Root cause analysis method and system and storage medium

Abstract

The invention discloses a root cause analysis method and system and a storage medium, and the method comprises the steps: obtaining a plurality of to-be-tested data of a to-be-analyzed service, generating a new service model based on the plurality of to-be-tested data, the to-be-tested data comprising base station data and user data, and generating a new service model based on the plurality of to-be-tested data; and dynamically updating the current service model into a target service model based on the model difference between the current service model and the new service model and / or the fluctuation information corresponding to the new service model, so as to perform root cause analysis on the to-be-analyzed service based on the target service model. Through the method, the accuracy of the root cause analysis result can be improved.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a root cause analysis method, system, and storage medium. Background Art

[0002] With the advancement of communication technology, communication networks have become increasingly complex, leading to new requirements for the stability of information transmission. Especially with the rapid development of 5G, with its massive connections and high speeds, the number of users and traffic has increased dramatically, and base station networks have become extremely complex. To improve the user experience, it is necessary to promptly identify various user issues. Promptly identifying various user issues requires rapid and accurate analysis of base station data, thereby analyzing user behavior at the base station (i.e., base station user models).

[0003] Currently, related technologies use a method of generating specific models based on fixed features. By processing specific features on the base station side, although this method has certain advantages in locating base station problems, it is not sensitive to user models. When user behavior changes, the root cause analysis of poor quality points using the specific model generated by the fixed features results in relatively low accuracy. Summary of the Invention

[0004] This application provides a root cause analysis method, system, and storage medium that can improve the accuracy of analysis results.

[0005] In a first aspect, the present application provides a root cause analysis method, the method comprising:

[0006] Acquire multiple test data of the service to be analyzed, and generate a new service model based on the multiple test data, the test data including base station data and user data;

[0007] Based on the model difference between the current business model and the new business model and / or the fluctuation information corresponding to the new business model, the current business model is dynamically updated to the target business model, so as to perform root cause analysis on the business to be analyzed based on the target business model.

[0008] In a second aspect, the present application provides a root cause analysis system, the root cause analysis system comprising a user model unit, a data reporting unit, and a root cause analysis unit;

[0009] A data reporting unit is used to obtain a plurality of test data of the service to be analyzed, the test data including base station data and user data;

[0010] A user model unit is configured to generate a new business model based on a plurality of test data; and dynamically update the current business model to a target business model based on a model difference between the current business model and the new business model and / or fluctuation information corresponding to the new business model;

[0011] The root cause analysis unit is used to perform root cause analysis on the business to be analyzed based on the target business model.

[0012] In a third aspect, the present application provides a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program can implement the above-mentioned method when executed by a processor.

[0013] The beneficial effects of the present application are: different from the existing technology, the present application generates a new business model through multiple test data including base station data and user data; then the current business model corresponding to the new business model is dynamically updated according to the model difference between the new business model and the current business model and / or the fluctuation information corresponding to the new business model, that is, the current business model can be adjusted and updated in real time according to the changes in base station data and changes in user behavior, so as to perform root cause analysis on the business to be analyzed based on the target business model obtained after dynamic update, improve the accuracy of the root cause analysis results, and thus better reflect the real behavior and business problems of users, and improve the limitations of the existing technology caused by the use of fixed features to generate specific models.

[0014] In addition, this application combines base station data and user data to generate a new business model, and obtains a user model based on the business model, so that when analyzing the corresponding business to be analyzed based on the new business model in the user model, it can have a more comprehensive understanding of the user's communication environment and usage, avoiding analysis deviations in the user model due to single data, thereby improving the accuracy of root cause identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without inventive efforts. Among them:

[0016] Figure 1 This is a flowchart of the first embodiment of the root cause analysis method provided by this application;

[0017] Figure 2 This is a flow chart of the second embodiment of the root cause analysis method provided by this application;

[0018] Figure 3 This is a flowchart of updating the initial user model set in the root cause analysis method provided by this application;

[0019] Figure 4 This is a flow chart of an embodiment of a root cause analysis system provided by the present application;

[0020] Figure 5This is a flow chart of another embodiment of the root cause analysis method provided by the present application;

[0021] Figure 6 It is a structural diagram of an embodiment of a computer-readable storage medium provided by this application. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It will be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only some, rather than all, structures related to the present application are shown in the drawings. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0023] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0024] With the advancement of communication technology, communication networks have become increasingly complex, leading to new requirements for the stability of information transmission. Especially with the rapid development of 5G, with its massive connections and high speeds, the number of users and traffic has increased dramatically, and base station networks have become extremely complex. To improve the user experience, it is necessary to promptly identify various user issues. Promptly identifying various user issues requires rapid and accurate analysis of base station data, thereby analyzing user behavior at the base station (i.e., base station user models).

[0025] Currently, related technologies use a method of generating specific models based on fixed features. By processing specific features on the base station side, although this method has certain advantages in locating base station problems, it is not sensitive to user models. When user behavior changes, the root cause analysis of poor quality points using the specific model generated by the fixed features results in relatively low accuracy.

[0026] Therefore, in order to solve the technical problem that the prior art uses a specific model generated by fixed features to perform root cause analysis of poor quality points, resulting in relatively low accuracy of the analysis results, the present application provides a root cause analysis method.

[0027] Figure 1 This is a flow chart of the first embodiment of the root cause analysis method provided by this application. Figure 1As shown, the method includes the following steps:

[0028] Step 110: Acquire multiple test data of the business to be analyzed, and generate a new business model based on the multiple test data.

[0029] The data to be measured includes base station data and user data. User data can specifically include the number of user accesses, Signal to Interference plus Noise Ratio (SINR), Reference Signal Received Power (RSRP), uplink traffic, and downlink traffic.

[0030] Base station data may specifically include physical resource block (PRB) utilization, control channel element (CCE) utilization, and path loss data. PRB utilization refers to the ratio of actually used resources to total resources in a physical resource block, and CCE utilization refers to the ratio of actually used resources to total resources in a control channel element.

[0031] Among them, generating a new business model based on multiple data to be tested can be to obtain the real-time data of the data to be tested in the time series (such as real-time characteristics) and the change data of the data to be tested in the time series (such as fluctuation characteristics) based on the data to be tested, and then obtain the new business model based on the real-time characteristics and fluctuation characteristics.

