Federal learning-based base station fault root cause positioning method, apparatus and device, and medium

Abstract

The invention provides a base station fault root cause positioning method and device based on federated learning, equipment and a medium, and the method comprises the steps: training an initial global model of each client through employing the historical alarm data and historical fault work order data of a local base station of each client, and obtaining a model parameter corresponding to each client; encrypting the model parameter corresponding to each client; the model parameter ciphertext corresponding to each client is sent to a central server, and the central server is used for carrying out weighted aggregation on the model parameter ciphertext corresponding to each client to obtain an aggregation parameter, carrying out iterative updating on the initial global model and distributing the updated global model to each client; receiving a global model distributed by the central server; and based on the global model, carrying out root cause positioning on the current fault of the target base station. According to the method, the efficiency and accuracy of base station fault root cause positioning are improved, the application range is wide, the data barrier is broken through, and the data privacy security can be protected.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a method, device, equipment and medium for locating the root cause of a base station failure based on federated learning. Background Art

[0002] The existing communication network is getting larger and larger, and the structure is becoming more and more complex. The maintenance quality and efficiency of wireless base stations are facing severe challenges. With the continuous development and progress of artificial intelligence, many base station operation and maintenance methods based on deep learning have emerged. In the existing deep learning-based wireless base station fault root cause location solutions, model training methods mainly include two types. One type relies on data provided by a specific data provider for model training, which has a smaller scope of application; the other type relies on data provided by multiple data holders for model training, which has a wider scope of application, but there are data privacy and security risks.

[0003] How to break through data barriers and protect data privacy and security while achieving fast and accurate operation and maintenance has become an urgent problem that major operators need to solve. Summary of the invention

[0004] The present invention provides a method, device, equipment and medium for locating the root cause of a base station fault based on federated learning, so as to solve the defects of the root cause locating method of a wireless base station fault based on deep learning in the prior art, which has a small scope of application and has potential risks to data privacy and security.

[0005] In a first aspect, the present invention provides a method for locating a root cause of a base station fault based on federated learning, which is applied to each of a plurality of clients, and the method comprises: Based on the FedNLP learning framework, the initial global model of each client is trained using the historical alarm data and historical fault work order data of the local base station of each client to obtain the model parameters corresponding to each client; Encrypting the model parameters corresponding to each client to obtain the ciphertext of the model parameters corresponding to each client; Sending the ciphertext of the model parameters corresponding to each client to the central server, the central server is used to perform weighted aggregation on the ciphertext of the model parameters corresponding to each client to obtain an aggregation parameter, iteratively update the initial global model based on the aggregation parameter to obtain a global model, and distribute the global model to each client; Receiving the global model distributed by the central server; Based on the global model, the root cause of the current fault of the target base station is located.

[0006] In some embodiments, the initial global model is constructed based on an Informer model, which includes a hidden layer, a fully connected layer, and a softmax activation function.

[0007] In some embodiments, the training process of the initial global model of each client includes: Obtaining historical alarm data and historical fault work order data of the local base station of each client; Preprocessing the historical alarm data and historical fault work order data to obtain a historical alarm title field; Segmenting the historical alarm title field to obtain a historical segmentation result; Determine the fault root cause type label corresponding to the historical word segmentation result; The historical word segmentation results are used as training data, and the fault root cause type labels corresponding to the historical word segmentation results are used as training labels, to train the initial global model of each client, and to iteratively update the parameters of the initial global model of each client.

[0008] In some embodiments, the word segmentation of the historical alarm title field includes: Determine whether the historical alarm title field is a pure English field. If so, use the first word segmentation tool to segment the historical alarm title field. Otherwise, determine whether the historical alarm title field is a mixed Chinese and English field. If the historical alarm title field is a mixed Chinese and English field, use the word segmentation model to segment the historical alarm title field. If the historical alarm title field is not a mixed Chinese and English field, use the second word segmentation tool to segment the historical alarm title field.

