Account division settlement optimization method for mobile terminal

Abstract

The invention provides a separate account settlement optimization method for a mobile terminal, and the method comprises the steps: obtaining the geographic position information, signal intensity data and movement track records of a plurality of mobile terminals, taking the data as an input parameter, and employing a clustering algorithm to carry out the grouping of the mobile terminals, and obtaining a plurality of mobile terminal groups; for each mobile terminal group, a main control mobile terminal in the group is determined by adopting a priority scoring model, the priority scoring model comprehensively considers the similarity of the signal intensity, the geographic position and the moving track of the mobile terminal, and the priority score of each mobile terminal is calculated; selecting the mobile terminal with the highest score as a master control mobile terminal; and when a certain mobile terminal group needs to initiate a paging request, the master control mobile terminal of the group firstly initiates paging on behalf of the group, and if the paging fails, other mobile terminals are tried to initiate paging in sequence according to a priority score sorting result of the other mobile terminals in the group until the paging succeeds or all the mobile terminals are traversed.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an account splitting and settlement optimization method for a mobile terminal. Background Art

[0002] In the split-account settlement scenario of mobile terminals, how to optimize the paging strategy to improve communication efficiency and resource utilization when multiple devices settle at the same time is a key technical issue. At present, the common practice is that each device independently initiates a paging request, which will lead to repeated paging and resource waste. If the paging requests of all devices are simply merged, the overall paging failure rate may increase due to differences in device location and signal strength. The core contradiction faced by paging optimization is how to balance the waste of resources caused by repeated paging and the paging failure rate caused by excessive merging. The ideal optimization solution needs to comprehensively consider multiple factors such as the device's geographical location, signal strength, and movement trajectory, and dynamically adjust the paging strategy. However, the location and signal quality of the device are changing at any time, and the optimization algorithm must be able to perceive these changes in real time and respond quickly. In addition, the movement trajectories of different devices may be correlated. Paging optimization needs to exploit this correlation and coordinate the paging order and time of different devices to avoid mutual interference. The multi-device collaborative paging optimization problem can be formally described as an optimal control problem of a random process. Its state variables include the location and signal strength of each device, and the control variable is the paging strategy of each device. The goal is to minimize a cost function that comprehensively considers repeated paging and paging failures. Solving this optimization problem requires real-time tracking of state variables, predicting their future change trends, and solving a stochastic dynamic programming problem. This places high demands on the computing and storage resources of mobile terminals, and also requires real-time information interaction and synchronization between mobile terminals, which poses new challenges to the design of wireless communication networks. Summary of the invention

[0003] The present invention provides an account splitting settlement optimization method for a mobile terminal, which mainly includes: Acquire basic information of multiple mobile terminals, the basic information including geographic location information, signal strength data, and movement trajectory records; use the basic information as input parameters, and use a clustering algorithm to group the multiple mobile terminals to obtain a plurality of mobile terminal groups; For each mobile terminal group, a multi-layer perceptron is used to construct a priority scoring model, and the priority scoring model is used to determine the master mobile terminal in each mobile terminal group; When a mobile terminal group needs to initiate a paging request, the master mobile terminal of the group first initiates paging on behalf of the group. If the paging fails, other mobile terminals in the group are ranked according to their priority scores, and then try to initiate paging in turn until the paging is successful or all mobile terminals are exhausted. When determining the paging order of each mobile terminal group, a scheduling algorithm based on a graph is adopted, each mobile terminal group is abstracted as a node of the graph, and the potential interference relationship between the mobile terminal groups is abstracted as directed edges between the nodes. The graph is divided by a minimum cut algorithm to obtain a number of subgraphs that do not interfere with each other. The mobile terminal groups in the subgraphs are paged in parallel, and the subgraphs are paged serially. Track the location changes and dynamic changes of signal strength of mobile terminals in real time. When it is detected that the location change of a mobile terminal exceeds the preset location change threshold or the signal strength change exceeds the preset signal strength change threshold, trigger incremental update and only adjust the affected mobile terminal grouping to avoid global regrouping. During the paging process, a reinforcement learning framework is used to continuously optimize the priority scoring model and grouping strategy of each mobile terminal. The success rate of each paging is used as a feedback reward. The model parameters are updated through the policy gradient algorithm, and the weight parameters in the priority scoring model are adjusted, so that the system can adaptively adjust the paging strategy based on historical experience.

