An intelligent problem analysis method, device, equipment and storage medium

Abstract

The embodiments of the present disclosure provide an intelligent problem analysis method, device, equipment and storage medium. The method includes: obtaining data to be analyzed and a problem analysis model obtained by pre-training; determining the input characteristic value corresponding to the problem analysis model according to the data to be analyzed; inputting the input characteristic value into the problem analysis model to determine the problem analysis result. According to the scheme provided by the embodiments of the present disclosure, the problem analysis is performed to obtain the analysis result based on the standardized input characteristic value corresponding to the model by using the intelligent problem analysis model obtained by pre-training, which can make full use of the analysis experience of previous experts, significantly improve the work efficiency of communication equipment in performing various types of problem analysis and positioning in testing or commercial use, and at the same time reduce the technical threshold for related problem analysis and positioning in the communication system.

Description

Technical Field

[0001] The present disclosure relates to, but is not limited to, the field of communications, and specifically to an intelligent problem analysis method, device, equipment, and storage medium. Background Art

[0002] As wireless communication technology enters the 5G era, commercial wireless communication networks operated by different operators in different countries and regions cover a variety of different standards, including 2G (such as GSM, CMDA), 3G (such as WCDMA, CMDA-2000, TD-SCDMA), 4G (also known as LTE), and 5G (also known as NR). Moreover, even for the same network standard, taking 4G and 5G as examples, the frequency bands and bandwidths supported by the networks of different operators in different countries and regions will be different for various reasons. Accordingly, in order to improve market competitiveness, mainstream commercial mobile phones on the modern market usually need to support different configurations of these different communication standards, frequency bands, and bandwidths at the same time. These requirements make the wireless communication systems in mobile phones more and more complex.

[0003] Before or even after a mobile phone is launched on the market, it is usually necessary to fully test and verify the performance of the wireless communication system, continuously eliminate defects, and optimize performance. This process includes internal laboratory testing, government certification testing, operator laboratory and field certification testing, as well as large-scale field testing in different countries and regions around the world. This process usually consumes huge manpower and material resources, and covers the entire process of R&D, testing, and commercial use. Analyzing and locating related problems that arise during the testing phase and the formal use of the equipment is a highly professional job that requires high technical personnel. They need to have theoretical knowledge in the field of wireless communications and practical R&D experience in related products. Under special conditions, they also need to have relevant knowledge and experience in different communication standards and modules. As a result, the analysis of related problems is highly dependent on personnel and is inefficient.

[0004] How to improve the automation and intelligence level of problem analysis in related systems and improve the efficiency of analysis work is an urgent problem to be solved in this field. Summary of the invention

[0005] The embodiments of the present disclosure provide an intelligent problem analysis method, apparatus, device and storage medium, which utilize a pre-trained intelligent problem analysis model to perform problem analysis based on standardized input feature values ​​corresponding to the model to obtain analysis results. This method can fully utilize the analysis experience of previous experts, significantly improve the work efficiency of communication equipment in performing various types of problem analysis and locating during testing or commercial use, and at the same time reduce the technical threshold for related problem analysis and locating in communication systems.

[0006] The present disclosure provides an intelligent problem analysis method, including:

[0007] Obtain the data to be analyzed and the pre-trained problem analysis model;

[0008] Determine the input feature value corresponding to the problem analysis model according to the data to be analyzed;

[0009] Inputting the input feature value into the problem analysis model to determine the problem analysis result;

[0010] The problem analysis model includes at least one of the following models:

[0011] The first problem analysis model is applicable to problems such as communication system parameter configuration errors, the second problem analysis model is applicable to problems such as communication system event execution order errors, the third problem analysis model is applicable to problems such as abnormal channel environment changes in the communication system, and the fourth problem analysis model is applicable to problems such as limited device processing capabilities.

[0012] The present disclosure also provides an intelligent problem analysis device, including:

[0013] An acquisition module, configured to acquire data to be analyzed and a pre-trained problem analysis model;

[0014] An analysis module is configured to determine an input feature value corresponding to the problem analysis model according to the data to be analyzed; input the input feature value into the problem analysis model to determine a problem analysis result;

[0015] The problem analysis model includes at least one of the following models:

[0016] The first problem analysis model is applicable to problems such as communication system parameter configuration errors, the second problem analysis model is applicable to problems such as communication system event execution order errors, the third problem analysis model is applicable to problems such as abnormal channel environment changes in the communication system, and the fourth problem analysis model is applicable to problems such as limited device processing capabilities.

[0017] The present disclosure also provides an electronic device, including:

[0018] one or more processors;

[0019] a storage device for storing one or more programs,

[0020] When the one or more programs are executed by the one or more processors, the one or more processors implement the intelligent problem analysis method as described in any embodiment of the present disclosure.

[0021] The embodiments of the present disclosure also provide a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the intelligent problem analysis method as described in any embodiment of the present disclosure is implemented.

[0022] Other aspects will be apparent upon reading and understanding the drawings and detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the solutions of the present application. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying any creative work.

[0024] Figure 1 A flow chart of an intelligent problem analysis method in an embodiment of the present disclosure;

[0025] Figure 2 A schematic diagram of an intelligent problem analysis method in an embodiment of the present disclosure;

[0026] Figure 3 A schematic diagram of a channel environment related parameter change in an embodiment of the present disclosure;

[0027] Figure 4 This is another schematic diagram of the change of channel environment related parameters in an embodiment of the present disclosure;

[0028] Figure 5 A schematic diagram of a capability threshold corresponding to a channel parameter function in an embodiment of the present disclosure;

[0029] Figure 6 This is a flow chart of another intelligent problem analysis method in an embodiment of the present disclosure;

[0030] Figure 7 It is a structural schematic diagram of an intelligent problem analysis device in an embodiment of the present disclosure.

[0031] The realization of the purpose, functional features and advantages of the present application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0033] It should be noted that all directional indications in the embodiments of the present disclosure (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0034] In addition, in the present disclosure, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0035] In the present disclosure, unless otherwise clearly specified and limited, the terms "connection", "fixation", etc. should be understood in a broad sense. For example, "fixation" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present application scheme can be understood according to specific circumstances.

[0036] In addition, the technical solutions between the various embodiments of the present disclosure can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0037] Before describing the specific embodiments, the abbreviations of the relevant terms involved in the present disclosure are explained as follows:

[0038]

[0039]

[0040]

[0041]

[0042] It can be seen that with the increasing complexity of wireless communication systems, the interweaving of multiple factors has made it increasingly difficult to analyze and locate various problems that occur during the testing or formal use of communication equipment. The requirements for the professional level and experience of technicians are getting higher and higher, and the training of such professional talents is relatively slow, resulting in a shortage of personnel. Therefore, when faced with problems, the efficiency of analysis and location is naturally low.

