Network performance prediction method and device, electronic equipment and storage medium

Abstract

This disclosure proposes a network performance prediction method, apparatus, electronic device, and storage medium, relating to the field of communication technology. The specific implementation scheme includes: obtaining indicator features corresponding to a target network performance indicator, the indicator features including reference values ​​for the target network performance indicator of at least one target cell under multiple time periods; obtaining ranking features of at least one target cell, wherein the ranking features indicate the ranking of at least one target cell in the indicator features; fusing the indicator features and the ranking features to obtain a fused feature; and using an indicator prediction model to predict the indicator value of the fused feature to obtain the predicted value of the target network performance indicator of at least one target cell under a target time period after multiple time periods. Therefore, based on deep learning technology, the predicted value of the target network performance indicator of at least one target cell under a target time period after multiple time periods can be obtained, which can improve the accuracy and reliability of the prediction results.

Description

Technical Field

[0001] This disclosure relates to the field of communication technology, and in particular to a network performance prediction method, apparatus, electronic device, and storage medium. Background Technology

[0002] Network performance prediction can be used to guide network planning, configuration, management and maintenance. It can help network administrators and operators to grasp the network performance of different regions and types in a timely manner, coordinate the allocation of network resources in a timely manner, and improve the quality requirements of different service levels.

[0003] In related technologies, network personnel can compare historical data of network performance indicators and use simple mathematical statistical methods such as sliding window regression analysis and trend analysis to evaluate and calculate the predicted values ​​of network performance indicators.

[0004] However, the accuracy of the predicted values ​​obtained by the above methods is relatively low, which may lead to resource waste or shortage during subsequent network resource allocation, thereby affecting the network's operating efficiency. Summary of the Invention

[0005] This disclosure provides a method, apparatus, electronic device, and storage medium for predicting network performance.

[0006] According to one aspect of this disclosure, a network performance prediction method is provided, the method comprising:

[0007] Obtain the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of at least one target cell under multiple time periods;

[0008] Obtain the ranking features of the at least one target cell, wherein the ranking features are used to indicate the ranking of the at least one target cell in the index features;

[0009] The indicator features and the ranking features are fused together to obtain the fused features;

[0010] The index prediction model is used to predict the index values ​​of the fusion features to obtain the predicted values ​​of the target network performance indicators of at least one target cell in the target time period after the multiple time periods.

[0011] Optionally, the step of using the indicator prediction model to predict the index values ​​of the fusion features to obtain the predicted values ​​of the target network performance indicators of the at least one target cell in the target time period after the multiple time periods includes:

[0012] Obtain the network parameters of the at least one target cell;

[0013] Extract association features related to the target network performance indicators from the network parameters of the at least one target cell;

[0014] The associated features and the fused features are concatenated to obtain the concatenated features;

[0015] The splicing features are input into the indicator prediction model to obtain the predicted values ​​of the target network performance indicators of the at least one target cell output by the indicator prediction model.

[0016] Optionally, the step of inputting the splicing features into the indicator prediction model to obtain the predicted value of the target network performance indicator corresponding to the at least one target cell output by the indicator prediction model includes:

[0017] The encoder in the index prediction model is used to encode the spliced ​​features to obtain the encoded features;

[0018] The decoder in the index prediction model is used to decode the encoded features to obtain the decoded features;

[0019] The fully connected layer in the indicator prediction model is used to predict the indicator values ​​of the decoded features to obtain the predicted values ​​of the target network performance indicators of the at least one target cell.

[0020] Optionally, the step of obtaining the indicator features corresponding to the target network performance indicator includes:

[0021] For any one of the at least one target cells, obtain the measurement value of the target network performance index corresponding to the target network performance index of the any one cell under the multiple time periods;

[0022] The measured values ​​of the target network performance indicators of any cell under the multiple time periods are preprocessed to obtain the reference values ​​of the target network performance indicators of any cell under the multiple time periods.

[0023] The indicator features are generated based on the reference values ​​corresponding to the target network performance indicators of the at least one target cell in the multiple time periods.

[0024] Optionally, generating the indicator features based on reference values ​​corresponding to the target network performance indicators of the at least one target cell in the multiple time periods includes:

[0025] For any one of the multiple time periods, an index vector corresponding to any one time period is generated based on the reference value corresponding to the target network performance index of the at least one target cell in the any one time period.

[0026] The indicator features are generated based on the indicator vectors of the multiple time periods.

[0027] Optionally, the preprocessing includes at least one of the following: deduplication, outlier removal, missing value backfilling, and standardization.

[0028] Optionally, the method further includes:

[0029] Obtain network parameters from multiple serving cells;

[0030] Based on the network parameters of the multiple serving cells, the multiple serving cells are divided into at least one category, wherein the similarity between the network parameters of the serving cells belonging to the same category is higher than the similarity threshold.

[0031] Determine the target category from the at least one category;

[0032] Each serving cell belonging to the target category is designated as the target cell.

[0033] According to another aspect of this disclosure, an electronic device is provided, the electronic device including a memory, a transceiver, and a processor;

[0034] A memory for storing computer programs; a transceiver for sending and receiving data under the control of the processor; and a processor for reading the computer programs from the memory and performing the following operations:

[0035] Obtain the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of at least one target cell under multiple time periods;

[0036] Obtain the ranking features of the at least one target cell, wherein the ranking features are used to indicate the ranking of the at least one target cell in the index features;

[0037] The indicator features and the ranking features are fused together to obtain the fused features;

[0038] The index prediction model is used to predict the index values ​​of the fusion features to obtain the predicted values ​​of the target network performance indicators of at least one target cell in the target time period after the multiple time periods.

[0039] Optionally, the processor is specifically configured to perform the following operations:

[0040] Obtain the network parameters of the at least one target cell;

[0041] Extract association features related to the target network performance indicators from the network parameters of the at least one target cell;

[0042] The associated features and the fused features are concatenated to obtain the concatenated features;

[0043] The splicing features are input into the indicator prediction model to obtain the predicted values ​​of the target network performance indicators of the at least one target cell output by the indicator prediction model.

[0044] Optionally, the processor is specifically configured to perform the following operations:

[0045] The encoder in the index prediction model is used to encode the spliced ​​features to obtain the encoded features;

[0046] The decoder in the index prediction model is used to decode the encoded features to obtain the decoded features;

[0047] The fully connected layer in the indicator prediction model is used to predict the indicator values ​​of the decoded features to obtain the predicted values ​​of the target network performance indicators of the at least one target cell.

[0048] Optionally, the processor is specifically configured to perform the following operations:

[0049] For any one of the at least one target cells, obtain the measurement value of the target network performance index corresponding to the target network performance index of the any one cell under the multiple time periods;

[0050] The measured values ​​of the target network performance indicators of any cell under the multiple time periods are preprocessed to obtain the reference values ​​of the target network performance indicators of any cell under the multiple time periods.

[0051] The indicator features are generated based on the reference values ​​corresponding to the target network performance indicators of the at least one target cell in the multiple time periods.

[0052] Optionally, the processor is specifically configured to perform the following operations:

[0053] For any one of the multiple time periods, an index vector corresponding to any one time period is generated based on the reference value corresponding to the target network performance index of the at least one target cell in the any one time period.

[0054] The indicator features are generated based on the indicator vectors of the multiple time periods.

[0055] Optionally, the preprocessing includes at least one of the following: deduplication, outlier removal, missing value backfilling, and standardization.

[0056] Optionally, the processor is specifically configured to perform the following operations:

[0057] Obtain network parameters from multiple serving cells;

[0058] Based on the network parameters of the multiple serving cells, the multiple serving cells are divided into at least one category, wherein the similarity between the network parameters of the serving cells belonging to the same category is higher than the similarity threshold.

[0059] Determine the target category from the at least one category;

[0060] Each serving cell belonging to the target category is designated as the target cell.

[0061] According to another aspect of this disclosure, a network performance prediction apparatus is provided, the method comprising:

[0062] The first acquisition module is used to acquire the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of at least one target cell in multiple time periods.

