Family customer broadband fault positioning method based on artificial neural network

Abstract

The invention particularly relates to a household customer broadband fault positioning method based on an artificial neural network. The method comprises the steps: collecting multi-modal data, preprocessing the collected data, constructing a dual-channel artificial neural network ANN model, extracting work order semantic features and time sequence abnormal features, dynamically fusing text features and numerical features, and outputting a fault classification result. Generating a fault category label according to the fusion feature, associating a historical maintenance knowledge base, and outputting a maintenance suggestion; and on the basis of the newly added work order data, updating the model parameters of the dual-channel artificial neural network ANN on line so as to adapt to the new fault type. According to the household customer broadband fault positioning method based on the artificial neural network, the broadband fault reason is predicted by fusing multi-modal data, the fault positioning efficiency and the solving efficiency are improved, the user experience is improved, and the market competitiveness of an operator is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent operation and maintenance of communication networks, and particularly relates to a method for locating broadband faults of home customers based on an artificial neural network. Background Art

[0002] Locating broadband faults of home customers is a systematic troubleshooting process, which needs to be gradually analyzed in combination with equipment status, network environment and error phenomena, and is crucial for the stable operation of home broadband services and improving customer satisfaction.

[0003] Most traditional processing methods generate work orders according to fault reports submitted by users and assign them to operation and maintenance personnel for processing. Traditional work order processing relies on the experience of operation and maintenance personnel, with a long fault location time (average processing time > 4 hours) and a misjudgment rate as high as 20%-25%.

[0004] At the same time, multi-source information such as work order text descriptions, equipment status logs, and network traffic data is not effectively integrated, resulting in insufficient data utilization, and key features are easily missed in single-dimensional analysis.

[0005] Limited by static rules, diagnostic systems based on decision trees or threshold rules are difficult to adapt to changes in complex network environments and have poor generalization ability.

[0006] Existing machine learning methods (such as SVM, random forest) have poor algorithm adaptability, low efficiency, and complex feature engineering (requiring manual definition of more than 100 feature fields) when dealing with high-dimensional heterogeneous data. Moreover, traditional methods cannot update the model dynamically and lack real-time performance, resulting in delayed identification of new types of faults (such as 5G interference, smart home device conflicts).

[0007] Statistics of a provincial operator in 2022 show that the proportion of duplicate work orders caused by incorrect fault location is 17%, and customer satisfaction has decreased by 12%.

[0008] In view of the problems of low efficiency and high misjudgment rate caused by the dependence of traditional work order processing on manual experience, the present invention proposes a method for locating broadband faults of home customers based on an artificial neural network. Summary of the Invention

[0009] In order to make up for the defects of the existing technology, the present invention provides a simple and efficient method for locating broadband faults of home customers based on an artificial neural network.

[0010] The present invention is realized by the following technical solutions:

[0011] A method for locating broadband faults of home customers based on an artificial neural network, comprising the following steps:

[0012] Step S1: Collect multimodal data, including customer trouble ticket texts, broadband device operation timing data, and network topology status information;

[0013] Step S2: Preprocess the collected data. Perform word segmentation and vectorization on the text data, and perform normalization and interpolation alignment on the timing data;

[0014] Step S3: Construct a dual-channel artificial neural network (ANN) model, including a text channel, a numerical channel, and a fusion layer;

[0015] The text channel uses Word2Vec and BiLSTM to extract the semantic features of the trouble tickets;

[0016] The numerical channel uses a 1D-CNN network to process the timing signals, extract the timing anomaly features, and capture the abnormal fluctuation patterns of the devices;

[0017] The fusion layer dynamically fuses the text features and numerical features based on the attention mechanism;

[0018] Step S4: The dual-channel artificial neural network (ANN) model outputs the fault classification results. Generate fault category labels based on the fused features, associate with the historical repair knowledge base, retrieve the top 5 historical similar trouble tickets with the highest similarity based on the Euclidean distance, output repair suggestions, and record the association relationship between the fault type and the device model and configuration parameters;

[0019] Step S5: Online update the parameters of the dual-channel artificial neural network (ANN) model based on the new trouble ticket data to adapt to the new fault types.

[0020] In the above Step S2, the preprocessing of the collected data includes:

[0021] Text cleaning, removing stop words and correcting misspelled words;

[0022] Numerical normalization, using Min-Max normalization to the [0, 1] interval;

[0023] Timing alignment, marking the abnormal data with a missing rate exceeding 20%, and linearly interpolating and complementing the missing data at a granularity of 5 minutes.