[0043] The present disclosure provides an intelligent problem analysis method. Figure 1 As shown, including:

[0044] Step 110, obtaining the data to be analyzed and the pre-trained problem analysis model;

[0045] Step 120, determining the input feature value corresponding to the problem analysis model according to the data to be analyzed;

[0046] Step 130, inputting the input feature value into the problem analysis model to determine the problem analysis result;

[0047] The problem analysis model includes at least one of the following models:

[0048] The first problem analysis model is applicable to problems such as communication system parameter configuration errors, the second problem analysis model is applicable to problems such as communication system event execution order errors, the third problem analysis model is applicable to problems such as abnormal channel environment changes in the communication system, and the fourth problem analysis model is applicable to problems such as limited device processing capabilities.

[0049] In some exemplary embodiments, the problem analysis model is pre-trained according to the following method:

[0050] According to the problem type applicable to the problem analysis model, data items for analyzing the problem type are obtained from the sample data to be analyzed as input feature values ​​of the training sample;

[0051] Marking the problem analysis results of the training samples;

[0052] The problem analysis model is trained using a plurality of the training samples.

[0053] In some exemplary embodiments, the problem analysis results of the training samples marked with the communication system parameter configuration error problem type include: the cause of the problem and a reasonable parameter value range.

[0054] In some exemplary embodiments, the problem analysis results of the training samples marked with the communication system event execution sequence error problem include: the cause of the problem and a reasonable event execution sequence.

[0055] In some exemplary embodiments, the problem analysis results of the training samples marked as abnormal communication system channel environment change problems include: the cause of the problem and a reasonable parameter variation range.

[0056] In some exemplary embodiments, the problem analysis results of the training samples marked as having limited device processing capability include: the cause of the problem.

[0057] In some exemplary embodiments, the first problem analysis model includes one of the following models or a combination of at least two of the following models: a Bayesian classifier model, a linear classifier model, a decision tree model, a support vector machine model, and a neural network model.

[0058] In some exemplary embodiments, the second problem analysis model includes one of the following models or a combination of at least two of the following models: a Markov chain and a recurrent neural network (RNN) model.

[0059] In some exemplary embodiments, the third problem analysis model includes one of the following models or a combination of at least two of the following models: a Markov chain, a recurrent neural network RNN ​​model, a deep neural network DNN model, and a convolutional neural network CNN model.

[0060] In some exemplary embodiments, the fourth problem analysis model includes one of the following models or a combination of at least two of the following models: a Bayesian classifier model, a linear classifier model, a decision tree model, a support vector machine model, and a deep neural network DNN model.

[0061] In some exemplary embodiments, the recurrent neural network includes: ordinary recurrent neural network (RNN), long short-term memory neural network (LSTM), gated recurrent unit neural network (GRU), etc.

[0062] It can be seen that the analysis model for the first problem belongs to a classification recognition model. According to the input feature value, the corresponding classification is determined to obtain the corresponding analysis result. The analysis model for the second problem belongs to a classification and sequence order recognition model. According to the input feature value, the corresponding classification is determined to obtain the corresponding analysis result. The analysis model for the third problem belongs to a classification and analysis recognition model with continuous values ​​of some parameters. According to the input feature value, the corresponding classification is determined to obtain the corresponding analysis result. The analysis model for the fourth problem belongs to a classification recognition model. According to the input feature value, the corresponding classification is determined to obtain the corresponding analysis result.

[0063] In some exemplary embodiments, the first problem analysis model and the fourth problem analysis model may use the same model, and train according to sample data including two corresponding types of problems to obtain a trained problem analysis model; or use two models of the same type, and train according to their respective corresponding sample data to obtain two trained problem analysis models of the same type; or use different types of models, and train according to their respective corresponding sample data to obtain two trained problem analysis models of different types. The selection of specific problem analysis models and training schemes are determined according to needs and are not limited to a specific method.

[0064] In some exemplary embodiments, the data to be analyzed includes: communication application layer data and communication protocol layer log data.

[0065] In some exemplary embodiments, the communication application layer data includes one or more of the following:

[0066] Service type, service problem type, time information, location information, operator information, communication network equipment information, channel environment information, equipment motion status information and communication system version information.

[0067] In some exemplary embodiments, the service type indicates the type of service currently being executed by the device, including but not limited to one or more of the following: data transmission rate test, web browsing, voice call, video call, online game, normal standby, and others.

[0068] In some exemplary embodiments, the service problem type indicates the type of service problem currently occurring in the device, including but not limited to one or more of the following: data transmission rate problems (uplink or downlink), bit error rate (Bit Error Rate, BER) or block error rate (Block Error Rate, BLER) problems (uplink or downlink), random access failure, wireless link failure problems, cell switching failure problems, voice call problems, paging problems, etc.

[0069] In some exemplary embodiments, the time information indicates the time when the data is collected or the time when the problem occurs; for example, year, month, day, hour, minute, second, etc.

[0070] In some exemplary embodiments, the location information indicates the location of the device, including but not limited to one or more of the following: longitude and latitude, country, city, laboratory, field, urban area, suburbia, etc.

[0071] In some exemplary embodiments, the operator information indicates operator information for improving wireless communication services.

[0072] In some exemplary embodiments, the communication network device information indicates device information that generates the data to be analyzed, including but not limited to one or more of the following: manufacturer, model, etc.

[0073] In some exemplary embodiments, the channel environment information includes but is not limited to one or more of the following: channel type, channel parameters, etc. The channel type includes but is not limited to one or more of the following: ideal channel, natural channel and artificial channel.

[0074] In some exemplary embodiments, the device motion state information includes but is not limited to one or more of the following: operation type, speed, etc., wherein the motion type includes but is not limited to one or more of the following: high-speed rail travel, car travel, bicycle riding, walking, and stationary.

[0075] In some exemplary embodiments, the communication system version information indicates version information of the communication system on the device generating the data to be analyzed, including but not limited to one or more of the following: software version, hardware version, etc.

[0076] In some exemplary embodiments, when the accurate content of one or more data items cannot be obtained, the one or more data items may not be included in the communication application layer data, or may be set to default values.

[0077] In some exemplary embodiments, the communication application layer data may be automatically collected by an application system on a device; or, all or part of the content may be manually collected in conjunction with a user operation interface, without being limited to a specific form.

[0078] In some exemplary embodiments, the communication application layer data is also referred to as system status information.

[0079] In some exemplary embodiments, the communication protocol layer log data includes one or more of the following:

[0080] Service Data Adaptation Protocol SDAP layer log data, non-access NAS layer log data, radio resource control RRC layer log data, packet data convergence protocol PDCP layer log data, radio link layer control RLC layer log data, media access MAC layer log data and physical PHY layer log data.

