Network switching method and apparatus, and terminal and storage medium

Abstract

A network switching method and apparatus, and a terminal and a storage medium, which belong to the technical field of network switching. The method comprises: acquiring current radio frequency fingerprint data, wherein the current radio frequency fingerprint data represents a radio frequency feature of a currently connected network (101); determining the current target subspace on the basis of a spatial attribute of each subspace in the current scenario and the current radio frequency fingerprint data (102); generating a network recommendation result on the basis of a communication semantic knowledge graph corresponding to the target subspace, wherein the communication semantic knowledge graph represents an association relationship between service data in different dimensions in the target subspace (103); and performing network switching on the basis of the current network connection and the network recommendation result (104).

Description

Network switching method, device, terminal and storage medium

[0001] This application claims priority to Chinese patent application number 202311713934.5, filed on December 13, 2023, entitled “Network switching method, device, terminal and storage medium,” the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The embodiments of the present application relate to the field of network switching technology, and in particular to a network switching method, device, terminal, and storage medium. Background Art

[0003] With the development of mobile internet, people's demand for network connectivity is growing. However, when choosing different network connection methods for different application scenarios, they often face problems such as network bandwidth, signal strength, and stability, which affect the user's network experience.

[0004] In related technologies, network switching is performed based on the network quality of the currently connected network detected by the terminal. For example, network quality is assessed using a series of indicators such as signal strength, signal-to-noise ratio, and bit error rate. When the signal strength falls below a strength threshold, a network scan is triggered. Once an available network with stronger signal strength is found, network switching is performed.

[0005] Obviously, in the related art, network switching attempts are only made when poor network quality is detected, resulting in delayed communication network switching and prone to obvious network freezes or even unavailability. Summary of the Invention

[0006] The embodiments of the present application provide a network switching method, device, terminal, and storage medium. The technical solution is as follows:

[0007] In one aspect, an embodiment of the present application provides a network switching method, the method comprising:

[0008] Obtaining current radio frequency fingerprint data, where the current radio frequency fingerprint data represents radio frequency characteristics of the currently connected network;

[0009] Determining a current target subspace based on spatial attributes of each subspace within the current scene and the current radio frequency fingerprint data, wherein the subspace is obtained by spatial clustering based on historical radio frequency fingerprint data, and the spatial attributes include historical radio frequency fingerprint features of the subspace;

[0010] Generate a network recommendation result based on the communication semantic knowledge graph corresponding to the target subspace, wherein the communication semantic knowledge graph represents the association relationship between network usage data of different dimensions in the target subspace;

[0011] Based on the current network connection status and the network recommendation result, network switching is performed.

[0012] On the other hand, an embodiment of the present application provides a network switching device, the device comprising:

[0013] A data acquisition module is used to acquire current radio frequency fingerprint data, wherein the current radio frequency fingerprint data represents the radio frequency characteristics of the currently connected network;

[0014] a spatial positioning module, configured to determine a current target subspace based on spatial attributes of each subspace within the current scene and the current RF fingerprint data, wherein the subspace is obtained by spatial clustering based on historical RF fingerprint data, and the spatial attributes include historical RF fingerprint features of the subspace;

[0015] A result generation module, configured to generate a network recommendation result based on a communication semantic knowledge graph corresponding to the target subspace, wherein the communication semantic knowledge graph represents the association relationship between network usage data of different dimensions in the target subspace;

[0016] The network switching module is used to switch the network based on the current network connection status and the network recommendation result.

[0017] On the other hand, an embodiment of the present application provides a terminal, which includes a processor and a memory, wherein the memory stores at least one computer instruction, and the at least one computer instruction is loaded and executed by the processor to implement the network switching method as described in the above aspects.

[0018] On the other hand, an embodiment of the present application provides a computer-readable storage medium, in which at least one computer instruction is stored. The at least one computer instruction is loaded and executed by a processor to implement the network switching method as described in the above aspects.

[0019] In another aspect, an embodiment of the present application provides a computer program product, comprising computer instructions stored in a computer-readable storage medium. A processor of a terminal reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the terminal to perform the network switching method provided in various optional implementations of the above aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] FIG1 shows a flow chart of a network switching method provided by an exemplary embodiment of the present application;

[0021] FIG2 shows an ontology structure diagram of a communication semantic knowledge graph provided by an exemplary embodiment of the present application;

[0022] FIG3 shows a structural diagram of a network entry type provided by an exemplary embodiment of the present application;

[0023] FIG4 shows a structural diagram of a network quality indicator provided by an exemplary embodiment of the present application;

[0024] FIG5 shows a flowchart of constructing a communication semantic knowledge graph provided by an exemplary embodiment of the present application;

[0025] FIG6 shows a flowchart of generating a subspace database provided by an exemplary embodiment of the present application;

[0026] FIG7 shows a schematic diagram of a subspace division scenario provided by an exemplary embodiment of the present application;

[0027] FIG8 shows a flow chart of radio frequency fingerprint data clustering provided by an exemplary embodiment of the present application;

[0028] FIG9 shows a flow chart of a network switching method provided by another exemplary embodiment of the present application;

[0029] FIG10 shows a flowchart of correcting network recommendation results provided by an exemplary embodiment of the present application;

[0030] FIG11 shows a flowchart of a network switching provided by an exemplary embodiment of the present application;

[0031] FIG12 shows a flowchart of generating a path list provided by an exemplary embodiment of the present application;

[0032] FIG13 shows a flowchart of generating a network recommendation list and model training provided by an exemplary embodiment of the present application;

[0033] FIG14 shows a flowchart of performing network switching provided by an exemplary embodiment of the present application;

[0034] FIG15 shows a flowchart of executing network switching provided by another exemplary embodiment of the present application;

[0035] FIG16 shows a flowchart of a network switching method provided by another exemplary embodiment of the present application;

[0036] FIG17 shows a flowchart of predicting network switching provided by an exemplary embodiment of the present application;

[0037] FIG18 shows a structural block diagram of a network switching device provided by an exemplary embodiment of the present application;

[0038] FIG19 shows a schematic structural diagram of a terminal provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0040] In related technologies, network switching is performed based on the network quality of the currently connected network detected by the terminal. For example, network quality is assessed using a series of indicators such as signal strength, signal-to-noise ratio, and bit error rate. When the signal strength falls below a strength threshold, a network scan is triggered. Once an available network with stronger signal strength is found, network switching is performed.

[0041] As can be seen, the related art only attempts to switch networks when it detects that the quality of the currently connected network is low. Therefore, before the network switch occurs, users will notice network lag, which affects their network experience. In complex multi-subspace and multi-network scenarios, as the user moves within the scene, the terminal needs to switch networks in a timely manner based on the current location. In other words, the network switching condition cannot be based solely on the network quality of the currently connected network.

[0042] Therefore, an embodiment of the present application proposes a network switching method, which obtains current RF fingerprint data and determines the current target subspace based on the current RF fingerprint data and the spatial attributes of each subspace in the current scene, thereby generating a network recommendation result based on the communication semantic knowledge graph corresponding to the target subspace, and then performing network switching based on the current network connection status and the network recommendation result, which can improve the efficiency of network switching and optimize the network experience.

[0043] Please refer to FIG1 , which shows a flow chart of a network switching method provided by an exemplary embodiment of the present application. The method includes the following steps:

[0044] Step 101: Acquire current radio frequency fingerprint data, where the current radio frequency fingerprint data represents radio frequency characteristics of the currently connected network.

[0045] In some embodiments, the terminal may acquire current radio frequency fingerprint data by initiating a WIFI scan or a cellular scan.

[0046] In the case of obtaining the current radio frequency fingerprint data based on WIFI scanning, the current radio frequency fingerprint data is the WIFI radio frequency fingerprint data, which may include the media access control (MAC) address, service set identifier (SSID), basic service set identifier (BSSID), received signal strength (RSSI), and data transmission round-trip time of the WIFI device.

[0047] In the case of obtaining the current radio frequency fingerprint data based on cellular scanning, the current radio frequency fingerprint data is the radio frequency data of the cellular network, which may include the radio access technology (RAT) corresponding to the cellular network, the cell identifier (Cell ID), the physical cell identifier (PCI), the frequency (Absolute Radio Frequency Channel Number, ARFCN), the cellular network code (Public Land Mobile Network Code), the network bandwidth (Bandwidth), the reference signal receiving power (RSRP), the reference signal receiving quality (RSRQ), and the signal to interference plus noise ratio (SINR), etc.

[0048] Step 102: Determine the current target subspace based on the spatial attributes of each subspace in the current scene and the current RF fingerprint data, wherein the subspace is obtained by spatial clustering based on the historical RF fingerprint data, and the spatial attributes include the historical RF fingerprint characteristics of the subspace.