[0024] In the dual-channel artificial neural network (ANN) model, the output dimension of the BiLSTM network in the text channel is 128;

[0025] The 1D-CNN in the numerical channel contains 3 convolutional kernels, with a kernel size of 3, a stride of 1, followed by a global max pooling layer;

[0026] The fusion layer is a 3-layer fully connected network (512→256→128 nodes), and the activation function is LeakyReLU.

[0027] The calculation formula of the attention mechanism is as follows:

[0028]

[0029] where Q is the text feature vector, K is the numerical feature vector, and d k takes a value of 128, and the calculation formula of the final fused feature is as follows:

[0030] Z = α × Q + (1 - α) × K

[0031] Z is the fused feature.

[0032] In step S5, when the parameters of the dual-channel artificial neural network ANN model are incrementally updated, the parameters of the first two layers of the model are frozen, and only the fully connected layer is fine-tuned;

[0033] And the learning rate is set to 0.0001, and the single incremental training time is less than 10 minutes.

[0034] A home customer broadband fault location system based on an artificial neural network is deployed on a cloud server and connected to the operator work order management system through an API interface to implement the above method, including:

[0035] A multi-source data acquisition module for accessing text, device, and network data of the operator work order system;

[0036] A preprocessing module for performing text cleaning, numerical standardization, and time series alignment;

[0037] A dual-channel ANN analysis engine with a built-in dual-channel artificial neural network ANN model for extracting work order semantic features and time series anomaly features, and dynamically fusing text features and numerical features based on the attention mechanism to obtain fused features;

[0038] A solution matching module for generating fault category labels according to the fused features, associating with the historical maintenance knowledge base, and outputting maintenance suggestions;

[0039] A historical maintenance knowledge base for recording historical work order information, including basic information, the association relationship between fault types and device models and configuration parameters, maintenance methods, and processing results.

[0040] The dual-channel ANN analysis engine supports the version management of the dual-channel artificial neural network ANN model and saves at least 3 historical versions for rollback;

[0041] At the same time, it supports real-time inference services to ensure that the single work order processing delay is less than 0.5 seconds.

[0042] The home customer broadband fault location system based on the artificial neural network of the present invention supports processing more than 100,000 work orders per day.

[0043] The beneficial effects of the present invention are as follows: The method for locating broadband faults of home customers based on artificial neural network solves the problems that the traditional method for locating broadband faults of home customers relies on the experience of operation and maintenance personnel, has a long time for fault location and troubleshooting, and a low accuracy rate. By accessing the device status logs and network topology information data, and through the dual-channel artificial neural network ANN model to fuse multi-modal data, the cause of broadband faults is predicted, the efficiency of fault location and solution is improved, the user experience is enhanced, and the market competitiveness of the operator is strengthened. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0045] Figure 1 It is a schematic diagram of the method for locating broadband faults of home customers based on artificial neural network of the present invention.

[0046] Figure 2 It is a schematic diagram of the processing method of the dual-channel artificial neural network ANN model of the present invention.

[0047] Figure 3 It is a schematic diagram of the system for locating broadband faults of home customers based on artificial neural network of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0049] An artificial neural network (ANN) is a machine learning model that simulates the structure of biological neurons and realizes autonomous learning and decision-making of complex data features through multi-layer non-linear transformations. The advantage of the artificial neural network (ANN) lies in its end-to-end learning ability, which can automatically capture complex associations in high-dimensional data and is suitable for efficient classification and root cause analysis of multi-source heterogeneous work orders in the home broadband scenario.

[0050] The TensorFlow framework is an open-source deep learning framework developed by Google, which supports the efficient construction, training, and deployment of artificial neural network models. Its core adopts a data flow graph computing mode, with features such as distributed training, GPU acceleration, and cross-platform compatibility, making it suitable for large-scale work order data processing.

[0051] The method for locating broadband faults of home customers based on artificial neural networks includes the following steps:

[0052] Step S1: Collect multi-modal data, including customer trouble ticket texts, broadband device operation timing data, and network topology status information;

[0053] Step S2: Preprocess the collected data, perform word segmentation and vectorization on the text data, and perform normalization and interpolation alignment on the timing data;

[0054] Step S3: Build a dual-channel artificial neural network ANN model, including a text channel, a numerical channel, and a fusion layer;

[0055] The text channel uses Word2Vec and BiLSTM to extract the semantic features of the work orders;

[0056] The numerical channel uses a 1D-CNN network to process the timing signals, extract the timing anomaly features, and capture the abnormal fluctuation patterns of the devices;

[0057] The fusion layer dynamically fuses the text features and numerical features based on the attention mechanism;

[0058] Step S4: The dual-channel artificial neural network ANN model outputs the fault classification results, generates fault category labels based on the fused features, associates with the historical repair knowledge base, retrieves the top 5 historical similar work orders with the highest similarity based on the Euclidean distance, outputs repair suggestions, and records the association relationship between the fault type and the device model and configuration parameters;

[0059] Step S5: Online update the parameters of the dual-channel artificial neural network ANN model based on the new work order data to adapt to new fault types.