[0081] In some exemplary embodiments, the communication protocol layer log data is also referred to as system configuration parameters.

[0082] In some exemplary embodiments, the communication protocol layer log data comes from a log file in the communication system. It can be known that the log file records various public configuration data and / or private configuration data of the device during the operation of the communication system, static configuration data and / or dynamic configuration data, and is an important source of information for problem analysis and location.

[0083] It can be seen that the communication protocol layer log data includes data at various levels corresponding to the communication protocol. Different wireless communication protocols correspond to different protocol levels, which is not limited to the aspects illustrated in the embodiments of the present disclosure.

[0084] In some exemplary embodiments, the log data of each communication protocol layer includes: public configuration data and / or proprietary configuration data.

[0085] In some exemplary embodiments, the physical PHY layer public configuration data includes but is not limited to one or more of the following:

[0086] Cell frequency band (Band);

[0087] Cell frequency (Frequency);

[0088] Cell bandwidth (Bandwidth);

[0089] Sub-Carrier Spacing (SCS)

[0090] The uplink and downlink configuration mode of the cell: TDD / FDD. If it is TDD, the specific uplink and downlink configuration information;

[0091] Uplink transmit power parameter semi-static configuration (such as the parameter P defined in 3GPP document 38.213) O_NOMINAL_PUSCH,f,c (j), P O_UE_PUSCH,b,f,c (j), α b,f,c (j) etc.).

[0092] In some exemplary embodiments, the proprietary configuration data in the log data of each communication protocol layer can more accurately reflect the real-time channel status and scheduling information, current working status, etc. of the UE.

[0093] For example, the physical PHY layer specific configuration data includes but is not limited to one or more of the following:

[0094] Downlink path loss (Pathloss) measured by the terminal;

[0095] Downlink reference signal receiving power (RSRP) measured by the terminal;

[0096] The downlink received signal strength indicator (RSSI) measured by the terminal;

[0097] The signal to interference plus noise ratio (SINR) of the downlink signal measured by the terminal;

[0098] Channel State Information (CSI) of the downlink signal measured by the terminal;

[0099] Doppler shift of the downlink signal measured by the terminal;

[0100] Doppler spread of the downlink signal measured by the terminal;

[0101] The terminal's uplink timing advance (TA);

[0102] Time domain scheduling information of uplink PUSCH channel;

[0103] Frequency domain scheduling information of uplink PUSCH channel;

[0104] Reference symbol parameter configuration for uplink PUSCH channel;

[0105] Modulation and Coding Scheme (MCS) of the uplink PUSCH channel;

[0106] Uplink PUSCH channel transmit power;

[0107] HARQ information of uplink PUSCH channel;

[0108] Time domain scheduling information of downlink PDSCH channel;

[0109] Frequency domain scheduling information of downlink PDSCH channel;

[0110] Reference symbol parameter configuration of downlink PDSCH channel;

[0111] Modulation and Coding Scheme (MCS) of the downlink PDSCH channel;

[0112] Decoding information of downlink PDCCH channel.

[0113] It should be noted that the data to be analyzed are generated by a device that has a problem and needs to be analyzed and located. The device may be a terminal device or a network-side device.

[0114] In some exemplary embodiments, the data to be analyzed may include data corresponding to one acquisition time point in each specific data item, or may include data corresponding to multiple acquisition time points.

[0115] In some exemplary embodiments, the data to be analyzed may also include a composite data item obtained by merging or combining multiple data. For example, based on the road loss data at multiple consecutive acquisition time points, the road loss continuous values ​​are combined in time order; for another example, based on the event information at multiple time points, the ordered event set is combined in time order. Based on this example, various other specific data items can be inferred, which is not limited to the aspects of the examples of the present disclosure.

[0116] In some exemplary embodiments, the composite data item can also be obtained by merging or combining multiple data in the original data to be analyzed according to the input requirements of the problem analysis model by the device performing the problem analysis in step 120, and is not limited to being generated by the source device where the problem occurs. It is flexibly selected according to the business environment and requirements for problem analysis positioning, and is not limited to a specific method.

[0117] In some exemplary embodiments, the data to be analyzed are recorded in a set format by the device where the problem occurs and needs to be analyzed and located, forming a standardized collection result of the data to be analyzed for subsequent problem analysis and location.

[0118] In some exemplary embodiments, the pre-trained analysis model can be loaded into the device and run locally on the device to obtain analysis results; optionally, the data to be analyzed can also be transmitted to a third-party device or platform via a data line or network, and then input into the pre-trained analysis model on the third-party device or platform to obtain analysis results. Specific aspects are flexibly selected according to the business environment and requirements of the problem analysis and positioning, and are not limited to a specific method.

[0119] It should be noted that the sample data to be analyzed used for pre-training of each problem analysis model is the standardized data to be analyzed collected when the corresponding type of problem occurs in each model before training. Based on each sample data to be analyzed, according to expert experience or analysis tools, one or more data items are extracted as the input feature values ​​of the training sample, and the corresponding problem analysis results are marked (that is, the output value of the training sample is marked). Based on the training samples of multiple marked problem analysis results, model training is performed according to the training method of the selected problem analysis model to obtain the trained problem analysis.

[0120] In some exemplary embodiments, obtaining a pre-trained problem analysis model in step 110 includes: determining a corresponding pre-trained problem analysis model based on a preliminary estimate of the type of problem currently occurring. That is, different problem types may correspond to the same problem analysis model or may correspond to different problem analysis models, specifically, the determination is based on the selection of the problem analysis model.

[0121] In some exemplary embodiments, the problem types include: communication system parameter configuration errors, communication system event execution sequence errors, communication system channel environment abnormal changes, and device processing capability limitations.

[0122] In some exemplary embodiments, the communication system parameter configuration error category refers to related problems caused by improper configuration of one or more system parameters. It should be noted that the improper system parameter configuration described in some exemplary embodiments is an improper configuration related to the current system comprehensive operation state, and is not a conclusion drawn by simply obtaining the current configuration of the system parameters and comparing it with a certain set static standard range. It needs to be comprehensively determined based on one or more other system status information and / or one or more other system configuration parameters.

[0123] For example, taking the analysis of the problem of too low uplink data transmission rate under additive white Gaussian noise channel conditions as an example, according to the analysis of experts or professional analysis tools, it is known that the causes of the problem may include: certain power parameters are too low, or the path loss in the use environment is too large.

