Electronic information data processing method based on big data
By constructing a multi-dimensional data acquisition system and a long short-term memory network, the problems of one-sidedness and prediction bias in existing communication quality assessments are solved, enabling real-time and accurate communication quality assessment and prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-08
- Publication Date
- 2026-03-10
AI Technical Summary
Existing technologies struggle to achieve real-time, accurate, and intelligent analysis when processing massive amounts of heterogeneous data. Furthermore, communication quality assessments often overlook key factors, leading to biased assessment results, delayed fault response, and significant prediction errors.
By constructing a multi-dimensional data acquisition system that integrates wireless air interface and geographic information, after data preprocessing, high-dimensional features are extracted from four dimensions: space, time, business and environment. Key features are screened using a hybrid correlation analysis method, and real-time anomaly identification and prediction are performed by combining long short-term memory networks, thereby constructing a communication quality assessment and prediction model.
It enables accurate assessment and real-time prediction of communication quality, dynamically identifies abnormal patterns, provides a basis for rapid troubleshooting, and improves predictive adaptability and assessment accuracy.
Smart Images
Figure CN121644320A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of communication engineering technology, and more specifically to a method for processing electronic information data based on big data. Background Technology
[0002] In the process of electronic information data processing, traditional technologies are limited by hardware performance and algorithm complexity, and mostly rely on single-machine processing mode. While they can cope with data of limited volume and single type, they struggle to handle massive heterogeneous data and have significant shortcomings in real-time performance, accuracy, and intelligent analysis. However, with the advent of the big data era, existing technologies, relying on the powerful computing power provided by distributed computing and cloud computing, combined with advanced algorithms such as machine learning and deep learning, can not only efficiently store and process massive multi-source data, but also deeply mine data value, realize intelligent assessment of communication quality, trend prediction, and accurate identification of anomalies, and provide strong support for decision-making in the field of electronic information.
[0003] However, existing technologies still have the following drawbacks:
[0004] Firstly, existing technologies typically rely solely on wireless signal strength (such as RSRP) or a single network metric (such as throughput) to assess communication quality, ignoring key factors such as spatial location, user behavior, and environmental interference, leading to one-sided assessment results.
[0005] Secondly, existing technologies mostly rely on threshold comparisons or offline analysis, which cannot capture dynamically changing abnormal patterns in real time, resulting in delayed fault response.
[0006] Third, existing predictive models (such as ARIMA) rely solely on static trends in historical data, ignoring short-term dynamic changes, resulting in large prediction biases. Summary of the Invention
[0007] In order to overcome the above-mentioned defects of the prior art, the present invention provides an electronic information data processing method based on big data to solve the problems existing in the background art.
[0008] This invention provides the following technical solution: a method for processing electronic information data based on big data, comprising:
[0009] S1: Acquire multi-source electronic information datasets and preprocess the acquired multi-source electronic information datasets;
[0010] S2: Extract features from the preprocessed multi-source electronic information dataset, perform correlation analysis on the extracted feature parameters, select feature parameters related to communication quality, and generate information communication feature vectors.
[0011] S3: Based on the generated information communication feature vector, perform a comprehensive analysis of the current communication operation status to identify whether there are any abnormal patterns in the communication operation status;
[0012] S4: Based on the analysis results of the current communication operation status, analyze and evaluate the communication quality within the preset time window to obtain the communication quality evaluation index, and evaluate the degree of excellence or inferiority of the current communication quality.
[0013] S5: Combine the current communication quality assessment index with the historical communication quality assessment index sequence recorded in the preset database to analyze the trend of communication quality change and calculate the communication quality change rate.
[0014] S6: Combine the current communication quality assessment index, the historical communication quality assessment index sequence, and the calculated communication quality change rate to construct a communication quality prediction model. Based on the output of the communication quality prediction model, predict the communication quality for a future period of time.
