Base station outage alarm prediction method and device

Abstract

The application provides a base station retreat service alarm prediction method and device, the method comprises the following steps: associating each type of historical cross-professional alarm of a base station with historical retreat service alarms of the base station, adding the historical retreat service alarms associated with each type of historical cross-professional alarm to a set of each type of historical cross-professional alarm; dividing the set into multiple alarm clusters according to the occurrence time of each alarm in the set, mining frequent items from the alarm clusters, matching the frequent items with each alarm cluster to obtain a matching result corresponding to each type of historical cross-professional alarm; performing word vector coding on each type of historical cross-professional alarm to obtain embedding features of each type of historical cross-professional alarm, inputting the matching result, embedding features and pre-acquired historical running state indicators corresponding to all types of historical cross-professional alarms into a retreat prediction model, and outputting a retreat service alarm prediction result of the base station. The application improves the accuracy and efficiency of base station retreat service alarm prediction.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and in particular to a method and apparatus for predicting base station outage alarms. Background Technology

[0002] Wireless network outage alarms are important output information for base station failures. Currently, research on outage alarm prediction is limited, with most focusing on post-outage maintenance strategies and lacking predictive research. Furthermore, inventions targeting alarm prediction are concentrated in other industries such as IT and power systems, lacking applicability to the wireless communication industry.

[0003] In current practical applications, traditional rule-based or manual judgment methods are mostly used. The original base station fault alarms are numerous and varied, requiring maintenance personnel to expend considerable effort on analysis. Currently, handling outage alarms mostly adopts a reactive, reactive approach, performing maintenance only after an alarm is generated. This inevitably results in the base station or cell being out of service during the maintenance period, significantly impacting customer perception and experience, and generating numerous complaints.

[0004] Current technologies lack proactive and predictive analysis of network problems. The primary analytical basis for handling service outages is base station alarm data. Alarms are predefined system anomaly information defined by system administrators or equipment vendors. When a device system malfunctions or malfunctions, it reports corresponding alarm information to the network management system; that is, the occurrence of an alarm generally indicates a fault. Alarm data can promptly reflect the current operating status of base stations. However, the occurrence of significant service outage alarms is often accompanied by minor alarms and performance fluctuations. Existing analysis methods rely solely on experience to coarsely estimate the risk of large-scale service outages, resulting in low predictive accuracy and difficulty in providing a scientific basis for inspection and hazard identification, and targeted scheduling of maintenance resources. Summary of the Invention

[0005] This invention provides a method and apparatus for predicting base station outage alarms, which solves the shortcomings of the prior art that relies on experience to predict base station outage alarms, resulting in inaccuracy and low efficiency, and improves the accuracy and efficiency of base station outage alarm prediction.

[0006] This invention provides a method for predicting base station outage alarms, comprising:

[0007] Associate each type of historical cross-professional alarm with the historical outage alarm of the base station, and add the historical outage alarm associated with each type of historical cross-professional alarm to the set of each type of historical cross-professional alarm.

[0008] The set is divided into multiple alarm clusters based on the occurrence time of each alarm in the set. Frequent items are extracted from the alarm clusters, and the frequent items are matched with each alarm cluster to obtain the matching results corresponding to each type of historical cross-professional alarm.

[0009] Word vector encoding is performed on each type of historical cross-professional alarm to obtain the embedding features of each type of historical cross-professional alarm. The matching results, embedding features, and pre-acquired historical operating status indicators corresponding to all types of historical cross-professional alarms are input into the service outage prediction model, and the service outage alarm prediction results of the base station are output.

[0010] The service termination prediction model is trained using cross-disciplinary alarm samples and service termination alarm samples as samples, and the corresponding service termination alarm prediction results as labels.

