjusting the identified outliers.

[0015] Optionally, in the third implementation manner of the second aspect of the present invention, the training module includes: a training unit for dividing the preprocessed network data set into a training data set and a test data set, and training each decision tree-LSTM model based on the training set data; an evaluation unit for evaluating each trained decision tree-LSTM model by using the test data set, and adjusting the decision tree-LSTM model based on the evaluation results.

[0016] Optionally, in the fourth implementation manner of the second aspect of the present invention, the training unit includes: an initialization subunit for initializing the weight parameters of each decision tree-LSTM model; a first input subunit for inputting the training set data into the decision tree-LSTM model, and performing forward propagation calculation according to the initialized weight parameters of the current decision tree-LSTM model to obtain a predicted value; a first calculation subunit for calculating the difference between the predicted value and the true value by using a loss function or the mean square error according to the prediction target to obtain a loss gradient; a propagation subunit for backpropagating the loss gradient, and optimizing the weight parameters of the current decision tree-LSTM model based on the loss gradient and by using an optimizer.

[0017] Optionally, in the fifth implementation manner of the second aspect of the present invention, the evaluation unit includes: a second input subunit, configured to input each data record in the test set data into the trained decision tree-LSTM model to obtain a predicted value; a comparison subunit, configured to compare the predicted value with the true value to determine whether the prediction is accurate, and count the total number of accurately predicted values; a second calculation subunit, configured to calculate an accuracy rate based on the total number of accurately predicted values and the total number of data records in the test set data. If the accuracy rate is less than a preset threshold, increase the number of hidden layers of the LSTM long short-term memory network and adjust the learning rate of the decision tree-LSTM model.

[0018] Optionally, in the sixth implementation manner of the second aspect of the present invention, the calling module includes: a construction unit, configured to construct a network fault diagnosis system based on a front-end and back-end separation architecture, and integrate the trained decision tree-LSTM model cluster into the network fault diagnosis system; a receiving unit, configured to use the network fault diagnosis system to receive user front-end requests and receive network data in real time; a calling unit, configured to call the trained decision tree-LSTM model cluster by the network fault diagnosis system to analyze and diagnose the network data received in real time, and output a diagnosis prediction result.

[0019] Optionally, the third aspect of the present invention provides a logistics center network fault automatic diagnosis device, including a memory and at least one processor. Computer-readable instructions are stored in the memory, and the memory and the at least one processor are interconnected by a line; the at least one processor calls the computer-readable instructions in the memory to enable the logistics center network fault automatic diagnosis device to execute each step of the above-mentioned logistics center network fault automatic diagnosis method.

[0020] Optionally, the fourth aspect of the present invention provides a computer-readable storage medium, in which computer-readable instructions are stored. When it runs on a computer, it enables the computer to execute each step of the above-mentioned logistics center network fault automatic diagnosis method.

[0021] In the technical solution provided by the present invention, by collecting the network data of the logistics center, the network data of the logistics center can be used as the basis for the model training for diagnosing and predicting the type of network faults. At the same time, the collected network data of the logistics center covers information of multiple dimensions, which can improve the accuracy of the model network fault diagnosis prediction. Secondly, the collected network data of the logistics center is preprocessed to provide a high-quality data basis for subsequent data analysis and model construction training, so as to avoid misleading the model to learn the wrong mode. After preprocessing the network data of the logistics center, samples are extracted from the preprocessed network data set according to the scale of the logistics center network and the complexity of the fault, forming a predetermined number of multiple sub-data sets, and a decision tree is constructed for each sub-data set to avoid the model from overfitting a single data mode and improve the accuracy of the model network fault diagnosis prediction; and when constructing each node classification of the decision tree, some features are randomly selected from the numerous dimensional features for splitting, which can promote the growth of the decision tree under different feature combinations, can mine the potential associations between the features, enhance the generalization ability of the model, and make the model more accurate in diagnosing and predicting the type of network faults. After building a predetermined number of decision trees, a long short-term memory network model is introduced at each node of the decision tree to build a decision tree-LSTM model cluster, so that the decision tree-LSTM model cluster has a strong processing capability for time series data and can capture the long-term dependencies in the logistics center network data. Moreover, based on the learning of the random selection features of the nodes by the long short-term memory network model, the feature segmentation points that can best distinguish different fault types can be selected, making the decision tree-LSTM model cluster more intelligent and accurate than the ordinary decision tree model. By calling the trained decision tree-LSTM model cluster, the real-time logistics center network data can be analyzed and diagnosed, so as to accurately diagnose and predict the type of network faults, provide decision support for operation and maintenance personnel, help operation and maintenance personnel quickly locate and solve network faults, and improve the network operation and maintenance efficiency of the logistics center. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A first flow chart of a method for automatic diagnosis of network faults in a logistics center provided by an embodiment of the present invention;

[0023] Figure 2 A second flow chart of the method for automatic diagnosis of network faults in a logistics center provided by an embodiment of the present invention;

[0024] Figure 3 A third flow chart of the method for automatic diagnosis of network faults in a logistics center provided by an embodiment of the present invention;

[0025] Figure 4 A fourth flow chart of the method for automatic diagnosis of network faults in a logistics center provided by an embodiment of the present invention;

[0026] Figure 5The fifth flowchart of the automatic diagnosis method for logistics center network faults provided by the embodiments of the present invention;

