Method and device for predicting user perception derogatory cell, equipment, medium and product
By combining the dual-domain model of user perception characteristics and wireless domain characteristics, predicting and optimizing the user perception derogatory cells, the problem that KPI indicators in the prior art cannot characterize user perception is solved, and the accurate identification of user perception derogatory cells is achieved and optimization.
Patent Information
- Application Number
- CN202411126035.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-16
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-08-16
AI Technical Summary
The existing wireless network optimization work relies on the KPI indicators of the wireless network and cannot effectively characterize the user's business perception, resulting in the inability to accurately predict and optimize the derogatory cells of user perception.
By obtaining the user's core domain documents and correlating them with the wireless domain data corresponding to the identity identification, the user's perceived characteristics and wireless domain characteristics are extracted, and a two-domain model is used to derogate the user's prediction to determine the derogatory cell. This method combines user-perceptual features and wireless domain features to train a dual domain model to improve prediction accuracy and explain the derogatory reasons.
It realizes a more accurate prediction of the user's perceived derogatory cells, and can identify wireless factors that lead to user's perceived derogatory cells, thereby carrying out targeted network optimization and improving user experience.
Smart Images

Figure CN119071803A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of communication technology, and in particular to a method, device, equipment, medium and product for predicting a user-perceived derogatory cell. Background Art
[0002] At present, wireless network optimization work is generally carried out based on the key performance indicators (KPIs) of wireless networks. These wireless network KPIs generally come from the operations and maintenance center (OMC) of wireless equipment, and are statistically obtained with the wireless coverage cell as the smallest geographical granularity and a certain time length (such as 15 minutes) as the time granularity.
[0003] However, the KPI indicators of wireless networks are statistical indicators of the cell dimension over a period of time, and generally cannot represent the service perception of users in the cell. Often, the KPI indicators of the cell wireless network are good, but there are still cases where users have poor perception.
[0004] Therefore, there is an urgent need for a method that can predict a cell where user perception is degraded and obtain the wireless factors that cause the user perception to be degraded, so as to optimize the wireless network of the cell. Summary of the invention
[0005] The present invention provides a method, device, equipment, medium and product for predicting a user-perceived degraded cell, so as to solve the defect in the prior art that wireless KPI indicators cannot represent the user's service perception.
[0006] The present invention provides a method for predicting a user-perceived derogatory cell, comprising: Obtain the user's core domain receipt; Based on the user's identity, the core network domain receipt is associated with the wireless network domain data corresponding to the identity to obtain a user broadband table, and user perception features and wireless network domain features are respectively extracted from the user broadband table; Inputting the user perception feature and the wireless network domain feature into a trained dual-domain model to predict derogatory users, obtaining derogatory users output by the dual-domain model, and determining derogatory features from the user perception feature and the wireless network domain feature, wherein the dual-domain model is trained based on user perception features, wireless network domain features, and derogatory labels of positive and negative sample users; Based on the derogatory characteristics of each cell passed by the derogatory user, a derogatory cell is determined from the cells.
[0007] According to the method for predicting a user-perceived derogatory cell provided by the present invention, the derogatory cell is determined from each cell based on the derogatory characteristics of each cell passed by the derogatory user, including: Determine, based on each derogatory feature in the user wide table of the derogatory user, a fused derogatory feature corresponding to the user wide table; The derogatory characteristics of each cell are compared with the integrated derogatory characteristics respectively, and the derogatory cell is determined from the cells according to the comparison result.
[0008] According to the method for predicting user-perceived derogatory cells provided by the present invention, the derogatory features of each cell are compared with the fused derogatory features respectively, and the derogatory cell is determined from the cells according to the comparison result, including: Compare the derogatory features of each cell with the fused derogatory features respectively, and obtain the statistical times of each cell under each derogatory feature according to the comparison results; Based on the statistical times of each cell under each derogatory feature, the derogatory cell is determined, and a wireless derogatory feature is determined from each derogatory feature, and the wireless derogatory feature is optimized.
[0009] According to the method for predicting a user-perceived derogatory cell provided by the present invention, the derogatory cell is determined from each cell based on the derogatory characteristics of each cell passed by the derogatory user, including: Based on the derogatory characteristics of each cell passed by the first derogatory user, a first derogatory cell is determined from the cells, wherein the first derogatory user is predicted by using a basic model based on the core network domain receipt of the user; Based on the derogatory characteristics of each cell passed by the second derogatory user, a second derogatory cell is determined from the cells, and the second derogatory user is predicted by using the dual-domain model based on the core network domain receipt and wireless network domain data of the first derogatory user; A final derogatory cell is determined based on the first derogatory cell and the second derogatory cell.
[0010] According to the method for predicting user-perceived derogatory cells provided by the present invention, extracting wireless network domain features from the user broadband table includes: Acquire field values of multiple granularities corresponding to each wireless network field of the wireless network domain data in the user broadband table in each document period; Determining, based on the field type of each wireless network field, a final field value of each wireless network field from the field values of the multiple granularities; Based on the final field value of each wireless network field, the wireless network domain feature is extracted.
[0011] According to the method for predicting user-perceived degraded cells provided by the present invention, the determining, based on the field type of each radio network field, the final field value of each radio network field from the field values of the multiple granularities includes: For the forward field type, the minimum value among the field values of the multiple granularities is used as the final field value; For the reverse field type, the maximum value among the field values of the multiple granularities is taken as the final field value.