[0049] Optionally, in order to meet the network needs of users in different spaces in complex scenarios, multiple candidate networks are usually set up. Therefore, in order to improve network switching efficiency, the terminal needs to be able to accurately locate the current space.

[0050] In a possible implementation, the terminal may determine the current target subspace according to the spatial attributes of each subspace in the current scene and the current radio frequency fingerprint data.

[0051] Optionally, each subspace in the scene may be obtained by spatial clustering based on historical radio frequency fingerprint data in the scene, and the spatial attribute of the subspace includes the historical radio frequency fingerprint feature of the subspace.

[0052] Optionally, the spatial attributes of the subspace may include not only historical radio frequency fingerprint features, but also historical network usage conditions, such as historical network usage time periods and historical network applications within the subspace.

[0053] Optionally, the terminal may store subspace information in the current scene through a subspace database. For example, the subspace database may store the number of subspaces in the current scene, subspace numbers, and space attributes.

[0054] In one possible implementation, after obtaining the current RF fingerprint data, the terminal can obtain the spatial attributes of each subspace in the current scene from the subspace database, where the spatial attributes include historical RF fingerprint features, so that the terminal can determine the current target subspace based on the current RF fingerprint data and each historical RF fingerprint data.

[0055] Step 103: Generate a network recommendation result based on the communication semantic knowledge graph corresponding to the target subspace. The communication semantic knowledge graph represents the association relationship between network usage data of different dimensions in the target subspace.

[0056] In some embodiments, considering that there may be multiple candidate networks or multiple network connection methods in the subspace, in order to improve the accuracy of network switching, after determining the current target subspace, the terminal can generate a network recommendation result based on the communication semantic knowledge graph corresponding to the target subspace, and then recommend each candidate network according to priority based on the network recommendation result.

[0057] Optionally, the communication semantic knowledge graph represents the associations between network usage data of different dimensions within the target subspace. Network usage data may include network usage time, network application, network application type, network type, signal strength, signal bandwidth, data transmission rate, application network usage feedback, etc., which is not limited in this embodiment of the present application.

[0058] In some embodiments, the terminal may find currently available candidate networks from the communication semantic knowledge graph corresponding to the target subspace according to the current network demand, and generate network recommendation results based on each candidate network.

[0059] Optionally, the network recommendation result may be the network quality of each candidate network, such as the current network signal strength of each candidate network, etc., or a network recommendation score for each candidate network, or other forms of network recommendation, which are not limited in the embodiments of the present application.

[0060] Step 104: Switch the network based on the current network connection status and the network recommendation result.

[0061] In some embodiments, after receiving the network recommendation result, the terminal needs to determine whether to switch networks based on the current network connection status. For example, if the currently connected network is better than the recommended candidate network, network switching will not be performed; if the recommended candidate network is better than the currently connected network, network switching will be performed.

[0062] In one possible implementation, the terminal may determine whether to switch networks by comparing the signal strength of the currently connected network with the signal strengths of each candidate network. If the signal strength of the currently connected network is lower than the signal strength of one of the candidate networks, the terminal may switch networks to the candidate network with higher signal strength.

[0063] In summary, in the embodiments of the present application, by obtaining the current RF fingerprint data and performing spatial positioning based on the spatial attributes of each subspace within the current scene and the current RF fingerprint data, the current target subspace is determined. Based on the communication semantic knowledge graph corresponding to the target subspace, a network recommendation result is generated, and network switching is performed based on the current network connection status and the network recommendation result. The solution provided by the embodiments of the present application can improve the efficiency of network switching within complex subspaces and optimize the network experience.

[0064] Optionally, based on the communication semantic knowledge graph corresponding to the target subspace, generate network recommendation results, including:

[0065] Based on the communication semantic knowledge graph, a path list corresponding to the target subspace is generated through graph traversal. The path list includes multiple connection relationship pairs consisting of different network requirements and network connection entrances.

[0066] Based on each connection relationship pair in the path list, the candidate network corresponding to the network connection entrance is predicted and scored to obtain the network confidence and network recommendation score corresponding to each candidate network;

[0067] Based on the network confidence and network recommendation score corresponding to each candidate network, a network recommendation result is generated.

[0068] Optionally, based on the communication semantic knowledge graph, a path list corresponding to the target subspace is generated by graph traversal, including:

[0069] Based on multiple network usage requirements, multiple requirement nodes corresponding to the network usage requirements are determined in the communication semantic knowledge graph, where the network usage requirements include at least one of network usage time, network usage application, and network application type;

[0070] Starting from the demand node, multiple network connection entrances that meet the demand node are determined by graph traversal in the communication semantic knowledge graph;

[0071] Based on the node path between the network demand and the network connection entrance, a connection relationship pair is formed;

[0072] Based on multiple connection relationship pairs formed by various network usage requirements and network connection entrances, a path list corresponding to the target subspace is generated.

[0073] Optionally, based on each connection relationship pair in the path list, the candidate networks corresponding to the network connection entry are predicted and scored to obtain the network confidence and network recommendation score corresponding to each candidate network, including:

[0074] Obtaining current network usage demand, where the current network usage demand includes at least one of current network usage time, current network application, and network application type;

[0075] Determine the path length corresponding to each connection relationship pair in the path list and the network quality of the candidate network;

[0076] Based on the current network demand, path length, and network quality of the candidate networks, each candidate network is predicted and scored to obtain the corresponding network confidence and network recommendation score. Among them, the network confidence is negatively correlated with the path length, and the network confidence is positively correlated with the network quality. The network recommendation score is negatively correlated with the path length, and the network recommendation score is positively correlated with the network quality.

[0077] Optionally, based on the communication semantic knowledge graph corresponding to the target subspace, network recommendation results are generated, which also includes:

[0078] Obtaining current network usage demand, where the current network usage demand includes at least one of current network usage time, current network application, and network application type;

[0079] The communication semantic knowledge graph corresponding to the current network demand and the target subspace is input into the network recommendation model, and the network recommendation model outputs the network confidence and network recommendation score corresponding to each candidate network;

[0080] Based on the network confidence and network recommendation score corresponding to each candidate network, a network recommendation result is generated.

[0081] Optionally, the method further includes:

[0082] Obtain sample network requirements, sample candidate networks, and network experience quality corresponding to the sample candidate networks;

[0083] The sample network requirements and the communication semantic knowledge graph corresponding to the target subspace are input into the network recommendation model, and the network recommendation model outputs the sample network confidence and sample network recommendation score corresponding to each sample candidate network;

[0084] Based on the sample network confidence, the sample network recommendation score, and the network experience quality corresponding to the sample candidate network, the network recommendation model is trained to obtain a trained network recommendation model.

[0085] Optionally, based on the network confidence and network recommendation score corresponding to each candidate network, a network recommendation result is generated, including:

[0086] Obtaining historical network connection information, where the historical network connection information includes at least one of the number of historical network connections, the duration of historical network connections, and the quality of historical network connections;

[0087] Based on the historical network connection situation, the network recommendation score corresponding to the candidate network is corrected to obtain the corrected network recommendation score;

[0088] Based on the network confidence corresponding to each candidate network and the corrected network recommendation score, a network recommendation result is generated.

[0089] Optionally, switch networks based on the current network connection status and network recommendation results, including:

[0090] Perform network screening based on the network recommendation results to obtain multiple networks to be switched, wherein the network confidence of the network to be switched is higher than the confidence threshold, and the network recommendation score of the network to be switched is higher than the network score of the currently connected network;

[0091] If the currently connected network is a WiFi network, the network to be switched is a cellular network, the network switch of the cellular network is on, and the network signal strength of the network to be switched meets the signal strength threshold, the network switching is performed;

[0092] When both the currently connected network and the network to be switched are WIFI networks and the network signal strength of the network to be switched meets the signal strength threshold, network switching is performed.

[0093] Optionally, the method further includes:

[0094] If the currently connected network is a cellular network, no network switching is performed;

[0095] If both the currently connected network and the network to be switched are Wi-Fi networks, and the network signal strength of the network to be switched does not meet the signal strength threshold, network switching will not be performed;

[0096] If the currently connected network is a Wi-Fi network, the network to be switched is a cellular network, and the cellular network switch is off, the network switching will not be performed;

[0097] If the currently connected network is a WiFi network, the network to be switched is a cellular network, the network switch of the cellular network is on, and the network signal strength of the network to be switched does not meet the signal strength threshold, network switching will not be performed.

[0098] Optionally, based on the spatial attributes of each subspace in the current scene and the current radio frequency fingerprint data, determining the current target subspace includes:

[0099] Determine, based on the spatial attributes of each subspace in the current scene and the current radio frequency fingerprint data, a first correlation coefficient between each subspace and the current radio frequency fingerprint data, where the first correlation coefficient represents a degree of correlation between the current radio frequency fingerprint data and the subspace;

[0100] Based on the first correlation coefficients between each subspace and the current radio frequency fingerprint data, the current target subspace is determined, and the first correlation coefficient corresponding to the target subspace is higher than the first correlation coefficients corresponding to other subspaces.