[0032] It should be noted that when a user has only one business to be analyzed, the new business model is equivalent to the user model. When a user has multiple businesses to be analyzed, multiple new business models are combined to form a user model.

[0033] Step 120: Based on the model difference between the current business model and the new business model and / or the fluctuation information corresponding to the new business model, the current business model is dynamically updated to the target business model, so that the root cause analysis of the business to be analyzed is performed based on the target business model.

[0034] The model difference may be a parameter change, an architecture change (such as a change in the number of features), or a feature value change between the current business model and the new business model.

[0035] The fluctuation information may be a characteristic value of a fluctuation feature on the new business model. For example, when the characteristic value of the fluctuation feature on the new business model exceeds a preset characteristic value, the current business model may be dynamically updated to the new business model.

[0036] In some embodiments, the fluctuation information may also be the fluctuation difference between the fluctuation characteristics of the new business model and the fluctuation characteristics of the user model set.

[0037] The target business model can be the current business model or a new business model. For example, when the model difference between the current business model and the new business model and / or the fluctuation information corresponding to the new business model exceeds a preset threshold, the current business model is dynamically updated to the new business model.

[0038] When the model difference between the current business model and the new business model and / or the fluctuation information corresponding to the new business model is smaller than a preset threshold, the current business model may be used as the target business model.

[0039] For example, if the current business model includes 10 A features and the new business model includes 10 A features and 2 B features, the current business model is dynamically updated to a target business model including 10 A features and 2 B features.

[0040] If the current business model includes 10 A features and 2 B features, the feature values ​​of the 10 A features are all n, and the feature values ​​of the 2 B features are all m; the new business model also includes 10 A features and 2 B features, the feature values ​​of the 10 A features are all n, and the feature values ​​of the 2 B features are all m-11, and the preset threshold is 10. Since the feature value difference of the B feature is 11, which exceeds the preset threshold 10, the current business model is dynamically updated to the target business model in which the feature values ​​of the 10 A features are all n, and the feature values ​​of the 2 B features are all m-11.

[0041] If the current business model includes 10 A features and 2 B features, the characteristic values ​​of the 10 A features are all n, and the characteristic values ​​of the 2 B features are all m; the new business model also includes 10 A features and 2 B features, the characteristic values ​​of the 10 A features are all n, and the characteristic values ​​of the 2 B features are all m-0.1, and the preset threshold is 10, then the current business model can be directly used as the target business model.

[0042] This embodiment generates a new business model by using multiple test data including base station data and user data; then dynamically updates the current business model corresponding to the new business model based on the model difference between the new business model and the current business model and / or the fluctuation information corresponding to the new business model. That is, the current business model can be adjusted and updated in real time according to changes in base station data and changes in user behavior, so as to perform root cause analysis on the business to be analyzed based on the target business model obtained after dynamic update, thereby improving the accuracy of the root cause analysis results, thereby better reflecting the user's real behavior and business problems, and improving the limitations of the existing technology caused by the use of fixed features to generate specific models.

[0043] In addition, this embodiment combines base station data and user data to generate a new business model, and obtains a user model based on the business model, so that when analyzing the corresponding business to be analyzed based on the new business model in the user model, it can more comprehensively understand the user's communication environment and usage, avoid analysis deviations in the user model due to single data, and thereby improve the accuracy of root cause identification.

[0044] In some embodiments, each piece of data to be tested corresponds to a plurality of quality difference points. Figure 2 , Figure 2 : This is a flow chart of the second embodiment of the root cause analysis method provided by this application. The method includes:

[0045] Step 210: Acquire a plurality of test data of the service to be analyzed, wherein the test data includes base station data and user data.

[0046] Step 220: sequentially determine the real-time characteristics of each test data in each corresponding quality difference point.

[0047] Step 220 may include the following process:

[0048] 1) Determine the sum of the mean values ​​of all the measured data in each quality difference point.

[0049] For example, the following formula (1) can be used to obtain the sum of the mean values ​​of all the measured data within the same quality difference point.

[0050]

[0051] Among them, n is the number of data to be tested, i represents the i-th data to be tested, d i (z) represents the i-th data to be tested in the z-th quality difference point, and d(z) represents the sum of the means of the n data to be tested in the z-th quality difference point.

[0052] 2) Based on the sum of the means, the real-time features corresponding to each data to be tested in each quality difference point are determined in turn.

[0053] For example, the following formula (2) can be used to determine the real-time characteristics corresponding to each data to be measured within the same quality difference point.

[0054]

[0055] Among them, d i (z) represents the i-th data to be measured in the z-th quality difference point, D i (z) represents the real-time feature corresponding to the i-th data to be tested in the z-th quality difference point.

[0056] 3) Traverse all the quality difference points corresponding to each data to be tested, and obtain the real-time features of each data to be tested in all the quality difference points.

[0057] For example, we will use all the test data obtained for base station ID 1 and cell ID 2. The number of poor quality points is n1, which are poor quality point 1, poor quality point 2, poor quality point 3, poor quality point 4, and poor quality point n1. The number of test data is 8, which are the number of user accesses, PRB utilization, CCE utilization, SINR, RSRP, path loss, uplink traffic, and downlink traffic. All the test data are shown in Table 1 below:

[0058] Table 1

[0059]

[0060]

[0061] Traversing all the quality difference points in Table 1, the real-time features of each test data in all the quality difference points can be obtained as follows:

[0062] First, determine the real-time feature D1(1) corresponding to the first test data d1(1) of quality difference point 1 (i.e., the number of user accesses in Table 1), and then determine the remaining test data in quality difference point 1 in turn, and obtain D2(1), D3(1), D4(1), D5(1), D6(1), D7(1), and D8(1); then determine the real-time features corresponding to the 8 test data in quality difference point 2, and obtain D1(2), D2(2), D3(2), D4(2), D5(2), D6(2), D7(2), and D8(2).

[0063] In a similar manner, the real-time features corresponding to the measured data of the remaining poor-quality points are obtained in sequence, and then the real-time feature table shown in Table 2 below is obtained.

[0064] Table 2

[0065]

[0066]

[0067] Step 230: Based on the real-time characteristics of each data to be measured and the preset time offset, the fluctuation characteristics of each data to be measured are obtained.

[0068] Step 230 may include the following process:

[0069] 1) Based on a preset time offset, sequentially determine the correlation value between each piece of data to be measured and the remaining pieces of data to be measured in the service to be analyzed.