[0009] In some embodiments, locating the root cause of the current fault of the target base station based on the global model includes: Obtain the current alarm data and current fault work order data of the target base station; Preprocessing the current alarm data and the current fault work order data to obtain a current alarm title field; Perform word segmentation on the current alarm title field to obtain a current word segmentation result; The current word segmentation result is input into the global model to obtain the current fault root cause type of the target base station output by the global model.

[0010] In a second aspect, the present invention further provides a method for locating the root cause of a base station failure based on federated learning, which is applied to a central server, and the method comprises: Receiving a ciphertext of a model parameter corresponding to each client sent by each client, wherein each client is used to execute any one of the above-described methods for locating a root cause of a base station failure based on federated learning; Performing weighted aggregation on the model parameter ciphertext corresponding to each client to obtain an aggregated parameter; Iteratively updating the initial global model based on the aggregation parameters to obtain a global model; The global model is distributed to each of the clients.

[0011] In a third aspect, a base station fault root cause location device based on federated learning is applied to each of a plurality of clients, including: A model training unit, configured to train an initial global model of each client based on a federated natural language processing (FedNLP) learning framework and using historical alarm data and historical fault work order data of a local base station of each client to obtain model parameters corresponding to each client; A parameter encryption unit, used to encrypt the model parameters corresponding to each client to obtain the ciphertext of the model parameters corresponding to each client; A first sending unit is used to send the model parameter ciphertext corresponding to each client to a central server, and the central server is used to perform weighted aggregation on the model parameter ciphertext corresponding to each client to obtain an aggregation parameter, iteratively update the initial global model based on the aggregation parameter to obtain a global model, and distribute the global model to each client; A first receiving unit, configured to receive the global model distributed by the central server; The fault root cause locating unit is used to locate the root cause of the fault of the target base station based on the global model.

[0012] In a fourth aspect, the present invention further provides a base station fault root cause location device based on federated learning, which is applied to a central server, comprising: A second receiving unit, configured to receive a ciphertext of model parameters corresponding to each client sent by each client, wherein each client is configured to execute any one of the above-described methods for locating a root cause of a base station failure based on federated learning; A parameter aggregation unit, used to perform weighted aggregation on the model parameter ciphertext corresponding to each client to obtain an aggregated parameter; An updating unit, configured to iteratively update the initial global model based on the aggregation parameters to obtain a global model; The second sending unit is used to distribute the global model to each client.

[0013] In a fifth aspect, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for locating the root cause of a base station fault based on federated learning as described in any one of the above-mentioned methods is implemented.

[0014] In a sixth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for locating the root cause of a base station fault based on federated learning as described in any one of the above.

[0015] In a seventh aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-described methods for locating the root cause of a base station fault based on federated learning.

[0016] The method, device, equipment and medium for locating the root cause of base station faults based on federated learning provided by the present invention adopt a federated natural language processing FedNLP learning framework, use the historical alarm data and historical fault work order data of the local base station of each client, train the initial global model of each client, obtain the model parameters corresponding to each client, use the federated learning framework to encrypt the model parameters corresponding to each client, send the ciphertext of the model parameters corresponding to each client to the central server, receive the updated global model distributed by the central server, and locate the root cause of the current fault of the target base station based on the global model, thereby improving the efficiency and accuracy of locating the root cause of base station faults, having a wide range of applications, breaking through data barriers, and being able to protect data privacy and security. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given to the drawings provided in the embodiments below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 is an architecture diagram of a base station fault root cause location system based on federated learning provided in an embodiment of the present invention; Figure 2 This is one of the flow charts of the method for locating the root cause of a base station failure based on federated learning provided in an embodiment of the present invention; Figure 3 It is a flowchart of a training process of a base station fault root cause classification model based on Informer provided in an embodiment of the present invention; Figure 4 is a flowchart of a training process of an initial global model of each client provided by an embodiment of the present invention; Figure 5 It is a schematic diagram of a process of segmenting a historical alarm title field provided by an embodiment of the present invention; Figure 6 This is a second flow chart of a method for locating the root cause of a base station failure based on federated learning provided in an embodiment of the present invention; Figure 7 This is one of the structural schematic diagrams of a base station fault root cause locating device based on federated learning provided in an embodiment of the present invention; Figure 8 This is a second structural diagram of a base station fault root cause location device based on federated learning provided in an embodiment of the present invention; Fig. 9 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0020] The terms "first", "second", etc. in the present invention are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present invention can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same type, and the number of objects is not limited. For example, the first object can be one or more.