[0004] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention discloses a method for optimizing account splitting and settlement for mobile terminals. The method obtains the geographic location, signal strength and movement trajectory data of multiple mobile terminals, groups the mobile terminals using a clustering algorithm, and determines the master mobile terminal of each group using a priority scoring model. When paging, the master mobile terminal is prioritized to initiate, and after failure, other mobile terminals are tried in sequence according to priority. The present invention also uses a graph-based scheduling algorithm to determine the paging order between groups, and realizes an optimized combination of parallel and serial paging. In addition, the present invention can track the status changes of mobile terminals in real time and perform incremental updates to avoid frequent global reorganization. By introducing a reinforcement learning framework, the present invention can continuously optimize the scoring model and grouping strategy to improve the paging success rate. The method significantly improves the efficiency and reliability of mobile terminal group paging, and is suitable for large-scale mobile communication scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1 The present invention is a flow chart of an account splitting and settlement optimization method for a mobile terminal.

[0006] Figure 2 A schematic diagram of an account splitting settlement optimization method for a mobile terminal according to the present invention.

[0007] Figure 3 It is another schematic diagram of an account splitting settlement optimization method for a mobile terminal according to the present invention. DETAILED DESCRIPTION

[0008] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0009] like Figure 1-3 In this embodiment, a method for optimizing split-account settlement for a mobile terminal may specifically include: S101. Obtain basic information of multiple mobile terminals, the basic information including geographic location information, signal strength data, and movement trajectory records; use the basic information as an input parameter, and adopt a clustering algorithm to group the multiple mobile terminals to obtain a plurality of mobile terminal groups.

[0010] The method comprises the steps of: obtaining geographic location information, signal strength data and movement trajectory records of at least two mobile terminals; preprocessing the geographic location information, using a Kalman filter algorithm to remove outliers and noise data, and obtaining cleaned location data; normalizing the signal strength data, using a maximum-minimum normalization method to uniformly map the signal strength data of different mobile terminals to the interval [0, 1], and obtaining normalized strength data; compressing the movement trajectory records, using a Douglas-Peucker algorithm to extract key location points including stop points and inflection points, and obtaining compressed trajectory data; and combining the cleaned location data, normalized strength data and compressed The trajectory data is constructed into a feature vector; the distance is used to measure the similarity and obtain the grouping result of the mobile terminals; for each mobile terminal group, the location data within the mobile terminal group is density clustered to obtain the typical location distribution of the mobile terminal group; for each mobile terminal group, the signal strength data within the mobile terminal group is statistically analyzed to obtain the signal strength distribution of the mobile terminal group; for each mobile terminal group, the trajectory data within the mobile terminal group is frequently mined to obtain the typical mobile trajectory pattern of the mobile terminal group; the location distribution, signal strength distribution and mobile trajectory pattern feature attributes of each mobile terminal group are integrated to construct a mobile terminal group portrait.

[0011] Specifically, firstly, the geographic location information, signal strength data and mobile trajectory records of multiple mobile terminals are obtained, for example, the longitude and latitude coordinates are obtained through GPS positioning, the signal strength value is obtained through the signal strength detection module, and the mobile trajectory points are obtained through the trajectory recording module. Then, the geographic location information is preprocessed, and the Kalman filter algorithm is used to remove outliers and noise data, for example, the longitude and latitude coordinates that exceed the reasonable range are regarded as outliers and removed, and the coordinate points with random fluctuations are regarded as noise and removed, so as to obtain the cleaned location data. Then, the signal strength data is normalized, and the maximum-minimum value normalization method is used to uniformly map the signal strength data of different mobile terminals to the range of 0 to 1, for example, the signal strength values ​​of -90dBm to -60dBm are mapped to the range of 0 to 1, so as to obtain the normalized strength data. The mobile trajectory records are compressed, and the Douglas-Peucker algorithm is used to extract the key location points of the stay points and inflection points, for example, the location points with a moving speed lower than 1m / s and a duration of more than 10 minutes are regarded as stay points, and the location points with a moving direction change of more than 45 degrees are regarded as inflection points, so as to obtain the compressed trajectory data. The cleaned location data, normalized intensity data and compressed trajectory data are constructed into feature vectors, for example, using a two-dimensional vector of longitude and latitude to represent location features, using a one-dimensional vector of signal strength to represent intensity features, and using a sequence of stop points and inflection points to represent trajectory features. The feature vector is input into the K-means clustering algorithm, and the similarity is measured by calculating the Euclidean distance between mobile terminals. For example, the sum of the Euclidean distances of the location vectors, intensity vectors and trajectory vectors of two mobile terminals is calculated as the similarity measure, and the number of clusters K is set to 3. The cluster center is iteratively optimized until convergence, and the grouping result of the mobile terminals is obtained. For each group, the location data in the group is clustered by DBSCAN density to obtain the typical location distribution of the group. For example, the density radius is set to 100 meters, the density threshold is set to 5, and the location area with a density exceeding the threshold is extracted as the typical location distribution. For each group, the signal strength data in the group is statistically analyzed to obtain the signal strength distribution of the group, for example, the mean, variance, quantile and other statistics of the signal strength are calculated, and the probability density curve and cumulative distribution curve of the signal strength are drawn. For each group, the PrefixSpan frequent sequence mining is performed on the trajectory data in the group to obtain the typical movement trajectory pattern of the group. For example, the frequency threshold is set to 20%, and the stay point sequence and inflection point sequence with a frequency exceeding the threshold are mined as typical trajectory patterns. Finally, the characteristic attributes such as the location distribution, signal strength distribution and movement trajectory pattern of each group are integrated to construct a mobile terminal group portrait, which is convenient for intuitively displaying the characteristic differences of each group and providing a reference for subsequent personalized services.