[0012] According to the method for predicting a user-perceived derogatory cell provided by the present invention, the step of constructing the user-perceived feature comprises: Obtain core network domain receipts of positive and negative sample users, and calculate statistics of each core network field in the core network domain receipts in each receipt period; Based on the difference in distribution of the statistics of each core network field of the positive and negative sample users, the user perception feature is mined from the core network fields.
[0013] According to the method for predicting user-perceived derogatory cells provided by the present invention, the step of constructing the wireless network domain features includes: Obtaining a user broadband table of positive and negative sample users, and calculating the statistics of each wireless network field of the wireless network domain data in the user broadband table in each bill period; Based on the difference in distribution of the statistics of each wireless network field of the positive and negative sample users, the wireless network domain features are mined from each wireless network field.
[0014] According to the method for predicting user-perceived derogatory cells provided by the present invention, the training step of the dual-domain model includes: Inputting the user perception features and wireless network domain features of the positive and negative sample users into an initial gradient boosting decision tree (GBDT) model to obtain the predicted derogation probabilities of the positive and negative sample users output by the initial GBDT model; Based on the difference between the predicted derogatory probability and the derogatory label, the initial GBDT model is trained to obtain the dual-domain model.
[0015] The present invention also provides a device for predicting a user-perceived derogatory cell, comprising: A receipt acquisition unit, used to acquire a user's core domain receipt; A feature extraction unit, configured to associate the core network domain receipt with the wireless network domain data corresponding to the identity identifier based on the user's identity identifier to obtain a user broadband table, and to extract user perception features and wireless network domain features from the user broadband table respectively; A user prediction unit, configured to input the user perception feature and the wireless network domain feature into a trained dual-domain model to predict a derogatory user, obtain a derogatory user output by the dual-domain model, and determine a derogatory feature from the user perception feature and the wireless network domain feature, wherein the dual-domain model is trained based on the user perception feature, wireless network domain feature, and derogatory label of positive and negative sample users; The cell determination unit is used to determine a derogatory cell from the cells based on the derogatory characteristics of the cells passed by the derogatory user.
[0016] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for predicting a user-perceived derogatory cell as described in any one of the above is implemented.
[0017] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for predicting a user-perceived derogatory cell as described in any one of the above is implemented.
[0018] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method for predicting a user-perceived derogatory cell as described in any one of the above is implemented.
[0019] The method, device, equipment, medium and product for predicting user-perceived derogatory cells provided by the present invention incorporate user perception characteristics and wireless network domain characteristics into the dual-domain model training. The use of this model can not only improve the accuracy of derogatory user prediction, but also obtain the derogatory characteristics in the explanation of the reasons for user derogation, and then determine the derogatory cells from each cell according to the derogatory characteristics of each cell passed by the derogatory user, thereby achieving more accurate prediction of the cell that the user perceives to be derogatory. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0021] Figure 1 It is a flow chart of the method for predicting user-perceived derogatory cells provided by the present invention.
[0022] Figure 2 This is one of the flow charts of the method for determining a derogatory cell provided by the present invention.
[0023] Figure 3This is the second flow chart of the method for determining a derogatory cell provided by the present invention.
[0024] Figure 4 It is a schematic diagram of the structure of the device for predicting user-perceived derogatory cells provided by the present invention.
[0025] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0026] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0027] Personal mobile Internet access is one of the main services of operators, with a vast user market. The quality of mobile Internet access and user experience are directly related to the reputation of operators. Poor Internet access experience may turn users into "derogatory users", that is, users who are dissatisfied with the operator's brand, products or services, which will bring adverse effects to operators that cannot be ignored.
[0028] According to the management and analysis of the quality of personal mobile Internet services by operators in recent years, the factors that affect the quality of the service and user perception are mainly wireless network problems, in addition to problems on the user's personal terminal side. Therefore, the work of operators to ensure service quality by improving network quality mainly falls on wireless network optimization. At present, wireless network optimization work is generally carried out with the KPI indicators of the wireless network as a guide. These wireless network KPI indicators generally come from the OMC of the wireless equipment, and are statistically obtained with the wireless coverage cell as the smallest regional granularity and a certain time length (such as 15 minutes) as the time granularity. Network optimization personnel pay attention to whether the wireless network KPI is deteriorated, and judge the cause of the deterioration based on experience. For example, the "RRC connection establishment success rate" is generally related to the connection status of the terminal and the base station, while the "RRC connection maximum number" is generally related to the load of the wireless network. If the KPI indicators are deteriorated, various optimization methods are used to optimize the network. After optimization, the corresponding KPI display returns to normal, which means that the optimization is effective and the problem is solved.
[0029] However, the KPI indicators of wireless networks are statistical indicators of the cell dimension over a period of time, and generally cannot represent the service perception of users in the cell. Often, the cell wireless KPI indicators are good, but there are still cases where users have poor perception. After all, the ultimate goal of operators to improve network and service quality is not to maintain the excellence of KPI indicators, but to ensure the real experience of users. Therefore, the network optimization department of the operator hopes to find the derogatory cells where users perceive poor quality and the network reasons that lead to the derogation, and then optimize them, so as to effectively improve the service experience and satisfaction of users.