[0063] The second acquisition module is used to acquire the ranking features of the at least one target cell, wherein the ranking features are used to indicate the ranking of the at least one target cell in the indicator features;

[0064] The fusion module is used to fuse the indicator features with the ranking features to obtain fused features;

[0065] The prediction module is used to predict the index value of the fusion feature using the index prediction model, so as to obtain the predicted value of the target network performance index of the at least one target cell in the target time period after the multiple time periods.

[0066] According to another aspect of this disclosure, a processor-readable storage medium is provided that stores a computer program for causing the processor to perform the aforementioned network performance prediction method.

[0067] According to another aspect of this disclosure, a computer program product is provided that, when an instruction processor in the computer program product is executed, performs the aforementioned network performance prediction method.

[0068] This disclosure offers the following technical advantages: By acquiring the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of at least one target cell under multiple time periods; acquiring the ranking features of at least one target cell, wherein the ranking features indicate the ranking of at least one target cell in the indicator features; fusing the indicator features and the ranking features to obtain fused features; and using an indicator prediction model to predict the indicator values ​​of the fused features to obtain the predicted values ​​of the target network performance indicators corresponding to at least one target cell under the target time period after multiple time periods. Therefore, based on deep learning technology, the predicted values ​​of the target network performance indicators corresponding to at least one target cell under the target time period after multiple time periods can be predicted, thereby improving the accuracy and reliability of the prediction results.

[0069] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0070] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0071] Figure 1 This is a flowchart illustrating a network performance prediction method provided in an embodiment of this disclosure;

[0072] Figure 2 This is a flowchart illustrating another network performance prediction method proposed in the embodiments of this disclosure;

[0073] Figure 3 This is a flowchart illustrating another network performance prediction method proposed in the embodiments of this disclosure;

[0074] Figure 4 This is a flowchart illustrating another network performance prediction method proposed in the embodiments of this disclosure;

[0075] Figure 5 This is a flowchart illustrating another network performance prediction method proposed in the embodiments of this disclosure;

[0076] Figure 6 This is a schematic diagram of the prediction process for network performance indicators in an embodiment of this disclosure;

[0077] Figure 7 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present disclosure;

[0078] Figure 8 This is a schematic diagram of the structure of a network performance prediction device provided in an embodiment of this disclosure. Detailed Implementation

[0079] The technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this disclosure.

[0080] In this embodiment of the disclosure, the term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0081] Currently, the network management center evaluates the network performance indicators of its 3G, 4G, and 5G base stations primarily through the following three methods:

[0082] The first method involves making a rough estimate of key network indicators based on the work experience of network engineers, thereby obtaining the predicted values ​​for those key indicators.

[0083] The second method involves network-side data center staff comparing historical network performance data with simple mathematical statistical methods such as sliding window regression analysis and trend analysis to evaluate and calculate predicted values ​​for network performance indicators.

[0084] The third approach involves network-side algorithm engineers building predictive models for specific cells based on machine learning algorithms such as the Autoregressive Integrated Moving Average (ARIMA) model and multiple regression.

[0085] However, the first empirical prediction method mentioned above is labor-intensive, has poor accuracy, and high manpower costs; the second simple statistical method is too simplistic in its statistical methods and relies too heavily on recent data, resulting in a lag in trend judgment and often failing to meet the need for rapid on-site problem resolution; the third single-index prediction method uses a single index for a single cell as the prediction unit, which has limited generalization ability. Therefore, large-scale deployment requires a large amount of resources, frequent model updates, and high operation and maintenance management requirements. In other words, due to the large number of base station cells under the jurisdiction of the network management center and the diversity of network performance indicators, the model's resource consumption is huge, which may lead to high operation and maintenance management complexity and low operating efficiency.

[0086] To address the aforementioned issues, this disclosure provides a network performance prediction method, apparatus, electronic device, and medium.

[0087] The network performance prediction method, apparatus, electronic device, and storage medium of this embodiment are described below with reference to the accompanying drawings.

[0088] Figure 1 This is a flowchart illustrating a network performance prediction method provided in an embodiment of this disclosure.

[0089] The network performance prediction method of this disclosure can be applied to any electronic device with computing capabilities. The electronic device can be a terminal device, a network device, etc.

[0090] The network device used here is an example of a base station. A base station may include multiple cells that provide services to terminal devices. Depending on the specific application, a base station may also be called an access point, or a device in the access network that communicates with wireless terminal devices through one or more sectors on the air interface, or other names. The network device can be used to exchange received air frames with Internet Protocol (IP) packets, acting as a router between the wireless terminal device and the rest of the access network, which may include an IP communication network. The network device can also coordinate the attribute management of the air interface. For example, the network equipment involved in this disclosure can be a Base Transceiver Station (BTS) in a Global System for Mobile communications (GSM) or Code Division Multiple Access (CDMA) system, a NodeB in a Wide-band Code Division Multiple Access (WCDMA) system, an evolved Node B (eNB or e-NodeB) in a long term evolution (LTE) system, a 5G base station (gNB) in a next-generation 5G network architecture, a Home evolved Node B (HeNB), a relay node, a femto, a pico, etc., and is not limited in this disclosure. In some network structures, the base station may include a Centralized Unit (CU) node and a Distributed Unit (DU) node, and the Centralized Unit and Distributed Unit may be geographically separated.

[0091] Terminal devices can be devices that provide voice and / or data connectivity to users, handheld devices with wireless connectivity, or other processing devices connected to a wireless modem. The name of the terminal device may differ in different systems; for example, in a 5G system, the terminal device can be called User Equipment (UE). Wireless terminal devices can communicate with one or more core networks (CNs) via a Radio Access Network (RAN). Wireless terminal devices can be mobile terminal devices, such as mobile phones (or "cellular" phones) and computers with mobile terminal devices, for example, portable, pocket-sized, handheld, computer-embedded, or vehicle-mounted mobile devices. They exchange voice and / or data with the RAN. Examples include Personal Communication Service (PCS) phones, cordless phones, Session Initiated Protocol (SIP) phones, Wireless Local Loop (WLL) stations, and Personal Digital Assistants (PDAs). Wireless terminal equipment can also be referred to as a system, subscriber unit, subscriber station, mobile station, mobile station, remote station, access point, remote terminal, access terminal, user terminal, user agent, or user device, but is not limited to these terms in the embodiments disclosed herein.

[0092] like Figure 1 As shown, the network performance prediction method may include the following steps:

[0093] Step 101: Obtain the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of at least one target cell in multiple time periods.

[0094] In this embodiment of the disclosure, the target network performance indicator can be any network performance indicator. For example, the target network performance indicator may include, but is not limited to: uplink PRB (Physical Resource Block), downlink PRB, radio connection success rate, RRC (Radio Resource Control) connection success rate, uplink traffic, downlink traffic, RSRP (Reference Signal Received Power), RSRQ (Reference Signal Receiving Quality), RRC establishment success rate, E-RAB (Evolved Radio Access Bearer) establishment success rate, call drop rate (or call interruption rate), handover success rate (such as cell handover success rate), etc.

[0095] In this embodiment of the disclosure, there is no limitation on the granularity of the time period division. For example, the length of the time period can be minutes, hours, days, etc.

[0096] In the embodiments of this disclosure, the number of target cells can be one or more, and this disclosure does not limit this.

[0097] In this embodiment of the disclosure, the reference value corresponding to the target network performance index may include at least one of the following: the measured value obtained by actually measuring the target network performance index, the processed value obtained by preprocessing the measured value, and the predicted value obtained by using an index prediction model to predict the index value.

[0098] In this embodiment of the disclosure, reference values ​​corresponding to the target network performance indicators of each target cell in multiple time periods can be obtained, and indicator features can be generated based on the above reference values.

[0099] As an example, after obtaining reference values ​​for the target network performance indicators of each target cell across multiple time periods, for any given time period, an indicator vector can be generated based on the reference values ​​for the target network performance indicators of each target cell in that given time period. This allows for the generation of indicator features based on the indicator vectors from multiple time periods. In other words, in this disclosure, the indicator features are two-dimensional, with one dimension representing the time period and the other representing the cell.

[0100] Step 102: Obtain the ranking features of at least one target cell, wherein the ranking features are used to indicate the ranking of at least one target cell in the indicator features.