[0015] Preferably, step S1 involves constructing a multi-dimensional data acquisition system that integrates wireless air interface and geographic information. This multi-dimensional data acquisition system collects user-level measurement report data from the wireless transmission network in real time through a signaling monitoring interface, including reference signal received power, signal-to-interference-plus-noise ratio, cell ID, and timestamp. Simultaneously, it obtains geographic fence and building vector data for the corresponding area through a GIS interface to obtain geographic information data. Finally, the collected user-level measurement report data and geographic information data are correlated to form a multi-source electronic information dataset that precisely binds wireless signal strength to spatial location.
[0016] Preprocessing operations are performed on the collected multi-source electronic information datasets, specifically including: data cleaning, data aggregation, data interpolation, and data normalization.
[0017] Preferably, step S2, based on the preprocessed multi-source electronic information dataset, performs composite feature extraction from four dimensions: space, time, service, and environment, constructing a high-dimensional feature parameter space. Specifically, the spatial dimension extracts the mean, variance, and coverage continuity index of signal strength within the grid; the temporal dimension extracts the hourly periodic components and trends of signal strength; the service dimension extracts user density and traffic load within the grid; and the environmental dimension extracts coding features such as building height, type, and whether it is an indoor scene. Subsequently, a hybrid correlation analysis method is employed, which integrates the maximum information coefficient and mutual information. The nonlinear correlation between each of the aforementioned high-dimensional feature parameters and the received power of the reference signal, serving as a communication quality benchmark, is calculated to obtain a first importance score. Simultaneously, the mutual information between each pair of feature parameters is calculated to quantify feature redundancy.
[0018] Finally, a dual screening criterion is set by combining the first importance score and feature redundancy. First, the first feature subset with a first importance score below a preset threshold is eliminated. Among the remaining features, a redundancy graph is constructed with each feature as a node and the mutual information value between features as the edge weight. Community detection clustering is performed on this redundancy graph, and features with mutual information values greater than the preset redundancy threshold are divided into the same cluster. Finally, within each cluster, only the feature with the highest first importance score is retained, and other features in the cluster are eliminated. All retained key feature parameters are vectorized and encoded to finally generate information communication feature vectors.
[0019] Preferably, in step S3, the real-time generated information communication feature vector is input into a spatiotemporal state prediction model constructed based on a long short-term memory network. This spatiotemporal state prediction model learns the time series dependence of historical feature vectors and outputs a predicted information communication feature vector for the communication operation state at the next moment. The multidimensional cosine similarity between the predicted information communication feature vector and the actual information communication feature vector at the current moment is calculated, and 1 is subtracted from the multidimensional cosine similarity value to obtain the state deviation, which serves as a quantitative indicator to measure the degree of state deviation.
[0020] The state deviation is compared with a preset deviation threshold, which is dynamically calculated based on historical data distribution. If the state deviation is greater than the preset deviation threshold, it is determined that there is an abnormal mode in the current communication operation state. The key feature parameters and abnormal type that caused the abnormal mode are further identified, and then an abnormal alarm is triggered. The analysis results containing the key feature parameters and abnormal type are used as input to execute the communication quality assessment process. If the state deviation is less than or equal to the preset deviation threshold, the current communication operation state is monitored and analyzed.
[0021] Preferably, step S4, based on the identification result of the abnormal mode in the current communication operation state, defines a preset time window centered on the time when the abnormal mode occurs, and extracts the information communication feature vectors for all times within the window; for the key feature parameters that cause the abnormality, calculate the statistical feature values of the key feature parameters within the entire time window, including the mean, peak value, and standard deviation; according to the abnormality type identified in S3, call a preset evaluation rule, which is used to assign corresponding weight coefficients to different statistical feature values; after normalizing each calculated statistical feature value, multiply it by its corresponding weight coefficient and sum it to obtain the communication quality evaluation index. The lower the communication quality evaluation index value, the more severe the degradation of communication quality within the time window.