[0124] For this type of problem, multiple sample data are collected in advance, and after marking the samples according to the analysis results, the selected first problem analysis model is trained to obtain a trained first problem analysis model, and the model is determined in steps 110-130, and the input feature values ​​corresponding to the model are further determined and input into the model to obtain the problem analysis results. The problem results marked by each training sample include: the cause of the problem and the reasonable parameter value range; or, also include: the unreasonable parameter value corresponding to the training sample.

[0125] In some exemplary embodiments, the communication system event execution order error class refers to related problems caused by improper execution order of one or more groups of system events. The events include: software events and / or hardware events. Among them, software events include: soft interrupts in programs, message transmissions, etc., and hardware events include: hard interrupts in hardware, the start of hardware tasks, the end of hardware tasks, etc. It should be noted that the improper execution order of system events described in some exemplary embodiments is an inappropriateness related to the current comprehensive operating state of the system. It is not a simple conclusion obtained by comparing the current execution order of related events with a set static standard order, but requires a comprehensive judgment based on one or more other system status information and / or one or more other system configuration parameters.

[0126] For this type of problem, multiple sample data are collected in advance, and after marking the samples according to the analysis results, the selected second problem analysis model is trained to obtain a trained second problem analysis model, and the model is determined in steps 110-130, and the input feature values ​​corresponding to the model are further determined and input into the model to obtain the problem analysis results. Among them, the problem results marked by each training sample include: the cause of the problem and the reasonable event execution sequence; or, also include: the unreasonable event execution sequence corresponding to the training sample.

[0127] In some exemplary embodiments, the abnormal change of the channel environment of the communication system refers to related problems caused by changes in the channel environment. For example, the Radio Link Failure (RLF) problem caused by the terminal moving to an area with poor signal coverage of the wireless communication cell can be determined by checking the changes in the downlink path loss, signal to interference and noise ratio, etc. measured by the terminal that can reflect the downlink signal quality of the cell. However, the radio link failure may also be caused by other reasons. Therefore, if it is necessary to determine that the problem is caused by poor cell signal coverage and exclude the radio link failure caused by other reasons, it is necessary to comprehensively analyze one or more other system status information and / or one or more other system configuration parameters.

[0128] For this type of problem, multiple sample data are collected in advance, and after marking the samples according to the analysis results, the selected third problem analysis model is trained to obtain a trained third problem analysis model, and the model is determined in steps 110-130, and the input feature values ​​corresponding to the model are further determined and input into the model to obtain the problem analysis results. Among them, the problem results marked by each training sample include: the cause of the problem and the reasonable parameter variation range; or, also include: the unreasonable parameter variation range corresponding to the training sample.

[0129] In some exemplary embodiments, the problem of limited device processing capability refers to related problems caused by limited device processing capability. For example, a common paging access problem may be caused by the fact that the received signal at the paging time exceeds the demodulation capability of the terminal during the paging access time, which leads to the failure of paging access. According to expert experience, it is necessary to conduct a comprehensive analysis based on multiple data items in the collected data to be analyzed in order to determine this reason, rather than other reasons; that is, it is necessary to make a comprehensive judgment based on multiple system status information and / or multiple system configuration parameters.

[0130] For this type of problem, multiple sample data are collected in advance, and after marking the samples according to the analysis results, the selected fourth problem analysis model is trained to obtain a trained fourth problem analysis model, and the model is determined in steps 110-130, and the input feature values ​​corresponding to the model are further determined and input into the model to obtain the problem analysis results. Among them, the problem results marked by each training sample include: the cause of the problem.

[0131] In some exemplary embodiments, the working process of the intelligent problem analysis method is as follows: Figure 2 As shown. Get the input feature value corresponding to the model, and input it into the problem analysis model to obtain the problem analysis result.

[0132] In some exemplary embodiments, taking the training and application of the first problem analysis model as an example:

[0133] The first problem analysis model is applicable to the problem of communication system parameter configuration errors. For example, it can be applied to the analysis of the problem of too low uplink data transmission rate under additive white Gaussian noise channel conditions. More specifically, it includes:

[0134] a. Some power parameters are too low;

[0135] b. The path loss in the communication environment is too large.

[0136] Analysis shows that the problem in case a is caused by the semi-static configuration parameter P of the uplink transmission power parameter. O_MONIMAL_PUSCH,f,c (j) (from 3GPP document 38.213) is caused by low configuration. After analyzing and solving this problem, it is necessary to form a training sample based on the analyzed data and the analysis results, extract the data items that are related to the analysis of the problem and need to be checked from the sample data to be analyzed that has been collected during the problem analysis process, form the input feature value of the test sample, and mark the problem analysis result corresponding to the test sample. Among them, the input feature value of the test sample includes multiple data items from the sample data to be analyzed, including data items in the communication application layer data and data items in the communication protocol layer log data; the problem analysis results include: the cause of the problem, a reasonable parameter value range, and the current unreasonable parameter value.

[0137] For example, a test sample in case a is shown in Table-1:

[0138] Table-1

[0139]

[0140] Accumulate multiple collection data and analysis results in case a to form multiple training samples.

[0141] Analysis shows that situation b is caused by excessive path loss in the communication environment. Similarly, after analyzing and solving this problem, it is necessary to form training samples based on the analyzed data and analysis results, extract data items related to the analysis of the problem and that need to be checked from the sample data to be analyzed that have been collected during the problem analysis process, form the input feature values ​​of the test sample, and mark the problem analysis results corresponding to the test sample. Among them, the input feature values ​​of the test sample include multiple data items from the sample data to be analyzed, including data items in the communication application layer data and data items in the communication protocol layer log data; the problem analysis results include: the cause of the problem, a reasonable parameter value range, and current inappropriate parameter values.

[0142] For example, a test sample in case b is shown in Table-2:

[0143] Table-2

[0144]

[0145] Accumulate multiple collection data and analysis results in case b to form multiple training samples.

[0146] For communication system parameter configuration errors, there are many other problems caused by parameter configuration errors. Similarly, according to the above examples, it can be known that historical collection data and analysis results are collected to form corresponding training samples. The input feature values ​​and problem analysis results of the training samples formed by different parameter configuration errors are different, and they can be determined independently.

[0147] The first problem analysis model is trained using all training samples, and the first problem model after training will be used for the analysis of new problems of this type that occur in the future. It can be understood that the first problem analysis model after training based on the existing training samples is a problem analysis model with intelligent classification and recognition capabilities. Problem analysis and location during daily testing or operation and maintenance based on this problem analysis model will greatly reduce the reliance on technical experts, and is based on an intelligent problem analysis model with expert experience, which effectively reduces the professional level requirements of technical personnel for related problem analysis and location work, and improves the work efficiency of problem analysis and location.