[0030] In the related technology, a trained model is usually used to infer the core network documents of existing network users to predict the perceived depreciation of target users, which has the following shortcomings: 1) In reality, network optimization is carried out based on wireless network KPI indicators. The model is built based on user perception, so the poor quality objects output by the model and the poor quality objects judged based on the wireless KPI indicator threshold do not necessarily correspond. If the poor quality cells output by the model do not have wireless indicators below the threshold, network optimization cannot be implemented.
[0031] 2) Since derogatory users are mobile, they may pass through multiple cells within a certain period. If in this case, every cell passed through is recorded as a derogatory cell, a certain deviation will be introduced.
[0032] In view of the above problems, an embodiment of the present invention provides a method for predicting a user-perceived derogatory cell, in which a core network domain receipt of the user is obtained; based on the user's identity identifier, the core network domain receipt is associated with the wireless network domain data corresponding to the identity identifier to obtain a user broadband table, and user perception features and wireless network domain features are respectively extracted from the user broadband table; the user perception features and the wireless network domain features are input into a trained dual-domain model to predict derogatory users, and the derogatory users output by the dual-domain model are obtained, and the derogatory features are determined from the user perception features and the wireless network domain features, wherein the dual-domain model is trained based on the user perception features, wireless network domain features, and derogatory labels of positive and negative sample users; based on the derogatory features of each cell passed by the derogatory user, the derogatory cell is determined from each cell.
[0033] In this embodiment, user perception features and wireless network domain features are incorporated into the dual-domain model training. The use of this model can not only improve the accuracy of the prediction of derogatory users, but also obtain the derogatory features in the explanation of the reasons for user derogation, and then determine the derogatory cells from each cell based on the derogatory features of each cell passed by the derogatory user, thereby predicting the cell where the user perceives derogation.
[0034] The embodiment of the present invention can be applied to the scenario of network optimization. The execution subject of the method can be an electronic device such as a terminal device, a computer, a server, a server cluster, or a specially designed user-perceived derogatory cell prediction device, or a prediction device set in the electronic device, which can be implemented by software, hardware, or a combination of the two.
[0035] Figure 1 FIG. 1 is a flow chart of a method for predicting a user-perceived derogatory cell provided by the present invention. Figure 1 As shown, the method comprises the following steps: Step 110, obtaining the user's core domain receipt.
[0036] Specifically, the user's core network domain receipt can be the full user receipt left in the core network domain. The core network domain receipt mainly includes the receipt of the core network user plane interface (collected by the base station and UPF side, such as N3 and N6 interfaces), and the relevant signaling plane interface (mainly N1 and N2 interfaces) receipt.
[0037] Step 120 , based on the user's identity, the core network domain receipt is associated with the wireless network domain data corresponding to the identity to obtain a user broadband table, and the user perception feature and the wireless network domain feature are respectively extracted from the user broadband table.
[0038] Specifically, considering that the deterioration of user experience is often related to the status of the wireless network and is rarely caused by problems on the core network side, incorporating wireless network information into the prediction model can improve the recognition accuracy of the model. The user's identity identifier can be, for example, a phone number. The core network domain receipt is associated with the corresponding wireless network domain data according to the user's phone number to obtain a user broadband table, that is, the user broadband table is based on the user's phone number, that is, it includes the user's core network domain receipt and the wireless network domain receipt.
[0039] Then, the user perception features and wireless network domain features are extracted from the user wide table, which may specifically be the feature values corresponding to each user perception feature and the feature values corresponding to each wireless network domain feature.
[0040] The extraction of wireless network domain features can be obtained through the following methods: Obtain the field values of multiple granularities corresponding to each wireless network field of the wireless network domain data in the user wide table in each document period; Determining a final field value of each wireless network field from field values of multiple granularities based on the field type of each wireless network field; Based on the final field value of each wireless network field, the wireless network domain features are extracted.
[0041] Specifically, the wireless network domain data may include multiple wireless network KPI indicators. The wireless network KPI indicators are generally obtained based on time granularity statistics. In this embodiment, for a user receipt with a periodic granularity, the cells that a single user passes through within the receipt period are extracted to obtain the wireless indicators of the corresponding time of these cells, that is, the multiple field values corresponding to the receipt period. Taking the wireless indicator with a granularity of 15 minutes as an example, a user passes through a cell within an hour and has a receipt with a duration of 35 minutes, which may correspond to 3 or 4 granular field values of this cell.
[0042] Then, based on the field type of each wireless network field, the final field value of each wireless network field can be determined from the field values of multiple granularities. Specifically, it includes: For the forward field type, the minimum value among the field values of multiple granularities is used as the final field value; for the reverse field type, the maximum value among the field values of multiple granularities is used as the final field value.
[0043] For example, for load-related fields, such as "maximum number of RRC connections" and "average downlink PRB utilization", the maximum value of these 3 or 4 values is taken; for service fields such as success rate, such as "QoS Flow establishment success rate" and "radio connection rate", the minimum value of these 3 or 4 values is taken. "That is to say, for the forward field, the larger the value, the better the perception, the minimum value of the field values of multiple granularities is taken; conversely, for the reverse field, the larger the value, the worse the perception, the maximum value of the field values of multiple granularities is taken. In this way, after mapping the wireless field values of multiple granularities corresponding to a user into one value, the characteristic value of the wireless network domain feature can be extracted.