[0022] Preferably, step S5 accesses a preset database, which stores historical communication quality assessment indices in chronological order and records the anomaly type and occurrence time corresponding to each index.
[0023] The system retrieves a historical communication quality assessment index sequence with the same anomaly type as the current one from a pre-defined database and uses it as the baseline data for the current analysis. The currently calculated communication quality assessment index is then appended to the end of this historical sequence to form an updated time series. Based on this time series, the system calculates the difference between the current communication quality assessment index and the previous historical communication quality assessment index that is immediately adjacent to the current moment in the time series, and divides this difference by the time interval between the two to obtain the communication quality change rate.
[0024] Preferably, step S6 fuses the current communication quality assessment index, the historical communication quality assessment index sequence, and the communication quality change rate to construct a multi-dimensional input feature vector. Then, the input feature vector is input into a pre-trained communication quality prediction model. This communication quality prediction model learns the time dependence and change patterns in historical data and combines them with the short-term dynamic trends reflected by the current state and the change rate to finally output the communication quality prediction index sequence for future time points.
[0025] The technical effects and advantages of this invention are as follows:
[0026] (1) By integrating wireless air interface data and geographic information, a dataset with precise “signal-space” binding is constructed, and high-dimensional features are extracted from four dimensions: space, time, service and environment to comprehensively characterize the factors affecting communication quality. By combining the maximum information coefficient and mutual information, features that are strongly correlated with communication quality and have low redundancy are selected to avoid information redundancy and improve the effectiveness of features.
[0027] (2) By learning the time dependence of historical feature vectors based on LSTM network, the communication status at the next moment can be predicted in real time, realizing the "prediction-comparison-early warning" closed loop; by calculating the similarity between the predicted vector and the actual vector, the deviation threshold can be dynamically set to accurately identify abnormal patterns, and when an abnormality is triggered, the feature parameters (such as a sudden increase in user density in a grid) and type (such as interference anomaly) that cause the abnormality can be automatically located, providing a basis for rapid troubleshooting.
[0028] (3) By defining a time window centered on the abnormal moment, and combining key feature statistics (mean, peak) with weight rules, an evaluation index reflecting the degree of short-term degradation is generated; by calculating the difference and time interval between the current evaluation index and the historical sequence, the communication quality change rate is dynamically updated to capture the trend turning point; the current index, historical sequence and change rate are integrated into an input vector, and the time dependence and short-term dynamics are learned by the pre-trained model to output the future prediction sequence and improve the prediction adaptability. Attached Figure Description
[0029] Figure 1 This is a diagram illustrating the method steps of the present invention. Detailed Implementation
[0030] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The electronic information data processing method based on big data involved in the present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0031] like Figure 1 The embodiment shown provides a method for electronic information data processing based on big data, including:
[0032] S1: Obtain multi-source electronic information datasets and preprocess the obtained multi-source electronic information datasets.
[0033] In this embodiment, step S1 constructs a multi-dimensional data acquisition system that integrates wireless air interface and geographic information. The multi-dimensional data acquisition system collects user-level measurement report data of the wireless transmission network in real time through the signaling monitoring interface, including reference signal received power, signal-to-interference-plus-noise ratio, cell ID, and timestamp. At the same time, it obtains geofence and building vector data of the corresponding area through the GIS interface to obtain geographic information data. Finally, the collected user-level measurement report data and geographic information data are associated to form a multi-source electronic information dataset that accurately binds wireless signal strength and spatial location.
[0034] Preprocessing operations are performed on the collected multi-source electronic information datasets, specifically including: data cleaning, data aggregation, data interpolation, and data normalization.