[0148] Accordingly, during testing or production operation and maintenance, steps 110-130 are executed to first obtain the data to be analyzed and determine to use the first problem analysis model that has been trained. Based on the data to be analyzed, the input feature values ​​corresponding to the model are determined, and the first problem analysis model is input to obtain the problem analysis results.

[0149] Wherein, according to the data to be analyzed, determining the input feature value corresponding to the model includes:

[0150] According to the input feature value set of various training samples of the model, the corresponding input feature value is determined.

[0151] For example, the training samples of the first problem analysis model include two types of training samples, Table-1 and Table-2, and the input feature values ​​corresponding to the model determined in step 120 are the union of the input feature values ​​of Table-1 and Table-2.

[0152] In some exemplary embodiments, the first problem analysis model is an intelligent analysis model belonging to classification, also known as an intelligent analysis algorithm, including but not limited to Bayesian classifiers, linear classifiers, decision trees, support vector machines, neural networks, etc. In the application stage of the trained model, the input feature values ​​are input into the trained first problem analysis model to obtain the corresponding problem analysis results, including the reasonable range of each abnormal parameter and the boundary of the abnormal range.

[0153] In some exemplary embodiments, a relevant regression model can be trained based on each training sample data to perform regression fitting on this dividing line. The regression model here can be any feasible algorithm, such as but not limited to polynomial regression, support vector regression, neural network regression, etc. The regression model can automatically output the critical value of the reasonable value of the abnormal parameter based on the input feature value corresponding to the current training sample. With this critical value, it can be used to guide the reconfiguration of the abnormal parameters, or to further find the cause of the problem based on the actual value of the abnormal parameter.

[0154] In some exemplary embodiments, taking the training and application of the second problem analysis model as an example:

[0155] The second problem analysis model is applicable to problems related to the incorrect execution order of events in a communication system, for example, analysis of problems caused by the incorrect execution order of certain events in the operation of a wireless communication system.

[0156] The analysis of this type of problem, in addition to the input feature values ​​shown in Table 1 and / or Table 2, also includes: one or more ordered event sets. For example, in a certain scenario, the correct execution order in these ordered event sets can be: {event A->event B->event C->event D}, or: {event A->event C->event B->event D}. However, the incorrect execution order that caused the exception is: {event A->event D->event B->event C}. Then the corresponding input feature values ​​in the process of analyzing this problem, in addition to including different data items similar to those in Table 1 and / or Table 2, also need to include the ordered event sets of these events; the corresponding problem analysis results include: the cause of the problem and a reasonable event execution order; or, also include: the current unreasonable event execution order.

[0157] After analyzing and solving this problem, it is necessary to form a training sample based on the analyzed data and analysis results, extract the data items related to the analysis of the problem and need to be checked from the sample data to be analyzed that has been collected during the problem analysis process, form the input feature value of the test sample, and mark the problem analysis result corresponding to the test sample. Among them, the input feature value of the test sample includes multiple data items from the sample data to be analyzed, including data items in the communication application layer data and data items in the communication protocol layer log data; the problem analysis results include: the cause of the problem, the reasonable event execution sequence, and the current unreasonable event execution sequence.

[0158] For example, the test samples in this case are shown in Table 3:

[0159] Table-3

[0160]

[0161]

[0162] Multiple collection data and analysis results in this case are accumulated to form multiple training samples.

[0163] Regarding the problem of incorrect execution order of communication system events, there are many other problems caused by errors in ordered event sets. Similarly, according to the above examples, we can know that historical collection data and analysis results are collected to form corresponding training samples. The input feature values ​​and problem analysis results of the training samples formed by different ordered event sets are different, and they can be determined independently.

[0164] The second problem analysis model is trained using all training samples, and the trained second problem model will be used for the analysis of new problems of this type that occur in the future. It can be understood that the second problem analysis model trained based on the existing training samples is a problem analysis model with intelligent classification and recognition capabilities. Problem analysis and location during daily testing or operation and maintenance based on this problem analysis model will greatly reduce the reliance on technical experts, and is based on an intelligent problem analysis model with expert experience, which effectively reduces the professional level requirements of technical personnel for related problem analysis and location work, and improves the work efficiency of problem analysis and location.

[0165] Accordingly, during testing or production operation and maintenance, steps 110-130 are executed to first obtain the data to be analyzed and determine to use the trained second problem analysis model. Based on the data to be analyzed, the input feature values ​​corresponding to the model are determined, and the second problem analysis model is input to obtain the problem analysis results.

[0166] Wherein, according to the data to be analyzed, determining the input feature value corresponding to the model includes:

[0167] According to the input feature value set of various training samples of the model, the corresponding input feature value is determined.

[0168] For example, the training samples of the second problem analysis model include multiple training samples similar to Table-3, then the input feature value corresponding to the model determined in step 120 is the union of the multiple input feature values ​​similar to Table-3.

[0169] In some exemplary embodiments, the second problem analysis model belongs to a classification and sequence order recognition model, also known as an intelligent analysis algorithm, including but not limited to Markov chains, recurrent neural networks and other design intelligent analysis algorithms, to analyze the order of these ordered events. The recurrent neural network here can be a common recurrent neural network (Recurrent Neural Network, RNN), a long short-term memory neural network (Long Short-Term Memory networks, LSTM), a gated recurrent unit neural network (Gated Recurrent Unit networks, GRU), etc. It can be understood that the trained second problem analysis model can be used to analyze and judge the order of these ordered events, distinguishing the correct order from the wrong order. If it is the wrong order, the correct order of related events can be given, that is, the order error of related events can be automatically corrected.

[0170] In some exemplary embodiments, taking the training and application of the third problem analysis model as an example:

[0171] The third problem analysis model is applicable to problems related to abnormal changes in the channel environment of the communication system, for example, analysis of the Radio Link Failure (RLF) problem caused when the terminal moves to an area with poor signal coverage of the wireless communication cell.

[0172] According to expert experience, when this type of wireless link failure occurs, in some cases (case c), the path loss will jump from a reasonable range to an abnormally high range, and the signal-to-interference-to-noise ratio may drop from a normal level to a relatively low level, such as Figure 3 shown.

[0173] In some cases (case d), it may be caused by temporary deterioration, such as temporary obstruction by the user's body, or temporary obstruction by other external objects (such as temporary obstruction by passing vehicles on the road), etc. After a period of time, there is no temporary obstruction, the channel returns to normal, and the terminal will re-establish a normal wireless connection; in this case, the path loss will temporarily jump from a reasonable range to an abnormally high range, and then fall back to the normal level after a period of time. At the same time, the signal-to-noise ratio may temporarily slip from a normal level to a relatively low level, and then rise to the normal level after a period of time, such as Figure 4 shown.