[0044] Step 130, input the user perception features and the wireless network domain features into the trained dual-domain model to predict the derogatory users, obtain the derogatory users output by the dual-domain model, and determine the derogatory features from the user perception features and the wireless network domain features. The dual-domain model is trained based on the user perception features, wireless network domain features, and derogatory labels of positive and negative sample users.
[0045] Specifically, after obtaining the characteristic values of the user perception feature and the wireless network domain feature, the characteristic values of the user perception feature and the wireless network domain feature are input into the trained dual-domain model to predict derogatory users, and the derogatory users output by the dual-domain model are obtained. Derogatory users are customers with poor Internet service experience. Derogatory features are determined from the user perception features and the wireless network domain features. Here, derogatory features are features that affect the user's Internet service experience. The top features that contribute the most to derogation, i.e., derogatory features, can be obtained through the treeSHAP algorithm.
[0046] It should be noted that the dual-domain model can be trained before this, and the dual-domain model is trained based on user perception features of positive and negative sample users, wireless network domain features, and derogatory labels.
[0047] In some embodiments, the training method of the dual-domain model includes: 1) Constructing user-perceived features: The core network domain receipts of positive and negative sample users are obtained, and the statistics of each core network field in the core network domain receipts within each document period are calculated; based on the difference in distribution of the statistics of each core network field of the positive and negative sample users, user perception features are mined from each core network field.
[0048] Specifically, complaining users are used as derogatory user labels, and non-complaining users are used as non-derogatory user labels. Complaining users are obtained as positive samples, and non-complaining users of equal proportion are used as negative samples. Positive and negative sample users are randomly mixed evenly, and then divided into training sets and validation sets according to the number and a certain ratio, for example, divided into two parts of 4:1, "4" as the training set and "1" as the validation set.
[0049] Obtain core network domain receipts for positive and negative sample users, mainly including receipts for core network user plane interfaces (collected on the base station and UPF sides, such as N3 and N6 interfaces), as well as relevant signaling plane interface (mainly N1 and N2 interfaces) receipts.
[0050] The core network domain receipts of the user are divided into time periods, where the time period can be, for example, one hour or 45 minutes, etc., and the embodiment of the present invention does not specifically limit this. The statistics of each core network field in all core network domain receipts within a receipt period granularity are calculated. For example, for the core network field "TCP layer RTT delay", the "TCP layer RTT average delay" of all core network domain receipts within this receipt period is calculated.
[0051] Then, compare the statistics of each core network field of the positive and negative samples to see if there is a significant difference in distribution. If yes, the core network field is constructed as a user-perceived feature, otherwise the core network field cannot be constructed as a user-perceived feature. This embodiment is based on the idea of big data, fully mines features from a numerical level, and constructs a feature library as large as possible.
[0052] In addition, based on considerations of limited computing power and model generalization, multiple algorithms, such as the Boruta algorithm, can be used to select an appropriate number of features from the constructed user perception feature library and truly incorporate them into the model as the final user perception features.
[0053] 2) Construct wireless domain features: The user broadband table of positive and negative sample users is obtained, and the statistics of each wireless network field of the wireless network domain data in the user broadband table in each document period are calculated; based on the difference in distribution of the statistics of each wireless network field of the positive and negative sample users, the wireless network domain features are mined from each wireless network field.
[0054] Based on the user broadband table, various statistics of all wireless network fields in the user broadband table within a document cycle are calculated. For example, for the field "maximum number of RRC connections", the average, maximum, minimum, median, variance, high-order moment and other statistics of this field in all documents within this cycle are calculated. Compare whether the statistics of positive and negative samples are significantly different in distribution. If so, they are constructed as wireless network domain features.
[0055] 3) Model training. The training steps of the dual-domain model include: inputting the user perception features and wireless network domain features of positive and negative sample users into the initial gradient boosting decision tree (GBDT) model to obtain the predicted derogation probability of the positive and negative sample users output by the initial GBDT model; based on the difference between the predicted derogation probability and the derogatory label, the initial GBDT model is trained to obtain the dual-domain model.
[0056] Specifically, after calculating the eigenvalues of the user perception features and the eigenvalues of the wireless network domain features of each user, the eigenvalues are input into the GBDT residual regression tree algorithm for training. The derogation probability of the positive sample is recorded as 1, and the derogation probability of the negative sample is recorded as 0. Multiple residual trees are used to calculate the approximate derogation probability. The validation set data is used for inference to obtain the user's derogation probability. Those greater than or equal to 0.5 are judged as poor quality, and those less than 0.5 are judged as non-poor quality. Combined with the sample labels in the validation set, the recall and precision of the model are calculated. Based on the comprehensive improvement of the recall and precision rates as the standard, the GBDT model is guided to be tuned and formed, and finally a dual-domain model of perception domain + wireless network is formed.
[0057] Step 140: Determine a derogatory cell from among the cells based on the derogatory characteristics of the cells that the derogatory user passes through.
[0058] Specifically, after obtaining the derogatory user and the derogatory characteristics through step 130, the derogatory cell can be determined from the cells based on the derogatory characteristics of the cells that the derogatory user passes through. Step 140 specifically includes: Step 141, based on each derogatory feature in the user wide table of the derogatory user, determining a fused derogatory feature corresponding to the user wide table; Step 142, compare the derogatory characteristics of each cell with the integrated derogatory characteristics respectively, and determine the derogatory cell from each cell according to the comparison result.