[0101] Optionally, the method further includes:

[0102] Collect historical network usage data corresponding to each subspace, including historical network usage time, historical network application, application type, historical radio frequency fingerprint data, historical protocol measurement data, and historical network experience data;

[0103] Perform data cleaning and statistics on historical network usage data to obtain processed historical network usage data;

[0104] Based on the processed historical network usage data, the communication semantic knowledge graph corresponding to each subspace is generated.

[0105] Optionally, based on the processed historical network usage data, a communication semantic knowledge graph corresponding to each subspace is generated, including:

[0106] Based on the processed historical network usage data, determining the feature dimensions corresponding to each historical network usage data;

[0107] Based on the characteristic dimensions corresponding to each historical network usage data, determine the correlation relationship between the historical network usage data of different dimensions;

[0108] Taking historical network usage data as nodes and the association relationships between historical network usage data as edges, we generate association paths between different historical network usage data and obtain the communication semantic knowledge graph corresponding to each subspace.

[0109] Optionally, the method further includes:

[0110] Collect historical RF fingerprint data corresponding to different scanning conditions in the current scenario;

[0111] When the amount of historical radio frequency fingerprint data reaches a quantity threshold, performing correlation coefficient calculations between each pair of historical radio frequency fingerprint data to obtain a second correlation coefficient corresponding to each piece of historical radio frequency fingerprint data;

[0112] Based on the second correlation coefficient corresponding to each historical radio frequency fingerprint data, spatial clustering is performed to obtain each subspace corresponding to the current scene.

[0113] In some embodiments, in order to improve the accuracy of generating network recommendation results, it is necessary to first improve the accuracy of the communication semantic knowledge graph corresponding to each subspace.

[0114] Optionally, the terminal may collect a large amount of historical network usage data corresponding to each subspace, and build a communication semantic knowledge graph based on the historical network usage data.

[0115] In a possible implementation, the terminal may first collect historical network usage data corresponding to each subspace. The historical network usage data may include historical network usage time, historical network usage applications, application types of historical network applications, historical radio frequency fingerprint data, historical protocol measurement data, and historical network usage experience data.

[0116] Optionally, the historical radio frequency fingerprint data may include network radio frequency fingerprint and network Layer 2 data. The network radio frequency fingerprint may be divided into cellular radio frequency fingerprint and WIFI radio frequency fingerprint. The network Layer 2 data may be divided into cellular Layer 2 data and WIFI Layer 2 data.

[0117] Among them, the cellular radio frequency fingerprint may include the radio access technology (Radio Access Technology, RAT) corresponding to the cellular network, cellular cell identifier (Cell ID), physical cell identifier (Physical Cell Identifier, PCI), frequency (Absolute Radio Frequency Channel Number, ARFCN), cellular network code (Public Land Mobile Network Code), network bandwidth (Bandwidth), signal received power (Reference Signal Receiving Power, RSRP), signal received quality (Reference Signal Receiving Quality, RSRQ) and signal to interference plus noise ratio (Signal to Interference plus Noise Ratio, SINR), etc.

[0118] Among them, cellular Layer 2 data may include download (Receive, RX) transmission rate, upload (Transmit, TX) transmission rate, uplink access permission (Uplink Grant), uploaded buffer data volume, buffer status report (Buffer Status Report, BSR), uplink block error rate (Block Error Rate, BLER), downlink BLER, etc.

[0119] Among them, the Wi-Fi radio frequency fingerprint may include the Media Access Control (MAC) address, Service Set Identifier (SSID), Basic Service Set ID (BSSID), Received Signal Strength Indicator (RSSI), and data transmission round-trip time of the Wi-Fi device.

[0120] Among them, Wi-Fi Layer 2 data may include RX transmission rate, TX transmission rate, channel utilization ratio, channel load strength, total TX success times, total TX retry times, total TX failure times, total RX success times, total current channel busy time, total current channel working time, etc.

[0121] Optionally, the historical protocol measurement data may be Transmission Control Protocol (TCP) / User Datagram Protocol (UDP) measurement data, which may include application package name, round-trip time (RTT), number of data packets, data packet size, number of retransmitted packets, retransmission rate, Domain Name System (DNS) query delay, etc.

[0122] Optionally, the historical network experience data may be user experience data, which may include application package name, user experience data of application feedback, user experience score data, etc.

[0123] In a possible implementation, after collecting the historical web usage data corresponding to each subspace, in order to improve the efficiency of constructing the knowledge graph, the terminal can also perform data cleaning and statistics on the historical web usage data to obtain processed historical web usage data.

[0124] Optionally, the terminal can filter historical network usage data based on network usage duration. For example, it can filter historical network usage data generated during shorter network connections. Optionally, the terminal can also filter other historical network usage data based on network experience data. For example, it can filter historical network usage data with a low user experience score.

[0125] Optionally, the terminal can also classify historical network usage data by network usage time period, and calculate the statistical values ​​of different network usage data in each network usage time period, including but not limited to mean, maximum value, minimum value, standard deviation, sample number, etc.

[0126] Furthermore, after processing the historical network usage data, the terminal can generate a communication semantic knowledge graph corresponding to each subspace based on the processed historical network usage data.

[0127] In one possible implementation, the terminal may first determine the characteristic dimensions corresponding to each piece of historical network usage data. For example, the characteristic dimensions can be divided into demand characteristics, network characteristics, and evaluation characteristics. The terminal then classifies the historical network usage data based on the characteristic dimensions corresponding to each piece of historical network usage data and determines the associations between historical network usage data of different dimensions. Thus, using each piece of historical network usage data as a node and the associations between historical network usage data as edges, the terminal generates association paths between different pieces of historical network usage data, thereby obtaining a communication semantic knowledge graph corresponding to each subspace.

[0128] In an illustrative example, demand characteristics may include network usage time, network applications, and network application types; network characteristics may include radio frequency fingerprint data and protocol measurement data; and evaluation characteristics may include network experience data. Thus, the terminal can determine the various feature entities and relationships based on the knowledge graph construction principle.

[0129] Schematically, as shown in Figure 2, the terminal can set the ontology structure of the communication semantic knowledge graph to be centered on network usage demand and network connection entry. Each network usage demand (Connection Demand) corresponds to a set of network usage time (Service Time Slot), network application, network application type, and subspace (Subspace). Each network connection entry corresponds to a set of network usage time, subspace, network entry type (Network Access Type), and network quality indicator (Network Access KPI).

[0130] Schematically, as shown in Figure 3, there are three types of network access: 5G network (5G Access Type), 4G network (4G Access Type), and Wi-Fi network (Wi-Fi Access Type). Different network types correspond to different network parameters. For example, 5G and 4G networks correspond to cellular cell identifiers, physical cell identifiers, frequencies, and cellular network codes; Wi-Fi networks correspond to basic service set identifiers, media access control addresses, and service set identifiers.

[0131] Schematically, as shown in Figure 4, there are three types of network quality indicators: 5G network quality (5G Access KPI), 4G network quality (4G Access KPI), and Wi-Fi network quality (Wi-Fi Access KPI). Each network quality indicator has different indicator parameters. For example, 5G network quality and 4G network quality have cellular signal strength, technical service quality indicator (QoS), and user experience indicator (QoE), respectively. Wi-Fi network quality has Wi-Fi signal strength, technical service quality indicator, and user experience indicator.

[0132] In some embodiments, considering that there may be multiple terminals in the same scene, that is, there are multiple terminal network usage data, for example, taking a family as an example, each family member corresponds to at least one terminal device, and different terminal devices correspond to different historical network usage data. Therefore, in order to improve the integrity of the construction of the communication semantic knowledge graph, the server can also construct the communication semantic knowledge graph corresponding to each subspace in the current scene.

[0133] In one possible implementation, the server obtains historical network usage data corresponding to each terminal in the scenario, integrates and compiles this data, and constructs a communication semantic knowledge graph corresponding to each subspace. Once the terminal determines the target subspace, it can obtain the communication semantic knowledge graph for that subspace from the server to generate network recommendation results.

[0134] Please refer to Figure 5, which shows a flowchart for constructing a communication semantic knowledge graph according to an exemplary embodiment of the present application. First, the terminal obtains the historical network usage data corresponding to each subspace in the current scenario and constructs communication semantic features. Then, by filtering, screening, and statistically processing the historical network usage data, the processed historical network usage data is obtained, and then a communication semantic knowledge graph is generated based on the processed historical network usage data.