[0021] Figure 1 The architecture diagram of the base station fault root cause location system based on federated learning provided in an embodiment of the present invention. Figure 1 As shown, a base station fault root cause location system based on federated learning is provided, which includes multiple clients (i.e., participants) and a central server. Different participants belong to different provinces / operators. Each parameter party stores local data and deploys a local model (i.e., a global model). The algorithm layer of the system includes multiple federated learning algorithms, such as Federated Averaging (FedAvg), Federated Optimization (FedOpt), Federated Proximal (FedProx), and Federated Nova (FedNova).

[0022] Figure 2 The present invention provides a method for locating the root cause of a base station fault based on federated learning. Figure 2 As shown in the figure, a method for locating the root cause of base station failure based on federated learning is provided, which is applied to Figure 1 For each of the multiple clients in the method, the method includes the following steps: step 210 to step 250. The method flow steps are only used as a possible implementation of the present invention.

[0023] Step 210: Based on the FedNLP learning framework, the initial global model of each client is trained using the historical alarm data and historical fault work order data of the local base station of each client to obtain the model parameters corresponding to each client.

[0024] Among them, the Federated Natural Language Processing (FedNLP) learning framework combines federated learning and natural language processing technologies, aiming to process and analyze natural language data in a distributed manner while protecting user privacy and data security. The goal of FedNLP is to achieve collaborative learning among all participants and use their own data for model training without the need to centralize the data in a central location.

[0025] Optionally, historical alarm data and historical fault work order data are processed by data cleaning, feature extraction, word segmentation, etc. to obtain historical word segmentation results, and the initial global model of each client is trained based on the historical word segmentation results and the fault root cause type labels corresponding to the historical word segmentation results.

[0026] It should be noted that each client is connected to the central server for communication, an initial global model is deployed on each client and the central server, and each client trains a local initial global model using local historical data.

[0027] It is understandable that each client performs model training locally and does not need to upload sensitive data to the central server, which can effectively protect user privacy and data security; the data of different clients is often diverse, and by training on multiple clients, the model can learn richer features and patterns, thereby improving its generalization ability in different scenarios; different clients can be trained at the same time, which improves the training efficiency of the overall model.

[0028] In some embodiments, the initial global model is built based on an Informer model, which includes a hidden layer, a fully connected layer, and a softmax activation function.

[0029] Among them, the Informer model introduces a multi-layer attention mechanism, which can better capture the long-distance dependencies in the text and automatically learn the deep feature representation of the text. It has strong generalization ability and can solve the problem of gradient disappearance or gradient explosion encountered by traditional recurrent neural networks when processing long sequences.

[0030] It should be noted that the Informer model allows the model to pay attention to only a small part of the query value / tensor when calculating the output of each position in the input sequence, and then further reduces the calculation through approximate sparsity measurement, which significantly reduces the amount of calculation and has higher efficiency when processing large-scale data sets.

[0031] Optionally, a base station fault root cause classification model is constructed based on the Informer model, and an initial global model is constructed based on the base station fault root cause classification model.

[0032] Optionally, fine-tune the Informer model by first adding a hidden layer to enable the model to generalize better to unseen data, then adding a fully connected layer on top of the model according to the number of fault categories to output the classification results of the text, and finally adding a softmax activation function after the output layer to convert the raw output of the model into a probability distribution.