[0174] For case c, after analyzing and solving this problem, it is necessary to form a training sample based on the analyzed data and the analysis results, extract the data items that are related to the analysis of the problem and need to be checked from the sample data to be analyzed that has been collected during the problem analysis process, form the input feature value of the test sample, and mark the problem analysis result corresponding to the test sample. Among them, in addition to the input feature values ​​similar to those shown in Table-1 and / or Table-2, the input feature value of the test sample also includes: continuous values ​​of one or more configuration parameters, that is, data items at multiple collection time points are merged or combined to obtain composite data items. For example, in case c, it also includes continuous values ​​of path loss and continuous values ​​of signal-to-interference-noise ratio. The corresponding problem analysis results include: the cause of the problem, a reasonable parameter variation range, and a currently unreasonable parameter variation range.

[0175] For example, a test sample in case c is shown in Table-4:

[0176] Table-4

[0177]

[0178]

[0179] Accumulate multiple collection data and analysis results in case c to form multiple training samples.

[0180] For case d, after analyzing and solving this problem, it is necessary to form a training sample based on the analyzed data and analysis results, extract the data items related to the analysis of the problem and need to be checked from the sample data to be analyzed that have been collected during the problem analysis process, form the input feature value of the test sample, and mark the problem analysis result corresponding to the test sample. Among them, the input feature value of the test sample is similar to that of case c, but the problem analysis result determined is different according to the specific value of the feature value.

[0181] For example, a test sample in case d is shown in Table-5:

[0182] Table-5

[0183]

[0184]

[0185] Accumulate multiple collection data and analysis results under d conditions to form multiple training samples.

[0186] For problems such as abnormal changes in the channel environment of the communication system, there are many other problems caused by changes in parameters beyond the range. Similarly, according to the above examples, it can be known that historical collection data and analysis results are collected to form corresponding training samples. The input feature values ​​and problem analysis results of the training samples formed by different abnormal parameter changes are different and can be determined independently.

[0187] The third problem analysis model is trained using all training samples, and the trained third problem model will be used for the analysis of new problems of this type that occur in the future. It can be understood that the third problem analysis model trained based on the existing training samples is a problem analysis model with intelligent classification and recognition capabilities. Problem analysis and location in daily testing or operation and maintenance processes based on this problem analysis model will greatly reduce reliance on technical experts, and will be based on an intelligent problem analysis model with expert experience, which effectively reduces the professional level requirements for technical personnel in related problem analysis and location work, and improves the work efficiency of problem analysis and location.

[0188] Accordingly, during testing or production operation and maintenance, steps 110-130 are executed, first obtaining the data to be analyzed and determining to use a trained third problem analysis model, determining the input feature values ​​corresponding to the model based on the data to be analyzed, and inputting the third problem analysis model to obtain a problem analysis result.

[0189] Wherein, according to the data to be analyzed, determining the input feature value corresponding to the model includes:

[0190] According to the input feature value set of various training samples of the model, the corresponding input feature value is determined.

[0191] For example, the training samples of the third problem analysis model include two types of training samples, Table-4 and Table-5, and the input feature values ​​corresponding to the model determined in step 120 are the union of the input feature values ​​of Table-4 and Table-5.

[0192] In some exemplary embodiments, the third problem analysis model belongs to the analysis and identification model of classification and continuous value of some parameters, also known as intelligent analysis algorithm, which can analyze the changes of the continuous time values ​​of these parameters and distinguish different reasons. If it is determined that the continuous time value of a parameter is abnormal, a prediction sample of the normal continuous time value of the parameter under the current conditions shall be given. In some exemplary embodiments, intelligent analysis algorithms such as Markov chain and recurrent neural network can be used to analyze these continuous values. The recurrent neural network here can be a common recurrent neural network (Recurrent Neural Network, RNN), a long short-term memory neural network (Long Short-Term Memory networks, LSTM), a gated recurrent unit neural network (Gated Recurrent Unit networks, GRU), etc. Optionally, other types of neural networks or analysis algorithms can also be used, such as basic deep neural networks (Deep Neural Networks, DNN), convolutional neural networks (Convolutional Neural Networks, CNN), etc.

[0193] In some exemplary embodiments, taking the training and application of the fourth problem analysis model as an example:

[0194] The fourth problem analysis model is applicable to problems with limited device processing capabilities.

[0195] For example, after the calling terminal completes the establishment of the calling network connection, because the called terminal cannot complete the corresponding establishment of the network connection, the calling terminal finally prompts the other party that the call cannot be connected, which is collectively referred to as the called party's paging access failure problem. The reason for the paging access failure may be that during the paging access time, the received signal at the paging moment exceeds the terminal's demodulation capability, etc. When analyzing this type of problem, it is necessary to consider multiple data items, including multiple communication application layer data and / or multiple communication protocol layer log data, and the corresponding problem analysis results include: the cause of the problem.

[0196] In a real mobile environment, the terminal receiver has different successful reception thresholds for signals sent by the base station under different channel conditions, calibrated by the signal-to-noise ratio (SNR). The successful reception threshold of the terminal receiver for signals sent by the base station is affected by various factors, including:

[0197] Downlink received signal signal-to-noise ratio (SNR)

[0198] Terminal moving speed

[0199] Time offset and frequency offset

[0200] Downlink receiving interference signal strength

[0201] Modulation coding parameters (such as modulation method, etc.)

[0202] ……etc

[0203] At the time of receiving the paging signal, the signal quality received by the terminal exceeds the terminal's ability to correctly demodulate and decode, resulting in the terminal being unable to demodulate the paging information sent by the base station.

[0204] After analyzing and solving this problem, it is necessary to form a training sample based on the analyzed data and the analysis results, extract the data items related to the analysis of the problem and need to be checked from the sample data to be analyzed that has been collected during the problem analysis process, form the input feature value of the test sample, and mark the problem analysis result corresponding to the test sample. Among them, the input feature value of the test sample includes multiple data items from the sample data to be analyzed, including data items in the communication application layer data and data items in the communication protocol layer log data; the problem analysis result includes: the cause of the problem.

[0205] For example, the test samples in this case are shown in Table-6:

[0206] Table-6

[0207]

[0208]

[0209] The data items may include communication application layer data items and communication protocol layer data items.

[0210] It should be noted that, in the expert analysis stage (training sample accumulation stage), a variety of specific schemes can be adopted to determine the analysis results of the above problems.