[0070] The correlation value can be calculated using the following formula (3).

[0071]

[0072] Among them, R is the correlation value, d n Indicates the nth data to be tested, d m Indicates the mth data to be tested, It represents the mean value of the nth tested data in all quality difference points. It represents the mean value of the mth data to be tested in all quality difference points, is the preset time offset, where the preset time offset Can be set manually to preset time offset The range of can be [1, n1].

[0073] For example, the number of the test data is 8, namely test data 1, test data 2, test data 3, test data 4, test data 5, test data 6, test data 7 and test data 8.

[0074] Then, the correlation values ​​between the test data 1 and the test data 2, the test data 3, the test data 4, the test data 5, the test data 6, the test data 7 and the test data 8 can be determined in sequence to obtain R12, R13, R14, R15, R16, R17 and R18.

[0075] Then determine the correlation values ​​between the test data 2 and the test data 1, test data 3, test data 4, test data 5, test data 6, test data 7 and test data 8 respectively, and obtain R21, R23, R24, R25, R26, R27 and R28.

[0076] The correlation values ​​between the test data 3, the test data 4, the test data 5, the test data 6, the test data 7 and the test data 8 and the remaining test data are determined in sequence using the same method.

[0077] 2) Based on the correlation value, relevant test data corresponding to each test data is screened out from the remaining test data of the business to be analyzed, and the real-time feature corresponding to each relevant test data is used as the causal feature corresponding to the test data corresponding to each relevant test data.

[0078] The obtained correlation value can be compared with a preset correlation value. If the correlation value is greater than or equal to the preset correlation value, the real-time feature corresponding to the test data involved in the correlation value is used as the causal feature. The preset correlation value can be set manually based on actual conditions, such as 0.6, 0.7, 0.8, etc. Generally speaking, the range of the correlation value is [0, 1].

[0079] Exemplarily, assuming that the preset correlation value is 0.7, the calculated values ​​are R12=0.2, R13=0.75, R14=0.6, R15=0.1, R16=0.3, R17=0.8, and R18=0.2.

[0080] Since R13=0.75 and R17=0.8 are both greater than the preset correlation value of 0.7, it can be considered that there is a causal relationship between the test data 1 and the test data 3, and there is a causal relationship between the test data 1 and the test data 7. Therefore, the multiple test data with causal relationships screened out are test data 1, test data 3 and test data 7, and the real-time features corresponding to test data 1, test data 3 and test data 7 are used as causal features.

[0081] 3) Based on all causal features corresponding to each data to be tested, multiple fluctuation features corresponding to each data to be tested are determined.

[0082] In some embodiments, the following process 3-1)-3-3) may be used to determine the fluctuation characteristics:

[0083] 3-1) Determine the effective time granularity of each causal feature.

[0084] The following steps can be used to determine the effective time granularity:

[0085] Step 21: Determine the first maximum value to be measured corresponding to the current data to be measured among all the poor-quality points, and the first time point corresponding to the first maximum value to be measured.

[0086] Step 22: Obtain relevant test data corresponding to the current test data.

[0087] Step 23: Determine the second maximum value to be measured corresponding to the relevant data to be measured among all the quality difference points and the second time point corresponding to the second maximum value to be measured.

[0088] Step 24: Determine the effective time granularity of the current causal feature based on the first time point and the second time point.

[0089] If the number of relevant data to be tested is 1, the time difference between the first time point and the second time point is used as the effective time granularity of the current causal feature.

[0090] For steps 21 to 24, the effective time granularity can be calculated using the following formula (4).

[0091]

[0092] in, is the effective time granularity corresponding to the i-th causal feature, i.e. the current causal feature, d i is the i-th data to be tested, i.e. the current data to be tested, is the first maximum value to be measured corresponding to the i-th measured data among all the quality difference points, d jis the jth data to be tested, i.e., the related data to be tested that has a causal relationship with the current data to be tested, T i For the first time point, is the second maximum value to be measured, T j The second time point.

[0093] It should be noted that if the number of relevant test data is at least two, the second time point corresponding to each relevant test data is obtained; the time difference between the first time point and multiple second time points is determined in turn; and the largest time difference is selected from all time differences as the effective time granularity of the current causal feature.

[0094] For example, if there is a causal relationship between the data to be tested 1 and the data to be tested 4, the data to be tested 5, and the data to be tested 6, then the data to be tested 4, the data to be tested 5, and the data to be tested 6 are related to the data to be tested 1, and the real-time features 1, real-time features 4, real-time features 5, and real-time features 6 corresponding to the data to be tested 1, the data to be tested 4, the data to be tested 5, and the data to be tested 6 are all used as causal features.

[0095] Assume that the first maximum value to be measured and the first time point corresponding to the measured data 1 are and T1, the second maximum values ​​to be measured corresponding to the test data 4, the test data 5, and the test data 6 are and The corresponding second time points are T4, T5, and T6 respectively.

[0096] Substitute T1, T4, T5, and T6 into the above formula 4 to calculate. Maximum, then As the effective time granularity of the current causal feature (i.e., causal feature 1).

[0097] 3-2) Based on the effective time granularity, determine the maximum causal relationship value corresponding to each data to be tested in all quality difference points in turn.

[0098] The maximum causal relationship value can be determined using the following formula (5):

[0099]

[0100] Among them, D imax represents the maximum causal relationship value of the i-th causal feature among all quality difference points, The range of valid time granularity.

[0101] 3-3) Based on the multiple maximum causal relationship values ​​corresponding to each data to be tested and the time corresponding to each maximum causal relationship value, determine the multiple fluctuation characteristics corresponding to each data to be tested.

[0102] The calculation of the fluctuation characteristics can be done using the following formula (6):

[0103]

[0104] Among them, D i (i) represents the i-th causal feature of the i-th quality difference point, W i (i) represents the fluctuation characteristics corresponding to the i-th causal characteristics of the i-th quality difference point, T i (i) represents the time point corresponding to the i-th causal feature of the i-th quality difference point, Represents the maximum causal relationship value D of the i-th causal feature among all quality difference points imax The corresponding time point.

[0105] Step 240: Generate a new business model based on the real-time features of the plurality of to-be-tested data and the fluctuation features of the plurality of to-be-tested data.