[0035] Specifically, the data cleaning operations include: time alignment of multi-source data based on timestamps, and identification and removal of signal strength outliers caused by terminal anomalies or measurement errors using an algorithm based on local outlier factors; the data aggregation operations include: spatial rasterization aggregation of massive user-level measurement report data using the Thiessen polygon algorithm, aggregating multiple signal data points within the same raster into a representative value for that raster to eliminate data redundancy and improve computational efficiency; the data interpolation operations include: for raster data missing due to insufficient sampling after aggregation, using Kriging interpolation, combined with signal values from neighboring rasteres, to perform spatial interpolation to smooth and fill data in signal coverage blind spots; and the data normalization operations include: Z-score standardization of parameters with different dimensions such as signal strength and signal quality after processing.
[0036] S2: Extract features from the preprocessed multi-source electronic information dataset, perform correlation analysis on the extracted feature parameters, select feature parameters related to communication quality, and generate information communication feature vectors.
[0037] In this embodiment, S2, based on the preprocessed multi-source electronic information dataset, performs composite feature extraction from four dimensions: space, time, service, and environment, constructing a high-dimensional feature parameter space. Specifically, the spatial dimension extracts the mean, variance, and coverage continuity index of signal strength within the grid; the temporal dimension extracts the hourly periodic components and trends of signal strength; the service dimension extracts user density and traffic load within the grid; and the environmental dimension extracts coding features such as building height, type, and whether it is an indoor scene. Subsequently, a hybrid correlation analysis method is employed, which integrates the maximum information coefficient and mutual information. The nonlinear correlation between each of the above high-dimensional feature parameters and the received power of the reference signal, serving as a communication quality benchmark, is calculated to obtain a first importance score. Simultaneously, the mutual information between each pair of feature parameters is calculated to quantify feature redundancy.
[0038] Finally, a dual screening criterion is set by combining the first importance score and feature redundancy. First, the first feature subset with a first importance score below a preset threshold is eliminated. Among the remaining features, a redundancy graph is constructed with each feature as a node and the mutual information value between features as the edge weight. Community detection clustering is performed on this redundancy graph, and features with mutual information values greater than the preset redundancy threshold are divided into the same cluster. Finally, within each cluster, only the feature with the highest first importance score is retained, and other features in the cluster are eliminated. All retained key feature parameters are vectorized and encoded to finally generate information communication feature vectors.
[0039] It's important to clarify that the hybrid correlation analysis method doesn't simply use two indicators side-by-side; instead, it employs a step-by-step quantification strategy. First, using the received power of the reference signal as the quantification benchmark for communication quality, the nonlinear correlation strength between each high-dimensional feature parameter and the RSRP is calculated using the maximum information coefficient (MIC), resulting in a first importance score between 0 and 1. This score directly reflects the predictive value of the feature for communication quality. Second, to address redundancy among features, the mutual information (MI) values between all pairwise feature parameters are calculated. These values are also normalized to... The interval is used to quantify the degree to which the information contained in one feature can be explained by another feature; a larger value indicates higher redundancy. Secondly, the specific execution process of the dual screening criterion is as follows: First, an importance threshold is set, and features with a first importance score below this threshold (such as features with weak impact on RSRP, like building height variance) are directly eliminated, forming a candidate feature set. Next, based on the candidate feature set, an undirected graph is constructed, where each node represents a feature, and the weight of the edge connecting two nodes is their mutual information value. Subsequently, a community detection algorithm (such as the Louvain algorithm) is applied to cluster the graph, setting a redundancy threshold (e.g., 0.7). When the mutual information value between two features is greater than this threshold, they are highly likely to be assigned to the same community (cluster), indicating that they carry highly overlapping information. Finally, within each identified cluster, only the feature with the highest first importance score is retained as the representative of that cluster, while all other redundant features within the cluster are eliminated. Through this series of specific operations, the final set of features is highly correlated with communication quality and has low redundancy. By vectorizing and encoding these key feature parameters in a predetermined order, a low-dimensional and efficient information communication feature vector can be generated.
[0040] S3: Based on the generated information communication feature vector, perform a comprehensive analysis of the current communication operation status to identify whether there are any abnormal modes in the communication operation status.