[0101] In this embodiment of the disclosure, ranking features can be generated based on the ranking of each target cell in the indicator features.

[0102] As an example, the cell identifiers (such as names) of each target cell can be encoded based on their ranking in the indicator features to obtain the ranking features. For instance, binary encoding, one-hot encoding, or other encoding algorithms can be used to encode the cell identifiers (such as names) of each target cell to obtain the ranking features.

[0103] Step 103: The indicator features and ranking features are fused to obtain the fused features.

[0104] In this embodiment of the disclosure, index features and ranking features can be fused to obtain fused features.

[0105] As an example, the index features can be added to the ranking features to obtain the fused features.

[0106] As another example, index features and ranking features can be concatenated to obtain a fused image.

[0107] As another example, the index features and ranking features can be concatenated to obtain intermediate features. Then, the intermediate features can be input into a convolutional layer to fuse them to obtain the fused features mentioned above.

[0108] Step 104: Use an indicator prediction model to predict the indicator values ​​of the fusion features, so as to obtain the predicted values ​​of the target network performance indicators of at least one target cell in the target time period after multiple time periods.

[0109] In this embodiment of the disclosure, the target time period can be a time period following multiple time periods. For example, taking the length of the time period as days as an example, the multiple time periods can be the 1st, 2nd, 3rd, 4th and 5th of a certain month, and the target time period can be the 6th, 7th, 8th, 9th, etc. of that month.

[0110] In this embodiment of the disclosure, an index prediction model can be used to predict the index values ​​of the fusion features in order to obtain the predicted values ​​of the target network performance indicators of at least one target cell in the target time period after multiple time periods.

[0111] For example, taking the length of a time period as a day, we can obtain reference values ​​(actually measured values ​​or preprocessed values) for the target network performance indicators of each target cell on the 1st, 2nd, 3rd, 4th, and 5th of a month. Based on these reference values, we can generate indicator features. Based on the ranking of each target cell in the indicator features, we can generate ranking features. We can then fuse the indicator features and the ranking features to obtain fused features. Thus, we can use an indicator prediction model to predict the indicator values ​​of the fused features and obtain the predicted values ​​for the target network performance indicators of each target cell on the 6th of that month.

[0112] Furthermore, the indicator prediction model can also continuously make predictions in a round-robin manner. For example, it can predict the target network performance indicators of each target cell in cell number 7 based on the reference values ​​(measured values, or processed values ​​obtained after preprocessing the measured values) of cells 2, 3, 4, and 5, and the predicted values ​​obtained by the model based on the reference value of cell number 6. Similarly, it can predict the target network performance indicators of each target cell in cell number 8 based on the reference values ​​(measured values, or processed values ​​obtained after preprocessing the measured values) of cells 3, 4, and 5, and the predicted values ​​obtained by the model based on the reference values ​​of cells 6 and 7, and so on.

[0113] The network performance prediction method of this disclosure involves obtaining indicator features corresponding to a target network performance indicator, wherein the indicator features include reference values ​​corresponding to the target network performance indicator of at least one target cell under multiple time periods; obtaining ranking features of at least one target cell, wherein the ranking features indicate the ranking of at least one target cell in the indicator features; fusing the indicator features and the ranking features to obtain a fused feature; and using an indicator prediction model to predict the indicator value of the fused feature to obtain the predicted value of the target network performance indicator of at least one target cell under a target time period after multiple time periods. Therefore, based on deep learning technology, the predicted value of the target network performance indicator of at least one target cell under a target time period after multiple time periods can be predicted, which can improve the accuracy and reliability of the prediction results.

[0114] To clearly illustrate how the indicator prediction model predicts the predicted values ​​of the target network performance indicators for each target cell in the above embodiments of this disclosure, this disclosure also proposes a network performance prediction method.

[0115] Figure 2 This is a flowchart illustrating another network performance prediction method proposed in an embodiment of this disclosure.

[0116] like Figure 2 As shown, the network performance prediction method may include the following steps:

[0117] Step 201: Obtain the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of at least one target cell in multiple time periods.

[0118] Step 202: Obtain the ranking features of at least one target cell, wherein the ranking features are used to indicate the ranking of at least one target cell in the indicator features.

[0119] Step 203: The indicator features and ranking features are fused to obtain the fused features.

[0120] The execution process of steps 201 to 203 can be found in the execution process of any embodiment of this disclosure, and will not be described in detail here.

[0121] Step 204: Obtain network parameters of at least one target cell.

[0122] In this embodiment of the disclosure, the network parameters of each target cell may include at least one of the following: the target cell's operational parameters (including antenna position, azimuth angle, transmit power, etc.), the O domain (operational domain, i.e., the data domain of the operation support system, including network data, such as signaling, alarms, faults, network resources, etc.), the B domain (business domain, i.e., the data domain of the business support system, including user data and business data, such as user consumption habits, terminal information, ARPU (Average Revenue Per User) grouping, business content, business target audience, etc.), and the M domain (management domain, i.e., the data domain of the management support system, including location information, such as crowd flow trajectory, map information, etc.).

[0123] As an example, network parameters may include, but are not limited to: the antenna location, frequency band (i.e., antenna transmission frequency band), scene type, base station type, azimuth angle, average daily service data, MR (Measurement Report), percentage of users of different levels, data sampling time, roof longitude, roof latitude, base station equipment manufacturer, and coverage area (i.e., cell coverage area).

[0124] Step 205: Extract the associated features related to the target network performance indicators from the network parameters of at least one target cell.

[0125] In this embodiment of the disclosure, after obtaining the network parameters corresponding to each target cell, the associated features related to the target network performance indicators can be extracted from the network parameters of each target cell.

[0126] For example, for each target cell, sub-features associated with the target network performance indicators can be extracted from the network parameters of the target cell and used as sub-features of the target cell. In this disclosure, the sub-features of each target cell can be spliced ​​or fused according to the ranking of each target cell in the indicator features to obtain associated features.

[0127] As an example, let's take the RRC connection success rate as the target network performance indicator. The network parameters associated with the RRC connection success rate can be the proportion of users at different levels, data sampling time, etc. The associated features can be generated based on the network parameters associated with the RRC connection success rate.

[0128] Step 206: Concatenate the associated features and the fused features to obtain the concatenated features.

[0129] In this embodiment of the disclosure, associated features and fused features can be concatenated to obtain concatenated features.

[0130] Step 207: Input the splicing features into the indicator prediction model to obtain the predicted value of the target network performance indicator of at least one target cell output by the indicator prediction model.

[0131] In this embodiment of the disclosure, the splicing features can be input into the index prediction model, and the index prediction model can be used to predict the index values ​​of the splicing features to obtain the predicted values ​​of the target network performance indicators of each target cell output by the index prediction model.

[0132] For example, if there are multiple target cells, say three, and the target cells are ranked as follows in terms of indicator features: cell 1, cell 2, and cell 3, the predicted values ​​output by the indicator prediction model are: the predicted value of the target network performance indicator for cell 1, the predicted value of the target network performance indicator for cell 2, and the predicted value of the target network performance indicator for cell 3.

[0133] The network performance prediction method of this disclosure involves: acquiring network parameters of at least one target cell; extracting correlation features associated with target network performance indicators from the network parameters of the at least one target cell; concatenating the correlation features with fused features to obtain concatenated features; and inputting the concatenated features into an indicator prediction model to obtain the predicted values ​​of the target network performance indicators of at least one target cell output by the indicator prediction model. Therefore, by predicting the predicted values ​​of the target network performance indicators of each target cell based on the correlation features and fused features associated with the target network performance indicators, the accuracy and reliability of the prediction results can be improved.

[0134] To clearly illustrate how the indicator prediction model predicts the predicted values ​​of the target network performance indicators for each target cell in any embodiment of this disclosure, this disclosure also proposes a network performance prediction method.

[0135] Figure 3 This is a flowchart illustrating another network performance prediction method proposed in an embodiment of this disclosure.

[0136] like Figure 3 As shown, the network performance prediction method may include the following steps:

[0137] Step 301: Obtain the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of at least one target cell in multiple time periods.

[0138] Step 302: Obtain the ranking features of at least one target cell, wherein the ranking features are used to indicate the ranking of at least one target cell in the indicator features.