[0059] Here, by querying the user broadband table of the derogatory user, all the cells that the derogatory user has passed through in the period can be obtained, and the value of the feature in each cell can be calculated, that is, the feature value of each derogatory feature can be obtained. The fused derogatory feature is obtained by fusing the feature values of each derogatory feature, which can reflect the average derogatory degree of the derogatory feature. For example, the fused derogatory feature can be obtained by weighted average of each derogatory feature in the user broadband table of the derogatory user.
[0060] Then, the derogatory features of each cell are compared with the integrated derogatory features, and the derogatory cells are determined from each cell according to the comparison results. Step 142 specifically includes: Step 142-1, comparing the derogatory characteristics of each cell with the integrated derogatory characteristics respectively, and obtaining the statistical times of each cell under each derogatory characteristic according to the comparison result; Step 142-2, based on the statistical times of each cell under each degradation feature, determine the degradation cell, determine the wireless degradation feature from each degradation feature, and optimize the wireless degradation feature.
[0061] Figure 2 FIG. 1 is one of the flow charts of the method for determining a derogatory cell provided by the present invention, such as Figure 2 As shown, for the cells A, B, and C that the derogatory user passes through, for derogatory feature 1, the derogatory feature 1 of each cell is compared with the integrated derogatory feature 1, and it is obtained that the derogatory feature 1 of cells B and C is worse than the integrated derogatory feature 1, that is, the statistical times of cells B and C can be increased by one; for derogatory feature 2, the derogatory feature 2 of each cell is compared with the integrated derogatory feature 2, and it is obtained that the derogatory feature 2 of cells A and B is worse than the integrated derogatory feature 2, that is, the statistical times of cells A and B can be increased by one. Similarly, for multiple derogatory features, comparisons are made respectively, and finally the statistical times of each cell are obtained, and the cell with a statistical time greater than the number threshold is regarded as a derogatory cell.
[0062] For the determination of wireless depreciation characteristics, the number of occurrences of TOP depreciation perception domain characteristics and depreciation wireless network domain characteristics corresponding to the depreciation cell can be calculated according to the depreciation characteristics corresponding to the depreciation users. Here, the depreciation perception domain characteristics refer to the user perception characteristics in the depreciation characteristics, and the depreciation wireless network domain characteristics refer to the wireless network domain characteristics in the depreciation characteristics. Among them, the depreciation wireless network domain characteristics with the most occurrences are used as wireless depreciation characteristics for reference in network optimization work, so as to optimize the wireless depreciation characteristics.
[0063] The method provided by the embodiment of the present invention compares the derogatory characteristics of multiple cells that the derogatory user passes through within a time period, and only when the derogatory characteristic of a cell is lower than the average value within this time period is it recorded as the derogatory cell corresponding to the user.
[0064] Based on any of the above embodiments, based on the derogatory characteristics of each cell passed by the derogatory user, determining the derogatory cell from each cell includes: Based on the derogatory characteristics of each cell that the first derogatory user passes through, a first derogatory cell is determined from each cell, and the first derogatory user is predicted by using a basic model based on the core network domain receipt of the user; Based on the derogatory characteristics of each cell that the second derogatory user passes through, a second derogatory cell is determined from each cell, and the second derogatory user is predicted by using a dual-domain model based on the core network domain receipt and wireless network domain data of the first derogatory user; Based on the first derogatory cell and the second derogatory cell, a final derogatory cell is determined.
[0065] Specifically, in actual applications, since the number of core network domain documents of all users is very large, it requires a lot of computing power to perform feature calculations after associating wireless domain data. This embodiment proposes a two-layer prediction method. Figure 3 FIG. 2 is a flow chart of the method for determining a derogatory cell provided by the present invention. Figure 3 As shown in the figure, firstly, the basic model is formed by training the user perception documents, and then the first derogatory user is predicted based on the user's core network domain documents using the basic model. That is, a round of pre-inference is performed on all user documents, and then users with a derogatory probability greater than a certain threshold are screened out as the first derogatory user. Based on the derogatory characteristics of each cell that the first derogatory user passes through, the first derogatory cell is determined from each cell.
[0066] On this basis, the core network domain receipt and wireless network domain data of the first derogatory user are associated, and the user perception characteristics and wireless network domain characteristics are calculated by the method described in the above embodiment, and the derogatory user is predicted using the dual-domain model to obtain the second derogatory user. Based on the derogatory characteristics of each cell passed by the second derogatory user, the second derogatory cell is determined from each cell.
[0067] It should be noted that the method for determining the first derogatory cell and the second derogatory cell here can refer to the introduction of the above embodiment, and will not be repeated here.
[0068] Then, the intersection of the first derogatory community and the second derogatory community is taken as the derogatory community finally outputted.
[0069] The method provided by the embodiment of the present invention superimposes a dual-domain model on the basic model to resolve the over-computation introduced by directly associating wireless network statistical indicators with core network domain user receipts in actual applications, and takes the intersection of the basic model devalued cell and the dual-domain model devalued cell to obtain a more accurate user-perceived devalued cell.
[0070] Based on any of the above embodiments, a method for predicting a user-perceived derogatory cell is provided, comprising: S1, dual-domain model training. Including: S11, constructing user perception features. Obtain core network domain receipts of positive and negative sample users, and calculate the statistics of each core network field in the core network domain receipts in each receipt period; based on the difference in distribution of the statistics of each core network field of the positive and negative sample users, mine the user perception features from each core network field.