[0041] In this embodiment, S3 inputs the real-time generated information communication feature vector into the spatiotemporal state prediction model constructed based on a long short-term memory network. The spatiotemporal state prediction model learns the time series dependency of historical feature vectors and outputs a predicted information communication feature vector for the communication operation state at the next moment. The multidimensional cosine similarity between the predicted information communication feature vector and the actual information communication feature vector at the current moment is calculated, and 1 is subtracted from the multidimensional cosine similarity value to obtain the state deviation, which is used as a quantitative indicator to measure the degree of state deviation.
[0042] The state deviation is compared with a preset deviation threshold, which is dynamically calculated based on historical data distribution. If the state deviation is greater than the preset deviation threshold, it is determined that there is an abnormal mode in the current communication operation state. The key feature parameters and abnormal type that caused the abnormal mode are further identified, and then an abnormal alarm is triggered. The analysis results containing the key feature parameters and abnormal type are used as input to execute the communication quality assessment process. If the state deviation is less than or equal to the preset deviation threshold, the current communication operation state is monitored and analyzed.
[0043] It's important to clarify that the spatiotemporal state prediction model built on Long Short-Term Memory (LSTM) networks learns the overall spatiotemporal evolution of normal communication patterns through unsupervised pre-training on unlabeled historical communication feature vector data. This allows it to accurately identify unknown anomalies deviating from normal patterns without requiring abnormal sample labeling. When the state deviation exceeds a preset deviation threshold, which is dynamically calculated based on the statistical distribution of state deviation data under historical normal states (e.g., mean and three standard deviations), the system, after determining an anomaly, doesn't just issue a general alarm but initiates an attribution analysis module to precisely pinpoint the root cause. The core of this attribution analysis lies in calculating the dimension-wise difference between the current actual feature vector and the predicted feature vector. Specifically, it first calculates the difference between the two vectors in each feature dimension; for example, calculating the difference between the current actual user density and the predicted user density. Then, these differences are sorted in descending order of absolute value, and the top N (e.g., the top 3) feature dimensions are identified as the "key feature parameters" causing the anomaly. For example, if the analysis results show that the difference between "user density" and "traffic load" is much greater than other features, it can be determined that the anomaly is caused by a sudden surge in service load. Based on the identified key feature parameters and their deviation direction (e.g., too high or too low), the system will map them to a predefined anomaly type library for classification. This anomaly type library is pre-built using expert knowledge and historical data, and contains a variety of typical communication anomaly patterns and their feature combinations, such as: "high load congestion type" (characterized by user density and traffic load significantly higher than predicted), "coverage hole type" (characterized by signal strength mean significantly lower than predicted and increased variance), and "interference surge type" (characterized by signal strength variance significantly increased and coverage continuity decreased), etc. By matching the current key feature parameters and their deviation patterns with templates in the library, the system can accurately identify the specific "anomaly type". Finally, the system will trigger an anomaly alarm. This alarm information will not only include the time and location of the anomaly, but will also explicitly attach the identified key feature parameters (e.g., user density, traffic load) and the specific anomaly type (e.g., high load congestion type). This structured analysis result was then used as input to the subsequent communication quality assessment process.
[0044] S4: Based on the analysis results of the current communication operation status, analyze and evaluate the communication quality within the preset time window to obtain the communication quality evaluation index, and evaluate the quality of the current communication.
[0045] In this embodiment, step S4, based on the identification result of an abnormal mode in the current communication operation state, defines a preset time window centered on the time when the abnormal mode occurs, and extracts the information communication feature vectors for all times within the window; for the key feature parameters that cause the abnormality, calculate the statistical feature values of the key feature parameters within the entire time window, including the mean, peak value, and standard deviation; according to the abnormality type identified in step S3, invoke a preset evaluation rule, which is used to assign corresponding weight coefficients to different statistical feature values; after normalizing each calculated statistical feature value, multiply it by its corresponding weight coefficient and sum it to obtain the communication quality evaluation index. The lower the communication quality evaluation index value, the more severe the degradation of communication quality within the time window.