[0027] Figure 6 The sixth flowchart of the automatic diagnosis method for logistics center network faults provided by the embodiments of the present invention;

[0028] Figure 7 The seventh flowchart of the automatic diagnosis method for logistics center network faults provided by the embodiments of the present invention;

[0029] Figure 8 The structural schematic diagram of the automatic diagnosis device for logistics center network faults provided by the embodiments of the present invention;

[0030] Figure 9 The structural schematic diagram of the automatic diagnosis equipment for logistics center network faults provided by the embodiments of the present invention. Specific embodiments

[0031] The embodiments of the present invention provide an automatic diagnosis method, device, equipment and storage medium for logistics center network faults. This method is used to improve the accuracy and robustness of network fault diagnosis. The method includes: preprocessing the collected logistics center network data; pre-using the bootstrap sampling method to extract samples from the preprocessed network data set to form a plurality of sub-data sets with a predetermined number. For each sub-data set, a decision tree is constructed. At each node of the decision tree, a part of the features are randomly selected from many dimensional features for splitting; traversing each decision tree, introducing a long short-term memory network model at each node of the decision tree to construct a decision tree-LSTM model cluster; training each decision tree-LSTM model in the decision tree-LSTM model cluster to obtain a trained decision tree-LSTM model cluster; calling the trained decision tree-LSTM model cluster to diagnose and predict the currently input logistics center network data, and outputting a diagnosis and prediction result.

[0032] The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and drawings of the present invention are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0033] For ease of understanding, the specific process of the embodiments of the present invention will be described below. Please refer to Figure 1 , the first embodiment of an automatic diagnosis method for network faults in a logistics center in the embodiments of the present invention includes:

[0034] S101. Preprocess the collected logistics center network data to obtain a preprocessed network data set.

[0035] Specifically, in this embodiment, the logistics center network data includes: system log data, network connectivity data, network traffic data, port and protocol data, device status data, application logs, and user behavior data. The system log data records resource usage conditions such as CPU usage rate, peak memory occupancy, and disk I / O busyness, etc., reflects the overall system operation load, and can be used to judge whether a network fault occurs due to system resource exhaustion. The network connectivity data includes Ping test data, Traceroute path tracing data, etc., and can accurately locate network interruption or high-latency nodes. For example, if it is found that the server CPU remains at 100% continuously for a certain period of time and the network latency increases sharply, there is a potential causal relationship between the two, which can be used to judge the type of network fault.

[0036] The network traffic data covers the peak traffic period and the data packet transmission rates between different regions and devices. By analyzing the correlation between the network traffic data and the network fault type, classification prediction of network fault types such as network congestion, device failure, DDoS attack, configuration error, etc. can be achieved. The port and protocol data records the open / closed status of each server port, protocol type and version. By analyzing the correlation between the port open / closed status, protocol type and version and the network fault type, classification prediction of network fault types such as service interruption, port scan attack, protocol vulnerability exploitation, configuration error, etc. can be achieved. For example, if a specific port used by the logistics warehousing management system is inexplicably closed, it will directly affect the data transmission of goods in and out of the warehouse, that is, a service interruption problem occurs.

[0037] The device status data includes status information such as device temperature, fan speed, port bandwidth utilization rate, etc., as well as the CPU and memory performance metrics of the device. For example, abnormal high temperature of the device may cause the device to freeze, thereby interrupting the network connection; while the long-term excessive port bandwidth utilization rate is related to the carrying capacity of the network link. Therefore, by analyzing the correlation between the device status data and the types of network faults, it is possible to achieve the classification prediction of network fault types such as device faults and poor network link carrying capacity. The application logs and user behavior data record the user operation process, error messages, as well as performance metrics such as system response time and throughput. For example, the delay in order processing may be caused by network problems hindering data interaction. Thus, by analyzing the collected application logs and user behavior data, the root cause of the network fault can be traced. For example, when employees in a certain area collectively report being unable to log in to the system, combined with the daily operation behavior data of logistics staff, including login time, operation frequency, data upload and download volume, etc., it is determined whether it is a local network problem or an application server fault.

[0038] Due to problems such as inconsistent data formats, missing data, and outliers in the network data of the logistics center, and these problems will affect subsequent data analysis and model construction. Therefore, after collecting the above-mentioned network data of the logistics center, preprocess the network data of the logistics center, convert the network data of the logistics center into a unified and standardized form, provide a high-quality data basis for subsequent data analysis and model construction training, and avoid misleading the model to learn wrong patterns.

[0039] S102. According to the network scale and fault complexity of the logistics center, pre-extract samples from the preprocessed network data set by using the bootstrap sampling method to form a plurality of sub-data sets with a predetermined number. For each sub-data set, construct a decision tree, and at each node of the decision tree, randomly select some features from many dimensional features for splitting.