[0012] S102: for each mobile terminal group, a multi-layer perceptron is used to construct a priority scoring model, and the priority scoring model is used to determine a master mobile terminal in each mobile terminal group.

[0013] The signal strength, geographic location and movement trajectory information of multiple mobile terminals are obtained; for each movement trajectory of the mobile terminal, two trajectories are sampled to obtain a sampling point sequence of equal length; the Euclidean distance matrix between the two sampling point sequences is calculated; the path with the minimum cost is found in the distance matrix, and the path with the minimum cost is used as the optimal match of the two trajectories; the similarity of the two trajectories is calculated according to the optimal match; for each mobile terminal, a weighted average model is used to calculate the comprehensive priority score of each mobile terminal according to the signal strength, geographic location and movement trajectory similarity, and the weight in the weighted average model is preset according to actual needs; the mobile terminal with the highest comprehensive priority score in each mobile terminal group is determined as the candidate master mobile terminal of the group; if there are multiple candidate master mobile terminals with the same comprehensive priority score, the signal strengths of the multiple candidate master mobile terminals are compared, and the candidate master mobile terminal with the highest signal strength is determined as the master mobile terminal of the group; the master mobile terminal information of each mobile terminal group is output, and the master mobile terminal information includes a mobile terminal identifier, a signal strength, a geographic location coordinate and a mobile terminal group to which it belongs.

[0014] Specifically, first, a multilayer perceptron network with three hidden layers is built, with 16, 8, and 4 nodes in the hidden layers, 1 node in the output layer, and ReLU as the activation function. The mean square error is used as the loss function, the Adam optimizer is used, and the learning rate is 0.1. The model is trained until convergence. The model outputs a priority score of 0-100. The signal strength data of multiple mobile terminals is obtained through the wireless communication module, for example, the RSSI value of each mobile terminal is obtained in dBm. At the same time, the geographical location coordinates of each mobile terminal are obtained through the GPS positioning module, for example, the longitude and latitude are obtained in degrees. Then, the historical movement trajectory data of each mobile terminal is obtained through the trajectory recording module, for example, a sequence of location points within a period of time is obtained. For the movement trajectory of each mobile terminal, the two trajectories are resampled using a linear interpolation algorithm to obtain a sequence of sampling points of equal length, for example, the trajectory is resampled into 100 equally spaced location points. Then, the Euclidean distance matrix between the two sampling point sequences is calculated, for example, the value of each element in the matrix is ​​calculated using the straight-line distance formula between two points. Then, find the path with the least cost in the distance matrix, for example, use the dynamic programming algorithm to find the diagonal path with the smallest cumulative distance, and use this path as the optimal match of the two trajectories. Calculate the similarity of the two trajectories based on the optimal match, for example, use the average distance value of the optimal match to measure the similarity, the smaller the distance, the higher the similarity. For each mobile terminal, use the weighted average algorithm to build a priority scoring model, calculate the comprehensive priority score, for example, set the signal strength weight to 5, the geographic location weight to 3, and the trajectory similarity weight to 2, and weight the three indicators to get the comprehensive score. The mobile terminal with the highest comprehensive score in each group is determined as the candidate master mobile terminal. If there are multiple candidate mobile terminals with the same score, compare their signal strengths and determine the mobile terminal with the highest signal strength as the master mobile terminal of the group. Finally, output the master mobile terminal information of each group, including mobile terminal ID, RSSI, longitude and latitude, and group ID, etc., for subsequent communication control and resource scheduling.