[0211] In some exemplary embodiments, the ability of the terminal to correctly demodulate and decode is closely related to the specific channel environment and coding and modulation parameters. The specific channel environment can be described by various specific data items (channel parameters) in the input feature values ​​shown in Table 6. Taking a single channel parameter as an example, under a specific channel environment, the threshold of the terminal's receiving demodulation and decoding ability can be expressed as a function of the channel parameter, such as Figure 5 As shown. Extended to a channel parameter group composed of multiple channel parameters, the threshold of the terminal's receiving demodulation and decoding capability can be expressed as a multivariate function of the channel parameter group, and correspondingly determine whether it exceeds the terminal's correct demodulation and decoding capability. More detailed judgment multivariate functions and judgment schemes are not further discussed in detail in the embodiments of this disclosure.

[0212] Multiple collection data and analysis results in this case are accumulated to form multiple training samples.

[0213] For problems related to limited equipment processing capabilities, in addition to the demodulation and decoding capabilities in the above examples, there are many other problems caused by exceeding one or more other processing capabilities. Similarly, based on the above examples, it can be known that historical acquisition data and analysis results are collected to form corresponding training samples. The input feature values ​​and problem analysis results of the training samples formed by the evaluation and analysis of different processing capabilities are different and can be determined independently.

[0214] The fourth problem analysis model is trained using all training samples, and the trained fourth problem model will be used for the analysis of new problems of this type that occur in the future. It can be understood that the fourth problem analysis model, which has been trained based on the existing training samples, is a problem analysis model with intelligent classification and recognition capabilities. Problem analysis and location during daily testing or operation and maintenance based on this problem analysis model will greatly reduce the reliance on technical experts, and is based on an intelligent problem analysis model with expert experience, which effectively reduces the professional level requirements of technical personnel for related problem analysis and location work, and improves the work efficiency of problem analysis and location.

[0215] Accordingly, during testing or production operation and maintenance, steps 110-130 are executed, first obtaining the data to be analyzed and determining to use the trained fourth problem analysis model, determining the input feature values ​​corresponding to the model based on the data to be analyzed, and inputting the fourth problem analysis model to obtain the problem analysis results.

[0216] Wherein, according to the data to be analyzed, determining the input feature value corresponding to the model includes:

[0217] According to the input feature value set of various training samples of the model, the corresponding input feature value is determined.

[0218] For example, the training samples of the fourth problem analysis model include multiple types of training samples similar to Table-6, then the input feature value corresponding to the model determined in step 120 is the union of multiple input feature values ​​similar to Table-6.

[0219] In some exemplary embodiments, the fourth problem analysis model belongs to a classification recognition model, also known as an intelligent analysis algorithm, whose input result is not an abnormal parameter value or an abnormal sequence of ordered events, but a conclusion that the processing capacity of the device is limited.

[0220] Taking the demodulation capability shown in Table 6 as an example, the demodulation capability is to obtain the basic downlink receiving signal demodulation capability by laboratory simulation and algorithm simulation of different channel environments, and to extract the signal demodulation success and failure by combining a large amount of actual test log information, and continuously and intelligently learn to improve the demodulation capability. For the scenario where there is no abnormality in paging, that is, the downlink signal reception and demodulation are successful, and for the scenario where the paging is abnormal, that is, the downlink signal reception and demodulation fail, it is necessary to intelligently improve the demodulation capability based on its channel and signal conditions. The fourth problem analysis model can analyze these parameters and demodulation success conditions, and continuously and intelligently analyze and optimize the terminal's receiving and demodulation capabilities. If it is determined that a demodulation decoding error occurs under a certain parameter, it is considered to exceed the terminal demodulation capability and can be used as an abnormal sample; if it is determined that a certain parameter is correct, it is considered to meet the demodulation capability and be used as a normal sample. Optionally, support vector machines, deep neural networks, etc. can be used to design intelligent analysis algorithms to learn and train the terminal demodulation capability thresholds under different channel conditions to obtain a trained analysis model for subsequent analysis of the same type of problems.

[0221] It can be seen that according to the embodiment of the present disclosure, an intelligent problem analysis solution is provided, and input feature values ​​are collected in a standardized manner according to the needs of the problem analysis model. In the model training stage, the required analysis data items and analysis results corresponding to various types of problems are recorded to form training samples, and a sufficient number of training samples are used to train the selected problem analysis model.

[0222] It should be noted that due to the limitations of the testing and application scenarios of related wireless communication products, for some problems, the amount or scenario coverage of the data collected during daily research and development may not meet the training requirements of the intelligent analysis system. Under this condition, we can actively design test scenarios and collect a large amount of specific data to ensure that the data volume and coverage can meet the training requirements of the intelligent analysis system.

[0223] After the problem analysis model has been fully trained and meets the performance requirements, it is implemented and promoted on a large scale through a computer program for subsequent analysis of communication system problems. The computer program here can be a single computer program that can be installed on a personal computer, an online program that can be deployed on a server or cloud and accessed through a personal computer, or an APP software installed on a smartphone, tablet computer, etc.

[0224] In some exemplary embodiments, when the problem analysis results obtained after executing steps 110-130 are correct, new training samples can be determined according to similar steps based on the current data to be analyzed and the problem analysis results. In the process of applying each problem analysis model, further iterative training is performed based on the new training samples to continuously improve its performance.

[0225] In some exemplary embodiments, if the problem analysis results obtained after executing steps 110-130 are incorrect, manual analysis is required, and new training samples are determined according to similar steps based on the current data to be analyzed and the correct manual problem analysis results, for further iterative training of each problem analysis model to continuously improve its performance.

[0226] The present disclosure also provides an intelligent problem analysis method. Figure 6 As shown, including:

[0227] Step 610, obtaining data to be analyzed and a plurality of pre-trained problem analysis models;

[0228] Step 620, determining the input feature value corresponding to each problem analysis model according to the data to be analyzed;

[0229] Step 630, inputting the input feature values ​​into corresponding problem analysis models in sequence, and determining problem analysis results respectively;

[0230] Step 640 , selecting the top N results in terms of confidence among the multiple problem analysis results as the problem analysis results of the data to be analyzed, where N is an integer greater than 0.

[0231] In some exemplary embodiments, N=1, that is, the problem analysis result corresponding to the current data to be analyzed is selected as the one with the highest confidence.

[0232] It can be understood that when the problem analysis model corresponding to the current problem cannot be determined, a parallel analysis method can be used to determine the problem analysis results separately, and finally select one or more analysis results with higher confidence.

[0233] The present disclosure also provides an intelligent problem analysis device 70, such as Figure 7 As shown, including:

[0234] An acquisition module 710 is configured to acquire data to be analyzed and a pre-trained problem analysis model;

[0235] The analysis module 720 is configured to determine the input feature value corresponding to the problem analysis model according to the data to be analyzed; input the input feature value into the problem analysis model to determine the problem analysis result;

[0236] The problem analysis model includes at least one of the following models:

[0237] The first problem analysis model is applicable to problems such as communication system parameter configuration errors, the second problem analysis model is applicable to problems such as communication system event execution order errors, the third problem analysis model is applicable to problems such as abnormal channel environment changes in the communication system, and the fourth problem analysis model is applicable to problems such as limited device processing capabilities.