[0071] S12, constructing wireless network domain features. Obtain user broadband tables of positive and negative sample users, and calculate the statistics of each wireless network field of wireless network domain data in the user broadband table in each bill period; based on the difference in distribution of the statistics of each wireless network field of positive and negative sample users, mine the wireless network domain features from each wireless network field.
[0072] S13, inputting the user perception features and wireless network domain features of the positive and negative sample users into the initial gradient boosting decision tree GBDT model to obtain the predicted derogation probability of the positive and negative sample users output by the initial GBDT model; based on the difference between the predicted derogation probability and the derogation label, training the initial GBDT model to obtain a dual-domain model.
[0073] S2, predicting the user's perceived derogatory cells. This includes: S21, obtaining the user's core domain receipt; S22, based on the user's identity, associating the core network domain receipt with the wireless network domain data corresponding to the identity to obtain a user broadband table, and extracting user perception features and wireless network domain features from the user broadband table respectively; S23, inputting the user perception features and the wireless network domain features into the trained dual-domain model to predict the derogatory users, obtaining the derogatory users output by the dual-domain model, and determining the derogatory features from the user perception features and the wireless network domain features; S24, based on the derogatory characteristics of each cell that the derogatory user passes through, determining the derogatory cell from each cell. Specifically including: Based on the derogatory features in the user broadband table of the derogatory users, the integrated derogatory features corresponding to the user broadband table are determined; the derogatory features of each cell are compared with the integrated derogatory features respectively, and the statistical times of each cell under each derogatory feature are obtained according to the comparison results; based on the statistical times of each cell under each derogatory feature, the derogatory cells are determined, and the wireless derogatory features are determined from the various derogatory features, and the wireless derogatory features are optimized.
[0074] Based on the depreciation characteristics of each cell passed by the first depreciating user, a first depreciating cell is determined from each cell, and the first depreciating user is predicted as a depreciating user by using a basic model based on the user's core network domain receipt; based on the depreciation characteristics of each cell passed by the second depreciating user, a second depreciating cell is determined from each cell, and the second depreciating user is predicted as a depreciating user by using a dual-domain model based on the core network domain receipt and wireless network domain data of the first depreciating user; based on the first depreciating cell and the second depreciating cell, a final depreciating cell is determined.
[0075] The method provided by the embodiment of the present invention incorporates wireless network domain data for model training, obtains wireless factors that affect the degraded user experience, and thus converges to the degraded cell in a more accurate way, obtains the influence of the main wireless indicators of the cell's perceived degraded, and makes the direction of the cell's optimization for user perception clearer, thereby bringing more effective improvement to user perception through cell optimization. In addition, this proposal proposes a two-layer model reasoning method, which further incorporates wireless domain features and uses a dual-domain model for secondary reasoning after inferring the range of network-perceived degraded users in the perception domain basic model. On the one hand, it can significantly reduce computing power overhead, and on the other hand, it can provide guidance for actual network optimization work through the wireless features that contribute most to degraded obtained from the model interpretation.
[0076] Based on any of the above embodiments, Figure 4 is a schematic diagram of the structure of the device for predicting user-perceived derogatory cells provided by the present invention, such as Figure 4 As shown, the device comprises: The receipt acquisition unit 410 is used to acquire the core network domain receipt of the user; The feature extraction unit 420 is used to associate the core network domain document with the wireless network domain data corresponding to the identity identifier based on the user's identity identifier to obtain a user broadband table, and extract user perception features and wireless network domain features from the user broadband table respectively; A user prediction unit 430 is used to input the user perception feature and the wireless network domain feature into a trained dual-domain model to predict a derogatory user, obtain a derogatory user output by the dual-domain model, and determine a derogatory feature from the user perception feature and the wireless network domain feature, wherein the dual-domain model is trained based on the user perception feature, wireless network domain feature, and derogatory label of positive and negative sample users; The cell determination unit 440 is configured to determine a derogatory cell from among the cells based on the derogatory characteristics of the cells that the derogatory user passes through.
[0077] In the device provided by the embodiment of the present invention, user perception features and wireless network domain features are incorporated into the dual-domain model training. The use of this model can not only improve the accuracy of the prediction of derogatory users, but also obtain the derogatory features in the explanation of the reasons for user derogation, and then determine the derogatory cells from each cell based on the derogatory features of each cell passed by the derogatory user, so as to predict the cell where the user perceives derogation.
[0078] Based on any of the foregoing embodiments, the cell determination unit is specifically configured to: Determine, based on each derogatory feature in the user wide table of the derogatory user, a fused derogatory feature corresponding to the user wide table; The derogatory characteristics of each cell are compared with the integrated derogatory characteristics respectively, and the derogatory cell is determined from the cells according to the comparison result.
[0079] Based on any of the foregoing embodiments, the cell determination unit is specifically configured to: Compare the derogatory features of each cell with the fused derogatory features respectively, and obtain the statistical times of each cell under each derogatory feature according to the comparison results; Based on the statistical times of each cell under each derogatory feature, the derogatory cell is determined, and a wireless derogatory feature is determined from each derogatory feature, and the wireless derogatory feature is optimized.