[0033] Figure 3 The flowchart of the training process of the base station fault root cause classification model based on Informer provided in the embodiment of the present invention is as follows. Figure 3 As shown in FIG. 1 , the training process of the base station fault root cause classification model based on Informer includes the following steps: Read the historical alarm data and fault work order data of the base station, process the data, build the training set, validation set and test set, label the training set, and obtain the labeled training set; A Chinese-English mixed general word segmentation method is used to segment the labeled training set, validation set, and test set, obtaining the first word segmentation result of the labeled training set, the second word segmentation result of the validation set, and the third word segmentation result of the test set; Based on the first word segmentation result, the base station fault root cause classification model based on Informer is trained, and based on the second word segmentation result, the base station fault root cause classification model based on Informer is verified; Based on the third word segmentation result, the base station fault root cause classification model based on Informer is tested to obtain the final base station fault root cause classification model.

[0034] Optionally, the base station fault root cause classification model based on Informer is fine-tuned to obtain a base station fault root cause classification model.

[0035] Optionally, feature selection is performed based on the first word segmentation result to obtain a first alarm feature, and a base station fault root cause classification model is trained based on the first alarm feature.

[0036] Optionally, feature selection is performed based on the second word segmentation result to obtain a second alarm feature, and the base station fault root cause classification model is verified based on the second alarm feature.

[0037] Optionally, the third word segmentation result is input into a base station fault root cause classification model to obtain a fault root cause classification prediction result output by the base station fault root cause classification model.

[0038] Step 220: Encrypt the model parameters corresponding to each client to obtain the ciphertext of the model parameters corresponding to each client.

[0039] Optionally, the model parameters corresponding to each client are encrypted using differential privacy encryption technology to obtain the ciphertext of the model parameters corresponding to each client.

[0040] Among them, differential privacy encryption technology combines differential privacy and encryption technology to ensure that query results do not leak any information about a single user; by adding noise, the differential privacy mechanism can ensure that even if an attacker knows certain information, he cannot determine whether a specific user is in the data set.

[0041] It can be understood that by encrypting the model parameters corresponding to each client, data security is improved.

[0042] Step 230: Send the ciphertext of the model parameters corresponding to each client to the central server. The central server is used to perform weighted aggregation on the ciphertext of the model parameters corresponding to each client to obtain the aggregation parameters, iteratively update the initial global model based on the aggregation parameters to obtain the global model, and distribute the global model to each client.

[0043] Optionally, the central server is used to determine the weight of the model parameter ciphertext corresponding to each client, and perform weighted aggregation on the model parameter ciphertext corresponding to each client according to the weight of the model parameter ciphertext corresponding to each client.

[0044] Optionally, the central server is used to encrypt the aggregation parameter to obtain the aggregation parameter ciphertext, and send the aggregation parameter ciphertext to each client.

[0045] It can be understood that by sending the ciphertext of the model parameters corresponding to each client to the central server, the amount of data transmission can be reduced, the efficiency of global model training can be improved, and the security of data transmission can be ensured.

[0046] Step 240: Receive the global model distributed by the central server.

[0047] Optionally, an aggregation parameter ciphertext sent by a central server is received, and based on the aggregation parameter ciphertext, a local initial global model is updated.

[0048] Step 250: Based on the global model, locate the root cause of the current fault of the target base station.

[0049] Optionally, the current alarm data and the current fault work order data of the target base station are obtained, the current alarm data and the current fault work order data are processed to obtain the characteristic data of the target base station, the characteristic data of the target base station is input into the global model, and the fault root cause type of the target base station output by the global model is obtained.

[0050] Optionally, clients in each province can use the global model to diagnose the cause of new faults without relying on local data on any single client, thereby improving the efficiency and accuracy of base station fault diagnosis while ensuring the privacy and security of operation and maintenance data in each province.

[0051] In an embodiment of the present invention, by adopting a federated natural language processing FedNLP learning framework, the initial global model of each client is trained using the historical alarm data and historical fault work order data of the local base station of each client to obtain the model parameters corresponding to each client, and the model parameters corresponding to each client are encrypted using the federated learning framework, and the ciphertext of the model parameters corresponding to each client is sent to the central server, and the updated global model distributed by the central server is received. Based on the global model, the root cause of the current fault of the target base station is located, which improves the efficiency and accuracy of locating the root cause of the base station fault, has a wide range of applications, breaks through data barriers, and can protect data privacy and security.