[0011] According to the base station outage alarm prediction method provided by the present invention, the categories of historical cross-disciplinary alarms include wireless professional alarms, transmission professional alarms and power and environment professional alarms;

[0012] Accordingly, associating each type of historical cross-professional alarm of the base station with the historical outage alarm of the base station includes:

[0013] If the wireless professional alarm and the historical outage alarm belong to the same network element, then the wireless professional alarm and the historical outage alarm will be associated.

[0014] If the transmission-specific alarm and the historical outage alarm belong to the same data center, then associate the transmission-specific alarm with the historical outage alarm.

[0015] If the power and environmental alarm and the historical service outage alarm belong to the same data center, then the power and environmental alarm and the historical service outage alarm will be associated.

[0016] According to a base station outage alarm prediction method provided by the present invention, the step of dividing the set into multiple alarm clusters based on the occurrence time of each alarm in the set includes:

[0017] Based on the DBSCAN density clustering algorithm, the occurrence time of each alarm in the set is clustered.

[0018] The alarm clusters are formed based on the alarms that occur at the same time within the same cluster.

[0019] According to a base station outage alarm prediction method provided by the present invention, the step of mining frequent items from the alarm cluster includes:

[0020] The PrefixSpan algorithm is used to extract frequent items from the alarm clusters corresponding to each type of historical cross-disciplinary alarm.

[0021] According to the base station outage alarm prediction method provided by the present invention, the step of performing word vector encoding on each type of historical cross-disciplinary alarm to obtain the embedding features of each type of historical cross-disciplinary alarm includes:

[0022] Based on the FastText model, word vector encoding is performed on each type of historical cross-disciplinary alarm to obtain the embedding features of each type of historical cross-disciplinary alarm.

[0023] According to the base station outage alarm prediction method provided by the present invention, the historical operating status indicators include the base station's board temperature, standing wave ratio, optical power, clock status, CP board load, radio frequency channel, and transmission quality.

[0024] The present invention also provides a base station outage alarm prediction device, comprising:

[0025] The association module is used to associate each type of historical cross-professional alarm of the base station with the historical outage alarm of the base station, and to add the historical outage alarm associated with each type of historical cross-professional alarm to the set of each type of historical cross-professional alarm.

[0026] The mining module is used to divide the set into multiple alarm clusters according to the occurrence time of each alarm in the set, mine frequent items from the alarm clusters, match the frequent items with each alarm cluster, and obtain the matching results corresponding to each type of historical cross-professional alarm.

[0027] The prediction module is used to encode word vectors for each type of historical cross-professional alarm, obtain the embedding features of each type of historical cross-professional alarm, input the matching results, embedding features and pre-acquired historical operating status indicators corresponding to all types of historical cross-professional alarms into the service outage prediction model, and output the service outage alarm prediction result of the base station.

[0028] The service termination prediction model is trained using cross-disciplinary alarm samples and service termination alarm samples as samples, and the corresponding service termination alarm prediction results as labels.

[0029] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the base station outage alarm prediction methods described above.

[0030] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the base station outage alarm prediction method as described above.

[0031] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the base station outage alarm prediction methods described above.

[0032] The base station outage alarm prediction method and apparatus provided by this invention constructs a mapping relationship between alarms and base station outages by correlating professional alarms with base station outage faults. Based on this mapping relationship, matching features based on cross-professional alarm association rules are generated, improving the accuracy and efficiency of base station outage alarm prediction. Based on the alarm matching features combined with the performance data features of the base station and the document vector features of the alarm itself, a comprehensive prediction of base station outage warnings is made, further improving the accuracy of base station outage alarm prediction. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0034] Figure 1 This is one of the flowcharts of the base station outage alarm prediction method provided by the present invention;

[0035] Figure 2 This is the second flowchart of the base station outage alarm prediction method provided by the present invention;

[0036] Figure 3 This is a schematic diagram of the FastText model in the base station outage alarm prediction method provided by the present invention;

[0037] Figure 4 This is a schematic diagram of the document vector generation process for alarm sequences in the base station outage alarm prediction method provided by the present invention;