[0060] In the above step S2, the preprocessing of the collected data includes:

[0061] Text cleaning, removing stop words and correcting spelling mistakes;

[0062] Numerical normalization, using Min-Max normalization to the [0, 1] interval;

[0063] Timing alignment, marking abnormal data with a missing rate exceeding 20%, and linearly interpolating and complementing the missing data at a granularity of 5 minutes.

[0064] In the described dual-channel artificial neural network (ANN) model, the output dimension of the BiLSTM network in the text channel is 128.

[0065] The 1D-CNN in the numerical channel contains 3 convolutional kernels with a kernel size of 3, a stride of 1, followed by a global max pooling layer.

[0066] The fusion layer is a 3-layer fully connected network (512→256→128 nodes), and the activation function is LeakyReLU.

[0067] The calculation formula of the attention mechanism is as follows:

[0068]

[0069] where Q is the text feature vector, K is the numerical feature vector, and d k takes a value of 128. The calculation formula of the final fused feature is as follows:

[0070] Z = α × Q + (1 - α) × K

[0071] Z is the fused feature.

[0072] In step S5, when the parameters of the dual-channel artificial neural network (ANN) model are incrementally updated, the parameters of the first two layers of the model are frozen, and only the fully connected layer is fine-tuned.

[0073] And the learning rate is set to 0.0001, and the single incremental training time is less than 10 minutes.

[0074] The home customer broadband fault location system based on the artificial neural network is deployed on a cloud server and interfaces with the operator work order management system through an API interface to implement the above method, including:

[0075] A multi-source data acquisition module for accessing text, device, and network data from the operator work order system.

[0076] A preprocessing module for performing text cleaning, numerical standardization, and time series alignment.

[0077] A dual-channel ANN analysis engine with a built-in dual-channel artificial neural network (ANN) model for extracting work order semantic features and time series anomaly features, and dynamically fusing text features and numerical features based on the attention mechanism to obtain fused features.

[0078] A solution matching module for generating fault category labels based on the fused features, associating with the historical repair knowledge base, and outputting repair suggestions.

[0079] A historical repair knowledge base for recording historical work order information, including basic information, the association relationship between fault types and device models and configuration parameters, repair methods, and processing results.

[0080] The dual-channel artificial neural network (ANN) model structure includes:

[0081] Input layer: with a dimension of 256 (128 dimensions for text + 128 dimensions for numerical values);

[0082] Hidden layer: 3 fully connected layers (512 → 256 → 128 nodes), and the activation function is LeakyReLU;

[0083] Output layer: Softmax classification (fault categories: 6 categories such as line faults, equipment faults, configuration errors, external interferences, etc.).

[0084] The dual-channel ANN analysis engine supports the version management of the dual-channel artificial neural network (ANN) model, and saves at least 3 historical versions for rollback;

[0085] Meanwhile, it supports real-time inference services, ensuring that the processing delay of a single work order is less than 0.5 seconds.

[0086] This home customer broadband fault location system based on artificial neural network supports processing more than 100,000 work orders per day.

[0087] Embodiment 1

[0088] The training and verification of the dual-channel artificial neural network (ANN) model are as follows:

[0089] 1). Data preparation:

[0090] Sample source: 100,000 home broadband work orders of a certain operator from 2021 to 2023 (with fault categories labeled);

[0091] Data division: training set (70%), validation set (15%), test set (15%).

[0092] 2). Model training:

[0093] Framework: TensorFlow 2.8;

[0094] Hyperparameters: learning rate 0.001, batch size 64, number of iterations 100;

[0095] Optimizer: AdamW (weight decay 0.01).

[0096] 3). Performance testing:

[0097] Accuracy rate: 90.7% for the test set;

[0098] Response time: the inference time for a single work order is < 2 seconds.

[0099] The input work order description is "The optical modem is online; the network cannot be connected, there is no large-scale group fault, and the broadband account is normal."

[0100] Output: Classified as "aging optical modem performance" (confidence level 92.3%), it is recommended to replace the device model HGU-XGPON.