[0238] The present disclosure also provides an intelligent problem analysis system, including:

[0239] A data collection subsystem configured to collect data to be analyzed according to a set standardized format;

[0240] The problem analysis subsystem is configured to obtain a pre-trained problem analysis model; determine an input feature value corresponding to the problem analysis model according to the data to be analyzed; input the input feature value into the problem analysis model to determine a problem analysis result;

[0241] The problem analysis model includes at least one of the following models:

[0242] The first problem analysis model is applicable to problems such as communication system parameter configuration errors, the second problem analysis model is applicable to problems such as communication system event execution order errors, the third problem analysis model is applicable to problems such as abnormal channel environment changes in the communication system, and the fourth problem analysis model is applicable to problems such as limited device processing capabilities.

[0243] In some exemplary embodiments, the intelligent problem analysis system further includes:

[0244] The model training subsystem is configured to obtain data items used for the problem type analysis from sample data to be analyzed as input feature values ​​of training samples; mark the problem analysis results of the training samples; and use multiple training samples to train the problem analysis model.

[0245] The present disclosure also provides an electronic device, including:

[0246] one or more processors;

[0247] a storage device for storing one or more programs,

[0248] When the one or more programs are executed by the one or more processors, the one or more processors implement the intelligent problem analysis method as described in any embodiment of the present disclosure.

[0249] The embodiments of the present disclosure also provide a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the intelligent problem analysis method as described in any embodiment of the present disclosure is implemented.

[0250] It can be seen that the solution provided by the embodiment of the present disclosure can form standardized training samples by constructing standardized collected problem data and problem analysis results, and form a growing large-scale experience sample library for problems generated during the development, testing, and application of communication systems and the results of their analysis and confirmation. The standardized sample library retains relevant experience in the problem-solving process in the form of empirical data, and is used as a training sample for intelligent analysis models (algorithms) to train and / or iteratively train various problem analysis models to continuously improve the accuracy of model output. In the process of testing and / or equipment operation and maintenance, the analysis and positioning of various problems that arise in the communication system can first rely on the trained intelligent problem analysis model, reducing the strong dependence on professional and technical personnel, which can significantly improve the efficiency of problem analysis and lower the technical threshold.

[0251] This solution uses an intelligent problem analysis model to collect standardized data to be analyzed and mark standardized problem analysis results to form training sample data, simulate the problem-solving experience implicit in the data to be analyzed, and use it to automatically analyze new problems, which can greatly improve the efficiency of problem analysis. Moreover, as the training samples continue to grow, in theory, the analysis performance of the intelligent problem analysis model will continue to improve.

[0252] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0253] The above description is only a preferred embodiment of the present application scheme, and does not limit the patent scope of the present application. All equivalent structural changes made based on the concept of the present application scheme and the contents of the present application specification and drawings, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. An intelligent problem analysis method, characterized in that: include: Obtain the data to be analyzed and the pre-trained problem analysis model; Determine the input feature value corresponding to the problem analysis model according to the data to be analyzed; Inputting the input feature value into the problem analysis model to determine the problem analysis result; The problem analysis model includes at least one of the following models: A first problem analysis model applicable to communication system parameter configuration error problems, a second problem analysis model applicable to communication system event execution sequence error problems, and a third problem analysis model applicable to communication system channel environment abnormal change problems; The problem analysis model is pre-trained according to the following method: According to the problem type applicable to the problem analysis model, data items for analyzing the problem type are obtained from the sample data to be analyzed as input feature values ​​of the training sample; Marking the problem analysis results of the training samples; Using a plurality of the training samples to train the problem analysis model; The problem analysis results of the training samples marked with the communication system parameter configuration error problem type include: the cause of the problem and the reasonable parameter value range; The problem analysis results of the training samples marked with the communication system event execution sequence error problem include: the cause of the problem and the reasonable event execution sequence; The problem analysis results of the training samples marked with abnormal communication system channel environment change problem type include: the cause of the problem and a reasonable parameter change range.

2. The method according to claim 1, characterized in that The first problem analysis model includes one of the following models or a combination of at least two of the following models: a Bayesian classifier model, a linear classifier model, a decision tree model, a support vector machine model, and a neural network model; The second problem analysis model includes one of the following models or a combination of at least two of the following models: a Markov chain and a recurrent neural network model; The third problem analysis model includes one of the following models or a combination of at least two of the following models: Markov chain, recurrent neural network model, deep neural network model and convolutional neural network model.

3. The method according to claim 1, characterized in that The data to be analyzed includes: communication application layer data and communication protocol layer log data.

4. The method according to claim 3, characterized in that The communication application layer data includes one or more of the following: Service type, service problem type, time information, location information, operator information, communication network equipment information, channel environment information, equipment motion status information, and communication system version information; The communication protocol layer log data includes one or more of the following: Service Data Adaptation Protocol SDAP layer log data, non-access NAS layer log data, radio resource control RRC layer log data, packet data convergence protocol PDCP layer log data, radio link layer control RLC layer log data, media access MAC layer log data and physical PHY layer log data.

5. An intelligent problem analysis device, characterized in that: include: An acquisition module, configured to acquire data to be analyzed and a pre-trained problem analysis model; An analysis module, configured to determine the input feature value corresponding to the problem analysis model according to the data to be analyzed; Inputting the input feature value into the problem analysis model to determine the problem analysis result; The problem analysis model includes at least one of the following models: A first problem analysis model applicable to communication system parameter configuration error problems, a second problem analysis model applicable to communication system event execution sequence error problems, and a third problem analysis model applicable to communication system channel environment abnormal change problems; The problem analysis model is pre-trained according to the following method: According to the problem type applicable to the problem analysis model, data items for analyzing the problem type are obtained from the sample data to be analyzed as input feature values ​​of the training sample; Marking the problem analysis results of the training samples; Using a plurality of the training samples to train the problem analysis model; The problem analysis results of the training samples marked with the communication system parameter configuration error problem type include: the cause of the problem and the reasonable parameter value range; The problem analysis results of the training samples marked with the communication system event execution sequence error problem include: the cause of the problem and the reasonable event execution sequence; The problem analysis results of the training samples marked with abnormal communication system channel environment change problem type include: the cause of the problem and a reasonable parameter change range.

6. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the intelligent problem analysis method as described in any one of claims 1-4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the intelligent problem analysis method as described in any one of claims 1 to 4 is implemented.

Images