[0038] Figure 5 This is a schematic diagram of the base station outage alarm prediction device provided by the present invention;

[0039] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0041] The following is combined with Figure 1 The present invention describes a base station outage alarm prediction method, comprising: step 101, associating each type of historical cross-professional alarm of the base station with the historical outage alarm of the base station, and adding the historical outage alarm associated with each type of historical cross-professional alarm to the set of each type of historical cross-professional alarm;

[0042] Outage alarms are the main type of alarms triggered by sudden base station failures. Once an outage occurs, the base station or cell will inevitably be out of service during maintenance, severely impacting user experience. Accurate and timely prediction of high-risk outage sites before an outage alarm occurs, and improved efficiency in identifying potential base station problems, have become crucial for reducing base station outage duration, improving network operation quality, and providing maintenance personnel with more comprehensive inspection guidance.

[0043] This embodiment associates historical cross-disciplinary alarms of the base station with historical outage alarms of the base station. The specific association method is not limited to this embodiment. The associated historical outage alarms are then added to the corresponding category of historical cross-disciplinary alarm set.

[0044] For example, you can select and associate historical cross-disciplinary alarms and historical service outage alarms from the previous 7 days. Each type of historical cross-disciplinary alarm and historical service outage alarm includes the following fields: alarm name, occurrence time, network element name, data center name, and the specialty it belongs to.

[0045] Step 102: Divide the set into multiple alarm clusters according to the occurrence time of each alarm in the set, extract frequent items from the alarm clusters, match the frequent items with each alarm cluster, and obtain the matching results corresponding to each type of historical cross-professional alarm.

[0046] This embodiment provides the following definitions:

[0047] Definition 1: An alarm event is defined as a tuple.<X,t> Where X is the alarm name and t is a timestamp indicating the time the alarm was generated.

[0048] Definition 2: An alarm sequence is defined as a triple. <s ts te>, where s is the set of alarm events, Ts and Te represent the time periods in which this sequence occurs, and s has a time order.

[0049] Definition 3: Transaction set. If an alarm sequence is likened to a transaction, then the set of all alarm sequences is called a transaction set.

[0050] The most important metric for measuring the correlation between alarms is the length of the time interval between their occurrences and their order of occurrence. Therefore, each alarm event includes the alarm name and the time it occurred.

[0051] Since alarm events include time, an alarm sequence is defined as a set of alarm events that occur within a certain period of time. This set can be regarded as a sample obtained by formatting alarm data and then performing association rule mining. Therefore, the transaction set is the sample set.

[0052] The alarms in each alarm set are clustered according to the time of occurrence of each alarm in the alarm set, so that the alarm events that occur within a certain period of time, i.e., the time window represented by Ts and Te, are regarded as an alarm cluster and constitute an alarm sequence.

[0053] In telecommunications networks, due to variations in network load and alarm threshold settings, a large number of frequent but short-lived alarms occur. A significant portion of this alarm data is redundant and useless, requiring filtering.

[0054] Optionally, alarms in each alarm cluster that have empty or obviously erroneous key attribute fields can be filtered. If the same alarm occurs frequently in each alarm cluster, only the alarm that occurred earliest in the alarm cluster is retained, and duplicate alarms are deleted.

[0055] Alarms with very short durations often resolve automatically without intervention. These alarms are not helpful in identifying the root cause and may even be disruptive; therefore, they can usually be filtered out. The minimum duration interval for filtering can be set to 2 minutes.

[0056] Optionally, the alarm cluster closest to the current time can be selected from the alarm clusters corresponding to each set for feature extraction.

[0057] Based on the obtained cross-disciplinary alarm transaction set, we further define Definition 4: Frequent terms, for alarm sequences <s tste>In a set of alarm events s, for any subset a∈s s, each alarm event in a...<X,t> The set of alarm names X is called a frequent item. If the length of a frequent item is N, it is called an N-item frequent set.