[0139] Step 303: The indicator features and ranking features are fused to obtain the fused features.

[0140] Step 304: Obtain network parameters of at least one target cell.

[0141] Step 305: Extract the associated features related to the target network performance indicators from the network parameters of at least one target cell.

[0142] Step 306: Concatenate the associated features and the fused features to obtain the concatenated features.

[0143] The execution process of steps 301 to 306 can be found in the execution process of any embodiment of this disclosure, and will not be described in detail here.

[0144] Step 307: Use the encoder in the index prediction model to encode the spliced ​​features to obtain the encoded features.

[0145] In this embodiment of the disclosure, an encoder in the index prediction model can be used to encode the spliced ​​features to obtain coded features.

[0146] As an example, let's take an index prediction model as an example of a model with a transformer as its basic structure. The encoder can include a self-attention layer and a feed-forward neural network. The role of the self-attention layer is to focus on other time points in the concatenated feature when it is input into the encoder at each time point. The role of the feed-forward neural network is to perform data pattern mining and information aggregation for the concatenated feature.

[0147] During the computation of the self-attention layer, three vectors (query vector Q, key vector K, and value vector V) are generated based on the encoder input, and multiple weight matrices are constructed accordingly. These weight matrices focus on different positions, forming a multi-headed attention mechanism. The three vectors are multiplied by their respective weight matrices to obtain weighted vectors, and then the weighted vectors are summed to complete the computation of the self-attention layer.

[0148] Step 308: Use the decoder in the index prediction model to decode the encoded features to obtain the decoded features.

[0149] In this embodiment of the disclosure, a decoder in the index prediction model can be used to decode the encoded features to obtain the decoded features.

[0150] As an example, let's take the indicator prediction model as an example of a model with a transformer as its basic structure. The decoder can include a self-attention layer, a feed-forward neural network, and an encoder-decoder attention layer, where the encoder-decoder attention layer is used to focus on the relevant parts related to the spliced ​​features.

[0151] Step 309: Use the fully connected layer in the indicator prediction model to predict the indicator values ​​of the decoded features, so as to obtain the predicted values ​​of the target network performance indicators of at least one target cell.

[0152] In this embodiment of the disclosure, a fully connected layer in the indicator prediction model can be used to predict the indicator values ​​of the decoded features in order to obtain the predicted values ​​of the target network performance indicators of each target cell.

[0153] The network performance prediction method of this disclosure employs an encoder in an indicator prediction model to encode spliced ​​features, obtaining encoded features; a decoder in the indicator prediction model to decode the encoded features, obtaining decoded features; and a fully connected layer in the indicator prediction model to predict indicator values ​​from the decoded features, thereby obtaining predicted values ​​for the target network performance indicators of at least one target cell. Thus, the encoder, decoder, and fully connected layer in the indicator prediction model can be used to effectively predict indicator values, obtaining predicted values ​​for the target network performance indicators of each target cell.

[0154] To clearly illustrate how the indicator features are obtained in any of the above embodiments, this disclosure also proposes a network performance prediction method.

[0155] Figure 4 This is a flowchart illustrating another network performance prediction method proposed in an embodiment of this disclosure.

[0156] like Figure 4 As shown, the network performance prediction method may include the following steps:

[0157] Step 401: For any cell in at least one target cell, obtain the measurement values ​​of the target network performance indicators of that cell in multiple time periods.

[0158] In this embodiment of the disclosure, for any cell among the target cells, the measurement values ​​of the target network performance indicators of that cell in multiple time periods can be obtained.

[0159] For example, taking the length of the time period as days, we can obtain the measurement values ​​of the target network performance indicators of any cell on the 1st, 2nd, 3rd, 4th and 5th of a certain month.

[0160] Step 402: Preprocess the measured values ​​of the target network performance indicators of any cell in multiple time periods to obtain the reference values ​​of the target network performance indicators of any cell in multiple time periods.

[0161] In this embodiment of the disclosure, the measured values ​​of the target network performance indicators of any cell under multiple time periods can be preprocessed to obtain the reference values ​​of the target network performance indicators of any cell under multiple time periods.

[0162] As an example, preprocessing including deduplication can remove redundant data from the measurements of a target performance metric for any given time period across multiple time periods.

[0163] As another example, taking preprocessing including outlier removal as an example, when there is a measurement value among the measurement values ​​corresponding to the target performance indicators in multiple time periods that does not match the historical fluctuation range, or the data format type of the measurement value is incorrect, or the encoding type of the measurement value is incorrect, in this disclosure, the above measurement value can be recorded as an outlier, thereby removing the outlier.

[0164] For example, for any cell, the mean and standard deviation of the measured values ​​of the target network performance index of that cell in multiple time periods can be calculated. If the difference between the measured value of the target network performance index of that cell in a certain time period and the above mean (such as the difference, the mean of the difference, the square of the difference, etc.) is greater than the standard deviation of a set multiple (such as 2 times, 3 times, 4 times, etc.), then the measured value of the target network performance index in that time period is recorded as an outlier, and thus the outlier can be removed.

[0165] As another example, taking preprocessing including missing value backfilling as an example, in order to ensure the continuity of the measured values ​​in time, a sliding average backfilling method with a set window length can be used to backfill the missing values. The set window length can be 2 time periods, 3 time periods, etc., and this disclosure does not limit it.

[0166] For example, the measured value v in time period i is denoted as v i If the measured value v is missing in time period t, a moving average model can be used to update v according to the following formula. t The possible values ​​of:

[0167] v t =β*vt-1 +(1-β)*θ t (1)

[0168] Where β∈[0,1), v t-1 Let θ be the value of the measured value v during the time interval t-1. t Set the sliding average value for the window length.

[0169] As another example, taking preprocessing including standardization as an example, in order to improve the prediction accuracy of the model, the measured values ​​of the target performance indicators at multiple time periods can be standardized.

[0170] For example, if x is the original measurement value, it can be standardized using the following formula:

[0171]

[0172] Where x′ represents the data after standardization of x, and μ and σ are the mean and standard deviation of the measured values ​​of the target performance index under multiple time periods, respectively.

[0173] It should be noted that the above example only illustrates that preprocessing includes one process. In actual applications, in order to improve the accuracy and reliability of the model prediction results, preprocessing may include multiple processes. For example, preprocessing may include multiple processes such as deduplication, outlier removal, missing value backfilling, and standardization. Alternatively, preprocessing may include other processes, such as normalization. This disclosure does not impose any limitations on this.

[0174] Step 403: Generate indicator features based on the reference values ​​of the target network performance indicators of at least one target cell in multiple time periods.

[0175] In this embodiment of the disclosure, indicator features can be generated based on the reference values ​​corresponding to the target network performance indicators of each target cell in multiple time periods.

[0176] As an example, after obtaining reference values ​​for the target network performance indicators of each target cell across multiple time periods, for any given time period, an indicator vector can be generated based on the reference values ​​for the target network performance indicators of each target cell in that given time period. This allows for the generation of indicator features based on the indicator vectors from multiple time periods. In other words, in this disclosure, the indicator features are two-dimensional, with one dimension representing the time period and the other representing the cell.

[0177] Step 404: Obtain the ranking features of at least one target cell, wherein the ranking features are used to indicate the ranking of at least one target cell in the indicator features.

[0178] Step 405: The indicator features and ranking features are fused to obtain the fused features.

[0179] Step 406: Use an index prediction model to predict the index values ​​of the fusion features, so as to obtain the predicted values ​​of the target network performance indicators of at least one target cell in the target time period after multiple time periods.

[0180] The execution process of steps 404 to 406 can be found in the execution process of any embodiment of this disclosure, and will not be described in detail here.

[0181] The network performance prediction method of this disclosure involves obtaining measured values ​​of target network performance indicators for any cell in at least one target cell across multiple time periods; preprocessing these measured values ​​to obtain reference values ​​for the target network performance indicators of any cell across multiple time periods; and generating indicator features based on the reference values ​​for the target network performance indicators of at least one target cell across multiple time periods. Therefore, by preprocessing the measured values ​​of target network performance indicators for at least one target cell across multiple time periods and generating indicator features based on the preprocessed measured values, the accuracy and reliability of the model prediction results can be improved.