[0135] In the above embodiment, by acquiring historical internet usage data of different dimensions, the associations between historical internet usage data are determined based on different feature dimensions to generate a communication semantic knowledge graph, thereby improving the efficiency of knowledge graph construction. Furthermore, the server constructs the communication semantic knowledge graph based on the historical internet usage data corresponding to each terminal in the current scenario, improving the integrity of the knowledge graph construction and facilitating more accurate generation of network recommendation results.

[0136] In some embodiments, in order to improve the accuracy of generating network recommendation results, in addition to optimizing the communication semantic knowledge graph, it is also necessary to improve the accuracy of spatial positioning, that is, the terminal first needs to accurately divide the various subspaces within the scene.

[0137] In a possible implementation, the terminal first needs to collect historical RF fingerprint data corresponding to different scanning conditions in the current scene, so that when the amount of historical RF fingerprint data collected reaches a quantity threshold, the terminal can calculate the correlation coefficient between each pair of historical RF fingerprint data to obtain a second correlation coefficient corresponding to each historical RF fingerprint data, and then perform spatial clustering based on the second correlation coefficient corresponding to each historical RF fingerprint data to obtain each subspace corresponding to the current scene.

[0138] Optionally, the calculation method of the second correlation coefficient may be a Pearson correlation coefficient calculation method, a cosine similarity calculation method, an adjusted cosine similarity calculation method, etc., which is not limited in the embodiment of the present application.

[0139] Optionally, the collection method of historical radio frequency fingerprint data can be set according to different scanning conditions, wherein the scanning conditions can be set according to the terminal status, screen status, network connection status and charging and discharging status.

[0140] For example, when the terminal is in mobile state with the screen on and connected to a WIFI network, it can obtain WIFI physical layer and data link layer (L2) data every 3 seconds, and obtain 1 second of cellular radio frequency fingerprint data after receiving the WIFI scan results. The collection time does not exceed 3 minutes.

[0141] For example, when the terminal is in a stationary state with the screen on and connected to a WIFI network, it can obtain WIFI L2 data every 60 seconds and obtain 1 second of cellular RF fingerprint data after receiving the WIFI scan result. The collection time does not exceed 5 minutes.

[0142] For another example, when the signal strength of Wi-Fi L2 is less than a strength threshold (e.g., -67dBm), the terminal may obtain Wi-Fi L2 data every 3 seconds and cellular radio frequency fingerprint data every 5 seconds, with the collection time not exceeding 3 minutes.

[0143] For another example, when the screen is on and the WIFI switch is turned from off to on, the terminal can obtain 1 second of cellular radio frequency fingerprint data after receiving the WIFI scan result based on the system WIFI scanning mechanism.

[0144] For another example, when the screen is on and the WIFI switch is turned on and off, the terminal can perform a WIFI scan and obtain 1 second of cellular radio frequency fingerprint data after receiving the WIFI scan result.

[0145] For example, when a screen on / off event or a charging event is triggered and the WIFI network is connected, the terminal collects the connection and disconnection time points of the WIFI network, as well as the Basic Service Set ID (BSSID), and collects the connection and disconnection time points of the cellular network during the connection to the WIFI network, as well as the cellular cell ID (Cell ID).

[0146] It should be noted that this application only provides a schematic description of the scanning conditions and collection methods for collecting radio frequency fingerprint data, and does not limit them.

[0147] In one possible implementation, during the collection of historical RF fingerprint data, the terminal may also utilize a filtering algorithm to filter the data and sequentially store the processed historical RF fingerprint data in a RF fingerprint database. Thus, when the amount of data in the RF fingerprint database reaches a threshold, the terminal may utilize a clustering algorithm to cluster the RF fingerprint data to form subspaces, and store the number of subspaces, spatial attributes, and space numbers within the current scene in the subspace database.

[0148] Please refer to Figure 6, which shows a flowchart of generating a subspace database provided by an exemplary embodiment of the present application. First, the terminal obtains radio frequency fingerprint data by starting WIFI scanning or cellular scanning, and preprocesses and filters the radio frequency fingerprint data, thereby storing the processed radio frequency fingerprint data in the radio frequency fingerprint database. Then, when the amount of data in the radio frequency fingerprint database reaches a threshold, that is, when the clustering condition is met, the terminal calculates the correlation coefficient between the radio frequency fingerprint data through a clustering algorithm, and clusters the radio frequency fingerprint data according to the correlation coefficient, that is, spatial clustering, thereby obtaining a spatial division result of the current scene, and storing each subspace in the current scene in the subspace database. Schematically, as shown in Figure 7, the current scene 701 generates 7 subspaces through spatial clustering.

[0149] Please refer to Figure 8, which shows a flowchart of RF fingerprint data clustering provided by an exemplary embodiment of the present application. First, in order to ensure that each RF fingerprint data is processed, each RF fingerprint data can be sample-labeled first, so that each RF fingerprint data is traversed in a loop and the accessed RF fingerprint data is relabeled. When traversing the first RF fingerprint data, it can be assumed to be the core point of a subspace, so that the RF fingerprint data is classified as a class C, and other RF fingerprint data that are correlated with the RF fingerprint data are obtained as a sample set N. In the sample set N, similarly, each RF fingerprint data in the set is traversed, and the RF fingerprint data with a high correlation with the core point are all classified into class C, until all RF fingerprint data in the sample set N are traversed, and then a new core point is re-determined, and so on, until all RF fingerprint data are traversed, thereby performing spatial division of the current scene according to the clustering results of the RF fingerprint data, that is, the various subspaces within the current scene can be obtained.

[0150] In the above embodiment, by collecting historical RF fingerprint data within the scene, spatial clustering is performed based on the correlation between the historical RF fingerprint data to obtain multiple subspaces corresponding to the scene, thereby improving the efficiency and accuracy of spatial clustering and scene division.

[0151] In some embodiments, after dividing and spatially clustering the scene to obtain the subspaces within the scene and generating the communication semantic knowledge graph corresponding to each subspace, the terminal can determine the target subspace and generate a network recommendation result based on the current RF fingerprint data.

[0152] Please refer to FIG9 , which shows a flow chart of a network switching method provided by another exemplary embodiment of the present application. The method includes the following steps:

[0153] Step 901: Acquire current radio frequency fingerprint data, where the current radio frequency fingerprint data represents radio frequency characteristics of the currently connected network.

[0154] The specific implementation of this step can refer to step 101, and this embodiment will not be described in detail here.

[0155] Step 902: Based on the spatial attributes of each subspace in the current scene and the current radio frequency fingerprint data, determine a first correlation coefficient between each subspace and the current radio frequency fingerprint data, where the first correlation coefficient represents the degree of association between the current radio frequency fingerprint data and the subspace.

[0156] In some embodiments, in order to improve the accuracy of spatial positioning, after obtaining the current RF fingerprint data, the terminal can also calculate the first correlation coefficient between the current RF fingerprint data and each subspace based on the spatial attributes of each subspace in the current scene and the current RF fingerprint data, thereby quantifying the degree of correlation between the current RF fingerprint data and each subspace.

[0157] Optionally, the first correlation coefficient represents the degree of correlation between the current radio frequency fingerprint data and the subspace, and the first correlation coefficient is positively correlated with the degree of correlation. The higher the degree of correlation, the greater the first correlation coefficient.

[0158] In a possible implementation, the terminal may calculate a correlation coefficient between the historical RF fingerprint data and the current RF fingerprint data based on the historical RF fingerprint data corresponding to each subspace, thereby serving as a first correlation coefficient between each subspace and the current RF fingerprint data.

[0159] Optionally, the calculation method of the first correlation coefficient may be a Pearson correlation coefficient calculation method, a cosine similarity calculation method, an adjusted cosine similarity calculation method, etc., which is not limited in the embodiment of the present application.

[0160] Step 903: Based on the first correlation coefficients between each subspace and the current radio frequency fingerprint data, the current target subspace is determined, and the first correlation coefficient corresponding to the target subspace is higher than the first correlation coefficients corresponding to other subspaces.

[0161] In some embodiments, after obtaining the first correlation coefficient between each subspace and the current radio frequency fingerprint data, the terminal can determine the current target subspace according to the correlation degree represented by each first correlation coefficient.

[0162] In a possible implementation, the terminal may determine the subspace corresponding to the maximum value of the first correlation coefficient as the target subspace based on the sorting results of the first correlation coefficients, that is, the first correlation coefficient corresponding to the target subspace is higher than the first correlation coefficients corresponding to other subspaces.

[0163] Step 904: Based on the communication semantic knowledge graph, a path list corresponding to the target subspace is generated through graph traversal. The path list includes multiple connection relationship pairs consisting of different network requirements and network connection entrances.

[0164] In some embodiments, after determining the current target subspace and obtaining the communication semantic knowledge graph corresponding to the target subspace, the terminal can traverse the communication semantic knowledge graph through graph traversal, and thereby generate a path list corresponding to the target subspace based on the traversal results.