[0058] Frequent item mining is performed on cross-disciplinary alarm transaction sets. Alarm combinations with the leading rules from the selected frequent items are matched against each alarm cluster. One-hot encoding is used to encode the matching results. If each alarm cluster includes the alarm combination, the match is successful, and the feature corresponding to that alarm cluster is set to 1; otherwise, it is set to 0, generating the matching results for each set.

[0059] If each set corresponds to m alarm clusters, then m-dimensional features are generated. If the number of alarm categories is 3, then 3*m-dimensional matching results are generated.

[0060] Step 103: Encode word vectors for each type of historical cross-professional alarm, obtain the embedding features of each type of historical cross-professional alarm, input the matching results, embedding features and pre-acquired historical operating status indicators corresponding to all types of historical cross-professional alarms into the service outage prediction model, and output the service outage alarm prediction result of the base station.

[0061] The service termination prediction model is trained using cross-disciplinary alarm samples and service termination alarm samples as samples, and the corresponding service termination alarm prediction results as labels.

[0062] All alarm sequences from the cross-disciplinary transaction set are used as an alarm prediction library. Alarms are cleaned by removing special characters and stop words, and then word vector encoding is performed on the alarms using a word vector model to obtain n-dimensional embedding features. The number of alarm categories is 3, generating 3*n-dimensional embedding features.

[0063] Based on the network element name of the base station and the same time dimension, the multi-source features extracted from the cross-professional alarm data, namely the matching results, embedded features and historical operating status indicators corresponding to the base station, are fused together.

[0064] The fusion results of multi-source features are used as input to the base station outage prediction model to predict whether the base station will trigger an alarm within a preset time period, such as 3 days.

[0065] Optionally, the service termination prediction model employs a convolutional neural network algorithm. It constructs a multi-layer convolutional neural network combined with two fully connected layers, and finally outputs the prediction result through softmax. Specifically, the convolutional neural network uses two layers of convolutional layers combined with two fully connected layers to construct a binary classifier. The loss function used is the cross-entropy loss function. The cross-entropy loss function for binary classification is as follows:

[0066]

[0067] Where n is the total number of training samples in a single batch, y i To predict the label of a sample, p i For the prediction y i The probability value. For the service termination prediction model, existing convolutional neural networks or deep convolutional neural network algorithms can be used for construction. The overall framework is as follows: Figure 2 As shown.

[0068] After training the model through multiple iterations, a model with good convergence to the samples is obtained based on its accuracy and loss function on the validation set. During prediction, the fusion result of the multi-source features of the base station is input into the convolutional neural network model to obtain the probability of outage alarms occurring within a preset time period in the future, thus obtaining the predicted probability value of the final base station failure.

[0069] This embodiment constructs a mapping relationship between alarms and base station outages by correlating professional alarms with base station outages. Based on this mapping relationship, matching features based on cross-professional alarm association rules are generated to improve the accuracy and efficiency of base station outage alarm prediction. Based on the alarm matching features combined with the base station's performance data features and the alarm's document vector features, a comprehensive prediction of base station outage warnings is made, further improving the accuracy of base station outage alarm prediction.

[0070] Based on the above embodiments, the categories of historical cross-disciplinary alarms in this embodiment include wireless professional alarms, transmission professional alarms, and power and environment professional alarms;

[0071] Optionally, historical cross-disciplinary alarms can be categorized based on their professional category, or based on both their professional category and the network element or data center to which they belong. Each type of alarm can then be treated as a separate alarm set.

[0072] When an alarm belongs to the wireless category, wireless alarms belonging to the same network element are grouped together. When an alarm belongs to the transmission or power / environment category, transmission or power / environment alarms belonging to the same equipment room are grouped together as one alarm set.

[0073] Accordingly, associating each type of historical cross-professional alarm of the base station with the historical outage alarm of the base station includes: if the wireless professional alarm and the historical outage alarm belong to the same network element, then associating the wireless professional alarm with the historical outage alarm.