[0182] It should be noted that, due to the large number of base stations and cells under the jurisdiction of the same network management center, different base stations or serving cells exhibit significant differences in service targets and network standards. If each serving cell is treated uniformly without differentiation, the neural network involved in training and prediction will experience slow convergence, low prediction accuracy, and poor generalization ability. Therefore, in this disclosure, to improve the accuracy and reliability of the model's prediction results, each serving cell can be divided into at least one category, allowing the model to predict indicator values ​​for serving cells belonging to the same category. The following section combines... Figure 5 The above process will be explained in detail.

[0183] Figure 5 This is a flowchart illustrating another network performance prediction method proposed in an embodiment of this disclosure.

[0184] like Figure 5 As shown, based on any of the above embodiments, the network performance prediction method may include the following steps:

[0185] Step 501: Obtain network parameters for multiple serving cells.

[0186] In this embodiment of the disclosure, multiple serving cells can be serving cells under the jurisdiction of the same network management center.

[0187] In this embodiment of the disclosure, the explanation of the network parameters can be found in the relevant description of step 204 in the above embodiments, and will not be repeated here.

[0188] Step 502: Based on the network parameters of multiple serving cells, divide the multiple serving cells into at least one category, wherein the similarity between the network parameters of serving cells belonging to the same category is higher than the similarity threshold.

[0189] The similarity threshold can be a pre-set threshold, which is a relatively high value, such as 70%, 80%, 90%, etc.

[0190] In this embodiment of the disclosure, multiple serving cells can be divided into at least one category based on their network parameters, wherein the similarity between the network parameters of serving cells belonging to the same category is higher than a similarity threshold.

[0191] As an example, multiple serving cells can be classified based on their network parameters using a classification algorithm to obtain at least one category. This classification algorithm can include, but is not limited to, decision trees, random forests, GBDT (Gradient Boosting Decision Tree), AdaBoost (Adaptive Boosting), XGBoost (Extreme Gradient Boosting), SVM (Support Vector Machine), KNN (K-Nearest Neighbor), Bayesian classification, logistic regression, and artificial neural networks.

[0192] As another example, clustering algorithms can be used to cluster multiple serving cells based on their network parameters, resulting in at least one category (or cluster). These clustering algorithms can include, but are not limited to, K-means, DBSCAN (Density-Based Spatial Clustering of Applications with Noise), GMM (Gaussian Mixed Model), hierarchical clustering, spectral clustering, and other clustering algorithms.

[0193] Step 503: Determine the target category from at least one category.

[0194] In the embodiments of this disclosure, a target category can be determined from various categories. For example, a category can be randomly selected from various categories as the target category. Alternatively, the number of serving cells included in each category can be determined, and the category with the largest number can be used as the target category. Or, the category with the smallest number can be used as the target category, and so on. This disclosure does not limit this.

[0195] Step 504: Select each serving cell belonging to the target category as the target cell.

[0196] In this embodiment of the disclosure, each serving cell belonging to the target category can be used as the target cell.

[0197] The network performance prediction method of this disclosure involves: acquiring network parameters of multiple serving cells; classifying the multiple serving cells according to their network parameters to obtain at least one category, wherein the similarity between the network parameters of serving cells belonging to the same category is higher than a similarity threshold; determining a target category from the at least one category; and designating each serving cell belonging to the target category as the target cell. Thus, by classifying multiple serving cells and grouping those with similar network parameters into the same category, the model can be used to predict the indicator values ​​of each serving cell within the same category, thereby improving the prediction accuracy of the model.

[0198] As an example, taking the indicator prediction model as an example of a model with Transformer as its basic structure, the specific process of network performance prediction can be divided into the following steps:

[0199] I. Data Acquisition and Preprocessing

[0200] 1. Data Acquisition

[0201] For multiple serving cells, network performance indicators and network parameters of each serving cell are obtained at multiple time periods. The network performance indicators of each serving cell include uplink PRB, downlink PRB, wireless connection success rate, RRC connection success rate, uplink traffic, downlink traffic, RSRP, RSRQ, RRC establishment success rate, E-RAB establishment success rate, call drop rate (or call interruption rate), and handover success rate (e.g., cell handover success rate). The network parameters of each serving cell include the operating parameters, O domain, B domain, and M domain of each serving cell. Specifically, the network parameters of each serving cell include the antenna location, frequency band (i.e., antenna transmission frequency band), scene type, base station type, azimuth angle, daily average service data, MR, user ratio of different levels, data sampling time, roof longitude, roof latitude, base station equipment manufacturer, and coverage area (i.e., cell coverage area).

[0202] 2. Data Preprocessing

[0203] The raw data obtained in the above steps can be transformed into a data type that is easier for the model to understand, thereby improving the accuracy of the model's prediction results. This includes, but is not limited to: removing duplicate values, missing values, and outliers; time-series completion and missing value backfilling; data scaling and feature construction.

[0204] ① Removal of duplicate, missing, and outlier values. Duplicate and outlier values ​​can be removed from the actual measurements of network performance indicators of each serving cell.

[0205] It is understandable that due to unavoidable hardware and software updates and upgrades for base stations and maintenance, as well as fault repairs, redundant data may be recorded within the same time period, and / or, data from several time periods may be null, i.e., duplicate and missing rows in the network management database. In this disclosure, duplicate values ​​can be directly removed and missing values ​​can be set to null to handle these situations.

[0206] Outliers refer to sudden increases or decreases that do not match the historical fluctuation range, incorrect data format types, and incorrect encoding conventions (i.e., incorrect data encoding leading to incorrect data format). In this disclosure, outliers can be directly removed.

[0207] ② Time series completion and missing value backfilling. Due to the removal of missing and outlier values, data will be empty in some time periods. In order to ensure the continuity of data in time series, a moving average with a fixed window length can be used for backfilling.

[0208] ③ Data scaling.

[0209] Data scaling and feature construction are key steps in feature engineering. They transform raw data into features that better describe potential problems to the model, thereby improving the model's accuracy in predicting unseen data. Data scaling can include logarithmic transformation and normalization.

[0210] Because the original data fluctuates significantly, it can lead to long model fitting times and poor prediction results. Data scaling can solve this problem by limiting the fluctuation range of the original data to a certain interval. In this disclosure, data scaling can be performed using standardization. Standardization transforms the original values ​​of each feature into normally distributed data with a mean of 0 and a standard deviation of 1. The standardized data are dimensionless values ​​and can improve the convergence speed of the model.

[0211] In this disclosure, after obtaining the measurement values ​​corresponding to the network performance indicators of each serving cell, the measurement values ​​under the same network performance indicator can be standardized so that the processed values ​​obtained after preprocessing each measurement value follow a normal distribution with a mean of 0 and a standard deviation of 1.

[0212] ④ Feature construction. Feature construction is based on the aggregation of information from engineering parameters, O-domain, B-domain, and M-domain data.

[0213] Because the measured values ​​of the same network performance metric exhibit strong periodicity, and because the engineering parameters, O-domain, B-domain, and M-domain of each serving cell also specifically reflect the service status of the corresponding base station, it is possible to extract correlation features related to the network performance metric from the network parameters of each serving cell. Constructing these correlation features can help uncover periodic patterns, leading to higher accuracy in model training and prediction. Furthermore, categorical features in the network parameters (such as base station type and scene type) need to be converted into continuous variables. For example, one-hot encoding can be used to transform categorical features into continuous variables.

[0214] As an example, each serving cell can establish a correspondence between time periods and reference values ​​(measured values ​​or processed values ​​obtained after preprocessing measured values) and associated features according to Table 1.

[0215] Table 1

[0216]

[0217] In Table 1, the indicators are abbreviations for network performance indicators. There are N network performance indicators, and each time period is a continuous time period.

[0218] II. Community Classification

[0219] Because a single network management center manages a large number of base stations and cells, and different base stations or serving cells differ significantly in terms of service targets and network standards, treating all serving cells the same without differentiation will result in slow convergence speed, low prediction accuracy, and poor generalization ability for the neural networks involved in training and prediction. Therefore, each serving cell can be classified into at least one category based on its network parameters, namely its industrial parameters, O-domain, B-domain, and M-domain.