[0080] Based on any of the foregoing embodiments, the cell determination unit is specifically configured to: Based on the derogatory characteristics of each cell passed by the first derogatory user, a first derogatory cell is determined from the cells, wherein the first derogatory user is predicted by using a basic model based on the core network domain receipt of the user; Based on the derogatory characteristics of each cell passed by the second derogatory user, a second derogatory cell is determined from the cells, and the second derogatory user is predicted by using the dual-domain model based on the core network domain receipt and wireless network domain data of the first derogatory user; A final derogatory cell is determined based on the first derogatory cell and the second derogatory cell.
[0081] Based on any of the above embodiments, the feature extraction unit is specifically used for: Acquire field values of multiple granularities corresponding to each wireless network field of the wireless network domain data in the user broadband table in each document period; Determining, based on the field type of each wireless network field, a final field value of each wireless network field from the field values of the multiple granularities; Based on the final field value of each wireless network field, the wireless network domain feature is extracted.
[0082] Based on any of the above embodiments, the feature extraction unit is specifically used for: For the forward field type, the minimum value among the field values of the multiple granularities is used as the final field value; For the reverse field type, the maximum value among the field values of the multiple granularities is taken as the final field value.
[0083] Based on any of the above embodiments, it also includes a feature construction unit, which is used to: Obtain core network domain receipts of positive and negative sample users, and calculate statistics of each core network field in the core network domain receipts in each receipt period; Based on the difference in distribution of the statistics of each core network field of the positive and negative sample users, the user perception feature is mined from the core network fields.
[0084] Based on any of the above embodiments, the feature construction unit is used to: Obtaining a user broadband table of positive and negative sample users, and calculating the statistics of each wireless network field of the wireless network domain data in the user broadband table in each bill period; Based on the difference in distribution of the statistics of each wireless network field of the positive and negative sample users, the wireless network domain features are mined from each wireless network field.
[0085] Based on any of the above embodiments, the device further includes a model training unit, which is specifically used to: Inputting the user perception features and wireless network domain features of the positive and negative sample users into an initial gradient boosting decision tree (GBDT) model to obtain the predicted derogation probabilities of the positive and negative sample users output by the initial GBDT model; Based on the difference between the predicted derogatory probability and the derogatory label, the initial GBDT model is trained to obtain the dual-domain model.
[0086] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5As shown, the electronic device may include: a processor (processor) 510, a communication interface (Communications Interface) 520, a memory (memory) 530 and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call the logic instructions in the memory 530 to execute a method for predicting a user-perceived derogatory cell, the method comprising: obtaining a core network domain receipt of the user; based on the identity of the user, associating the core network domain receipt with the wireless network domain data corresponding to the identity to obtain a user broadband table, and extracting user perception features and wireless network domain features from the user broadband table respectively; inputting the user perception features and the wireless network domain features into a trained dual-domain model to predict derogatory users, obtaining derogatory users output by the dual-domain model, and determining derogatory features from the user perception features and the wireless network domain features, wherein the dual-domain model is trained based on user perception features, wireless network domain features, and derogatory labels of positive and negative sample users; and determining derogatory cells from the cells based on the derogatory features of the cells passed by the derogatory user.
[0087] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0088] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program 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 method for predicting user-perceived derogatory cells provided by the above methods, the method including: obtaining a core network domain receipt of a user; based on the identity of the user, associating the core network domain receipt with the wireless network domain data corresponding to the identity to obtain a user broadband table, and extracting user perception features and wireless network domain features from the user broadband table respectively; inputting the user perception features and the wireless network domain features into a trained dual-domain model to predict derogatory users, obtaining derogatory users output by the dual-domain model, and determining derogatory features from the user perception features and the wireless network domain features, wherein the dual-domain model is trained based on user perception features, wireless network domain features, and derogatory labels of positive and negative sample users; and determining derogatory cells from the cells based on the derogatory features of the cells passed by the derogatory user.
[0089] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the method for predicting user-perceived derogatory cells provided by the above-mentioned methods, the method comprising: obtaining a core network domain receipt of a user; based on the identity of the user, associating the core network domain receipt with the wireless network domain data corresponding to the identity to obtain a user broadband table, and extracting user perception features and wireless network domain features from the user broadband table respectively; inputting the user perception features and the wireless network domain features into a trained dual-domain model to predict derogatory users, obtaining derogatory users output by the dual-domain model, and determining derogatory features from the user perception features and the wireless network domain features, wherein the dual-domain model is trained based on user perception features, wireless network domain features, and derogatory labels of positive and negative sample users; and determining derogatory cells from the cells based on the derogatory features of the cells passed by the derogatory user.
[0090] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0091] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for predicting user-perceived derogatory cells, It is characterized in that include: Obtain the user's core domain receipt; Based on the user's identity, the core network domain receipt is associated with the wireless network domain data corresponding to the identity to obtain a user broadband table, and user perception features and wireless network domain features are respectively extracted from the user broadband table; Inputting the user perception feature and the wireless network domain feature into a trained dual-domain model to predict derogatory users, obtaining derogatory users output by the dual-domain model, and determining derogatory features from the user perception feature and the wireless network domain feature, wherein the dual-domain model is trained based on user perception features, wireless network domain features, and derogatory labels of positive and negative sample users; Based on the derogatory characteristics of each cell passed by the derogatory user, a derogatory cell is determined from the cells.