[0165] Optionally, the path list includes multiple connection relationship pairs consisting of different network usage requirements and network connection entries. For example, network usage requirements include network usage time, network application, network application type, etc., and network parameters corresponding to the network connection entry may include network type, network quality indicators, etc. If the network application is the same but the network usage time is different, it may correspond to different network connection entries, that is, different network types and network quality indicators.

[0166] In one possible implementation, the terminal may first determine multiple demand nodes corresponding to the network usage requirements in the communication semantic knowledge graph based on a variety of different network usage requirements, and then, starting from the demand node, traverse other nodes connected to the demand node through graph traversal in the communication semantic knowledge graph, thereby determining multiple network connection entrances that meet the demand node, and then, based on the node path between the network usage requirement and the network connection entrance, form a connection relationship pair, and generate a path list corresponding to the target subspace based on the multiple connection relationship pairs formed by each network usage requirement and the network connection entrance.

[0167] The network usage requirements may include at least one of network usage time, network application, and network application type. If the network application is the same but the network usage time is different, it may correspond to different demand nodes, thereby forming different node paths with different network connection entrances.

[0168] Optionally, the process of generating a path list based on the communication semantic knowledge graph can also be performed by the server. In one possible implementation, after the server generates the communication semantic knowledge graph corresponding to each subspace based on the historical RF fingerprint data uploaded by each terminal device in the scene, the server can further traverse the communication semantic knowledge graph, starting from each demand node, and forming a node path with different network connection entrances to obtain a path list. Then, during the process of network switching by the terminal, the path list corresponding to each subspace can be directly obtained from the server, or the path list corresponding to the target subspace can be directly obtained.

[0169] Step 905 : Based on each connection relationship pair in the path list, the candidate networks corresponding to the network connection entry are predicted and scored to obtain the network confidence and network recommendation score corresponding to each candidate network.

[0170] In some embodiments, after obtaining multiple connection relationship pairs in the path list, the terminal can predict and score the candidate networks corresponding to the network connection entrances according to the current network demand, thereby determining the network confidence and network recommendation score corresponding to each candidate network.

[0171] The network recommendation score represents the availability of a candidate network, while the network confidence represents the accuracy of the network recommendation score. In one possible implementation, the terminal can determine the probability of reaching the target node from different node paths to the network connection portal based on the target node corresponding to the current network demand. Based on the path probability corresponding to each connection relationship, the terminal can determine the network confidence and network recommendation score for each candidate network.

[0172] In one possible implementation, the terminal first obtains the current network usage demand, which may include at least one of the current network usage time, the current network application, and the network application type, and determines the path length corresponding to each connection relationship in the path list and the network quality of the candidate network. Then, the terminal predicts and scores each candidate network based on the current network usage demand, the path length, and the network quality of the candidate network, and obtains the network confidence and network recommendation score corresponding to each candidate network.

[0173] Among them, network confidence is negatively correlated with path length, network confidence is positively correlated with network quality, network recommendation score is negatively correlated with path length, and network recommendation score is positively correlated with network quality.

[0174] Step 906: Obtain current network usage requirements, where the current network usage requirements include at least one of current network usage time, current network application, and network application type.

[0175] In some embodiments, in order to improve the efficiency of determining network recommendation results, the terminal can also determine the network confidence and network recommendation score of each candidate network based on the network recommendation model, so that the terminal needs to first obtain the current network usage demand, wherein the current network usage demand includes at least one of the current network usage time, the current network application, and the network application type.

[0176] In step 907, the current network demand and the communication semantic knowledge graph corresponding to the target subspace are input into the network recommendation model, and the network recommendation model outputs the network confidence and network recommendation score corresponding to each candidate network.

[0177] In some embodiments, the terminal inputs the current network demand and the communication semantic knowledge graph corresponding to the target subspace into the network recommendation model, and outputs the network confidence and network recommendation score corresponding to each candidate network through the network recommendation model.

[0178] In a possible implementation, the terminal may also input the current network demand and the path list corresponding to the target subspace into the network recommendation model, and output the network confidence and network recommendation score corresponding to each candidate network through the network recommendation model.

[0179] Optionally, in order to improve the output efficiency of the network recommendation model, the terminal also needs to train the network recommendation model first. In one possible implementation, the terminal first obtains the sample network requirements, the sample candidate networks, and the network experience quality corresponding to the sample candidate networks, and inputs the sample network requirements and the communication semantic knowledge graph corresponding to the target subspace into the network recommendation model. The network recommendation model outputs the sample network confidence and sample network recommendation score corresponding to each sample candidate network, and then trains the network recommendation model based on the sample network confidence, sample network recommendation score, and network experience quality corresponding to the sample candidate networks to obtain a trained network recommendation model.

[0180] In a possible implementation, in order to reduce the data processing pressure of the terminal during the model training process, the network recommendation general model can be trained through the server based on the general sample network usage requirements to obtain the trained network recommendation general model, so that the terminal can directly obtain the path list corresponding to the target subspace and the network recommendation general model from the server side, and perform personalized training on the network recommendation general model based on the individual sample network usage requirements to obtain the corresponding network recommendation model.

[0181] Step 908: Generate a network recommendation result based on the network confidence and network recommendation score corresponding to each candidate network.

[0182] In some embodiments, after obtaining the network confidence and network recommendation score corresponding to each candidate network, the terminal may sort the candidate networks according to the network confidence and network recommendation score, and generate a network recommendation result.

[0183] In one possible implementation, considering that the network recommendation results are predicted and scored based on objective standards, in order to make the network recommendation results more in line with actual historical network usage, the terminal can also obtain historical network connection status, which includes at least one of the number of historical network connections, the duration of historical network connections, and the quality of historical network connections. Based on the historical network connection status, the network recommendation score corresponding to the candidate network is corrected to obtain a corrected network recommendation score, and a network recommendation result is generated based on the network confidence corresponding to each candidate network and the corrected network recommendation score.

[0184] Schematically, as shown in FIG10 , historical network connection status can reflect the user's network usage preferences, and the terminal can determine the user's network usage preferences through intent recognition based on the user's manual network switching operations, semantic behaviors, and semantic scenarios, and thus modify the network recommendation scores of each candidate network based on the network usage preferences to obtain the final network recommendation result.

[0185] Step 909 , performing network screening based on the network recommendation result to obtain multiple networks to be switched, wherein the network confidence of the network to be switched is higher than the confidence threshold, and the network recommendation score of the network to be switched is higher than the network score of the currently connected network.

[0186] In some embodiments, considering that the network recommendation results include the network confidence and network recommendation score of each candidate network in the path list, and the network recommendation scores of some candidate networks are obviously low, in order to improve the efficiency of subsequent network switching, the terminal can also first perform network screening based on the network recommendation results to determine multiple networks to be switched.

[0187] Optionally, the terminal may set a confidence threshold to filter candidate networks whose network confidence is lower than the confidence threshold, so that the network confidence of the network to be switched is higher than the confidence threshold.

[0188] Optionally, the terminal may also determine the network score of the currently connected network based on the network quality of the currently connected network, and filter candidate networks whose network recommendation scores are lower than the network score of the currently connected network, so that the network recommendation score of the network to be switched is higher than the network recommendation score of the currently connected network.

[0189] Optionally, the terminal can also determine the score difference between the network recommendation score of each candidate network and the network score of the currently connected network by setting a score difference threshold, and filter the candidate networks whose score difference is lower than the score difference threshold, so that the network score difference between the network to be switched and the currently connected network is greater than the score difference threshold.

[0190] Step 910: When the currently connected network is a WiFi network, the network to be switched is a cellular network, the network switch of the cellular network is on, and the network signal strength of the network to be switched meets the signal strength threshold, network switching is performed.

[0191] In some embodiments, after obtaining multiple networks to be switched, the terminal may determine whether to perform network switching based on the network types and network signal strengths of the currently connected network and the network to be switched.

[0192] In a possible implementation, when the currently connected network is a WIFI network, the network to be switched is a cellular network, the network switch of the cellular network is on, and the network signal strength of the network to be switched meets the signal strength threshold, the terminal can perform network switching.

[0193] In a possible implementation, when the currently connected network is a WIFI network, the network to be switched is a cellular network, and the network switch of the cellular network is in an off state, the terminal does not perform network switching.

[0194] In a possible implementation, when the currently connected network is a WIFI network, the network to be switched is a cellular network, the network switch of the cellular network is on, and the network signal strength of the network to be switched does not meet the signal strength threshold, the terminal does not perform network switching.

[0195] Step 911: When both the currently connected network and the network to be switched are WIFI networks and the network signal strength of the network to be switched meets the signal strength threshold, network switching is performed.