[0074] Associate wireless professional alarms and historical outage alarms belonging to the same network element, and add the associated historical outage alarms to the alarm set where the wireless professional alarms are located.

[0075] If the transmission-specific alarm and the historical outage alarm belong to the same data center, then associate the transmission-specific alarm with the historical outage alarm.

[0076] Associate transmission-related alarms and historical outage alarms belonging to the same data center, and add the associated historical outage alarms to the alarm set where the transmission-related alarms are located.

[0077] If the power and environmental alarm and the historical service outage alarm belong to the same data center, then the power and environmental alarm and the historical service outage alarm will be associated.

[0078] Associate power and environmental alarms and historical service outage alarms belonging to the same data center, and add the associated historical service outage alarms to the alarm set where the power and environmental alarms are located.

[0079] The alarms in each alarm set are sorted in chronological order of occurrence to form a wireless outage alarm dataset, a transmission outage alarm dataset, and an environmental outage alarm dataset.

[0080] Based on the above embodiments, the method of dividing the set into multiple alarm clusters according to the occurrence time of each alarm in the set in this embodiment includes: clustering the occurrence time of each alarm in the set based on the DBSCAN density clustering algorithm; and forming the alarm cluster according to the alarms corresponding to the occurrence time in the same cluster.

[0081] Considering that alarms often cluster over time when faults occur, density-based clustering is used to divide time windows. DBSCAN is a density-based clustering algorithm that generally assumes that categories can be determined by the density of sample distributions. The timestamps of alarm events can be calculated using Euclidean distance to determine the density of distributions, thus dividing the data into different categories, each representing an alarm sequence.

[0082] DBSCAN describes the density of a sample set based on a set of neighborhoods. The parameters (∈∈, MinPts) describe the density of the sample distribution within the neighborhood. Here, ∈∈ describes the neighborhood distance threshold for a given sample, and MinPts describes the threshold for the number of samples in the neighborhood of a given sample with a distance of ∈∈.

[0083] For example, for 6 months of alarm data across different disciplines, by setting ∈∈ to 1 hour and MinPts to 2, we obtain the wireless outage alarm transaction set and the transmission outage alarm transaction set; by setting ∈∈ to 5 hours and MinPts to 2, we obtain the power environment outage alarm transaction set. These cross-discipline transaction sets are then used for association rule mining. An example of real-time cross-discipline associated alarm clusters is shown in Table 1.

[0084] Table 1 Examples of Real-Time Cross-Disciplinary Related Alarm Clusters

[0085]

[0086] Based on the above embodiments, the method of mining frequent items from the alarm clusters in this embodiment includes: mining frequent items from the alarm clusters corresponding to each type of historical cross-disciplinary alarms based on the PrefixSpan algorithm.

[0087] By dividing frequent items into preceding and following items, rule preceding items and rule following items are formed, and the correlation between alarms is measured by support, confidence and lift.

[0088] Support represents the probability that frequent sequences of consecutive terms in a rule both appear in the same transaction set. The formula is:

[0089] Support(X→Y)=P(X,Y) / P(I)=num(X∩Y) / num(I)

[0090] Where num(I) represents the number of transactions in the transaction set, X represents the preceding term of the rule, Y represents the following term of the rule, and num(X∩Y) represents the number of times the preceding and following terms of the rule appear simultaneously in the transaction set.

[0091] Confidence level represents the probability that, given the occurrence of the precondition X, the association rule X→Y implies Y. It indicates the likelihood that Y will also occur in an itemset where X occurs; that is, the proportion of instances where X and Y occur simultaneously to instances where only X occurs. The formula is:

[0092] Confidence(X→Y)=P(X,Y) / P(X)=num(X∩Y) / num(X);

[0093] Lift represents the ratio of the probability of Y occurring given X to the probability of Y occurring alone. Lift reflects the correlation between X and Y in an association rule. A lift > 1 and higher values ​​indicate a stronger positive correlation, while a lift < 1 and lower values ​​indicate a stronger negative correlation. A lift = 1 indicates no correlation, meaning they are independent. The formula is as follows:

[0094] Lift(X→Y)=P(Y|X) / P(Y).