[0220] It should be noted that when dividing the serving cells, a supervised classification algorithm with clear cell categories can be used to divide the serving cells, or a clustering algorithm without clear categories can be used to divide the serving cells.

[0221] After classifying the data into categories, a separate indicator prediction model can be built for each category. Based on the relevant data of each serving cell in that category, indicator features, ranking features, and correlation features corresponding to each network performance indicator can be constructed. The indicator prediction model for that category can be trained based on the above features. The trained indicator prediction model can be deployed to the network management center to predict indicator values ​​for each serving cell in the corresponding category.

[0222] It should be noted that after the relocation, construction, or demolition of base station cells within the network management center, the updated serving cells can be categorized to obtain updated categories. Furthermore, the relevant data of each serving cell under each updated category can be used to retrain the corresponding category's indicator prediction model to improve the model's prediction performance.

[0223] III. Model Construction

[0224] For each category, a separate indicator prediction model can be built. This indicator prediction model can employ an encoder-decoder network, such as the Transformer model. The Transformer model abandons the traditional recurrent neural network approach of extracting sequence information, innovatively utilizing a self-attention mechanism to achieve fast parallelism, thus overcoming the slow training limitation of recurrent neural networks. Furthermore, both the encoder and decoder in the Transformer model can be built into deep networks, leveraging the characteristics of DNN (Deep Neural Networks) models to improve the accuracy of the model's prediction results.

[0225] For each category, when training the Transformer model corresponding to that category, the serving cells under that category need to be constructed in a specific way. The first construction method is: the arrangement rules of each network performance indicator should be consistent, and the network performance indicators of the same serving cell should be clustered together. This method can ensure that the temporal rules of the same serving cell are consistent. The second construction method is: the network performance indicators of each serving cell are clustered together, and the arrangement rules of the serving cells within each cluster are consistent. This method can help the model quickly discover the data patterns of similar indicators.

[0226] As an example, based on Table 1, there can be multiple serving cells under the same category. Therefore, the data of each serving cell under the same category participating in the training can be constructed in a specific way: the first construction method mentioned above, that is, the arrangement rules of each network performance index are consistent, and the network performance index of the same serving cell is a cluster. The constructed data can be shown in Table 2. This method can ensure that the timing rules of the same serving cell are consistent.

[0227] Table 2

[0228]

[0229] Among them, there are X serving cells under the same category, N network performance indicators, and m time periods, and each time period is a continuous time period.

[0230] As an example, the second construction method mentioned above, in which the same network performance index of each serving cell is grouped into a cluster and the arrangement rules of the serving cells within each cluster are consistent, as shown in Table 3, will help the transformer model to quickly discover data patterns of similar indices.

[0231] Table 3

[0232]

[0233] An encoder-decoder network based on the Transformer architecture can include an encoder and a decoder. The encoder and decoder have similar structures, both including a self-attention layer and a feed-forward neural network. The difference lies in the decoder, which also includes an encoder-decoder attention layer to focus on relevant parts associated with the concatenated features. The self-attention layer focuses on other time points of the concatenated feature as input by the encoder at each time point. The feed-forward neural network performs data pattern mining and information aggregation for the concatenated features.

[0234] Taking an encoder consisting of N encoding units and a decoder consisting of M decoding units in an indicator prediction model as an example, the process of network performance prediction can be as follows: Figure 6 As shown:

[0235] (1) For each time period, the index vectors corresponding to each serving cell (referred to as the target cell in this disclosure) under the same category can be generated according to Table 2 or Table 3; repeating the above actions, the index features corresponding to multiple time periods can be obtained.

[0236] (2) Using binary coding, one-hot coding and other coding algorithms, the cell identifier of each target cell is encoded according to the ranking of each target cell in the indicator features, and the ranking features of each target cell are obtained. Thus, the indicator features and ranking features in (1) are fused to obtain the fused features.

[0237] (3) The correlation features related to network performance indicators in the network parameters of each target cell are concatenated with the fusion features in (3) to obtain concatenated features, which are then used as input samples for the Transformer model.

[0238] (4) Construct an encoding network (i.e., encoder) and a decoding network (i.e. decoder) with N encoding units and M decoding units respectively. The encoder includes a multi-head attention layer and a feedforward neural network, and the decoder includes a multi-head attention layer, an encoding and decoding attention layer and a feedforward neural network.

[0239] (5) Finally, the output sequence is obtained through at least one fully connected layer. The output sequence includes the predicted values ​​of the network performance indicators of each target cell under a target time period after multiple time periods.

[0240] In this disclosure, a loss function can be generated based on the difference between the output sequence of the model and the labeled sequence corresponding to the input sample. The value of the loss function is positively correlated with the difference, thus allowing the model to be trained based on the value of the loss function. For example, the model can be trained based on the value of the loss function to minimize its value.

[0241] For example, taking days as the length of a time period, we can obtain the measured values ​​of the network performance indicators of each target cell on the 1st, 2nd, 3rd, 4th, and 5th of a certain month. Based on the measured values ​​of the network performance indicators of each target cell on the 1st, 2nd, 3rd, and 4th, we can generate input samples. Based on the measured values ​​of the network performance indicators of each target cell on the 5th, we can generate a labeled sequence corresponding to the input samples. The model can then predict the predicted values ​​of the network performance indicators of each target cell on the 5th, resulting in an output sequence. Thus, we can train the model based on the difference between the output sequence and the labeled sequence.

[0242] Thus, the Transformer model completes its training.

[0243] In summary, by continuously learning the historical time-series data characteristics of multi-dimensional indicators from the network management center, and mining the internal trends and patterns of the data, a long-term indicator prediction model can be constructed. This model can serve the scenario of simultaneous prediction of multiple indicators across multiple service cells in the future over multiple time periods. The indicator prediction model can employ the Transformer model, which is a seq2seq (sequence-to-sequence) encoding and decoding deep neural network.

[0244] Compared to ARIMA and RNN (Recurrent Neural Network) models, Transformer-based time series prediction offers several advantages: First, its multi-head attention mechanism can uncover and learn data patterns from different perspectives. Simultaneously, fusing the temporal positions of each serving cell with indicator features through encoding facilitates the neural network's identification, memorization, and learning. Second, the Transformer model can simultaneously target multiple serving cells and continuously predict multiple indicators across multiple time periods, greatly simplifying multi-indicator prediction scenarios. Therefore, the Transformer model provides more accurate and efficient predictions, significantly reducing resource consumption and operational management complexity. It can accurately grasp the changing patterns of network performance indicators in future time periods, ensuring the network's continuous, stable, and efficient operation.

[0245] IV. Network Performance Prediction

[0246] Based on the Transformer model trained in step three, it can be deployed to the network management center. After deployment, the model will continuously predict each target cell under the same category in a round-robin manner. The input data for each round is constructed in the manner of (1) to (3) in step three.

[0247] The network performance prediction method based on the embodiments of this disclosure can be summarized from the following three key points:

[0248] 1) For multiple service cells and multiple network performance indicators, an indicator prediction model based on Transformer is adopted. Based on the time period and the ranking of service cells, indicator features, ranking features and correlation features are constructed.

[0249] 2) The Transformer-based seq2seq model can achieve multi-time-period synchronous prediction of multiple network performance indicators. This model can not only learn external related features (such as correlation features), but also establish and integrate the ranking features of each serving cell, which can improve the prediction accuracy and fault tolerance of the model.

[0250] 3) Supervised classification and clustering algorithms are used to establish cell category attribution for the serving cells under test in the network management center, which makes the prediction of network performance indicators of each serving cell under the same category in the same model more accurate.

[0251] The network performance prediction method based on the embodiments of this disclosure has advantages in practical application scenarios in at least several aspects:

[0252] a) Based on the Transformer model, the system can mine patterns in historical time series data to achieve synchronous prediction of performance indicators of multiple service cells and multiple networks. This reduces system resource consumption, lowers operation and maintenance complexity, improves prediction timeliness and accuracy, and effectively ensures decision-making reference for the network management center.