[0052] Figure 4 The flowchart of the training process of the initial global model of each client provided by the embodiment of the present invention is as follows. Figure 4 As shown, in some embodiments, the training process of the initial global model of each client includes: Step 410: Obtain historical alarm data and historical fault work order data of the local base station of each client; Step 420: pre-process the historical alarm data and the historical fault work order data to obtain the historical alarm title field; Step 430: Segment the historical alarm title field to obtain a historical segmentation result; Step 440: determine the fault root cause type label corresponding to the historical word segmentation result; Step 450: Use the historical word segmentation results as training data, use the fault root cause type labels corresponding to the historical word segmentation results as training labels, train the initial global model of each client, and iteratively update the parameters of the initial global model of each client.

[0053] Optionally, the historical alarm data and historical fault work order data are preprocessed to obtain an initial current alarm title field, and the initial historical alarm title field is processed by removing blanks, deduplication, aggregation, etc. to obtain a processed historical alarm title field.

[0054] Optionally, regular matching is performed on the processed historical alarm title field to retain only Chinese characters, uppercase and lowercase English letters, and Arabic numerals to obtain the historical alarm title field.

[0055] Figure 5 The flowchart of the embodiment of the present invention provides a method for segmenting the historical alarm title field. Figure 5 As shown, in some embodiments, the historical alarm title field is segmented, including: Determine whether the historical alarm title field is a pure English field. If so, use the first word segmentation tool to segment the historical alarm title field. Otherwise, determine whether the historical alarm title field is a mixed field of Chinese and English. If the historical alarm title field is a mixed field of Chinese and English, use the word segmentation model to segment the historical alarm title field. If the historical alarm title field is not a mixed field of Chinese and English, use the second word segmentation tool to segment the historical alarm title field.

[0056] Optionally, the word segmentation model is a sequence-to-sequence (seq2seq) pre-trained model, which uses a self-attention mechanism and an encoding-decoding architecture to understand context information and generate word segmentation results.

[0057] In some embodiments, based on the global model, root cause location of the current fault of the target base station includes: Obtain the current alarm data and current fault work order data of the target base station; Preprocess the current alarm data and the current fault work order data to obtain the current alarm title field; Perform word segmentation on the current alarm title field to obtain the current word segmentation result; The current word segmentation result is input into the global model to obtain the current fault root cause type of the target base station output by the global model.

[0058] Optionally, the current alarm data and the current fault work order data are preprocessed to obtain an initial current alarm title field, and the initial current alarm title field is processed by removing blanks, de-duplication, aggregation, etc. to obtain a processed current alarm title field.

[0059] Optionally, regular expression matching is performed on the processed current alarm title field to retain only Chinese characters, uppercase and lowercase English letters, and Arabic numerals, and the types of characters to be retained are determined. According to the different types of characters, corresponding word segmentation tools or word segmentation models are used to segment them.

[0060] Figure 6 The second flowchart of the method for locating the root cause of a base station failure based on federated learning provided in an embodiment of the present invention. Figure 6 As shown, a method for locating the root cause of a base station fault based on federated learning is provided, which is applied to a central server. The method includes the following steps: steps 610 to 640. The steps of the method flow are only a possible implementation of the present invention.

[0061] Step 610: receiving a ciphertext of model parameters corresponding to each client sent by each client, and each client is used to execute any of the above-mentioned methods for locating the root cause of a base station failure based on federated learning; Step 620: Perform weighted aggregation on the model parameter ciphertext corresponding to each client to obtain an aggregated parameter; Step 630: iteratively update the initial global model based on the aggregation parameters to obtain a global model; Step 640: Distribute the global model to each client.

[0062] Optionally, the weight of the model parameter ciphertext corresponding to each client is based on the weight of the model parameter ciphertext corresponding to each client, and the model parameter ciphertext corresponding to each client is weightedly aggregated.