[0095] Frequent items of strong rules are filtered using support, confidence, and lift. Optionally, the minimum support is set to 100, the minimum confidence to 0.7, and the minimum lift to 1. Strong rules indicate a strong correlation between the alarm combination pattern of the preceding item and the outage alarm; therefore, the alarm combination pattern of the preceding item of a strong rule can be extracted as a feature. Examples of the filtered frequent items are shown in Table 2.

[0096] Table 2 Examples of Frequent Items

[0097]

[0098]

[0099] Based on the above embodiments, the step of performing word vector encoding on each type of historical cross-disciplinary alarm to obtain the embedding features of each type of historical cross-disciplinary alarm in this embodiment includes: performing word vector encoding on each type of historical cross-disciplinary alarm based on the FastText model to obtain the embedding features of each type of historical cross-disciplinary alarm.

[0100] The architecture of FastText is similar to that of CBOW in word2vec. The CBOW architecture takes 2d context alarms as input to alarm w(t), and outputs alarm w(t) after passing through a hidden layer. The structure of the FastText model is as follows: Figure 3 As shown.

[0101] FastText doesn't simply input alarms into the neural network; it also statistically analyzes their character-level n-gram information. For example, the character-level 3-gram for an alarm about exceeding the optical power limit is <input light, input light power, optical power, power exceeding limit, rate exceeding limit>. A set G of alarm n-grams consists of the alarm and its n-gram. Word embeddings are performed on all alarms and their character-level n-grams based on the alarm corpus, resulting in alarm word vectors and alarm character-level n-gram word vectors. By adding n-grams, character-level semantic information can be learned, and better representations can be learned for words that appear less frequently.

[0102] When FastText is used for classification tasks, it averages the word vectors and character-level n-gram word vectors of the trained document to obtain the document vector. For an alarm sequence, it averages the word vectors of all alarms and their character-level n-grams to obtain the document vector of the alarm sequence, which serves as the embedding feature of the alarm sequence. Figure 4 As shown, for an alarm sequence, the alarm word vectors or alarm n-gram word vectors are represented as c1, c2, c3, and c4. The word vectors are summed and averaged to obtain the embedding representation of the alarm sequence.

[0103] Based on the above embodiments, the historical operating status indicators in this embodiment include the base station's board temperature, standing wave ratio, optical power, clock status, CP (Cyclic Prefix) board load, radio frequency channel, and transmission quality.

[0104] Using MML (Man-Machine Language) commands, the OMC (Operation and Maintenance Center) command platform is used to periodically obtain historical operational status indicator data derived from the daily operation of the base station.

[0105] The historical operational status index data of the base station at the hourly granularity are correlated and aggregated. After further data cleaning, standardization, and normalization, the data is used as the historical operational status index data of the base station at a certain moment.

[0106] There are a total of k historical operational status indicators for the base station. The preprocessed data of each historical operational status indicator is used as the feature indicator of that indicator. Therefore, the historical operational status indicators of the base station have a total of k dimensions.

[0107] The base station outage alarm prediction device provided by the present invention is described below. The base station outage alarm prediction device described below can be referred to in correspondence with the base station outage alarm prediction method described above.

[0108] like Figure 5 As shown, the device includes an association module 501, a mining module 502, and a prediction module 503, wherein:

[0109] The association module 501 is used to associate each type of historical cross-professional alarm of the base station with the historical outage alarm of the base station, and to add the historical outage alarm associated with each type of historical cross-professional alarm to the set of each type of historical cross-professional alarm.

[0110] The mining module 502 is used to divide the set into multiple alarm clusters according to the occurrence time of each alarm in the set, mine frequent items from the alarm clusters, match the frequent items with each alarm cluster, and obtain the matching results corresponding to each type of historical cross-professional alarm.