[0253] b) Supervised classification and clustering algorithms were used to automatically determine the cell category, enabling Transformer to learn from similar cells with similar features within the same model, thus ensuring the accuracy of the model and effectively reducing system bias.

[0254] c) By integrating indicator features, correlation features, and ranking features, the types of input sample features have been improved, which has further enhanced the accuracy of model prediction.

[0255] d) Using big data analytics to replace on-site testing can improve the work efficiency of network personnel and automatically update indicator results, making the network operation more efficient.

[0256] To implement the above embodiments, this disclosure also provides a source access network device.

[0257] Figure 7 This is a schematic diagram of the structure of a source access network device provided according to an embodiment of the present disclosure.

[0258] like Figure 7 As shown, the base station may include a transceiver 700, a processor 710, and a memory 720, wherein:

[0259] Transceiver 700 is used to receive and send data under the control of processor 710.

[0260] Among them, Figure 7 In this context, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits together, represented by one or more processors (processor 710) and memory (memory 720). The bus architecture can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. The transceiver 700 can be multiple elements, including transmitters and receivers, providing units for communicating with various other devices over transmission media, including wireless channels, wired channels, optical fibers, etc. The processor 710 is responsible for managing the bus architecture and general processing, and the memory 720 can store data used by the processor 710 during operation.

[0261] The processor 710 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor can also adopt a multi-core architecture.

[0262] The processor 710 calls a computer program stored in memory and performs the following operations:

[0263] Obtain the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of at least one target cell under multiple time periods;

[0264] Obtain the ranking features of at least one target cell, wherein the ranking features are used to indicate the ranking of at least one target cell in the index features;

[0265] The indicator features and ranking features are fused together to obtain the fused features;

[0266] An index prediction model is used to predict the index values ​​of the fusion features in order to obtain the predicted values ​​of the target network performance index of at least one target cell in the target time period after multiple time periods.

[0267] Alternatively, as another embodiment, the processor 710 is specifically configured to perform the following operations:

[0268] Obtain network parameters for at least one target cell;

[0269] Extract the associated features related to the target network performance indicators from the network parameters of at least one target cell;

[0270] The associated features and the fused features are concatenated to obtain the concatenated features;

[0271] The spliced ​​features are input into the indicator prediction model to obtain the predicted value of the target network performance indicator of at least one target cell output by the indicator prediction model.

[0272] Alternatively, as another embodiment, the processor 710 is specifically configured to perform the following operations:

[0273] The encoder in the index prediction model is used to encode the spliced ​​features to obtain the encoded features;

[0274] The decoder in the index prediction model is used to decode the encoded features to obtain the decoded features;

[0275] The fully connected layer in the index prediction model is used to predict index values ​​for the decoded features, so as to obtain the predicted values ​​of the target network performance index corresponding to at least one target cell.

[0276] Alternatively, as another embodiment, the processor 710 is specifically configured to perform the following operations:

[0277] For any cell in at least one target cell, obtain the measurement values ​​of the target network performance indicators for that cell in multiple time periods;

[0278] Preprocess the measured values ​​of the target network performance indicators of any cell in multiple time periods to obtain the reference values ​​of the target network performance indicators of any cell in multiple time periods.

[0279] Based on the reference values ​​of the target network performance indicators of at least one target cell in multiple time periods, indicator features are generated.

[0280] Alternatively, as another embodiment, the processor 710 is specifically configured to perform the following operations:

[0281] For any given time period among multiple time periods, an index vector corresponding to any given time period is generated based on the reference value corresponding to the target network performance index of at least one target cell in that time period.

[0282] Indicator features are generated based on indicator vectors from multiple time periods.

[0283] Alternatively, as another embodiment, the preprocessing includes at least one of the following: deduplication, outlier removal, missing value backfilling, and standardization.

[0284] Alternatively, as another embodiment, the processor 710 is specifically configured to perform the following operations:

[0285] Obtain network parameters from multiple serving cells;

[0286] Based on the network parameters of multiple serving cells, the multiple serving cells are divided into at least one category, wherein the similarity between the network parameters of serving cells belonging to the same category is higher than the similarity threshold.

[0287] Determine the target category from at least one category;

[0288] Each service cell belonging to the target category will be designated as the target cell.

[0289] It should be noted that the access network device provided in this embodiment can achieve the above-mentioned functions. Figures 1 to 5All method steps implemented in the method embodiment can achieve the same technical effect. Therefore, the parts that are the same as those in the method embodiment and their beneficial effects will not be described in detail here.

[0290] With the above Figures 1 to 5 Corresponding to the network performance prediction method provided in the embodiments, this disclosure also provides a network performance prediction device. Since the network performance prediction device provided in the embodiments of this disclosure is similar to the one described above... Figures 1 to 5 The network performance prediction method provided in the embodiments corresponds to the network performance prediction device provided in the embodiments of this disclosure, and will not be described in detail in the embodiments of this disclosure.

[0291] To implement the above embodiments, this disclosure also proposes a network performance prediction device.

[0292] Figure 8 This is a schematic diagram of the structure of a network performance prediction device provided in an embodiment of this disclosure.

[0293] like Figure 8 As shown, the network performance prediction device 800 may include: a first acquisition module 801, a second acquisition module 802, a fusion module 803, and a prediction module 804.

[0294] The first acquisition module 801 is used to acquire the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of at least one target cell under multiple time periods.

[0295] The second acquisition module 802 is further configured to acquire the ranking features of at least one target cell, wherein the ranking features are used to indicate the ranking of at least one target cell in the indicator features.

[0296] The fusion module 803 is used to fuse indicator features and ranking features to obtain fused features.

[0297] The prediction module 804 is used to predict the index values ​​of the fusion features using an index prediction model, so as to obtain the predicted values ​​of the target network performance indicators of at least one target cell in the target time period after multiple time periods.

[0298] Optionally, in one possible implementation of this disclosure, the prediction module 804 is specifically used to: obtain network parameters of at least one target cell; extract association features associated with target network performance indicators from the network parameters of at least one target cell; concatenate the association features with the fused features to obtain concatenated features; and input the concatenated features into the indicator prediction model to obtain the predicted values ​​of the target network performance indicators of at least one target cell output by the indicator prediction model.

[0299] Optionally, in one possible implementation of this disclosure, the prediction module 804 is specifically used to: encode the spliced ​​features using the encoder in the indicator prediction model to obtain coded features; decode the coded features using the decoder in the indicator prediction model to obtain decoded features; and predict the indicator values ​​of the decoded features using the fully connected layer in the indicator prediction model to obtain predicted values ​​corresponding to the target network performance indicators of at least one target cell.

[0300] Optionally, in one possible implementation of this disclosure, the first acquisition module 801 is specifically configured to: acquire, for any cell among at least one target cell, the measurement values ​​corresponding to the target network performance indicators of any cell in multiple time periods; preprocess the measurement values ​​corresponding to the target network performance indicators of any cell in multiple time periods to obtain reference values ​​corresponding to the target network performance indicators of any cell in multiple time periods; and generate indicator features based on the reference values ​​corresponding to the target network performance indicators of at least one target cell in multiple time periods.

[0301] Optionally, in one possible implementation of this disclosure, the first acquisition module 801 is specifically used to: for any time period among multiple time periods, generate an index vector corresponding to any time period based on the reference value corresponding to the target network performance index of at least one target cell in any time period; and generate index features based on the index vectors of multiple time periods.

[0302] Optionally, in one possible implementation of the embodiments of this disclosure, the preprocessing includes at least one of the following: deduplication, outlier removal, missing value backfilling, and standardization.

[0303] Optionally, in one possible implementation of this disclosure, the network performance prediction device 800 may further include:

[0304] The third acquisition module is used to acquire network parameters of multiple serving cells.

[0305] The segmentation module is used to segment multiple serving cells based on their network parameters to obtain at least one category, wherein the similarity between the network parameters of serving cells belonging to the same category is higher than a similarity threshold.

[0306] The determination module is used to determine the target category from at least one category.

[0307] The selection module is used to select each serving cell belonging to the target category as the target cell.

[0308] It should be noted that the network performance prediction device provided in this embodiment can achieve the above-mentioned... Figures 1 to 5All method steps implemented in the method embodiment can achieve the same technical effect. Therefore, the parts that are the same as those in the method embodiment and their beneficial effects will not be described in detail here.