[0046] It's important to clarify that the pre-defined evaluation rule is essentially a multi-dimensional weighted scoring model. Its core lies in assigning differentiated weights to key characteristic parameters and their statistical feature values under different anomaly types. Specifically, the rule predefines a weight configuration table. For example, for "high-load congestion" anomalies, the rule assigns higher weights to the key characteristic parameters "user density" and "traffic load," while other characteristic parameters have zero or very low weights. At the statistical feature value level, since peak values better reflect the severity of congestion, the weight coefficient for "peak value" is higher than that for "mean" and "standard deviation." Conversely, for "coverage hole" anomalies, the rule significantly increases the weight of "mean signal strength," as it directly reflects the overall coverage level. In the specific calculation, the system first normalizes the extracted statistical feature values (such as peak user density and mean signal strength), for example, by using min-max normalization to map them to... The system first defines an interval to eliminate the influence of unit dimensions. Then, based on the identified anomaly type (such as "high load congestion type"), the system calls the corresponding weight coefficient from the preset evaluation rules, multiplies the normalized statistical feature value with its respective weight coefficient, and sums all the product results to finally obtain a comprehensive communication quality evaluation index.
[0047] S5: Combine the current communication quality assessment index with the historical communication quality assessment index sequence recorded in the preset database to analyze the trend of communication quality change and calculate the communication quality change rate.
[0048] In this embodiment, step S5 accesses a preset database, which is used to store historical communication quality evaluation indices in chronological order and records the anomaly type and occurrence time corresponding to each index.
[0049] The system retrieves a historical communication quality assessment index sequence with the same anomaly type as the current one from a pre-defined database and uses it as the baseline data for the current analysis. The currently calculated communication quality assessment index is then appended to the end of this historical sequence to form an updated time series. Based on this time series, the system calculates the difference between the current communication quality assessment index and the previous historical communication quality assessment index that is immediately adjacent to the current moment in the time series, and divides this difference by the time interval between the two to obtain the communication quality change rate.
[0050] It is important to clarify that retrieving historical communication quality assessment index sequences that are identical to the current anomaly type is a key design element of this step, aiming to achieve a homogeneous benchmark comparison. Different types of anomalies (such as "high load congestion" and "coverage hole") have drastically different impact mechanisms and degrees of degradation on communication quality, and their assessment indices also differ significantly in magnitude and fluctuation range. If different types of indices are treated the same for trend analysis, the conclusions will be distorted and meaningless. Therefore, by filtering by anomaly type and constructing a historical assessment index sequence containing only similar events, an accurate and comparable reference benchmark can be provided for the current event. Based on this, the currently calculated assessment index is appended to the end of this sequence, forming an updated, homogeneous time series data. Subsequently, when calculating the "communication quality change rate," the specific formula used is: Communication Quality Change Rate = (Current Communication Quality Assessment Index - Historical Communication Quality Assessment Index) / Time Interval. Here, "previous historical index" specifically refers to the assessment index corresponding to the previous similar anomaly event immediately preceding the current moment in this homogeneous time series, not the index at any arbitrary point in time.
[0051] S6: Combine the current communication quality assessment index, the historical communication quality assessment index sequence, and the calculated communication quality change rate to construct a communication quality prediction model. Based on the output of the communication quality prediction model, predict the communication quality for a future period of time.
[0052] In this embodiment, step S6 fuses the current communication quality assessment index, the historical communication quality assessment index sequence, and the communication quality change rate to construct a multi-dimensional input feature vector. Then, the input feature vector is input into a pre-trained communication quality prediction model. The communication quality prediction model learns the time dependence and change patterns in historical data and combines the short-term dynamic trends reflected by the current state and the change rate to finally output the communication quality prediction index sequence for future time points.