[0106] For example, the eight test data in the above example are used for illustration. If the real-time features corresponding to six of the test data have a causal relationship, the fluctuation features of the six test data are obtained. Then, the new business model can be obtained using the following formula (7):

[0107]

[0108] Among them, U type(u) For new business models, D n (i) represents the nth real-time feature corresponding to the i-th quality difference point, W k (i) represents the kth fluctuation feature corresponding to the i-th quality difference point.

[0109] Among them, formula 7 can be simplified to obtain formula (8):

[0110] U type(u) =[D n∈(1,8) (n),W k∈(1,6) (k)] Formula (8).

[0111] D n∈(1,8) (n) represents the nth real-time feature obtained by summing n1 quality difference points, W k∈(1,6) (k) represents the kth fluctuation feature obtained by summing n1 quality difference points.

[0112] Step 250: Based on the model difference between the current business model and the new business model and / or the fluctuation information corresponding to the new business model, the current business model is dynamically updated to the target business model, so that the root cause analysis of the business to be analyzed is performed based on the target business model.

[0113] In some embodiments, after generating a new business model based on multiple data to be tested, the causal characteristics and fluctuation characteristics of the new business model can be matched with the causal characteristics and fluctuation characteristics of the current business model respectively to obtain the model difference between the new business model and the current business model.

[0114] For example, the new business model is The current business model is Since the values ​​of the causal features and the real-time features are the same, we can match D1 and D'1 and find their difference, and match W1 and W'1 and find their difference. Where i is greater than or equal to k, when i is greater than k, the W position in the Dith row of the matrix is ​​filled with 0.

[0115] It should be noted that when comparing causal features, if the real-time features in the new business model are D1-D8, where D1 and D4 are causal features, it is only necessary to match and subtract D1 from D'1 of the current business model, and to match and subtract D4 from D'4 of the current business model to obtain the model difference between the new business model and the current business model.

[0116] In some embodiments, after obtaining the new service model of the user, service analysis can be performed on multiple users included in the cell where the user is located (i.e., the preset cell), where each user includes multiple services to be analyzed. If the characteristics of the same service of different users are highly similar, a target user model set for the preset cell can be generated, so that batch updates can be performed on multiple users in the preset cell based on the target user model set. For details, please refer to Figure 3 The following steps:

[0117] Step 310: Determine a new business model corresponding to the business to be analyzed of each user, and determine a feature model set corresponding to each new business model based on the new business model and a preset relationship expression to obtain a feature model set.

[0118] Among them, the feature model set includes all feature model sets, and the feature model set can be a tree structure or a multi-dimensional matrix structure. For example, the feature model set can be

[0119] Among them, U type(u) For the new business model, W k ′ is the new business model U type(u) Fluctuation characteristics within, D n ′ is the new business model U type(u) Real-time features within It is a preset relationship.

[0120] Step 320: Generate an end-user model set based on the feature model set.

[0121] Step 320 may specifically include the following steps:

[0122] Step 321: Determine the first user and the first feature model set corresponding to the first user in the feature model set based on the user identification number and the generation time number of the feature model set, and use the first feature model set as the initial user model set.

[0123] Among them, the user identification number can be the user identification serial number on the base station side, such as the first user, the second user and the third user, etc., and the generation time number of the feature model set can be the completion time serial number of the feature model set reported in the log, such as the first feature model set, the second feature model set and the third feature model set, etc.

[0124] Step 322: perform a difference comparison between the next feature model set in the feature model set and the initial user model set in sequence.

[0125] Step 323: Based on the comparison result and the initial user model set, a target user model set is obtained.

[0126] For example, if the preset cell includes 4 users, such as user 1, user 2, user 3, and user 4, and each user includes 4 services to be analyzed, namely service 1, service 2, service 3, and service 4, then the feature model set corresponding to each service model is as shown in Table 3 below:

[0127] Table 3

[0128] User 1 User 2 User 3 User 4 Business 1 T11 T21 T31 T41 Business 2 T12 T22 T32 T42 Business 3 T13 T23 T33 T43 Business 4 T14 T24 T34 T44

[0129] Among them, the feature model set includes all feature model sets T11, T12...T44.

[0130] The first feature model set T11 corresponding to the first user is used as the initial user model set T0. The next feature model set T12 in the feature model set is compared with T0 for difference. Based on the comparison result and the initial user model set T0, the target user model set is obtained. For example, based on the comparison result, the initial user model set T0 can be updated according to the feature value in T12 to obtain the target user model set T1.

[0131] Step 324: Determine whether the next feature model set in the feature model set is the last feature model set.

[0132] If not, go to step 325; if so, go to step 326.

[0133] Step 325: Use the target user model set as the new initial user model set and return to step 322.

[0134] Among them, it can be determined whether the next feature model set T12 in the feature model set is the last feature model set. If not, the target user model set T1 is used as the new initial user model set, and step 322 is repeated until the next feature model set in the feature model set is the last feature model set.

[0135] Specifically, for example, the next feature model set T13 in the feature model set is compared with T1 for difference to obtain the target user model set T2, and the target user model set T2 is used as the new initial user model set; the next feature model set T14 is compared with T2 to obtain the target user model set T3, the next feature model set T21 is compared with T3 to obtain the target user model set T4...the next feature model set T44 is compared with T14 to obtain the target user model set T15.

[0136] Step 326: Use the target user model set as the final user model set.

[0137] For example, the next feature model set T44 in the feature model set is the last feature model set, and the target user model set T15 obtained by comparing T44 with T14 is used as the final user model set.

[0138] In some embodiments, the various differences include fluctuation characteristic differences and preset relationship differences. Based on the comparison results and the initial user model set, a target user model set is obtained, which may specifically include the following situations:

[0139] The first method: If the comparison result shows that all differences are greater than the preset threshold value, the initial user model set is updated based on the next feature model set in the feature model set to obtain the target user model set.

[0140] For example, the feature model set of user 1's business type 1 Set the initial user model set T0=T11.

[0141] Feature model set of user 1's business type 2

[0142] Assume that the difference between the W' value in the feature model set of user 1's service type 1 and the W' value in the feature model set of user 1's service type 2 is greater than the preset threshold value, and If the difference is greater than the preset threshold, the initial user model set T0 is updated to obtain the target user model set:

[0143] in, This indicates that service type 1 and service type 2 of user 1 share a W1' value and a W2' value, and the W1' value and the W2' value are the values ​​of service type 2.