[0063] It can be understood that by weighted aggregation of the ciphertext of the model parameters of each client, it can be ensured that the impact of data from different clients on the global model is reasonably reflected. In particular, when there are differences in the amount and quality of data from the clients, iterative updating of the initial global model based on the aggregation parameters can accelerate the convergence of the global model, help improve the performance of the global model, make it more adaptable in more scenarios, and be suitable for large-scale distributed learning scenarios with good scalability.

[0064] The following describes a base station fault root cause locating device based on federated learning provided in an embodiment of the present invention. The base station fault root cause locating device based on federated learning described below and the base station fault root cause locating method based on federated learning described above can refer to each other.

[0065] Figure 7 One of the structural diagrams of the base station fault root cause location device based on federated learning provided in an embodiment of the present invention is as follows: Figure 7 As shown, a base station fault root cause location device based on federated learning is provided, which is applied to each client of multiple clients. The base station fault root cause location device based on federated learning 700 includes: The model training unit 710 is used to train the initial global model of each client based on the federated natural language processing FedNLP learning framework by using the historical alarm data and historical fault work order data of the local base station of each client to obtain the model parameters corresponding to each client; The parameter encryption unit 720 is used to encrypt the model parameters corresponding to each client to obtain the ciphertext of the model parameters corresponding to each client; The first sending unit 730 is used to send the model parameter ciphertext corresponding to each client to the central server, and the central server is used to perform weighted aggregation on the model parameter ciphertext corresponding to each client to obtain an aggregation parameter, iteratively update the initial global model based on the aggregation parameter to obtain a global model, and distribute the global model to each client; A first receiving unit 740 is used to receive a global model distributed by a central server; The fault root cause locating unit 750 is used to locate the root cause of the fault of the target base station based on the global model.

[0066] Optionally, the initial global model is built based on an Informer model, which includes a hidden layer, a fully connected layer, and a softmax activation function.

[0067] Optionally, the training process of the initial global model of each client includes: Obtain historical alarm data and historical fault work order data of each client's local base station; Pre-process the historical alarm data and historical fault work order data to obtain the historical alarm title field; Perform word segmentation on the historical alarm title field to obtain historical word segmentation results; Determine the fault root cause type label corresponding to the historical word segmentation result; The historical word segmentation results are used as training data, and the fault root cause type labels corresponding to the historical word segmentation results are used as training labels. The initial global model of each client is trained, and the parameters of the initial global model of each client are iteratively updated.

[0068] Optionally, perform word segmentation on the historical alarm title field, including: Determine whether the historical alarm title field is a pure English field. If so, use the first word segmentation tool to segment the historical alarm title field. Otherwise, determine whether the historical alarm title field is a mixed field of Chinese and English. If the historical alarm title field is a mixed field of Chinese and English, use the word segmentation model to segment the historical alarm title field. If the historical alarm title field is not a mixed field of Chinese and English, use the second word segmentation tool to segment the historical alarm title field.

[0069] Optionally, based on the global model, root cause location of the current fault of the target base station is performed, including: Obtain the current alarm data and current fault work order data of the target base station; Preprocess the current alarm data and the current fault work order data to obtain the current alarm title field; Perform word segmentation on the current alarm title field to obtain the current word segmentation result; The current word segmentation result is input into the global model to obtain the current fault root cause type of the target base station output by the global model.

[0070] Figure 8 The second structural diagram of the base station fault root cause location device based on federated learning provided by an embodiment of the present invention is as follows: Figure 8 As shown, a base station fault root cause location device based on federated learning is provided, which is applied to a central server. The base station fault root cause location device based on federated learning 800 includes: The second receiving unit 810 is used to receive the ciphertext of the model parameters corresponding to each client sent by each client, and each client is used to execute any of the above-mentioned base station fault root cause location methods based on federated learning; The parameter aggregation unit 820 is used to perform weighted aggregation on the model parameter ciphertext corresponding to each client to obtain an aggregated parameter; An updating unit 830, configured to iteratively update the initial global model based on the aggregation parameters to obtain a global model; The second sending unit 840 is used to distribute the global model to each client.