[0015] S103. When a mobile terminal group needs to initiate a paging request, the master mobile terminal of the group first initiates paging on behalf of the group. If the paging fails, other mobile terminals in the group are sorted according to their priority scores, and then try to initiate paging in turn until the paging is successful or all mobile terminals are traversed.

[0016] A1. Obtain pre-configured mobile terminal group information and determine the master mobile terminal and standby mobile terminal in each group; A2. When a mobile terminal group needs to initiate a paging request, determine whether the master mobile terminal in the group is available; if the master mobile terminal is available, the master mobile terminal initiates a paging request on behalf of the group; if the master mobile terminal is not available, obtain preset priority information of the standby mobile terminals in the group; A3. Sort the standby mobile terminals in descending order according to the priority to obtain a sorting result; A4. According to the sorting result, sort the standby mobile terminals in descending order according to the priority higher than the preset threshold. The standby mobile terminal with the target value attempts to initiate a paging request and waits for a response; if a response is received from the target mobile terminal within the preset time threshold, the paging request is judged to be successful and the paging process is terminated; if a response is not received from the target mobile terminal within the preset time threshold, the paging request is judged to have failed; A5. If there are still standby mobile terminals that have not been attempted to be paged, select the next standby mobile terminal with a priority higher than the preset threshold and repeat steps A1-A4; A6. If paging is still unsuccessful after traversing all standby mobile terminals in the group, the paging request of the mobile terminal group is judged to have failed and the paging process is terminated.

[0017] Specifically, first read the mobile terminal group information from the pre-configured database, for example, there are 10 groups, each group contains 5-10 mobile terminals. For each group, use the above-mentioned priority scoring model to calculate the comprehensive priority score of each mobile terminal, determine the mobile terminal with the highest score as the master mobile terminal, determine the mobile terminal with the second highest score as the first standby mobile terminal, and so on. When group 1 needs to initiate paging to the target mobile terminal, first determine the status flag of the master mobile terminal A. If it is 1, it means it is available, and A directly initiates the paging request. If the status flag of A is 0, read the priority array of the standby mobile terminal, for example, {B:8, C:7, D:6, E:5}, and set the priority threshold to 6. Arrange the standby mobile terminals in descending order according to the priority to obtain the sequence {B, C, D}. First, B sends a paging request to the target mobile terminal and starts a 2-second timer at the same time. If a response from the target mobile terminal is received within 2 seconds, B is marked as the new master mobile terminal, the priority array is updated, and the paging process ends. If no response is received after 2 seconds, the next standby mobile terminal C is traversed and the above process is repeated. If no response is received after traversing {B, C, D}, the entire group 1 is marked as a paging failure and reported to the network management system, triggering the exception handling mechanism. Finally, the paging result is written to the database, the status information of the relevant mobile terminal is updated, and a paging task is completed, waiting for the next paging request to arrive.

[0018] S104. When determining the paging order of each mobile terminal group, a graph-based scheduling algorithm is used. Each mobile terminal group is abstracted as a node of the graph. The potential interference relationship between the mobile terminal groups is abstracted as directed edges between the nodes. The graph is divided by a minimum cut algorithm to obtain a number of sub-graphs that do not interfere with each other. The mobile terminal groups within the sub-graphs are paged in parallel, and the sub-graphs are paged serially.

[0019] According to the mobile terminal grouping information, an initial interference graph is constructed. Each node in the initial interference graph represents a mobile terminal group, and the directed edges between nodes represent the potential interference relationship between groups. The maximum flow minimum cut algorithm is used to sub-divide the initial interference graph into several sub-graphs that do not interfere with each other. The minimum cut indicates that the interference between sub-graphs is minimal, ensuring that the parallel paging within the sub-graph will not interfere with each other. For each sub-graph obtained by division, all mobile terminal groups within it are added to the same parallel paging queue. The groups in the parallel paging queue can perform paging operations at the same time without interference. The topological order of the sub-graph is obtained through the depth-first search algorithm. The topological order represents the dependencies between the sub-graphs. According to the dependency relationship, the subgraph that completes paging first is arranged in front, and the one that completes later is arranged in the back; according to the topological order, the serial paging scheme between subgraphs is determined to ensure that the paging between subgraphs will not affect each other; in the parallel paging queue of each subgraph, the polling scheduling algorithm is used to perform paging operations on each mobile terminal group in turn. The polling scheduling adopts a cyclic method to allocate paging time slices to each group in turn until the queue is empty; in the serial paging process, if all groups of the current subgraph have completed paging, then according to the topological order relationship, jump to the next subgraph and start its parallel paging queue; through a recursive method, parallel paging and serial paging are continuously executed until the paging queues of all subgraphs are empty, completing the entire paging scheduling process.