[0144] The second method: If the comparison result shows that all differences are less than the preset threshold value, the initial user model set is used as the target user model set.

[0145] For example, if the W' value in the feature model set of user 1's service type 1 is close to the W' value in the feature model set of user 1's service type 2, and h' is close, that is, the fluctuation feature difference and the preset relationship difference are both less than the preset threshold value, if the difference between the W' value in T and the W' value in T12 is less than the preset threshold value, and The difference between is less than the preset threshold, then the target user model set is T1=

[0146]

[0147] in, This indicates that service type 1 and service type 2 of user 1 share the same W1' value and W2' value, and the W1' value and W2' value are the values ​​in service type 1.

[0148] Among them, W4' is a unique feature of business type 2. Therefore, it is necessary to add W4' to the initial user model set to obtain the target user model set. If business type 2 does not have unique features, the initial user model set T0 is directly used as the target user model set.

[0149] The third method: If the comparison result shows that one of the differences is less than the preset threshold value, the next feature model set of the feature model set is added to the initial user model set to obtain the target user model set.

[0150] For example, if the comparison result is that the fluctuation feature difference is greater than the preset threshold value and the preset relationship difference is less than the preset threshold value; or the fluctuation feature difference is less than the preset threshold value and the preset relationship difference is greater than the preset threshold value, then the next feature model set in the feature model set is added to the initial user model set to obtain the target user model set.

[0151] For example, the difference between the W' value in the feature model set of user 1's service type 1 and the W' value in the feature model set of user 1's service type 2 is greater than the preset threshold value, and If the difference is less than the preset threshold, the feature model set of service type 2 is added to the initial user model set, and the target user model set is obtained as

[0152] All business models of user 1 are traversed in sequence, and the initial user model set T0 is updated according to the above comparison process.

[0153] For example, user 2’s business model 1 generates the following feature model set: The target user model set obtained after user 1 updates the initial user model set is

[0154] Then take T1 as the new initial user model set, compare T21 with T1, and if the difference between the W' values ​​and The difference between is less than the preset threshold, then the target user model set is

[0155] in, This indicates that service type 1 and service type 2 of user 1 and user 2 both share the same W1' value and W2' value, and the W1' value and W2' value are the values ​​in service type 1.

[0156] If the difference between the W' values ​​and The difference between the two values ​​is greater than the preset threshold, then the target user model set is T2=

[0157]

[0158] in, This indicates that service type 1 and service type 2 of user 1 and user 2 both share the same W1' value and W2' value, and the W1' value and W2' value are the values ​​in service type 1.

[0159] If the difference in W' value is greater than the preset threshold, and The difference between the values ​​of W' and W' is less than the preset threshold, or the difference between the values ​​of W' and W' is less than the preset threshold, and If the difference between T21 and T21 is greater than the preset threshold, T21 is added to the initial user model set, and the target user model set is obtained as

[0160] In the above process of comparing the fluctuation feature difference and the preset relationship difference of all feature model sets contained in the feature model set with the initial user model set T0, if the difference is greater than the preset threshold value, it means that the correlation is enhanced, and the initial user model set T0 can be updated or a new feature model set can be added to the initial user model set T0.

[0161] In some embodiments, the first target fluctuation feature that has been updated in the target user model set can be determined based on the comparison result; the second target fluctuation feature corresponding to the first target fluctuation feature in the new business models of multiple users ranked before the current user is obtained; and the second target fluctuation feature is updated based on the first target fluctuation feature.

[0162] For example, after updating the initial user model set, the target user model set is

[0163] Assuming that the user is the third user, and the first target fluctuation characteristics updated in the target user model set T2 are W1' and W2', then obtain the second target fluctuation characteristics (W1 and W2) corresponding to the first target fluctuation characteristics (W1' and W2') in the new business models of multiple users (first user and second user) that are ranked before the current user (third user), and synchronously update the fluctuation characteristic W1 in the first user and the second user to the value of W1', and synchronously update the fluctuation characteristic W2 in the first user and the second user to the value of W2'.

[0164] The first target fluctuation feature W1 ′ and the second target fluctuation feature W1 correspond to the same data to be measured, and the first target fluctuation feature W2 ′ and the second target fluctuation feature W2 correspond to the same data to be measured.

[0165] The preset threshold value can be set manually according to actual conditions.

[0166] In some embodiments, if the user is an edge user of a preset cell, the neighboring cell fluctuation characteristics corresponding to the user can be obtained first; then a new business model is generated based on the real-time characteristics, fluctuation characteristics of the cell where the user is located, and the neighboring cell fluctuation characteristics corresponding to the user.

[0167] For example, the eight test data in the above example are used for illustration. If the real-time features corresponding to six of the test data have a causal relationship, the fluctuation features of the six test data are obtained. Then, the new business model can be obtained using the following formula (9):

[0168]

[0169] Among them, U type(u) For new business models, D n (i) represents the nth real-time feature corresponding to the i-th quality difference point, W k (i) represents the kth fluctuation feature corresponding to the i-th quality difference point, W k '(i) represents the fluctuation characteristics of the kth neighboring area corresponding to the i-th quality difference point.

[0170] Among them, formula 9 can be simplified to obtain formula (10):

[0171] U type(u) =[D n∈(1,8) (n),W k∈(1,6) (k),W' k∈(1,6) (k)] Formula (10).

[0172] D n∈(1,8) (n) represents the nth real-time feature obtained by summing n1 quality difference points, Wk∈(1,6) (k) represents the kth fluctuation characteristic obtained by summing n1 quality difference points, W' k∈(1,6) (k) represents the fluctuation characteristics of the kth neighborhood obtained by summing the n1 quality difference points.

[0173] It should be noted that the characteristic value of the neighboring area fluctuation feature can be the same as or different from the characteristic value of the fluctuation feature. Considering that the fluctuation information of the neighboring areas will be relatively similar and for the convenience of calculation, this embodiment can set the characteristic value of the neighboring area fluctuation feature to the same characteristic value as the fluctuation feature.

[0174] In some embodiments, the target business model is a new business model, and performing root cause analysis on the business to be analyzed based on the target business model may specifically include the following steps:

[0175] Step 41: Obtain a first matrix based on all the test data of the new business model and the test data corresponding to the causal characteristics.