[0040] Specifically, since the amount of network data in the logistics center may be large, directly processing all the network data will result in a large amount of calculation and possible overfitting problems. For example, for the complex network of a large logistics park, it may be necessary to build 50-100 decision trees, while for a small logistics site, the logistics center network is relatively simple and 20-30 decision trees need to be built. Therefore, in this embodiment, samples are first extracted from the pre-processed network data set according to the scale of the logistics center network and the complexity of the fault, forming a predetermined number of multiple sub-data sets, and a decision tree is constructed for each sub-data set. When extracting samples, a self-service sampling method is used to extract samples from the massive pre-processed network data set with replacement, so that each sub-data set is of moderate size and has a certain degree of randomness. For example, the preprocessed network dataset contains 10,000 data records, 8,000 of which are randomly selected each time, and repeated multiple times to form multiple different sub-datasets. This ensures that each sub-dataset can not only reflect the overall data feature distribution, but also increase the diversity and generalization ability of the model, laying the foundation for the subsequent construction of multiple decision tree-LSTM models, avoiding the model from overfitting a single data pattern, and improving the accuracy of the model network fault diagnosis prediction.

[0041] When constructing each node split of the decision tree, some features are randomly selected from many dimensional features for splitting, such as 5-8 data types randomly selected from system log data, network connectivity data, network traffic data, port and protocol data, device status data, application logs and user behavior data. This random selection promotes the growth of decision trees under different feature combinations, so it can mine potential associations between features and enhance the generalization ability of the model. For example, when determining the root cause of network delay problems, a split may be based on randomly selected features such as system CPU usage, network link traffic, and application server response time, so that a unique decision branch path can be constructed for diagnosing and predicting network fault types.

[0042] S103, traversing each decision tree, introducing a long short-term memory network model at each node of the decision tree, and constructing a decision tree-LSTM model cluster, wherein the long short-term memory network model is used to determine the optimal segmentation point for some features randomly selected from the current node.

[0043] Specifically, in this embodiment, since network fault data often has temporal characteristics, for example, the state changes of network devices, fluctuations in traffic, etc. all occur along with the time series. When processing data, the ordinary decision tree model usually makes splitting decisions based on the features of the current node, and it is difficult to directly capture the long-term dependencies in the data. For example, network faults may be related to factors such as abnormal network traffic and device state changes several hours or even days ago. It is difficult for the decision tree to effectively incorporate this long-span information into fault prediction. Therefore, in order to improve the accuracy of network fault type diagnosis and prediction, a long short-term memory network model is introduced at each node of each decision tree. The long short-term memory network model is used to capture the long-term dependencies in the data with its powerful processing ability for time series data. Taking network traffic data as an example, the traffic fluctuates periodically with the peak and trough periods of business. The long short-term memory network model can capture the dependencies between the previous and current traffic moments, accurately judge whether the current traffic change trend is abnormal, and thus improve the accuracy of network fault type diagnosis and prediction.

[0044] Moreover, when splitting nodes, based on the learning of the features randomly selected by the long short-term memory network model for the nodes, the feature split points that can best distinguish different fault types are selected, making the decision tree - LSTM model cluster more intelligent and accurate than the ordinary decision tree model.

[0045] S104. Train each decision tree - LSTM model in the decision tree - LSTM model cluster to obtain the trained decision tree - LSTM model cluster.

[0046] Specifically, in this embodiment, through model training, the decision tree - LSTM model cluster learns the relationship between the logistics center network data and network faults, enabling the decision tree - LSTM model cluster to diagnose and predict whether there are network faults and possible network fault types based on the input network data.

[0047] S105. Invoke the trained decision tree - LSTM model cluster to perform diagnostic prediction on the currently input logistics center network data and output the diagnostic prediction result.

[0048] Specifically, in this embodiment, the trained decision tree - LSTM model cluster is used to analyze and diagnose the real-time logistics center network data, so as to accurately diagnose and predict the network fault type, provide decision support for the operation and maintenance personnel, help the operation and maintenance personnel quickly locate and solve network faults, and improve the logistics center network operation and maintenance efficiency.

[0049] This embodiment provides a method for automatic diagnosis of network faults in a logistics center. By collecting network data of the logistics center, the network data of the logistics center can be used as the basis for model training for diagnosing and predicting network fault types. At the same time, the collected network data of the logistics center covers information of multiple dimensions, which can improve the accuracy of model network fault diagnosis prediction. Secondly, the collected network data of the logistics center is preprocessed, and the network data of the logistics center is converted into a unified and standardized form, providing a high-quality data basis for subsequent data analysis and model construction training, and avoiding misleading the model to learn wrong patterns. After preprocessing the network data of the logistics center, samples are extracted from the preprocessed network data set according to the scale of the logistics center network and the complexity of the fault, forming a predetermined number of multiple sub-data sets, and a decision tree is constructed for each sub-data set, laying the foundation for the subsequent construction of multiple decision tree-LSTM models, avoiding the model from overfitting a single data mode, and improving the accuracy of model network fault diagnosis prediction; and when constructing each node classification of the decision tree, some features are randomly selected from the numerous dimensional features for splitting, which can promote the growth of the decision tree under different feature combinations, can mine the potential associations between features, enhance the generalization ability of the model, and make the model more accurate in diagnosing and predicting network fault types. After building a predetermined number of decision trees, a long short-term memory network model is introduced at each node of the decision tree, thereby building a decision tree-LSTM model cluster with multiple decision tree-LSTM models, so that the decision tree-LSTM model cluster has a strong processing capability for time series data and can capture the long-term dependencies in the logistics center network data. Moreover, based on the learning of the random selection features of the nodes by the long short-term memory network model, the feature segmentation points that can best distinguish different fault types can be selected, making the decision tree-LSTM model cluster more intelligent and accurate than the ordinary decision tree model. By calling the trained decision tree-LSTM model cluster, the real-time logistics center network data can be analyzed and diagnosed, so as to accurately diagnose and predict the type of network faults, provide decision support for operation and maintenance personnel, help operation and maintenance personnel quickly locate and solve network faults, and improve the network operation and maintenance efficiency of the logistics center.