[0196] In a possible implementation, when the currently connected network and the network to be switched are both WIFI networks and the network signal strength of the network to be switched meets a signal strength threshold, the terminal can perform network switching.

[0197] In a possible implementation, when the currently connected network and the network to be switched are both WIFI networks, and the network signal strength of the network to be switched does not meet the signal strength threshold, the terminal does not perform network switching.

[0198] In a possible implementation, when the currently connected network is a cellular network, the terminal does not perform network switching.

[0199] Please refer to Figure 11, which shows a flowchart of network switching provided by an exemplary embodiment of the present application. After obtaining the network recommendation results, the terminal first screens the candidate networks based on the network confidence and confidence threshold of each candidate network. After obtaining the screened candidate networks, the terminal calculates the score difference between the network recommendation score of the candidate network and the network score of the currently connected network, and further screens the candidate networks based on the score difference and the difference threshold to obtain the network to be switched. Then, based on the network type of the currently connected network and the network to be switched, it is determined whether to perform network switching. When the currently connected network is a cellular network, network switching will not be performed; when the currently connected network and the network to be switched are both WIFI networks, and the network signal strength of the network to be switched is higher than the signal strength threshold, network switching will be performed; when the currently connected network and the network to be switched are both WIFI networks, and the network signal strength of the network to be switched is lower than the signal strength threshold, network switching will not be performed; when the currently connected network is a WIFI network, the network to be switched is a cellular network, the network switch of the cellular network is on, and the network signal strength of the network to be switched meets the signal strength threshold, network switching will be performed; when the currently connected network is a WIFI network, the network to be switched is a cellular network, and the network switch of the cellular network is off, network switching will not be performed; when the currently connected network is a WIFI network, the network to be switched is a cellular network, the network switch of the cellular network is on, and the network signal strength of the network to be switched does not meet the signal strength threshold, network switching will not be performed.

[0200] In the above embodiment, spatial positioning accuracy is improved by calculating the first correlation coefficient between the current RF fingerprint data and each subspace within the scenario and performing spatial positioning based on the first correlation coefficient. Furthermore, by traversing the communication semantic knowledge graph of the target subspace, the connection relationship pairs between various network requirements and network connection portals are determined, thereby determining the network recommendation score for each candidate network based on the current network requirement, thereby optimizing the efficiency and accuracy of network recommendations. Furthermore, by correcting the network recommendation score based on historical network connection status, the rationality and adaptability of the network recommendation results are increased, thereby optimizing the user's network experience.

[0201] Please refer to FIG. 12 , which shows a flowchart of generating a path list provided by an exemplary embodiment of the present application. The process can be executed by a terminal or a server.

[0202] When the process is executed by the server, the server first needs to obtain the historical network usage data stored by the terminal, and construct communication semantic features based on the historical network usage data, and then obtain the processed historical network usage data by filtering, screening and statistical processing of the historical network usage data, and then construct a communication semantic knowledge graph based on the historical network usage data and communication semantic feature data. In this way, the terminal can directly query and obtain the communication semantic knowledge graph corresponding to the current scenario from the server, and determine the connection relationship pairs between each network usage demand and the network connection entrance by traversing the communication semantic knowledge graph, thereby generating a path list.

[0203] Please refer to FIG13 , which shows a flowchart of generating a network recommendation list and model training provided by an exemplary embodiment of the present application.

[0204] Considering that model training requires a large amount of data, the terminal can first perform data detection on the communication semantic knowledge graph to determine whether the model training conditions are met. If the model training conditions are met, the model training process can be started. First, by traversing the communication semantic knowledge graph, the sample feature data required for model training is generated, and then the currently stored model data is obtained from the server. When the model data is the latest version of the data, the network recommendation model is trained based on the sample feature data. If an abnormality occurs during the training process, the training time needs to be readjusted according to the current state machine logic. Furthermore, after the model training is completed, in the process of actually applying the model for network recommendation, the terminal can also retrain the model based on data with network experience feedback, thereby improving the output accuracy of the model. And after the model training is completed, the terminal can directly store the network recommendation list and model data.

[0205] Please refer to FIG14 , which shows a flowchart of performing network switching provided by an exemplary embodiment of the present application.

[0206] In actual network switching scenarios, the terminal can determine whether to initiate a network switching service based on home fence information, subspace changes, service status changes, and time period changes. When initiating a network switch, the terminal can directly retrieve the network recommendation scores of each candidate network from the stored network recommendation list and modify the corresponding network recommendation scores of the candidate networks based on user preferences to generate a network recommendation result. If no candidate network information is available from the network recommendation list, the terminal will need to invoke the network recommendation model data and, using the network recommendation model, output the corresponding network recommendation scores of each candidate network. This network recommendation score will then be modified based on user preferences to generate a network recommendation result.

[0207] Please refer to FIG15 , which shows a flowchart of executing network switching provided by another exemplary embodiment of the present application.

[0208] First, the terminal determines the current scene, such as home, through the positioning subsystem. When it is determined that the terminal is at home, the terminal further obtains the information required for network recommendation, including the subspace positioning result, the current system time, the current network recommendation model, and the currently running application. After obtaining all the information, it further determines whether the network recommendation conditions are met. The network recommendation conditions include at least one of subspace changes, time period changes, and foreground application status changes. When the network recommendation conditions are met, the terminal executes the network recommendation algorithm process, outputs the network recommendation score of the candidate network through the network recommendation model, and sorts and filters it, thereby passing the network recommendation result to the recommendation sub-function, and the network switching unit determines whether to perform network switching based on the network recommendation result.

[0209] Please refer to FIG16 , which shows a flow chart of a network switching method provided by yet another exemplary embodiment of the present application.

[0210] Among them, the overall process of network switching can be divided into two stages: training and application verification.

[0211] First, in the training phase, a large amount of historical RF fingerprint data is collected. When the amount of historical RF fingerprint data reaches a quantity threshold, spatial clustering is performed based on the historical RF fingerprint data to obtain the subspaces within the scene, and the results are stored in the subspace database. Furthermore, based on the historical network usage data and historical RF fingerprint data, a communication semantic knowledge graph corresponding to each subspace is constructed, and through graph traversal, multiple connection relationship pairs between network usage requirements and network connection entrances are determined, thereby generating a path list and storing it in the path list database.

[0212] During the application verification phase, after the terminal receives the positioning request and obtains the current RF fingerprint data, it filters the RF fingerprint data and, if the current RF fingerprint data is valid, determines the current target subspace through the positioning algorithm based on the current RF fingerprint data and the spatial attributes of each subspace. Then, based on the current network demand and the connection relationships in the path list, the terminal determines the network recommendation result for the corresponding candidate network, and performs network switching based on the currently connected network and the network recommendation result.

[0213] Please refer to Figure 17, which shows a flowchart of predictive network switching provided by an exemplary embodiment of the present application. To optimize the network experience, the terminal needs to monitor the network status in real time, predict the current network usage, and determine whether to make a network recommendation. If a network recommendation is required, the terminal needs to make a network recommendation based on the communication semantic knowledge graph corresponding to the target subspace and the current network demand, thereby optimizing the current wireless network communication status.

[0214] Please refer to Figure 18, which shows a structural block diagram of a network switching device provided by an exemplary embodiment of the present application. The device includes:

[0215] The data acquisition module 1801 is used to acquire current radio frequency fingerprint data, where the current radio frequency fingerprint data represents the radio frequency characteristics of the currently connected network;

[0216] A spatial positioning module 1802 is configured to determine a current target subspace based on spatial attributes of each subspace within the current scene and the current RF fingerprint data, wherein the subspace is obtained by spatial clustering based on historical RF fingerprint data, and the spatial attributes include historical RF fingerprint features of the subspace;

[0217] A result generation module 1803 is configured to generate a network recommendation result based on a communication semantic knowledge graph corresponding to the target subspace, wherein the communication semantic knowledge graph represents the association relationship between network usage data of different dimensions in the target subspace;

[0218] The network switching module 1804 is configured to perform network switching based on the current network connection status and the network recommendation result.

[0219] Optionally, the result generation module 1803 includes:

[0220] a list generating unit, configured to generate a path list corresponding to the target subspace by graph traversal based on the communication semantic knowledge graph, wherein the path list includes a plurality of connection relationship pairs consisting of different network requirements and network connection entrances;

[0221] a scoring unit, configured to predict and score the candidate networks corresponding to the network connection entries based on each connection relationship pair in the path list, and obtain a network confidence and a network recommendation score corresponding to each candidate network;

[0222] The result generating unit is configured to generate the network recommendation result based on the network confidence and network recommendation score corresponding to each candidate network.