2. The method for predicting user-perceived derogatory cells according to claim 1, It is characterized in that The step of determining a derogatory cell from the cells based on the derogatory characteristics of the cells passed by the derogatory user comprises: Determine, based on each derogatory feature in the user wide table of the derogatory user, a fused derogatory feature corresponding to the user wide table; The derogatory characteristics of each cell are compared with the integrated derogatory characteristics respectively, and the derogatory cell is determined from the cells according to the comparison result.
3. The method for predicting user-perceived derogatory cells according to claim 2, It is characterized in that The comparing the derogatory features of each cell with the integrated derogatory feature respectively, and determining the derogatory cell from each cell according to the comparison result, comprises: Compare the derogatory features of each cell with the fused derogatory features respectively, and obtain the statistical times of each cell under each derogatory feature according to the comparison results; Based on the statistical times of each cell under each derogatory feature, the derogatory cell is determined, and a wireless derogatory feature is determined from each derogatory feature, and the wireless derogatory feature is optimized.
4. The method for predicting user-perceived derogatory cells according to claim 1, It is characterized in that The step of determining a derogatory cell from the cells based on the derogatory characteristics of the cells passed by the derogatory user comprises: Based on the derogatory characteristics of each cell passed by the first derogatory user, a first derogatory cell is determined from the cells, wherein the first derogatory user is predicted by using a basic model based on the core network domain receipt of the user; Based on the derogatory characteristics of each cell passed by the second derogatory user, a second derogatory cell is determined from the cells, and the second derogatory user is predicted by using the dual-domain model based on the core network domain receipt and wireless network domain data of the first derogatory user; A final derogatory cell is determined based on the first derogatory cell and the second derogatory cell.
5. The method for predicting user-perceived derogatory cells according to claim 1, It is characterized in that The wireless network domain features are extracted from the user broadband table, including: Acquire field values of multiple granularities corresponding to each wireless network field of the wireless network domain data in the user broadband table in each document period; Determining, based on the field type of each wireless network field, a final field value of each wireless network field from the field values of the multiple granularities; Based on the final field value of each wireless network field, the wireless network domain feature is extracted.
6. The method for predicting user-perceived derogatory cells according to claim 5, It is characterized in that The determining, based on the field type of each wireless network field, a final field value of each wireless network field from the field values of the multiple granularities comprises: For the forward field type, the minimum value among the field values of the multiple granularities is used as the final field value; For the reverse field type, the maximum value among the field values of the multiple granularities is taken as the final field value.
7. The method for predicting user-perceived derogatory cells according to any one of claims 1 to 6, It is characterized in that The step of constructing the user perception feature comprises: Obtain core network domain receipts of positive and negative sample users, and calculate statistics of each core network field in the core network domain receipts in each receipt period; Based on the difference in distribution of the statistics of each core network field of the positive and negative sample users, the user perception feature is mined from the core network fields.
8. The method for predicting user-perceived derogatory cells according to any one of claims 1 to 6, It is characterized in that The steps of constructing the wireless network domain characteristics include: Obtaining a user broadband table of positive and negative sample users, and calculating the statistics of each wireless network field of the wireless network domain data in the user broadband table in each bill period; Based on the difference in distribution of the statistics of each wireless network field of the positive and negative sample users, the wireless network domain features are mined from each wireless network field.
9. The method for predicting a user-perceived derogatory cell according to any one of claims 1 to 6, It is characterized in that The training steps of the dual-domain model include: Inputting the user perception features and wireless network domain features of the positive and negative sample users into an initial gradient boosting decision tree (GBDT) model to obtain the predicted derogation probabilities of the positive and negative sample users output by the initial GBDT model; Based on the difference between the predicted derogatory probability and the derogatory label, the initial GBDT model is trained to obtain the dual-domain model.
10. A device for predicting a user's perception of a derogatory cell, It is characterized in that include: A receipt acquisition unit, used to acquire a user's core domain receipt; A feature extraction unit, configured to associate the core network domain receipt with the wireless network domain data corresponding to the identity identifier based on the user's identity identifier to obtain a user broadband table, and to extract user perception features and wireless network domain features from the user broadband table respectively; A user prediction unit, configured to input the user perception feature and the wireless network domain feature into a trained dual-domain model to predict a derogatory user, obtain a derogatory user output by the dual-domain model, and determine a derogatory feature from the user perception feature and the wireless network domain feature, wherein the dual-domain model is trained based on the user perception feature, wireless network domain feature, and derogatory label of positive and negative sample users; The cell determination unit is used to determine a derogatory cell from the cells based on the derogatory characteristics of the cells passed by the derogatory user.
11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the method for predicting a user-perceived derogatory cell according to any one of claims 1 to 9 is implemented.
12. A non-transitory computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the method for predicting a user-perceived derogatory cell according to any one of claims 1 to 9 is implemented.
13. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the method for predicting a user-perceived derogatory cell according to any one of claims 1 to 9 is implemented.
Citation Information
Patent Citations
Data processing method and device, electronic equipment and storage medium
CN113468394A
Method for predicting derogatory users in tariff package based on user data
CN114004307A
Positioning method and device for poor user perception quality, electronic equipment and storage medium
CN115802375A
Mobile communication network quality optimization method and device
CN116963110A
Radio access network congestion management method and apparatus, and congestion strategy management method and system
WO2014146502A1