[0050] See also Figure 2 A second embodiment of a method for automatic diagnosis of a logistics center network fault in an embodiment of the present invention includes:

[0051] S201, collecting logistics center network data;

[0052] S202, preprocessing missing values and abnormal values in the collected logistics center network data to obtain a preprocessed network data set, wherein the preprocessed network data set includes normal network data of the logistics center and faulty network data of the logistics center.

[0053] Specifically, in this embodiment, first, the logistics center network data is collected. For example, for system log data, which usually comes from logistics center servers, operating systems, databases, etc., log management tools can be used to collect the running logs of logistics center servers, operating systems, databases, etc. in real time. For network traffic data, which usually comes from devices such as core switches, routers, and firewalls, traffic data can be collected through the NetFlow / IPFIX protocol, and source IP, destination IP, port number, traffic size, etc. are recorded. For device status data, which usually comes from hardware devices such as routers, switches, and servers, device status parameters such as device temperature, fan speed, and port bandwidth utilization can be obtained through the SNMP protocol. For application logs and user behavior data, which usually come from the logistics management system, relevant data can be extracted through API interfaces or file systems. After the logistics center network data is collected, the missing values and outliers in the collected logistics center network data are preprocessed to transform the logistics center network data into a unified and standardized form. For the preprocessed network data set, including normal logistics center network data and faulty logistics center network data, by learning from the normal logistics center network data and faulty logistics center network data, the accuracy of network fault diagnosis and prediction can be improved. For example, in the operation data of logistics equipment, some devices may experience short-term performance fluctuations during specific working periods, but this is a normal phenomenon. The decision tree-LSTM model will not misjudge this normal fluctuation as a fault through learning from normal data, thus effectively reducing the false alarm rate.

[0054] Please refer to Figure 3 , the third embodiment of an automatic logistics center network fault diagnosis method in the embodiments of the present invention includes:

[0055] S301. Use a data analysis tool to check the collected logistics center network data to check whether there are missing values in each field of the logistics center network data;

[0056] S302. Based on the data characteristics of the missing values, use the mean filling method or the linear interpolation method to process the missing values;

[0057] S303. Use the three-standard-deviation principle method or a rule-based anomaly detection method to identify outliers in the logistics center network data, and correct and adjust the identified outliers.

[0058] Specifically, in this embodiment, first, use data analysis tools (such as the Pandas library in Python, R language, etc.) to import the collected logistics center network data into the analysis environment, traverse each field in the data, and for each field, check whether there are missing values (for example, in Python, missing values are usually represented by NaN; in R, they are represented by NA). Then, analyze the data characteristics of the fields with missing values, and based on the data characteristics, use the mean filling method or the linear interpolation method to process the missing values. For example, for the occasionally missing resource usage record data in the system log data, use the mean filling method to replace the missing values with the average values within a certain time window before and after this period; if it is the device status data that is missing, combine the historical operation rules of the device and the data of the same type of device during the same period, and fill it up by the linear interpolation method (for example, when the port bandwidth utilization rate of a certain switch at a certain moment is missing, refer to the data trends in the previous and next 5 minutes for reasonable interpolation). Use the three - standard - deviation principle (3σ principle) method or the rule - based anomaly detection method to identify outliers in the logistics center network data. For example, use the three - standard - deviation principle method to identify the abnormally high values in the network traffic data, and judge whether it is a real traffic peak or a data collection error. If it is an error value, smooth and correct it according to the adjacent normal data; for the abnormal values of the device temperature, first check the possibility of hardware failure. If it is not a hardware problem, then use the rule - based anomaly detection method and combine factors such as the environmental temperature and device load to adjust it within the empirical threshold range.

[0059] Please refer to Figure 4 , the fourth embodiment of an automatic logistics center network fault diagnosis method in the embodiments of the present invention includes:

[0060] S401. Divide the pre - processed network data set into a training data set and a test data set, and train each decision tree - LSTM model based on the training set data;

[0061] S402. Use the test data set to evaluate each trained decision tree - LSTM model, and adjust the decision tree - LSTM model based on the evaluation results.

[0062] Specifically, in this embodiment, the preprocessed network dataset is divided into a training dataset and a test dataset. The training dataset is used to train each decision tree-LSTM model so that the model can learn the patterns and regularities in the data. Only through sufficient training can each decision tree-LSTM model have a certain prediction ability and show certain performance in subsequent evaluations. The test dataset is used to evaluate each trained decision tree-LSTM model, and the performance of the decision tree-LSTM model is measured by calculating various evaluation metrics (such as accuracy, recall rate, mean square error, etc.). If the evaluation result is not satisfactory, it indicates that the model may not have learned accurate enough patterns during the training process, or the structure and parameter settings of the model are unreasonable, and the decision tree-LSTM model needs to be optimized and adjusted in a timely manner. By continuously optimizing the performance of the decision tree-LSTM model, the optimal model parameters and structure can be found, and the prediction accuracy and stability of the model can be improved.