[0111] The prediction module 503 is used to perform word vector encoding on each type of historical cross-professional alarm, obtain the embedding features of each type of historical cross-professional alarm, input the matching results, embedding features and pre-acquired historical operating status indicators corresponding to all types of historical cross-professional alarms into the service outage prediction model, and output the service outage alarm prediction result of the base station.

[0112] The service termination prediction model is trained using cross-disciplinary alarm samples and service termination alarm samples as samples, and the corresponding service termination alarm prediction results as labels.

[0113] This embodiment constructs a mapping relationship between alarms and base station outages by correlating professional alarms with base station outages. Based on this mapping relationship, matching features based on cross-professional alarm association rules are generated to improve the accuracy and efficiency of base station outage alarm prediction. Based on the alarm matching features combined with the base station's performance data features and the alarm's document vector features, a comprehensive prediction of base station outage warnings is made, further improving the accuracy of base station outage alarm prediction.

[0114] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 can call logic instructions in the memory 630 to execute a base station outage alarm prediction method. This method includes: associating each type of historical cross-disciplinary alarm with the historical outage alarms of the base station; adding the historical outage alarms associated with each type of historical cross-disciplinary alarm to a set of each type of historical cross-disciplinary alarm; dividing the set into multiple alarm clusters according to the occurrence time of each alarm in the set; mining frequent items from the alarm clusters; matching the frequent items with each alarm cluster to obtain the matching result corresponding to each type of historical cross-disciplinary alarm; performing word vector encoding on each type of historical cross-disciplinary alarm to obtain the embedding features of each type of historical cross-disciplinary alarm; inputting the matching results, embedding features, and pre-acquired historical operating status indicators corresponding to all types of historical cross-disciplinary alarms into an outage prediction model; and outputting the base station outage alarm prediction result.

[0115] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0116] On the other hand, the present invention also provides a computer program product, which includes a computer program that 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 outage alarm prediction method provided by the above methods. The method includes: associating each type of historical cross-professional alarm of the base station with the historical outage alarm of the base station; adding the historical outage alarms associated with each type of historical cross-professional alarm to a set of each type of historical cross-professional alarm; dividing the set into multiple alarm clusters according to the occurrence time of each alarm in the set; mining frequent items from the alarm clusters; matching the frequent items with each alarm cluster to obtain the matching result corresponding to each type of historical cross-professional alarm; performing word vector encoding on each type of historical cross-professional alarm to obtain the embedding features of each type of historical cross-professional alarm; inputting the matching results, embedding features, and pre-acquired historical operating status indicators corresponding to all types of historical cross-professional alarms into the outage prediction model; and outputting the outage alarm prediction result of the base station.

[0117] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the base station outage alarm prediction method provided by the above methods. The method includes: associating each type of historical cross-professional alarm of the base station with the historical outage alarm of the base station; adding the historical outage alarms associated with each type of historical cross-professional alarm to a set of each type of historical cross-professional alarm; dividing the set into multiple alarm clusters according to the occurrence time of each alarm in the set; mining frequent items from the alarm clusters; matching the frequent items with each alarm cluster to obtain the matching result corresponding to each type of historical cross-professional alarm; performing word vector encoding on each type of historical cross-professional alarm to obtain the embedding features of each type of historical cross-professional alarm; inputting the matching results, embedding features, and pre-acquired historical operating status indicators corresponding to all types of historical cross-professional alarms into an outage prediction model; and outputting the outage alarm prediction result of the base station.