[0053] It is important to explain that this is achieved by constructing a multi-dimensional input feature vector that comprehensively reflects the state of communication quality. Specifically, this input feature vector contains at least three parts: the first part is a "historical communication quality assessment index sequence" that has been aligned and normalized over a time window. For example, the assessment indices of the top N similar abnormal events at the current moment are selected as features reflecting long-term periodic fluctuations and historical baselines. The second part is the "current communication quality assessment index," which represents the immediate state of communication quality. The third part is the "communication quality change rate," which serves as a key dynamic feature, quantifying the short-term trend and speed of quality change. In this way, the input vector integrates long-term historical dependencies, the current immediate state, and future short-term trends, providing rich and comprehensive information for the predictive model. Furthermore, pre-trained communication quality prediction models typically employ recurrent neural networks or temporal Transformer models. The training process is as follows: First, a large historical training dataset is constructed, where each sample contains a historical evaluation index sequence, a current evaluation index, a rate of change, and a corresponding future true evaluation index sequence (as a label). Then, the model is supervised learning using this dataset. The model learns the complex nonlinear time dependencies in historical data by minimizing the error (e.g., root mean square error) between the predicted and true sequences, and learns how to infer future evolution paths based on the current state and trends. After training, the model can be used for online prediction. The input feature vectors, constructed in real-time, are input into the model, and it outputs a predicted sequence of communication quality evaluation indices for a future period (e.g., the next few hours or several anomalous periods).
[0054] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
[0055] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A big data-based electronic information data processing method, characterized by, Comprise: S1: Obtain multi-source electronic information data set, and pre-process the obtained multi-source electronic information data set; S2: Feature extraction is carried out on the pre-processed multi-source electronic information data set, and correlation analysis is carried out on the extracted feature parameters, and the feature parameters related to the communication quality are screened out to generate information communication feature vector; S3: According to the generated information communication feature vector, the current communication running state is comprehensively analyzed, and whether the communication running state exists abnormal mode is identified; S4: According to the analysis result of the current communication running state, the communication quality in the preset time window is analyzed and evaluated, the communication quality evaluation index is obtained, and the advantages and disadvantages of the current communication quality are evaluated; S5: Combine the current communication quality evaluation index with the historical communication quality evaluation index sequence recorded in the preset database, analyze the change trend of the communication quality, and calculate the communication quality change rate; S6: Combine the current communication quality evaluation index, the historical communication quality evaluation index sequence and the calculated communication quality change rate to construct a communication quality prediction model, and predict the communication quality in a future period of time according to the output result of the communication quality prediction model.
2. The electronic information data processing method based on big data according to claim 1, characterized in that, The S1 constructs a multi-dimensional data acquisition system integrating wireless air interface and geographic information; the multi-dimensional data acquisition system collects user-level measurement report data of wireless transmission network in real time through signaling monitoring interface, including reference signal receiving power, signal to interference plus noise ratio, cell ID and time stamp, and obtains geographic fence and building vector data of corresponding area through GIS interface to obtain geographic information data; Finally, the collected user-level measurement report data and geographic information data are associated to form a multi-source electronic information data set that accurately binds wireless signal strength and spatial position; The pre-processing operation is carried out on the collected multi-source electronic information data set, specifically including data cleaning, data aggregation, data interpolation and data normalization.