[0063] Please refer to Figure 5 , the fifth embodiment of an automatic network fault diagnosis method for a logistics center in the embodiment of the present invention includes:

[0064] S501. Initialize the weight parameters of each decision tree-LSTM model;

[0065] S502. Input the training set data into the decision tree-LSTM model, and perform forward propagation calculation according to the initialized weight parameters of the current decision tree-LSTM model to obtain a predicted value;

[0066] S503. Calculate the difference between the predicted value and the true value according to the prediction target using a loss function or mean square error to obtain a loss gradient;

[0067] S504. Backpropagate the loss gradient, and optimize the weight parameters of the current decision tree-LSTM model based on the loss gradient and using an optimizer.

[0068] Specifically, in this embodiment, clarify the specific structure of each decision tree-LSTM model, including the number of neurons in the LSTM layer, the depth of the decision tree, and the number of nodes, etc., and then assign initial values to each layer of the decision tree-LSTM model to initialize the weight parameters of the decision tree-LSTM model, so that the LSTM layer can learn in the direction of effectively capturing features at the initial stage of training, thereby improving the accuracy of network fault diagnosis prediction. Moreover, the decision tree-LSTM model is close to the area of the optimal solution at the initial stage of training, thereby accelerating the convergence speed. For tasks such as network fault diagnosis prediction that require processing a large amount of data, fast convergence means that the model training can be completed in a shorter time, improving the training efficiency of the model.

[0069] After initializing the weight parameters of the decision tree-LSTM model, the training set data is converted into a format suitable for input to the decision tree-LSTM model and dimension adjustment is performed according to the requirements of the model; after the data conversion, the training set data is input into the decision tree-LSTM model, and forward propagation calculation is performed according to the structure of the model and the initialized weight parameters, passing through the LSTM layer and the decision tree part in sequence, to obtain the predicted value. Here, the predicted value can be a predicted value of network performance metrics (such as network bandwidth, latency, packet loss rate, etc.), or it can be a predicted network fault type, the probability of the fault occurring, the severity of the fault, etc.

[0070] Then, according to the nature of the prediction target, a suitable loss function or calculation method is selected to calculate the difference between the predicted value and the true value. Here, the true value can be the true network fault type label, or it can be the true predicted value of network performance metrics. For the prediction of network fault types, a cross-entropy loss function can be used and based on the predicted probability of the network fault type and the true network fault type label, the loss gradient can be calculated; for predicting network performance metrics, the mean squared error can be used to calculate the mean squared error loss between the predicted value and the true value, and the loss gradient can be calculated with the help of the automatic differentiation mechanism.

[0071] After calculating the loss gradient, the loss gradient is backpropagated, and the optimizer updates the weight parameters of the current decision tree-LSTM model according to the loss gradient, thus completing one training iteration of the decision tree-LSTM model. By continuously repeating steps S501 - S504 in this way, the decision tree-LSTM model is evaluated until it converges or reaches the preset number of training epochs. And after each update of the weight parameters of the decision tree-LSTM model, the gradient needs to be cleared to avoid the impact of gradient accumulation on the next iteration of the model.

[0072] Specifically, the Adam optimizer can be selected. In the initial stage of training the decision tree-LSTM model, the error between the predicted value obtained by the model and the true value is also relatively large, and the magnitude of the gradient is often relatively large. The Adam optimizer will automatically adjust the learning rate according to these larger gradient information, making the learning rate relatively large, so that the parameters can be updated with a relatively large step size, quickly moving in the direction of the loss function decreasing, and accelerating the convergence speed of the model in the initial stage of training, rapidly reducing the loss value; as the model training progresses, the model gradually converges, and the parameters approach the optimal value. At this time, the magnitude of the gradient will gradually become smaller. The Adam optimizer can sense this change and automatically reduce the learning rate. A smaller learning rate can make the step size of parameter update smaller, avoiding the parameters oscillating back and forth near the optimal value due to too large a step size, or even missing the optimal value, thus achieving a stable optimization process.

[0073] Please refer to Figure 6, the sixth embodiment of an automatic diagnosis method for network faults in a logistics center in the embodiments of the present invention includes:

[0074] S601. Input each data record in the test set data into the trained decision tree-LSTM model to obtain a predicted value;

[0075] S602. Compare the predicted value with the true value to determine whether the prediction is accurate, and count the total number of accurately predicted values;

[0076] S603. Calculate the accuracy rate based on the total number of accurately predicted values and the total number of data records in the test set data. If the accuracy rate is less than the preset threshold, increase the number of hidden layers of the LSTM long short-term memory network and adjust the learning rate of the decision tree-LSTM model.

[0077] Specifically, in this embodiment, each data record in the test set data, like the data records in the training set data, first undergoes format conversion processing (such as normalization processing) to ensure that the test set data meets the input requirements of the decision tree-LSTM model; then each data record after format conversion processing is input into the already trained decision tree-LSTM model. The decision tree-LSTM model performs forward propagation calculations based on the patterns and weights it has learned to obtain a predicted value. The type of the predicted value can be the same as the type of the aforementioned predicted value. Then the predicted value is compared with the true value to determine whether the prediction is accurate, and the total number of accurately predicted values is counted.