[0176] The first matrix can be:

[0177]

[0178] The first matrix d n (n1) Arrange all the data to be measured in the preset cell, d k (n1) Arrange the test data corresponding to the causal characteristics.

[0179] In some embodiments, if the neighboring cell fluctuation feature is added, the first matrix is ​​as follows:

[0180]

[0181] Among them, d' k (n1) is the neighboring cell data to be measured that affects the preset cell.

[0182] Step 42: Obtain a second matrix based on the real-time features and the fluctuation features in the new business model.

[0183]

[0184] Among them, D n (m) is arranged into real-time features in the new business model, W k (m) Arranged as the various fluctuation characteristics in the new business model.

[0185] In some embodiments, if the neighboring cell fluctuation feature is added, the second matrix is ​​as follows:

[0186]

[0187] Among them, W' k(m) is arranged as the fluctuation characteristics of the neighboring area.

[0188] Step 43: Perform matrix multiplication on the first matrix, the second matrix, and the preset number of root cause types to obtain the probability of each poor quality point in the preset number of root cause types.

[0189] The probability can be calculated using the following formula (11):

[0190] p=[n+k] T *[m][n+k] Formula (11).

[0191] Where p is the probability, T is the transpose, [n+k] T is the first matrix, [n+k] is the second matrix, [n+k] T For the specific value of [n+k], please refer to the above related content.

[0192] In some embodiments, if the neighboring cell fluctuation characteristics are added, the probability can be calculated using the following formula (12):

[0193] p=[n+2k] T *[m][n+2k] Formula (12).

[0194] Where p is the probability, T is the transpose, [n+2k] T is the first matrix, [n+2k] is the second matrix, [n+2k] T For the specific value of [n+2k], please refer to the above related content.

[0195] Step 44: Select the root cause with the highest probability as the root cause result of each quality difference point.

[0196] For example, quality difference point 1 has 5 root causes such as m1, m2, m3, m4, and m5, and the corresponding probabilities are p11, p12, p13, p14, and p15 respectively. If p13 is the largest, then m3 is the root cause result of quality difference point 1.

[0197] Step 45: Select the root cause with the most repetitions from the root cause results of all poor quality points as the root cause analysis result of the business to be analyzed.

[0198] For example, there are 10 quality difference points, among which the root cause result of quality difference point 1 is m3; the root cause result of quality difference point 2 is m1; the root cause result of quality difference point 3 is m3; the root cause result of quality difference point 4 is m3; the root cause result of quality difference point 5 is m2; the root cause result of quality difference point 6 is m3; the root cause result of quality difference point 7 is m3; the root cause result of quality difference point 8 is m2; the root cause result of quality difference point 9 is m3; the root cause result of quality difference point 10 is m3.

[0199] Since m3 has the largest number of repetitions, m3 is selected as the root cause analysis result of the business to be analyzed.

[0200] In combination with the above embodiments, the present application further provides a root cause analysis system, which includes a user model unit, a data reporting unit, and a root cause analysis unit;

[0201] The data reporting unit is used to obtain a plurality of test data of the service to be analyzed, the test data including base station data and user data;

[0202] The model difference between the preset threshold value type and the new business model and / or the fluctuation information corresponding to the new business model is used to dynamically update the current business model to the target business model;

[0203] The root cause analysis unit is used to perform root cause analysis on the business to be analyzed based on the target business model.

[0204] See Figure 4 The root cause analysis system 10 includes a user model unit 100, a data reporting unit 200 and a root cause analysis unit 300, wherein the user model unit 100 includes a model generation module 101, a model update module 102 and a view module 103, combined with Figure 5 , explaining the main functions provided by each module and the interactions between modules.

[0205] (1) The model generation module 101 provides a user model generation function.

[0206] a. Analyze the reported base station data and user data (i.e., the data to be tested) to obtain the real-time characteristics, causal characteristics, and effective time granularity of the data to be tested at each quality difference point.

[0207] b. Determine the fluctuation characteristics corresponding to the causal characteristics based on the effective time granularity and characteristic peak value.

[0208] c. Generate a complete user model based on real-time characteristics and fluctuation characteristics. When the user is an edge user, a complete user model can be generated by combining the real-time characteristics and fluctuation characteristics of the cell where the user is located and the fluctuation characteristics of the neighboring cells.

[0209] (2) The model updating module 102 provides a model correction function, and corrects the new business model based on the root cause analysis results.

[0210] (3) The view module 103 provides view management functions, as follows.

[0211] a. Provides a first-level perception and presentation function for quality difference analysis. First-level perception directly displays whether base stations and cells have poor quality, including but not limited to the number of poor quality events and the root cause of the poor quality.

[0212] b. Provides secondary root cause demarcation and presentation for quality degradation analysis. Secondary demarcation visually presents detailed root causes and their distribution for base stations and cells. This includes, but is not limited to, high load and poor coverage.

[0213] c. Provides a three-level in-depth presentation of root causes for poor quality analysis. This intuitively displays the root causes of base stations and cells. This includes, but is not limited to, HARQ retransmissions (not limited to high RI), scheduling packet accumulation (not limited to QoS selection failure), and more.

[0214] The data reporting unit 200 mainly provides the following functions:

[0215] a. Provide data and report it to the user model unit 100 to generate and update the user model.

[0216] b. Provide data and report it to the root cause analysis unit 300 for root cause analysis.

[0217] The root cause analysis unit 300 mainly provides the following functions:

[0218] (1) Provide root cause analysis task interface functionality. Receive task requests (including but not limited to network element ID, base station ID, etc.) and send task responses (including network element ID, base station ID, task progress, etc.) through the task interface. The task interface includes but is not limited to API and configuration file methods.

[0219] (2) Providing a root cause result generation function to generate a root cause result based on user model data and correlation data, wherein the correlation data can be the user data to be tested and the base station data.

[0220] a. Obtain the new service model, base station data, and user data, and perform matrix multiplication between each piece of data to be tested and the new service model data, i.e. [n+2k] T *[m][n+2k], obtain the root cause probabilities of m root causes, and select the root cause with the highest probability as the root cause of the poor quality record (i.e., the poor quality point). From the root cause results of all poor quality points, select the root cause with the most recurrences as the root cause analysis result for the service to be analyzed. From all the user's services to be analyzed, select the root cause with the most recurrences as the root cause for the user.