[0223] Optionally, the list generating unit is configured to:

[0224] Based on multiple network usage requirements, determining multiple requirement nodes corresponding to the network usage requirements in the communication semantic knowledge graph, where the network usage requirements include at least one of network usage time, network usage application, and network usage application type;

[0225] Starting from the demand node, determining multiple network connection entrances that meet the demand node by graph traversal in the communication semantic knowledge graph;

[0226] forming the connection relationship pair based on the node path between the network usage demand and the network connection entrance;

[0227] Based on the plurality of connection relationship pairs formed by the various network usage requirements and the network connection entrances, the path list corresponding to the target subspace is generated.

[0228] Optionally, the scoring unit is used to:

[0229] Obtaining current network usage demand, where the current network usage demand includes at least one of current network usage time, current network application, and network application type;

[0230] Determine the path length corresponding to each connection relationship pair in the path list and the network quality of the candidate network;

[0231] Based on the current network demand, the path length, and the network quality of the candidate network, each candidate network is predicted and scored to obtain a network confidence and a network recommendation score corresponding to each candidate network, wherein the network confidence is negatively correlated with the path length, the network confidence is positively correlated with the network quality, the network recommendation score is negatively correlated with the path length, and the network recommendation score is positively correlated with the network quality.

[0232] Optionally, the result generation module 1803 further includes:

[0233] a demand acquisition unit, configured to acquire a current network usage demand, wherein the current network usage demand includes at least one of a current network usage time, a current network application, and a network application type;

[0234] an output unit, configured to input the current network demand and the communication semantic knowledge graph corresponding to the target subspace into a network recommendation model, and output the network confidence and network recommendation score corresponding to each candidate network through the network recommendation model;

[0235] The result generating unit is further configured to generate the network recommendation result based on the network confidence and network recommendation score corresponding to each candidate network.

[0236] Optionally, the device further includes:

[0237] A demand acquisition module is used to obtain sample network requirements, sample candidate networks, and network experience quality corresponding to the sample candidate networks;

[0238] an output module, configured to input the sample network requirements and the communication semantic knowledge graph corresponding to the target subspace into the network recommendation model, and output the sample network confidence and sample network recommendation score corresponding to each sample candidate network through the network recommendation model;

[0239] A training module is used to train the network recommendation model based on the sample network confidence, the sample network recommendation score and the network experience quality corresponding to the sample candidate network to obtain the trained network recommendation model.

[0240] Optionally, the result generating unit is further configured to:

[0241] Acquiring historical network connection status, wherein the historical network connection status includes at least one of the number of historical network connections, the duration of historical network connections, and the quality of historical network connections;

[0242] Based on the historical network connection status, the network recommendation score corresponding to the candidate network is corrected to obtain the corrected network recommendation score;

[0243] The network recommendation result is generated based on the network confidence corresponding to each candidate network and the revised network recommendation score.

[0244] Optionally, the network switching module 1804 is configured to:

[0245] Performing network screening based on the network recommendation result to obtain multiple networks to be switched, wherein the network confidence of the networks to be switched is higher than a confidence threshold, and the network recommendation score of the networks to be switched is higher than the network score of the currently connected network;

[0246] When the currently connected network is a WiFi network, the network to be switched is a cellular network, the network switch of the cellular network is on, and the network signal strength of the network to be switched meets a signal strength threshold, network switching is performed;

[0247] When both the currently connected network and the network to be switched are WIFI networks and the network signal strength of the network to be switched meets a signal strength threshold, network switching is performed.

[0248] Optionally, the network switching module 1804 is further configured to:

[0249] When the currently connected network is a cellular network, no network switching is performed;

[0250] If both the currently connected network and the network to be switched are WiFi networks, and the network signal strength of the network to be switched does not meet the signal strength threshold, network switching is not performed;

[0251] If the currently connected network is a WiFi network, the network to be switched is a cellular network, and the network switch of the cellular network is in the off state, no network switching is performed;

[0252] If the currently connected network is a WiFi network, the network to be switched is a cellular network, the network switch of the cellular network is on, and the network signal strength of the network to be switched does not meet the signal strength threshold, network switching is not performed.

[0253] Optionally, the spatial positioning module 1802 is configured to:

[0254] Determining, based on the spatial attributes of each subspace in the current scene and the current radio frequency fingerprint data, a first correlation coefficient between each subspace and the current radio frequency fingerprint data, wherein the first correlation coefficient represents a degree of association between the current radio frequency fingerprint data and the subspace;

[0255] Based on the first correlation coefficients between each subspace and the current radio frequency fingerprint data, the current target subspace is determined, and the first correlation coefficient corresponding to the target subspace is higher than the first correlation coefficients corresponding to other subspaces.

[0256] Optionally, the device further includes:

[0257] A first data collection module is configured to collect historical network usage data corresponding to each subspace, wherein the historical network usage data includes historical network usage time, historical network applications, application types of historical network applications, historical radio frequency fingerprint data, historical protocol measurement data, and historical network experience data;

[0258] A data processing module, configured to clean and compile the historical network usage data to obtain processed historical network usage data;

[0259] The graph generation module is used to generate a communication semantic knowledge graph corresponding to each subspace based on the processed historical network usage data.

[0260] Optionally, the graph generation module is further used to:

[0261] Determining, based on the processed historical web usage data, feature dimensions corresponding to each piece of historical web usage data;

[0262] Determining, based on characteristic dimensions corresponding to each historical web usage data, correlation relationships between the historical web usage data of different dimensions;

[0263] Taking the historical network usage data as nodes and the association relationships between the historical network usage data as edges, association paths between different historical network usage data are generated to obtain the communication semantic knowledge graph corresponding to each subspace.

[0264] Optionally, the device further includes:

[0265] The second data acquisition module is used to collect historical radio frequency fingerprint data corresponding to different scanning conditions in the current scene;

[0266] a coefficient calculation module, configured to calculate the correlation coefficients between each pair of the historical radio frequency fingerprint data when the amount of the historical radio frequency fingerprint data reaches a quantity threshold, to obtain a second correlation coefficient corresponding to each piece of historical radio frequency fingerprint data;

[0267] The spatial clustering module is used to perform spatial clustering based on the second correlation coefficient corresponding to each historical radio frequency fingerprint data to obtain each subspace corresponding to the current scene.

[0268] In summary, in the embodiments of the present application, by obtaining the current RF fingerprint data and performing spatial positioning based on the spatial attributes of each subspace within the current scene and the current RF fingerprint data, the current target subspace is determined. Based on the communication semantic knowledge graph corresponding to the target subspace, a network recommendation result is generated, and network switching is performed based on the current network connection status and the network recommendation result. The solution provided by the embodiments of the present application can improve the efficiency of network switching within complex subspaces and optimize the network experience.

[0269] Please refer to FIG19 , which shows a schematic structural diagram of a terminal provided by an exemplary embodiment of the present application.

[0270] The terminal 1900 can execute the network switching method of the above embodiment, such as a smartphone, smartwatch, vehicle-mounted terminal, tablet computer, laptop computer, desktop computer, Bluetooth headset, etc. The terminal 1900 may also be referred to as user equipment, portable terminal, or other names. The terminal 1900 may also include one or more of the following components: a processor 1910 and a memory 1920.

[0271] Optionally, the processor 1910 uses various interfaces and lines to connect various parts of the entire electronic device, and performs various functions of the electronic device and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 1920, and calling data stored in the memory 1920. Optionally, the processor 1910 can be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), and programmable logic array (PLA). The processor 1910 can integrate one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), a neural network processing unit (NPU), and a baseband chip. Among them, the CPU mainly processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing the content required to be displayed on the touch screen; the NPU is used to implement artificial intelligence (AI) functions; and the baseband chip is used to handle wireless communication. It is understandable that the above-mentioned baseband chip may not be integrated into the processor 1910, but may be implemented by a separate chip.

[0272] The memory 1920 may include a random access memory (RAM) or a read-only memory (ROM). Optionally, the memory 1920 includes a non-transitory computer-readable storage medium. The memory 1920 may be used to store instructions, programs, codes, code sets, or instruction sets. The memory 1920 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function, instructions for implementing the various method embodiments described above, and the data storage area may store data created based on the use of the electronic device.

[0273] In addition, those skilled in the art will understand that the structure of the electronic device shown in the above figures does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figures, or combine certain components, or arrange the components differently.

[0274] The present application also provides a computer-readable storage medium, wherein the computer-readable storage medium stores at least one instruction, and the at least one instruction is loaded and executed by a processor to implement the method described in the above embodiment. Optionally, the computer-readable storage medium may include: ROM, RAM, solid-state drives (SSDs), or optical disks. Among them, RAM may include resistance random access memory (ReRAM) and dynamic random access memory (DRAM).

[0275] The present application also provides a computer program product, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the network switching method provided in various optional implementations of the above aspects.