[0078] As an example, assume that there is a test set data record [0.7, 0.05, 0.6], which represents that the network bandwidth utilization rate is 70%, the packet loss rate is 5%, and the device load rate is 60%. After normalization processing, it is input into the trained decision tree-LSTM model, and the model outputs a predicted value of 0.9, that is, it is predicted that the probability of a delay fault occurring in this network state is 90%. If the true value corresponding to this test set data record is 1 (indicating that a delay fault has indeed occurred), according to the rule of predicting a fault when the probability is greater than 0.5, this test set data record is accurately predicted, and the counter is incremented by 1. Assume that there are 100 data records in the test set data. After judging each record, the final value of the counter is 80, that is, the total number of accurately predicted values is 80.

[0079] After counting the total number of accurately predicted values, use the formula: accuracy = total number of accurately predicted values / total number of data records in the test set data to calculate the accuracy of the decision tree-LSTM model in predicting the test set data. Then, compare the calculated accuracy with a preset threshold, which is a standard value set according to business requirements and experience. If the comparison shows that the accuracy is less than the preset threshold, it indicates that the performance of the decision tree-LSTM model needs to be improved. At this time, it is necessary to increase the learning ability of the decision tree-LSTM model for time series features in the data and adjust the learning rate of the decision tree-LSTM model.

[0080] Please refer to Figure 7 , the seventh embodiment of an automatic network fault diagnosis method for a logistics center in an embodiment of the present invention includes:

[0081] S701. Build a network fault diagnosis system based on a front-end and back-end separation architecture, and integrate the trained decision tree-LSTM model cluster into the network fault diagnosis system;

[0082] S702. Use the network fault diagnosis system to receive user front-end requests and receive network data in real time;

[0083] S703. The network fault diagnosis system calls the trained decision tree-LSTM model cluster to analyze and diagnose the network data received in real time, and outputs a diagnosis prediction result.

[0084] Specifically, in this embodiment, a network fault diagnosis system is built based on a front-end and back-end separation architecture. Among them, a user interaction interface is developed at the front end to facilitate operation and maintenance personnel to input network fault-related query information and view network fault diagnosis reports; the back end is docked with the existing network management system and monitoring platform of the logistics center, so that network fault data can be collected in real time from multiple data sources to ensure the timeliness of diagnosis, and the trained decision tree-LSTM model cluster is integrated at the back end. When performing network fault diagnosis, the back end calls the trained decision tree-LSTM model cluster to output network fault diagnosis results and related confidence scores, and the back end presents the diagnosis results in an intuitive and easy-to-understand chart and text form on the front-end interface, such as a fault type distribution map, a ranking of fault occurrence probabilities, repair suggestions, etc., to assist operation and maintenance personnel in quickly locating and solving network faults, improving the network operation and maintenance efficiency of the logistics center and reducing labor costs.

[0085] The above describes the automatic network fault diagnosis method for a logistics center in an embodiment of the present invention. Next, the device in the embodiment of the invention will be described. Please refer to Figure 8 , the implementation manner of the automatic network fault diagnosis device for a logistics center in an embodiment of the present invention includes:

[0086] A preprocessing module 81 for preprocessing the collected logistics center network data to obtain a preprocessed network data set. The logistics center network data includes system logs, network traffic, device status, application logs, and user behavior data;

[0087] An extraction module 82 for pre - sampling from the preprocessed network data set according to the logistics center network scale and fault complexity by using the bootstrap sampling method in advance to form multiple sub - data sets with a predetermined number. For each sub - data set, a decision tree is constructed. At each node of the decision tree, a part of the features are randomly selected from many dimensional features for splitting;

[0088] A construction module 83 for traversing each decision tree and introducing a long short - term memory network model at each node of the decision tree to construct a decision tree - LSTM model cluster. The long short - term memory network model is used to determine the best split point for a part of the features randomly selected at the current node;

[0089] A training module 84 for training each decision tree - LSTM model in the decision tree - LSTM model cluster to obtain a trained decision tree - LSTM model cluster;

[0090] An invocation module 85 for invoking the trained decision tree - LSTM model cluster to perform diagnostic prediction on the currently input logistics center network data and output a diagnostic prediction result.

[0091] In this embodiment, the preprocessing module 81 includes: a collection unit 811 for collecting logistics center network data; a preprocessing unit 812 for preprocessing the missing values and outliers in the collected logistics center network data to obtain a preprocessed network data set. The preprocessed network data set includes normal logistics center network data and faulty logistics center network data.

[0092] In this embodiment, the preprocessing unit 812 includes: an inspection subunit 8121 for inspecting the collected logistics center network data through a data analysis tool to check whether there are missing values in each field of the logistics center network data; a processing subunit 8122 for processing the missing values by using the mean filling method or the linear interpolation method based on the data characteristics of the missing values; an identification subunit 8123 for identifying outliers in the logistics center network data by using the three - standard - deviation principle method or a rule - based anomaly detection method, and correcting and adjusting the identified outliers.

[0093] In this embodiment, the training module 84 includes: a training unit 841, configured to divide the preprocessed network data set into a training data set and a test data set, and train each decision tree-LSTM model based on the training set data; an evaluation unit 842, configured to evaluate each trained decision tree-LSTM model using the test data set, and adjust the decision tree-LSTM model based on the evaluation results.