[0118] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0119] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.< / s> < / s>

Claims

1. A method for predicting base station outage alarms, characterized in that, include: Associate each type of historical cross-professional alarm with the historical outage alarm of the base station, and add the historical outage alarm associated with each type of historical cross-professional alarm to the set of each type of historical cross-professional alarm. The set is divided into multiple alarm clusters based on the occurrence time of each alarm in the set. Frequent items are extracted from the alarm clusters. The alarm combination of the rule preceding item in the frequent items is matched with each alarm cluster. If the alarm cluster includes the alarm combination, the match is successful. The matching result corresponding to each type of historical cross-professional alarm is obtained. Word vector encoding is performed on each type of historical cross-professional alarm to obtain the embedding features of each type of historical cross-professional alarm. The matching results, embedding features and pre-acquired historical operating status indicators corresponding to all types of historical cross-professional alarms are fused together, and the fusion result is input into the service outage prediction model to output the service outage alarm prediction result of the base station. The service termination prediction model is trained using cross-disciplinary alarm samples and service termination alarm samples as samples, and the corresponding service termination alarm prediction results as labels.

2. The base station outage alarm prediction method according to claim 1, characterized in that, The categories of historical cross-disciplinary alarms include wireless professional alarms, transmission professional alarms, and power and environmental professional alarms; Accordingly, associating each type of historical cross-professional alarm of the base station with the historical outage alarm of the base station includes: If the wireless professional alarm and the historical outage alarm belong to the same network element, then the wireless professional alarm and the historical outage alarm will be associated. If the transmission-specific alarm and the historical outage alarm belong to the same data center, then associate the transmission-specific alarm with the historical outage alarm. If the power and environmental alarm and the historical service outage alarm belong to the same data center, then the power and environmental alarm and the historical service outage alarm will be associated.

3. The base station outage alarm prediction method according to claim 1, characterized in that, The step of dividing the set into multiple alarm clusters based on the occurrence time of each alarm in the set includes: Based on the DBSCAN density clustering algorithm, the occurrence time of each alarm in the set is clustered. The alarm clusters are formed based on the alarms that occur at the same time within the same cluster.

4. The base station outage alarm prediction method according to claim 1, characterized in that, The process of mining frequent items from the alarm cluster includes: The PrefixSpan algorithm is used to extract frequent items from the alarm clusters corresponding to each type of historical cross-disciplinary alarm.

5. The base station outage alarm prediction method according to any one of claims 1-4, characterized in that, The step of performing word vector encoding on each type of historical cross-disciplinary alarm to obtain the embedding features of each type of historical cross-disciplinary alarm includes: Based on the FastText model, word vector encoding is performed on each type of historical cross-disciplinary alarm to obtain the embedding features of each type of historical cross-disciplinary alarm.

6. The base station outage alarm prediction method according to any one of claims 1-4, characterized in that, The historical operating status indicators include the base station's board temperature, standing wave ratio, optical power, clock status, CP board load, radio frequency channel, and transmission quality.

7. A base station outage alarm prediction device, characterized in that, include: The association module is used to associate each type of historical cross-professional alarm of the base station with the historical outage alarm of the base station, and to add the historical outage alarm associated with each type of historical cross-professional alarm to the set of each type of historical cross-professional alarm. The mining module is used to divide the set into multiple alarm clusters according to the occurrence time of each alarm in the set, mine frequent items from the alarm clusters, match the alarm combination of the rule preceding item in the frequent items with each alarm cluster, and if the alarm cluster includes the alarm combination, the match is successful, and obtain the matching result corresponding to each type of historical cross-professional alarm. The prediction module is used to encode word vectors for each type of historical cross-professional alarm, obtain the embedding features of each type of historical cross-professional alarm, fuse the matching results, embedding features and pre-acquired historical operating status indicators corresponding to all types of historical cross-professional alarms, and input the fusion result into the service outage prediction model to output the service outage alarm prediction result of the base station. The service termination prediction model is trained using cross-disciplinary alarm samples and service termination alarm samples as samples, and the corresponding service termination alarm prediction results as labels.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the base station outage alarm prediction method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the base station outage alarm prediction method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the base station outage alarm prediction method as described in any one of claims 1 to 6.

Images