[0020] Specifically, first, an initial interference graph is constructed according to the group information of the mobile terminal. For example, there are 10 groups, and each group is regarded as a node in the graph. According to the interference relationship between the groups, directed edges are added between the nodes, and the weight of the edge represents the interference intensity. Then, the interference graph is divided by the maximum flow minimum cut algorithm. For example, the Edmonds-Karp algorithm is used to repeatedly find augmenting paths on the residual graph and calculate the maximum flow between nodes until no augmenting path is found. Finally, the graph is divided into several non-interfering subgraphs according to the minimum cut, and the subgraphs can be paged in parallel. Next, for each subgraph, all the groups in it are added to a parallel paging queue. Then, the topological order of the subgraph is obtained through depth-first search. The nodes with in-degree 0 are visited first, and then the adjacent nodes are visited recursively until all nodes are visited. The topological order represents the dependency relationship between the subgraphs, and the subgraph that completes the paging first is ranked first. The serial paging scheme between the subgraphs is determined according to the topological order to ensure that the subgraph that is paged first will not affect the subgraph that is paged later. In the parallel paging queue of each subgraph, a polling scheduling algorithm is used, such as time slice rotation, to allocate a 10-millisecond time slice to each group, and paging is performed in sequence until the queue is empty. In serial paging, if the current subgraph has completed paging, it jumps to the next subgraph according to the topological order and starts its parallel paging queue. The entire paging scheduling process is completed by continuously executing parallel and serial paging recursively until the queues of all subgraphs are empty. This solution cleverly uses graph algorithms to realize the automatic decomposition and scheduling of paging tasks, maximizes paging efficiency, and avoids interference. It is an efficient and feasible technical solution.

[0021] S105. Track the location changes and dynamic changes of signal strength of mobile terminals in real time. When it is detected that the location change of a mobile terminal exceeds a preset location change threshold or the signal strength change exceeds a preset signal strength change threshold, trigger an incremental update and only adjust the affected groups to avoid global regrouping.

[0022] The real-time location information and signal strength information of the mobile terminal are obtained; the real-time location information of the mobile terminal is compared with a preset location threshold, and the signal strength information is compared with a preset signal strength threshold; if the change of the real-time location information of the mobile terminal exceeds the preset location threshold or the change of the signal strength information exceeds the preset signal strength threshold, an incremental update is triggered; according to the current location and signal strength of the mobile terminal, the affected groups are determined using a pre-established location-signal strength-group mapping table; the K-means clustering algorithm is used to re-cluster the affected groups based on the location and signal strength of the mobile terminal to obtain new grouping results; the new grouping results are compared with the original global grouping results one by one, and if the number of mobile terminals or the location center of the group changes significantly, it is determined to be a group that needs to be adjusted; according to the group that needs to be adjusted, the mapping relationship between the mobile terminal and the base station is updated, and each group is mapped to the base station with the best signal strength to obtain an updated global grouping scheme; the updated global grouping scheme is packaged as an incremental update instruction, and sent to the corresponding mobile terminal and base station through the mobile network to notify them to update the local group cache to complete the incremental update.