[0094] In this embodiment, the training unit 841 includes: an initialization subunit 8411, configured to initialize the weight parameters of each decision tree-LSTM model; a first input subunit 8412, configured to input the training set data into the decision tree-LSTM model, and perform forward propagation calculation according to the initialized weight parameters of the current decision tree-LSTM model to obtain a predicted value; a first calculation subunit 8413, configured to calculate the difference between the predicted value and the true value using a loss function or mean square error according to the prediction target to obtain a loss gradient; a propagation subunit 8414, configured to backpropagate the loss gradient, and optimize the weight parameters of the current decision tree-LSTM model based on the loss gradient and using an optimizer.

[0095] In this embodiment, the evaluation unit 842 includes: a second input subunit 8421, configured to input each data record in the test set data into the trained decision tree-LSTM model to obtain a predicted value; a comparison subunit 8422, configured to compare the predicted value with the true value, determine whether the prediction is accurate, and count the total number of accurately predicted values; a second calculation subunit 8423, configured to calculate an accuracy rate based on the total number of accurately predicted values and the total number of data records in the test set data. If the accuracy rate is less than a preset threshold, increase the number of hidden layers of the LSTM long short-term memory network and adjust the learning rate of the decision tree-LSTM model.

[0096] In this embodiment, the invocation module 85 includes: a construction unit 851, configured to construct a network fault diagnosis system based on a front-end and back-end separation architecture, and integrate the trained decision tree-LSTM model cluster into the network fault diagnosis system; a receiving unit 852, configured to use the network fault diagnosis system to receive user front-end requests and receive network data in real time; an invocation unit 853, configured to use the network fault diagnosis system to invoke the trained decision tree-LSTM model cluster, analyze and diagnose the network data received in real time, and output a diagnostic prediction result.

[0097] In this embodiment, by collecting the network data of the logistics center, the network data of the logistics center can be used as the basis for model training for diagnosing and predicting the type of network faults. At the same time, the collected network data of the logistics center covers information of multiple dimensions, which can improve the accuracy of the model network fault diagnosis prediction. Secondly, the collected network data of the logistics center is preprocessed to provide a high-quality data basis for subsequent data analysis and model construction training, and avoid misleading the model to learn the wrong mode. After preprocessing the network data of the logistics center, samples are extracted from the preprocessed network data set according to the scale of the logistics center network and the complexity of the fault, forming a predetermined number of multiple sub-data sets, and a decision tree is constructed for each sub-data set to avoid the model from overfitting a single data mode and improve the accuracy of the model network fault diagnosis prediction; and when constructing each node classification of the decision tree, some features are randomly selected from the numerous dimensional features for splitting, which can promote the growth of the decision tree under different feature combinations, can mine the potential associations between features, enhance the generalization ability of the model, and make the model more accurate in diagnosing and predicting the type of network faults. After building a predetermined number of decision trees, a long short-term memory network model is introduced at each node of the decision tree to build a decision tree-LSTM model cluster, so that the decision tree-LSTM model cluster has a strong processing capability for time series data and can capture the long-term dependencies in the logistics center network data. Moreover, based on the learning of the random selection features of the nodes by the long short-term memory network model, the feature segmentation points that can best distinguish different fault types can be selected, making the decision tree-LSTM model cluster more intelligent and accurate than the ordinary decision tree model. By calling the trained decision tree-LSTM model cluster, the real-time logistics center network data can be analyzed and diagnosed, so as to accurately diagnose and predict the type of network faults, provide decision support for operation and maintenance personnel, help operation and maintenance personnel quickly locate and solve network faults, and improve the network operation and maintenance efficiency of the logistics center.

[0098] Figure 8 The logistics center network fault automatic diagnosis device shown does not constitute a limitation on the logistics center network fault automatic diagnosis device, and can implement the various steps of the logistics center network fault automatic diagnosis method provided by the above-mentioned method embodiments.

[0099] above Figure 8 The automatic diagnosis device for network faults of a logistics center in an embodiment of the present invention is described in detail from the perspective of modular functional entities. The automatic diagnosis device for network faults of a logistics center in an embodiment of the present invention is described in detail from the perspective of hardware processing.

[0100] Figure 9FIG. 0 is a schematic structural diagram of an automatic diagnostic device for logistics center network failures provided by an embodiment of the present invention. The device 90 may vary greatly due to different configurations or performances, and may include one or more processors (central processing units, CPUs) 910 (for example, one or more processors) and a memory 920, and one or more storage media 930 for storing application programs 933 or data 932 (for example, one or more mass storage devices). Among them, the memory 920 and the storage media 930 may be transient storage or persistent storage. The program stored in the storage media 930 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the device 90. Further, the processor 910 may be configured to communicate with the storage media 930 and execute a series of instruction operations in the storage media on the device 90.

[0101] The device 90 may further include one or more power supplies 940, one or more wired or wireless network interfaces 950, one or more input / output interfaces 960, and / or one or more operating systems 931, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc.

[0102] An embodiment of the present invention further provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium, and when the instructions are run on a computer, the computer is caused to execute the steps of the method for automatically diagnosing logistics center network failures.