[0071] It should be noted here that the base station fault root cause location device based on federated learning provided in an embodiment of the present invention can implement all the method steps implemented in the above-mentioned base station fault root cause location method embodiment based on federated learning, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as those of the method embodiment will not be described in detail here.

[0072] Fig. 9 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Fig. 9As shown, the electronic device may include: a processor (processor) 910 , a communication interface (Communications Interface) 920 , a memory (memory) 930 and a communication bus 940 , wherein the processor 910 , the communication interface 920 , and the memory 930 communicate with each other through the communication bus 940 . The processor 910 can call the logic instructions in the memory 930 to execute a method for locating the root cause of a base station fault based on federated learning, which includes: based on the federated natural language processing FedNLP learning framework, using the historical alarm data and historical fault work order data of the local base station of each client, training the initial global model of each client to obtain the model parameters corresponding to each client; encrypting the model parameters corresponding to each client to obtain the model parameter ciphertext corresponding to each client; sending the model parameter ciphertext corresponding to each client to the central server, the central server is used to weightedly aggregate the model parameter ciphertext corresponding to each client to obtain the aggregation parameters, iteratively update the initial global model based on the aggregation parameters to obtain the global model, and distribute the global model to each client; receiving the global model distributed by the central server; based on the global model, root locating the current fault of the target base station.

[0073] In addition, the logic instructions in the above-mentioned memory 930 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0074] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the base station fault root cause location method based on federated learning provided by the above methods, the method including: based on the federated natural language processing FedNLP learning framework, using the historical alarm data and historical fault work order data of the local base station of each client, training the initial global model of each client to obtain the model parameters corresponding to each client; encrypting the model parameters corresponding to each client to obtain the model parameter ciphertext corresponding to each client; sending the model parameter ciphertext corresponding to each client to a central server, the central server is used to perform weighted aggregation on the model parameter ciphertext corresponding to each client to obtain the aggregation parameters, iteratively update the initial global model based on the aggregation parameters to obtain the global model, and distribute the global model to each client; receiving the global model distributed by the central server; based on the global model, root cause location of the current fault of the target base station.

[0075] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the base station fault root cause location method based on federated learning provided by the above-mentioned methods, the method comprising: based on the federated natural language processing FedNLP learning framework, using the historical alarm data and historical fault work order data of the local base station of each client, training the initial global model of each client to obtain the model parameters corresponding to each client; encrypting the model parameters corresponding to each client to obtain the model parameter ciphertext corresponding to each client; sending the model parameter ciphertext corresponding to each client to a central server, the central server is used to perform weighted aggregation on the model parameter ciphertext corresponding to each client to obtain the aggregation parameters, iteratively update the initial global model based on the aggregation parameters to obtain the global model, and distribute the global model to each client; receiving the global model distributed by the central server; based on the global model, performing root cause location of the current fault of the target base station.

[0076] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0077] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit of the technical solutions of the embodiments of the present invention.

Claims

1. A method for locating the root cause of a base station failure based on federated learning, characterized in that: Applied to each of a plurality of clients, the method comprises: Based on the FedNLP learning framework, the initial global model of each client is trained using the historical alarm data and historical fault work order data of the local base station of each client to obtain the model parameters corresponding to each client; Encrypting the model parameters corresponding to each client to obtain the ciphertext of the model parameters corresponding to each client; Sending the ciphertext of the model parameters corresponding to each client to the central server, the central server is used to perform weighted aggregation on the ciphertext of the model parameters corresponding to each client to obtain an aggregation parameter, iteratively update the initial global model based on the aggregation parameter to obtain a global model, and distribute the global model to each client; Receiving the global model distributed by the central server; Based on the global model, the root cause of the current fault of the target base station is located.