[0276] The above description is merely an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A network switching method, the method comprising: Acquire current radio frequency fingerprint data, where the current radio frequency fingerprint data represents radio frequency characteristics of the currently connected network; Determine the current target subspace based on the spatial attributes of each subspace in the current scene and the current radio frequency fingerprint data, wherein the subspace is obtained by spatial clustering based on the historical radio frequency fingerprint data, and the spatial attributes include the historical radio frequency fingerprint features of the subspace; Generate a network recommendation result based on the communication semantic knowledge graph corresponding to the target subspace, wherein the communication semantic knowledge graph represents the association relationship between network usage data of different dimensions in the target subspace; Based on the current network connection status and the network recommendation result, network switching is performed.

2. The method according to claim 1, wherein: The generating of network recommendation results based on the communication semantic knowledge graph corresponding to the target subspace includes: Based on the communication semantic knowledge graph, a path list corresponding to the target subspace is generated through graph traversal, wherein the path list includes a plurality of connection relationship pairs consisting of different network requirements and network connection entrances; Based on each connection relationship pair in the path list, predict and score the candidate network corresponding to the network connection entrance to obtain the network confidence and network recommendation score corresponding to each candidate network; The network recommendation result is generated based on the network confidence and network recommendation score corresponding to each candidate network.

3. The method according to claim 2, wherein: The generating a path list corresponding to the target subspace by graph traversal based on the communication semantic knowledge graph includes: Based on multiple network usage requirements, multiple requirement nodes corresponding to the network usage requirements are determined in the communication semantic knowledge graph, where the network usage requirements include at least one of network usage time, network usage application, and network usage application type; Starting from the demand node, a plurality of network connection entrances that meet the demand node are determined by graph traversal in the communication semantic knowledge graph; Based on the network demand and the node path between the network connection entrance, forming the connection relationship pair; Based on various network usage requirements and multiple connection relationship pairs formed by network connection entrances, the path list corresponding to the target subspace is generated.

4. The method according to claim 2, wherein: The predicting and scoring of the candidate networks corresponding to the network connection entry based on each connection relationship pair in the path list to obtain the network confidence and network recommendation score corresponding to each candidate network includes: Acquire current network usage demand, where the current network usage demand includes at least one of current network usage time, current network application, and network application type; Determine the path length corresponding to each connection relationship pair in the path list and the network quality of the candidate network; Based on the current network demand, the path length, and the network quality of the candidate network, each candidate network is predicted and scored to obtain a network confidence and a network recommendation score corresponding to each candidate network, wherein the network confidence is negatively correlated with the path length, the network confidence is positively correlated with the network quality, the network recommendation score is negatively correlated with the path length, and the network recommendation score is positively correlated with the network quality.

5. The method according to any one of claims 1 to 4, wherein: The generating of network recommendation results based on the communication semantic knowledge graph corresponding to the target subspace further includes: Acquire current network usage demand, where the current network usage demand includes at least one of current network usage time, current network application, and network application type; Input the current network demand and the communication semantic knowledge graph corresponding to the target subspace into a network recommendation model, and output the network confidence and network recommendation score corresponding to each candidate network through the network recommendation model; The network recommendation result is generated based on the network confidence and network recommendation score corresponding to each candidate network.

6. The method according to claim 5, wherein: The method further comprises: Obtaining sample network requirements, sample candidate networks, and network experience quality corresponding to the sample candidate networks; Input the sample network demand and the communication semantic knowledge graph corresponding to the target subspace into the network recommendation model, and output the sample network confidence and sample network recommendation score corresponding to each sample candidate network through the network recommendation model; Based on the sample network confidence, the sample network recommendation score, and the network experience quality corresponding to the sample candidate network, the network recommendation model is trained to obtain the trained network recommendation model.

7. The method according to claim 2 or 5, wherein: The generating the network recommendation result based on the network confidence and network recommendation score corresponding to each candidate network includes: Acquire historical network connection status, where the historical network connection status includes at least one of the number of historical network connections, the duration of historical network connections, and the quality of historical network connections; Based on the historical network connection status, the network recommendation score corresponding to the candidate network is corrected to obtain a corrected network recommendation score; The network recommendation result is generated based on the network confidence corresponding to each candidate network and the corrected network recommendation score.

8. The method according to claim 2 or 5, wherein: The performing network switching based on the current network connection status and the network recommendation result includes: Performing network screening based on the network recommendation result to obtain multiple networks to be switched, wherein the network confidence of the network to be switched is higher than the confidence threshold, and the network recommendation score of the network to be switched is higher than the network score of the currently connected network; When the currently connected network is a WIFI network, the network to be switched is a cellular network, the network switch of the cellular network is on, and the network signal strength of the network to be switched meets the signal strength threshold, network switching is performed; When the currently connected network and the network to be switched are both WIFI networks and the network signal strength of the network to be switched meets the signal strength threshold, network switching is performed.

9. The method according to claim 8, wherein: The method further comprises: When the currently connected network is a cellular network, no network switching is performed; If both the currently connected network and the network to be switched are WIFI networks, and the network signal strength of the network to be switched does not meet the signal strength threshold, network switching is not performed; When the currently connected network is a WIFI network, and the network to be switched is a cellular network, and the network switch of the cellular network is in the off state, no network switching is performed; When the currently connected network is a WIFI network, and the network to be switched is a cellular network, and the network switch of the cellular network is on, and the network signal strength of the network to be switched does not meet the signal strength threshold, network switching is not performed.

10. The method according to any one of claims 1 to 9, wherein: The determining the current target subspace based on the spatial attributes of each subspace in the current scene and the current radio frequency fingerprint data includes: Based on the spatial attributes of each subspace in the current scene and the current radio frequency fingerprint data, determine a first correlation coefficient between each subspace and the current radio frequency fingerprint data, wherein the first correlation coefficient represents a degree of association between the current radio frequency fingerprint data and the subspace; Based on the first correlation coefficient between each subspace and the current radio frequency fingerprint data, the current target subspace is determined, and the first correlation coefficient corresponding to the target subspace is higher than the first correlation coefficients corresponding to other subspaces.

11. The method according to any one of claims 1 to 10, wherein: The method further comprises: Collecting historical network usage data corresponding to each subspace, the historical network usage data includes historical network usage time, historical network usage applications, application types of historical network usage applications, historical radio frequency fingerprint data, historical protocol measurement data, and historical network usage experience data; Performing data cleaning and statistics on the historical network usage data to obtain processed historical network usage data; Based on the processed historical network usage data, a communication semantic knowledge graph corresponding to each subspace is generated.

12. The method according to claim 11, wherein: The generating of the communication semantic knowledge graph corresponding to each subspace based on the processed historical network data includes: Determining feature dimensions corresponding to each piece of historical web usage data based on the processed historical web usage data; Determining, based on the characteristic dimensions corresponding to each historical web usage data, the correlation relationship between the historical web usage data of different dimensions; The historical network usage data is used as a node, and the association relationship between the historical network usage data is used as an edge to generate an association path between different historical network usage data, and obtain the communication semantic knowledge graph corresponding to each subspace.

13. The method according to any one of claims 1 to 12, wherein: The method further comprises: Collect historical RF fingerprint data corresponding to different scanning conditions in the current scene; When the amount of the historical radio frequency fingerprint data reaches a quantity threshold, calculating correlation coefficients between each pair of the historical radio frequency fingerprint data to obtain a second correlation coefficient corresponding to each piece of historical radio frequency fingerprint data; Based on the second correlation coefficients corresponding to the historical radio frequency fingerprint data, spatial clustering is performed to obtain subspaces corresponding to the current scene.

14. A network switching device, comprising: A data acquisition module, used to acquire current radio frequency fingerprint data, wherein the current radio frequency fingerprint data represents the radio frequency characteristics of the currently connected network; A spatial positioning module, configured to determine a current target subspace based on spatial attributes of each subspace in a current scene and the current radio frequency fingerprint data, wherein the subspace is obtained by spatial clustering based on historical radio frequency fingerprint data, and the spatial attributes include historical radio frequency fingerprint features of the subspace; A result generation module, used to generate a network recommendation result based on the communication semantics knowledge graph corresponding to the target subspace, wherein the communication semantics knowledge graph represents the association relationship between network usage data of different dimensions in the target subspace; The network switching module is used to switch the network based on the current network connection status and the network recommendation result.

15. A terminal, comprising a processor and a memory; the memory stores at least one computer instruction, and the at least one computer instruction is used to be executed by the processor to implement the network switching method according to any one of claims 1 to 13.

16. A computer-readable storage medium, wherein at least one computer instruction is stored in the computer-readable storage medium, and the at least one computer instruction is loaded and executed by a processor to implement the network switching method according to any one of claims 1 to 13.

17. A computer program product, comprising computer instructions, wherein the computer instructions are stored in a computer-readable storage medium; a processor of a terminal reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the terminal executes the network switching method as described in any one of claims 1 to 13.

Images