[0221] b. Calculate the root causes of all user model data in the same cell, and take the root cause with the largest number of root causes as the root cause of the cell.

[0222] (3) Provides in-depth root cause analysis. Based on the root cause results and the time of occurrence, the system processes the scheduling information within the base station within the time range of the poor quality event to generate further causes, including but not limited to HARQ retransmission (not limited to high RI), scheduling packet accumulation (not limited to QoS selection failure), etc.

[0223] Based on the above root cause analysis system, refer to Figure 5 The root cause analysis method provided in this application mainly includes the following steps:

[0224] S11: Acquire base station data and user data (i.e., data to be tested) at poor-quality points.

[0225] S12: Traverse data by cell.

[0226] S13: Determine causal relationship characteristics through time-lagged correlation analysis.

[0227] S14: Determine real-time features through cross-correlation analysis.

[0228] S15: Determine the fluctuation characteristics within the effective time period based on the causal relationship characteristics and the real-time characteristics.

[0229] Among them, the causal characteristics and effective time granularity are obtained according to the time lag correlation, and the fluctuation characteristics corresponding to the causal characteristics are determined.

[0230] S16: Determine the fluctuation characteristics within the effective time period of the corresponding neighboring area.

[0231] Among them, step S16 can be selected according to the situation. If the user is an edge user of the preset cell, the neighboring cell fluctuation characteristics corresponding to the user can be obtained; if the user is not an edge user of the preset cell, this step can be omitted.

[0232] S17: Finally generate a complete user model.

[0233] The business model to be analyzed is generated based on the real-time characteristics and fluctuation characteristics, and then all the business models to be analyzed of the same user are combined into a user model.

[0234] S18: Dynamically update the user model based on feature-level correlation.

[0235] Among them, the preset relationship expression can represent the feature-level correlation. For example, based on each new business model and the preset relationship expression, the feature model set corresponding to each new business model is obtained in turn, and one is selected from all feature model sets as the initial user model set. Usually, the first feature model set of the first user is selected as the initial user model set. Then, each feature model set is iteratively compared with the initial user model set, and the initial user model set is processed based on the comparison results to obtain the target user model set.

[0236] In the process of comparing the difference between the remaining feature model sets and the initial user model set according to the user sequence number, the new business models of multiple users ranked before the current user are batch updated based on the comparison results to achieve dynamic update of user models according to feature-level relevance.

[0237] S19: Perform root cause analysis based on the user model data and the correlation of the root causes to generate the root causes.

[0238] The root cause analysis process may refer to the relevant contents of the above embodiment.

[0239] S20: Aggregate the root causes and generate a root cause analysis view.

[0240] The root cause analysis method and system provided above can be applied to campuses, residential areas, transportation stations, scenic spots, stadiums, and other places. For specific application scenarios, please refer to the following.

[0241] Application Scenario 1: Automatically analyze the root causes of poor quality and high load on 5G base stations and cells within the range of traffic stations.

[0242] 1) Users in the cell send data reporting tasks through the API, including but not limited to periodic reporting and event-triggered reporting.

[0243] 2) The data reporting unit 200 receives the task request and sends the base station data and user data to the user model unit.

[0244] 3) The user model unit 100 receives base station data and user data (i.e., data to be tested), traverses the cells corresponding to the quality difference points, obtains real-time features based on current correlation analysis, obtains causal features and effective time granularity based on time lag correlation, determines the fluctuation features corresponding to the causal features, and generates a user model based on the real-time features and fluctuation features.

[0245] 4) After receiving the model update request, the root cause analysis unit 300 updates the current user model, generates a root cause result based on the user model and base station data, and reports the root cause result to the view module 103 of the user model unit 100.

[0246] 5) The data view module receives the data presentation request and presents the root cause data in a hierarchical manner.

[0247] See Figure 6 , Figure 6 1 is a schematic diagram of the structure of an embodiment of a computer-readable storage medium provided by the present application. The computer-readable storage medium 70 is used to store a computer program 71. When the computer program 71 is executed by a processor, it is used to implement the following method steps:

[0248] Acquire multiple test data of the service to be analyzed, and generate a new service model based on the multiple test data, the test data including base station data and user data;

[0249] Based on the model difference between the current business model and the new business model and / or the fluctuation information corresponding to the new business model, the current business model is dynamically updated to the target business model, so as to perform root cause analysis on the business to be analyzed based on the target business model.

[0250] It can be understood that when the computer program 71 is executed by the processor, it is also used to implement the technical solution of any embodiment in the present application.

[0251] In the several embodiments provided in this application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of modules or units is merely a logical functional division. In actual implementation, other division methods may be used. For example, two units or components may be combined or integrated into another system, or some features may be ignored or not implemented.

[0252] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across two network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0253] In addition, each functional unit in each embodiment of the present application may be integrated into a processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The above-mentioned integrated units may be implemented in the form of hardware or software functional units.

[0254] If the integrated units in the above other embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or 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 enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the various implementation methods of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0255] The above is only an implementation method of the present application and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the description and drawings of this application, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A root cause analysis method, characterized in that: The method comprises: Acquire a plurality of test data of a service to be analyzed, and generate a new service model based on the plurality of test data, wherein the test data includes base station data and user data; Based on the model difference between the current business model and the new business model and / or the fluctuation information corresponding to the new business model, the current business model is dynamically updated to a target business model, so as to perform a root cause analysis on the business to be analyzed based on the target business model.

2. The method according to claim 1, characterized in that Generating a new business model based on the plurality of data to be tested includes: sequentially determining the real-time characteristics of each of the test data in each corresponding quality difference point, wherein each of the test data corresponds to a plurality of quality difference points; Obtaining a fluctuation characteristic of each of the data to be measured based on a real-time characteristic of each of the data to be measured and a preset time offset; The new business model is generated based on a plurality of real-time features of the data to be tested and a plurality of fluctuation features of the data to be tested.

3. The method according to claim 2, characterized in that The step of sequentially determining the real-time characteristics of each of the test data at each corresponding poor quality point includes: Determine the sum of the mean values ​​of all the data to be measured in each of the poor quality points; Based on the sum of the mean values, sequentially determining the real-time features corresponding to each of the data to be measured in each of the poor quality points; All the quality difference points corresponding to each of the data to be tested are traversed to obtain the real-time features of each of the data to be tested in all the quality difference points.