3. The electronic information data processing method based on big data according to claim 2, characterized in that, The S2 is based on the pre-processed multi-source electronic information data set, and carries out composite feature extraction from four dimensions of space, time, service and environment to construct a high-dimensional feature parameter space, wherein the spatial dimension extracts the mean, variance and coverage continuity index of signal strength in the grid, the time dimension extracts the hour-level periodic component and change trend of signal strength, the service dimension extracts the user density and traffic load in the grid, and the environment dimension extracts the building height, type and coding features whether it is an indoor scene; then, a hybrid correlation analysis method is used, which combines the maximum information coefficient and mutual information, to calculate the nonlinear correlation degree between the above high-dimensional feature parameters and the reference signal receiving power as the communication quality benchmark to obtain the first importance score, and calculate the mutual information between the feature parameters to quantify the feature redundancy; Finally, a double screening standard is set by combining the first importance score and feature redundancy. First, the first feature subset with a first importance score lower than a preset threshold is removed. A redundancy relationship graph is constructed by taking each feature as a node and the mutual information value between features as the weight of the edge. Community discovery clustering is performed on the redundancy relationship graph, and features with a mutual information value greater than a preset redundancy threshold are divided into the same cluster. Finally, in each cluster, only the feature with the highest first importance score is retained, and other features in the cluster are removed. The remaining key feature parameters are vectorized and encoded to generate an information communication feature vector.
4. The electronic information data processing method based on big data according to claim 3, characterized in that, The S3 inputs the real-time generated information communication feature vector into a spatio-temporal state prediction model based on a long short-term memory network. The spatio-temporal state prediction model learns the time sequence dependency of the historical feature vector and outputs a predicted information communication feature vector of the next time communication running state. The multi-dimensional cosine similarity between the predicted information communication feature vector and the actual information communication feature vector at the current time is calculated, and 1 minus the multi-dimensional cosine similarity value is obtained as the state deviation, which is a quantitative index for measuring the state deviation degree. The state deviation is compared with a preset deviation threshold, and the preset similarity threshold is dynamically calculated based on historical data distribution. If the state deviation is greater than the preset deviation threshold, it is determined that the current communication running state has an abnormal mode, and the key feature parameters and the abnormal type causing the abnormal mode are further identified. Then, an abnormal alarm is triggered, and the analysis results including the key feature parameters and the abnormal type are input to perform a communication quality evaluation process. If the state deviation is less than or equal to the preset deviation threshold, the current communication running state is continuously monitored and analyzed.
5. The big data-based electronic information data processing method according to claim 4, characterized in that, The S4 sets a preset time window centered on the time when the abnormal mode occurs according to the identification result of the current communication running state, and extracts the information communication feature vectors at all times within the window. For the key feature parameters causing the abnormality, the statistical characteristic values of the key feature parameters within the entire time window are calculated, including the mean, peak value and standard deviation. According to the abnormal type identified by S3, a preset evaluation rule is called, which is used to assign corresponding weight coefficients to different statistical characteristic values. After normalization processing, the calculated statistical characteristic values are multiplied by the corresponding weight coefficients and summed to obtain a communication quality evaluation index. The lower the communication quality evaluation index value, the more serious the degradation of communication quality within the time window.
6. The electronic information data processing method based on big data according to claim 5, characterized in that, The S5 accesses a preset database for storing historical communication quality evaluation indexes in chronological order and records the abnormal type and occurrence time corresponding to each index. The same historical communication quality evaluation index sequence as the current abnormal type is retrieved from the preset database as the baseline data for the current analysis. The current calculated communication quality evaluation index is appended to the end of the historical sequence to form an updated time sequence; based on the time sequence, a communication quality change rate is obtained by calculating the difference between the current communication quality evaluation index and the previous historical communication quality evaluation index closest to the current time in the time sequence, and dividing the difference by the time interval between the two.
7. The electronic information data processing method based on big data according to claim 6, characterized in that, The S6 performs feature fusion on the current communication quality evaluation index, the historical communication quality evaluation index sequence, and the communication quality change rate to construct a multi-dimensional input feature vector; then, the input feature vector is input into a pre-trained communication quality prediction model, the communication quality prediction model learns the time dependence and change pattern in the historical data, and combines the short-term dynamic trend reflected by the current state and the change rate, and finally outputs a communication quality prediction index sequence at a future time point.