[0103] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, device, or unit can refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0104] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0105] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or equivalently replace some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An automatic diagnosis method for network faults in a logistics center, characterized in that The automatic diagnosis method for the logistics center network failure includes: Preprocessing the collected logistics center network data to obtain a preprocessed network data set. The logistics center network data includes system logs, network traffic, device status, application logs, and user behavior data; According to the logistics center network scale and the complexity of the failure, the bootstrap sampling method is used in advance to extract samples from the preprocessed network data set to form multiple sub-data sets with a predetermined number. For each sub-data set, a decision tree is constructed. At each node of the decision tree, a part of the features is randomly selected from many dimensional features for splitting; Traverse each decision tree, introduce a long short-term memory network model at each node of the decision tree, and construct a decision tree-LSTM model cluster. The long short-term memory network model is used to determine the best splitting point for a part of the features randomly selected at the current node; Train each decision tree-LSTM model in the decision tree-LSTM model cluster to obtain a trained decision tree-LSTM model cluster; Call the trained decision tree-LSTM model cluster to perform diagnostic prediction on the currently input logistics center network data and output the diagnostic prediction result.

2. The automatic diagnosis method for network faults of a logistics center according to claim 1, characterized in that The preprocessing of the collected logistics center network data includes: Collect the logistics center network data; Preprocess the missing values and outliers in the collected logistics center network data to obtain a preprocessed network data set. The preprocessed network data set includes normal logistics center network data and faulty logistics center network data.

3. The automatic diagnosis method for logistics center network faults according to claim 2, wherein The preprocessing of the missing values and outliers in the collected logistics center network data includes: Use a data analysis tool to check the collected logistics center network data to check whether there are missing values in each field of the logistics center network data; Based on the data characteristics of the missing values, use the mean filling method or the linear interpolation method to process the missing values; Use the three-standard-deviation principle method or the rule-based anomaly detection method to identify the outliers in the logistics center network data, and correct and adjust the identified outliers.

4. The automatic diagnosis method for logistics center network faults according to claim 1, characterized in that The training of each decision tree-LSTM model in the decision tree-LSTM model cluster to obtain a trained decision tree-LSTM model cluster includes: Divide the preprocessed network data set into a training data set and a test data set, and train each decision tree-LSTM model based on the training set data; Use the test data set to evaluate each trained decision tree-LSTM model, and adjust the decision tree-LSTM model based on the evaluation result.

5. The automatic diagnosis method for logistics center network faults according to claim 4, characterized in that The training of each decision tree-LSTM model based on the training set data includes: Initialize the weight parameters of each decision tree-LSTM model; Input the training set data into the decision tree-LSTM model, and perform forward propagation calculation according to the initialized weight parameters of the current decision tree-LSTM model to obtain a predicted value; Calculate the difference between the predicted value and the true value using a loss function or the mean squared error according to the prediction target to obtain a loss gradient; Backpropagate the loss gradient, and optimize the weight parameters of the current decision tree-LSTM model based on the loss gradient and using an optimizer.

6. The automatic diagnosis method for network faults of a logistics center according to claim 4, wherein Use the test data set to evaluate each trained decision tree-LSTM model, and adjust the decision tree-LSTM model based on the evaluation results: Input each data record in the test set data into the trained decision tree-LSTM model to obtain predicted values; Compare the predicted values with the true values to determine whether the prediction is accurate, and count the total number of accurately predicted values; Based on the total number of accurately predicted values and the total number of data records in the test set data, calculate the accuracy rate. If the accuracy rate is less than the preset threshold, increase the number of hidden layers of the LSTM long short-term memory network and adjust the learning rate of the decision tree-LSTM model.

7. The automatic logistics center network fault diagnosis method according to claim 1, characterized in that Call the trained decision tree-LSTM model cluster to diagnose and predict the currently input logistics center network data, and the output diagnosis and prediction results include: Construct a network fault diagnosis system based on the front-end and back-end separation architecture, and integrate the trained decision tree-LSTM model cluster into the network fault diagnosis system; Use the network fault diagnosis system to receive user front-end requests and receive network data in real time; The network fault diagnosis system calls the trained decision tree-LSTM model cluster to analyze and diagnose the network data received in real time, and outputs diagnosis and prediction results.

8. An automatic diagnosis device for network faults in a logistics center, characterized in that, Including: A preprocessing module for preprocessing the collected logistics center network data to obtain a preprocessed network data set. The logistics center network data includes: system logs, network traffic, device status, application logs, and user behavior data; An extraction module for pre-using the bootstrap sampling method to extract samples from the preprocessed network data set according to the logistics center network scale and fault complexity, to form a plurality of sub-data sets of a predetermined number. For each sub-data set, construct a decision tree, and at each node of the decision tree, randomly select some features from many dimensional features for splitting; A construction module for traversing each decision tree, introducing a long short-term memory network model at each node of the decision tree, and constructing a decision tree-LSTM model cluster. The long short-term memory network model is used to determine the best splitting point for some features randomly selected at the current node; A training module for training each decision tree-LSTM model in the decision tree-LSTM model cluster to obtain a trained decision tree-LSTM model cluster; A calling module for calling the trained decision tree-LSTM model cluster to diagnose and predict the currently input logistics center network data, and outputting diagnosis and prediction results.

9. An automatic diagnosis device for network faults in a logistics center, characterized in that, Including a memory and at least one processor, and computer-readable instructions are stored in the memory; The at least one processor calls the computer-readable instructions in the memory to execute each step of the automatic logistics center network fault diagnosis method according to any one of claims 1-7.

10. A computer-readable storage medium having computer-readable instructions stored thereon, characterized in that, When the computer-readable instructions are executed by a processor, each step of the automatic diagnosis method for the logistics center network failure as described in any one of claims 1-7 is implemented.

Images