4. The method according to claim 2, characterized in that The step of obtaining the fluctuation characteristics of each of the data to be measured based on the real-time characteristics of each of the data to be measured and the preset time offset includes: Based on the preset time offset, sequentially determining a correlation value between each of the to-be-tested data and the remaining to-be-tested data in the to-be-analyzed service; Based on the correlation value, relevant test data corresponding to each of the test data are screened out from the remaining test data of the business to be analyzed, and the real-time feature corresponding to each of the relevant test data is used as the causal feature corresponding to each of the relevant test data; Based on all the causal features corresponding to each of the data to be tested, multiple fluctuation features corresponding to each of the data to be tested are determined.

5. The method according to claim 4, characterized in that The determining, based on all causal features corresponding to each of the data to be tested, a plurality of fluctuation features corresponding to each of the data to be tested, includes: determining the effective time granularity of each of the causal features; Based on the effective time granularity, sequentially determining the maximum causal relationship value corresponding to each of the to-be-tested data in all the quality difference points; Based on the multiple maximum causal relationship values ​​corresponding to each of the to-be-tested data and the time corresponding to each of the maximum causal relationship values, multiple fluctuation features corresponding to each of the to-be-tested data are determined.

6. The method according to claim 5, characterized in that Determining the effective time granularity of each causal feature includes: Determine a first maximum value to be measured of the current data to be measured among all the poor-quality points, and a first time point corresponding to the first maximum value to be measured; Acquire relevant data to be tested corresponding to the current data to be tested; Determine a second maximum value to be measured of the relevant data to be measured among all the quality difference points and a second time point corresponding to the second maximum value to be measured; The effective time granularity of the current causal feature is determined based on the first time point and the second time point.

7. The method according to claim 6, characterized in that The determining the effective time granularity of the current causal feature based on the first time point and the second time point includes: If the number of the relevant data to be tested is 1, the time difference between the first time point and the second time point is used as the effective time granularity of the current causal feature.

8. The method according to claim 6, characterized in that The determining of the effective time granularity of the current causal feature based on the first time point and the second time point further includes: If the number of the relevant data to be tested is at least two, obtaining a second time point corresponding to each of the relevant data to be tested; sequentially determining time differences between the first time point and a plurality of the second time points; The largest time difference is selected from all the time differences as the effective time granularity of the current causal feature.

9. The method according to any one of claims 1 to 8, characterized in that After generating a new business model based on the plurality of data to be tested, the method further includes: The causal characteristics and fluctuation characteristics of the new business model are matched with the causal characteristics and fluctuation characteristics of the current business model respectively, and the difference is calculated to obtain the model difference between the new business model and the current business model.

10. The method according to any one of claims 2 to 8, characterized in that: The preset cell includes multiple users, each of the users includes multiple services to be analyzed, and the method further includes: Determine a new business model corresponding to the business to be analyzed of each of the users, and determine a feature model set corresponding to each of the new business models based on the new business model and a preset relationship, to obtain a feature model set, wherein the feature model set includes all the feature model sets; Based on the feature model set, an end-user model set is generated.

11. The method according to claim 10, characterized in that Generating a final user model set based on the feature model set includes: Determine a first user and a first feature model set corresponding to the first user in the feature model set based on a user identification number and a generation time number of the feature model set, and use the first feature model set as an initial user model set; Comparing the difference between the next feature model set in the feature model set and the initial user model set in sequence; Based on the comparison result and the initial user model set, a target user model set is obtained; Determining whether the next feature model set in the feature model set is the last feature model set; If not, the target user model set is used as a new initial user model set, and the process returns to the step of sequentially comparing the difference between the next feature model set in the feature model set and the initial user model set; If so, the target user model set is used as the final user model set.

12. The method according to claim 11, characterized in that The step of obtaining a target user model set based on the comparison result and the initial user model set includes: If the comparison result shows that all the differences are greater than the preset threshold value, the initial user model set is updated based on the next feature model set in the feature model set to obtain the target user model set, wherein the differences include the fluctuation feature difference and the preset relationship difference; If the comparison result shows that all differences are less than the preset threshold value, the initial user model set is used as the target user model set; If the comparison result shows that one of the differences is smaller than the preset threshold value, the next feature model set in the feature model set is added to the initial user model set to obtain the target user model set.

13. The method according to claim 11, characterized in that The method further comprises: Determining a first target fluctuation feature that is updated in the target user model set based on the comparison result; Obtaining a second target fluctuation feature corresponding to the first target fluctuation feature in new business models of multiple users ranked before the current user; The second target fluctuation feature is updated based on the first target fluctuation feature.

14. The method according to any one of claims 11 to 13, characterized in that: The method further comprises: If the user is an edge user of the preset cell, obtaining the neighboring cell fluctuation characteristics corresponding to the user; A new service model is generated based on the real-time characteristics, the fluctuation characteristics, and the neighboring cell fluctuation characteristics.

15. The method according to any one of claims 2 to 8, characterized in that: The target business model is the new business model, and performing root cause analysis on the business to be analyzed based on the target business model includes: Obtaining the distribution probability of all the poor quality points in a preset number of root cause types; Selecting the root cause with the greatest probability from the distribution probability as the root cause result of each quality difference point; The root cause with the greatest number of repetitions is selected from the root cause results of all the poor quality points as the root cause analysis result of the service to be analyzed.

16. The method according to claim 15, characterized in that The obtaining of the distribution probability of all the poor quality points in a preset number of root cause types includes: Obtaining a first matrix based on all the to-be-tested data of the new business model and the to-be-tested data corresponding to the causal feature; Obtaining a second matrix based on each of the real-time characteristics and each of the fluctuation characteristics in the new business model; The first matrix, the second matrix, and a preset number of root cause types are matrix-multiplied to obtain the distribution probability of all the poor quality points in the preset number of root cause types.

17. A root cause analysis system, characterized in that: The root cause analysis system includes a user model unit, a data reporting unit and a root cause analysis unit; The data reporting unit is used to obtain a plurality of test data of the service to be analyzed, wherein the test data includes base station data and user data; The user model unit is used to generate a new business model based on the plurality of data to be tested; and dynamically updating the current business model to a target business model based on a model difference between the current business model and the new business model and / or fluctuation information corresponding to the new business model; The root cause analysis unit is configured to perform a root cause analysis on the business to be analyzed based on the target business model.

18. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 16 can be implemented.

Images