[0101] Embodiment 2

[0102] Add a new fault type of "Smart Home Wi-Fi conflict", and perform incremental updates on the model parameters of the dual-channel artificial neural network ANN model. The process is as follows:

[0103] 1). Input 500 new work orders (including cases of interference from smart speakers and cameras);

[0104] 2). Freeze the first two layers of the original model and fine-tune the last fully connected layer (learning rate 0.0001);

[0105] 3). The accuracy of the updated model in identifying new categories reaches 89.6%.

[0106] The above-described embodiments are only one of the specific implementation manners of the present invention. Any ordinary changes and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for locating broadband faults of home customers based on an artificial neural network, characterized in that: It includes the following steps: Step S1: Collect multi-modal data, including customer trouble ticket texts, broadband device operation timing data, and network topology status information; Step S2: Preprocess the collected data, perform word segmentation and vectorization on the text data, and perform normalization and interpolation alignment on the timing data; Step S3: Construct a dual-channel artificial neural network ANN model, including a text channel, a numerical channel, and a fusion layer; The text channel uses Word2Vec and BiLSTM to extract the semantic features of the work orders; The numerical channel uses a 1D-CNN network to process the timing signals, extract the timing anomaly features, and capture the abnormal fluctuation patterns of the devices; The fusion layer dynamically fuses the text features and numerical features based on the attention mechanism; Step S4: The dual-channel artificial neural network ANN model outputs the fault classification results, generates fault category labels according to the fusion features, associates with the historical maintenance knowledge base, retrieves the top 5 historical similar work orders with the highest similarity based on the Euclidean distance, outputs the maintenance suggestions, and records the association relationships between the fault types, device models, and configuration parameters; Step S5: Online update the parameters of the dual-channel artificial neural network ANN model based on the new work order data to adapt to the new fault types.

2. The method for locating broadband faults of home customers based on artificial neural network according to claim 1, wherein: In step S2, when preprocessing the collected data, it includes: Text cleaning, removing stop words and correcting spelling mistakes; Numerical normalization, using Min-Max normalization to the [0, 1] interval; Timing alignment, marking the abnormal data with a missing rate exceeding 20%, and linearly interpolating and filling the missing data at a granularity of 5 minutes.

3. The method for locating broadband faults of home customers based on artificial neural network according to claim 1, characterized in that: In the dual-channel artificial neural network ANN model, the output dimension of the BiLSTM network in the text channel is 128; The 1D-CNN in the numerical channel contains 3 convolutional kernels, the kernel size is 3, the stride is 1, and it is followed by a global max pooling layer; The fusion layer is a 3-layer fully connected network, and the activation function is LeakyReLU.

4. The method for locating broadband faults of home customers based on artificial neural network according to claim 1, characterized in that: The calculation formula of the attention mechanism is: Among them, Q is the text feature vector, K is the numerical feature vector, and d k takes the value of 128, and the calculation formula for the final fused feature is as follows: Z = α × Q + (1 - α) × K Z is the fusion feature.

5. The method for locating broadband faults of home customers based on artificial neural network according to claim 1, wherein: In step S5, when incrementally updating the parameters of the dual-channel artificial neural network ANN model, freeze the parameters of the first two layers of the model and only fine-tune the fully connected layer; And set the learning rate to 0.0001, and the single incremental training time is less than 10 minutes.

6. A home customer broadband fault location system based on an artificial neural network, characterized in that: Deployed on the cloud server, connected to the operator's work order management system through the API interface, and used to implement the above method, including: A multi-source data collection module, used to access the text, device, and network data of the operator's work order system; A preprocessing module, which performs text cleaning, numerical standardization, and timing alignment; A dual-channel ANN analysis engine, which is built with a dual-channel artificial neural network ANN model, used to extract the semantic features of work orders and timing anomaly features, and dynamically fuse the text features and numerical features based on the attention mechanism to obtain the fusion features; A solution matching module, used to generate fault category labels according to the fusion features, associate with the historical maintenance knowledge base, and output maintenance suggestions; A historical maintenance knowledge base, used to record historical work order information, including basic information, the association relationships between fault types, device models, and configuration parameters, maintenance methods, and processing results.

7. The home customer broadband fault location system based on artificial neural network according to claim 6, characterized in that: The dual-channel ANN analysis engine supports the version management of the dual-channel artificial neural network (ANN) model and saves at least 3 historical versions for rollback; At the same time, it supports real-time inference services to ensure that the processing delay of a single work order is less than 0.5 seconds.

8. The home customer broadband fault location system based on an artificial neural network according to claim 6, characterized in that: It supports processing more than 100,000 work orders per day.

Images