2. The method for locating the root cause of a base station failure based on federated learning according to claim 1, characterized in that: The initial global model is constructed based on an Informer model, which includes a hidden layer, a fully connected layer, and a softmax activation function.

3. The method for locating the root cause of a base station failure based on federated learning according to claim 1, characterized in that: The training process of the initial global model of each client includes: Obtaining historical alarm data and historical fault work order data of the local base station of each client; Preprocessing the historical alarm data and historical fault work order data to obtain a historical alarm title field; Segmenting the historical alarm title field to obtain a historical segmentation result; Determine the fault root cause type label corresponding to the historical word segmentation result; The historical word segmentation results are used as training data, and the fault root cause type labels corresponding to the historical word segmentation results are used as training labels, to train the initial global model of each client, and to iteratively update the parameters of the initial global model of each client.

4. The method for locating the root cause of a base station failure based on federated learning according to claim 3, characterized in that: The word segmentation of the historical alarm title field includes: Determine whether the historical alarm title field is a pure English field. If so, use the first word segmentation tool to segment the historical alarm title field. Otherwise, determine whether the historical alarm title field is a mixed Chinese and English field. If the historical alarm title field is a mixed Chinese and English field, use the word segmentation model to segment the historical alarm title field. If the historical alarm title field is not a mixed Chinese and English field, use the second word segmentation tool to segment the historical alarm title field.

5. The method for locating the root cause of a base station failure based on federated learning according to any one of claims 2 to 4, characterized in that: The root cause location of the current fault of the target base station based on the global model includes: Obtain the current alarm data and current fault work order data of the target base station; Preprocessing the current alarm data and the current fault work order data to obtain a current alarm title field; Perform word segmentation on the current alarm title field to obtain a current word segmentation result; The current word segmentation result is input into the global model to obtain the current fault root cause type of the target base station output by the global model.

6. A method for locating the root cause of a base station failure based on federated learning, characterized in that: Applied to a central server, the method comprises: Receiving a ciphertext of a model parameter corresponding to each client sent by each client, wherein each client is used to execute the method for locating a root cause of a base station failure based on federated learning according to any one of claims 1 to 5; Performing weighted aggregation on the model parameter ciphertext corresponding to each client to obtain an aggregated parameter; Iteratively updating the initial global model based on the aggregation parameters to obtain a global model; The global model is distributed to each of the clients.

7. A base station fault root cause location device based on federated learning, characterized in that: Applies to each of the multiple clients, including: A model training unit, used in a federated natural language processing FedNLP learning framework, trains an initial global model of each client using historical alarm data and historical fault work order data of a local base station of each client to obtain model parameters corresponding to each client; A parameter encryption unit, used to encrypt the model parameters corresponding to each client to obtain the ciphertext of the model parameters corresponding to each client; A first sending unit is used to send the model parameter ciphertext corresponding to each client to a central server, and the central server is used to perform weighted aggregation on the model parameter ciphertext corresponding to each client to obtain an aggregation parameter, iteratively update the initial global model based on the aggregation parameter to obtain a global model, and distribute the global model to each client; A first receiving unit, configured to receive the global model distributed by the central server; The fault root cause locating unit is used to locate the root cause of the fault of the target base station based on the global model.

8. A base station fault root cause location device based on federated learning, characterized in that: Applied to the central server, including: A second receiving unit, configured to receive a ciphertext of model parameters corresponding to each client sent by each client, wherein each client is configured to execute the method for locating a root cause of a base station fault based on federated learning according to any one of claims 1 to 5; A parameter aggregation unit, used to perform weighted aggregation on the model parameter ciphertext corresponding to each client to obtain an aggregated parameter; An updating unit, configured to iteratively update the initial global model based on the aggregation parameters to obtain a global model; The second sending unit is used to distribute the global model to each client.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for locating the root cause of a base station failure based on federated learning as described in any one of claims 1 to 6 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for locating the root cause of a base station failure based on federated learning as described in any one of claims 1 to 6 is implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for locating the root cause of a base station failure based on federated learning as described in any one of claims 1 to 6 is implemented.

Images