[0309] It should be noted that the division of units in the embodiments of this disclosure is illustrative and only represents one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units.

[0310] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, 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. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network-side device, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0311] It should be noted that the apparatus provided in this embodiment can implement all the method steps implemented in the above method embodiment and can achieve the same technical effect. Therefore, the parts and beneficial effects that are the same as those in the method embodiment will not be described in detail here.

[0312] On the other hand, embodiments of this disclosure also provide a processor-readable storage medium storing a computer program for causing a processor to execute the present disclosure. Figures 2 to 5 The method shown in the embodiment.

[0313] The processor-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic storage (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO)), optical storage (e.g., CD, DVD, BD, HVD), and semiconductor storage (e.g., ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)).

[0314] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0315] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-executable instructions. These computer-executable instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0316] These processor-executable instructions may also be stored in a processor-readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the processor-readable memory produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0317] These processors can execute instructions that can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0318] Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from its spirit and scope. Therefore, if such modifications and variations fall within the scope of the claims of this disclosure and their equivalents, this disclosure is also intended to include such modifications and variations.

Claims

1. A method for predicting network performance, characterized in that, The method includes: Obtain the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of multiple target cells under multiple time periods; Based on the ranking of the multiple target cells in the indicator features, the cell identifiers of the multiple target cells are encoded to obtain the ranking features of the multiple target cells, wherein the ranking features are used to indicate the ranking of the multiple target cells in the indicator features; The indicator features and the ranking features are fused together to obtain the fused features; The index prediction model is used to predict the index values ​​of the fusion features to obtain the predicted values ​​of the target network performance indicators of the multiple target cells in the target time period after the multiple time periods. The acquisition of the indicator features corresponding to the target network performance indicators includes: For any one of the plurality of target cells, obtain the measurement value of the target network performance index corresponding to the target network performance index of the any one cell under the plurality of time periods; The measured values ​​of the target network performance indicators of any cell under the multiple time periods are preprocessed to obtain the reference values ​​of the target network performance indicators of any cell under the multiple time periods. For any one of the multiple time periods, an index vector corresponding to that time period is generated based on the reference values ​​corresponding to the target network performance indicators of the multiple target cells in that any one time period. The indicator features are generated based on the indicator vectors of the multiple time periods.

2. The method according to claim 1, characterized in that, The step of using the indicator prediction model to predict the indicator values ​​of the fusion features, so as to obtain the predicted values ​​of the target network performance indicators of the multiple target cells in the target time period after the multiple time periods, includes: Obtain the network parameters of the multiple target cells; From the network parameters of the multiple target cells, extract the association features that are associated with the target network performance indicators; The associated features and the fused features are concatenated to obtain the concatenated features; The splicing features are input into the indicator prediction model to obtain the predicted values ​​of the target network performance indicators of the multiple target cells output by the indicator prediction model.

3. The method according to claim 2, characterized in that, The step of inputting the splicing features into the indicator prediction model to obtain the predicted values ​​of the target network performance indicators of the multiple target cells output by the indicator prediction model includes: The encoder in the index prediction model is used to encode the spliced ​​features to obtain the encoded features; The decoder in the index prediction model is used to decode the encoded features to obtain the decoded features; The fully connected layer in the indicator prediction model is used to predict the indicator values ​​of the decoded features, so as to obtain the predicted values ​​of the target network performance indicators of the multiple target cells.

4. The method according to claim 1, characterized in that, The preprocessing includes at least one of the following: deduplication, outlier removal, missing value backfilling, and standardization.

5. The method according to any one of claims 1-4, characterized in that, The method further includes: Obtain network parameters from multiple serving cells; Based on the network parameters of the multiple serving cells, the multiple serving cells are divided into at least one category, wherein the similarity between the network parameters of the serving cells belonging to the same category is higher than the similarity threshold. Determine the target category from the at least one category; Each serving cell belonging to the target category is designated as the target cell.

6. An electronic device, characterized in that, The electronic device includes a memory, a transceiver, and a processor; A memory for storing computer programs; a transceiver for sending and receiving data under the control of the processor; and a processor for reading the computer programs from the memory and performing the following operations: Obtain the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of multiple target cells under multiple time periods; Based on the ranking of the multiple target cells in the indicator features, the cell identifiers of the multiple target cells are encoded to obtain the ranking features of the multiple target cells, wherein the ranking features are used to indicate the ranking of the multiple target cells in the indicator features; The indicator features and the ranking features are fused together to obtain the fused features; The index prediction model is used to predict the index values ​​of the fusion features to obtain the predicted values ​​of the target network performance indicators of the multiple target cells in the target time period after the multiple time periods. Specifically, the processor is used to perform the following operations: For any one of the plurality of target cells, obtain the measurement value of the target network performance index corresponding to the target network performance index of the any one cell under the plurality of time periods; The measured values ​​of the target network performance indicators of any cell under the multiple time periods are preprocessed to obtain the reference values ​​of the target network performance indicators of any cell under the multiple time periods. For any one of the multiple time periods, an index vector corresponding to that time period is generated based on the reference values ​​corresponding to the target network performance indicators of the multiple target cells in that any one time period. The indicator features are generated based on the indicator vectors of the multiple time periods.

7. The electronic device according to claim 6, characterized in that, The processor is specifically used to perform the following operations: Obtain the network parameters of the multiple target cells; From the network parameters of the multiple target cells, extract the association features that are associated with the target network performance indicators; The associated features and the fused features are concatenated to obtain the concatenated features; The splicing features are input into the indicator prediction model to obtain the predicted values ​​of the target network performance indicators of the multiple target cells output by the indicator prediction model.

8. The electronic device according to claim 7, characterized in that, The processor is specifically used to perform the following operations: The encoder in the index prediction model is used to encode the spliced ​​features to obtain the encoded features; The decoder in the index prediction model is used to decode the encoded features to obtain the decoded features; The fully connected layer in the indicator prediction model is used to predict the indicator values ​​of the decoded features, so as to obtain the predicted values ​​of the target network performance indicators of the multiple target cells.

9. The electronic device according to claim 6, characterized in that, The preprocessing includes at least one of the following: deduplication, outlier removal, missing value backfilling, and standardization.

10. The electronic device according to any one of claims 6-9, characterized in that, The processor is also used to perform the following operations: Obtain network parameters from multiple serving cells; Based on the network parameters of the multiple serving cells, the multiple serving cells are divided into at least one category, wherein the similarity between the network parameters of the serving cells belonging to the same category is higher than the similarity threshold. Determine the target category from the at least one category; Each serving cell belonging to the target category is designated as the target cell.

11. A network performance prediction device, characterized in that, The device includes: The first acquisition module is used to acquire the indicator features corresponding to the target network performance indicators, wherein the indicator features include reference values ​​corresponding to the target network performance indicators of multiple target cells in multiple time periods. The second acquisition module is further configured to encode the cell identifiers of the plurality of target cells according to the order of the plurality of target cells in the indicator features, thereby obtaining the order features of the plurality of target cells, wherein the order features are used to indicate the order of the plurality of target cells in the indicator features; The fusion module is used to fuse the indicator features with the ranking features to obtain fused features; The prediction module is used to predict the index values ​​of the fusion features using the index prediction model, so as to obtain the predicted values ​​of the target network performance indicators of the multiple target cells in the target time period after the multiple time periods. Specifically, the first acquisition module is used for: For any one of the plurality of target cells, obtain the measurement value of the target network performance index corresponding to the target network performance index of the any one cell under the plurality of time periods; The measured values ​​of the target network performance indicators of any cell under the multiple time periods are preprocessed to obtain the reference values ​​of the target network performance indicators of any cell under the multiple time periods. For any one of the multiple time periods, an index vector corresponding to that time period is generated based on the reference values ​​corresponding to the target network performance indicators of the multiple target cells in that any one time period. The indicator features are generated based on the indicator vectors of the multiple time periods.

12. A processor-readable storage medium, characterized in that, The processor-readable storage medium stores a computer program that causes the processor to perform the method of claims 1-5.

Images