[0023] Specifically, the mobile terminal collects GPS location and signal strength RSSI every 5 seconds and reports them to the location management server. The server memory maintains a trigger with a location change threshold of 10 meters and a signal strength change threshold of 3dB. Once the location of the mobile terminal changes by more than 10 meters or the signal strength changes by more than 3dB, the trigger will generate an incremental update event. The event carries the new location and new signal strength of the mobile terminal. The hash algorithm is used to quickly find the location-signal strength-group mapping table and locate the affected group. Then, the K-means clustering algorithm is used, and the cluster center K=5 is set. The longitude and latitude of the mobile terminal location and the signal strength RSSI are used as feature vectors to re-cluster the affected groups and obtain several new clusters. The new clusters are compared with the original groups one by one, and the change rate of the number of mobile terminals in the cluster and the drift distance of the cluster center are calculated. If the number change rate exceeds 20% or the center drift distance exceeds 15 meters, it is determined that the original group corresponding to the cluster needs to be adjusted. When adjusting, the drifting mobile terminal is first removed, and then the newly added mobile terminal is inserted into the cluster. For the groups that need to be adjusted, update their mapping relationship with the base station, traverse each mobile terminal, select the base station with the largest signal strength RSSI as the service base station, and establish a new group-base station mapping table. Finally, package the updated grouping scheme into incremental update instructions, compress it with Protobuf, and distribute it to the corresponding mobile terminals and base stations through message queues to notify them to update the local group cache. The incremental update instructions use JSON format and contain fields such as group ID, mobile terminal list, and base station ID. Through this incremental update mechanism, it is possible to quickly respond to changes in mobile terminal location and signal strength, dynamically optimize the grouping scheme, and thus improve paging efficiency and success rate.

[0024] S106. During the paging process, a reinforcement learning framework is used to continuously optimize the priority scoring model and grouping strategy of each mobile terminal. The success rate of each paging is used as a feedback reward. The model parameters are updated through a policy gradient algorithm, and the weight parameters in the priority scoring model are adjusted, so that the system can adaptively adjust the paging strategy based on historical experience.

[0025] Relevant data in the paging process are obtained, including the location information, moving speed and power attributes of the mobile terminal, and the historical paging success rate; the relevant data are cleaned and normalized preprocessed, and key features are extracted as the input of the model; the mobile terminal priority scoring model is trained to predict its priority score according to the mobile terminal attributes; the mobile terminals are grouped according to the priority scores of the mobile terminals using the K-means clustering algorithm, and mobile terminals with similar priority scores are divided into the same group; in each paging process, a specific mobile terminal group is selected for paging according to a preset strategy, and the preset strategy includes randomly selecting a mobile terminal group with a preset probability and selecting a mobile terminal group with the highest priority score with a preset probability; the success rate of each paging is recorded as a reward feedback for reinforcement learning, and the policy gradient algorithm is used to update the parameters of the mobile terminal priority scoring model; through continuous data collection and reinforcement learning, the mobile terminal priority scoring model and paging strategy are continuously optimized to adapt to changes in actual paging scenarios and improve the overall paging success rate.

[0026] Specifically, during the paging process, the data is cleaned and normalized through attribute data such as GPS location, moving speed, and power reported by the mobile terminal, as well as related data such as historical paging success rate. When cleaning, outliers with GPS location beyond the reasonable range (such as latitude and longitude beyond -90~90 degrees), moving speed less than 0 or greater than 100m / s, and power less than 0 or greater than 100% are eliminated. When normalizing, the maximum and minimum values ​​are normalized for continuous variables, and unique hot encoding is used for discrete variables. Four key features of mobile terminal location, moving speed, power, and historical paging success rate are extracted as model inputs. Based on priority scores, the K-means clustering algorithm is used to divide mobile terminals into 5 groups, and the Euclidean distance is used to measure similarity during clustering. When paging, a group is randomly selected with a probability of 20%, and the group with the highest average priority score is selected with a probability of 80%. Assuming that the second group is selected for a certain paging, which contains 100 mobile terminals, and 90 of them are successfully awakened, the paging success rate is 90%. The success rate is used as reward feedback and substituted into the policy gradient algorithm to update the model parameters and group selection probability. The policy gradient uses the Monte Carlo method to estimate the gradient and is weighted by the reward value with a learning rate of 0.01. Through continuous iterative optimization, the model can accurately predict the priority based on the attributes of the mobile terminal, and the group selection strategy can adapt to different scenarios, thereby improving the overall paging success rate.

[0027] The description of the above embodiments is only used to help understand the technical solutions and core ideas of the present application. Ordinary technical personnel in this field should understand that they can still modify the technical solutions recorded in the aforementioned embodiments, or replace some of the technical features therein with equivalents. However, these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for optimizing account splitting and settlement for a mobile terminal, characterized in that: The method comprises: Acquire basic information of multiple mobile terminals, the basic information including geographic location information, signal strength data, and movement trajectory records; use the basic information as input parameters, and use a clustering algorithm to group the multiple mobile terminals to obtain a plurality of mobile terminal groups; For each mobile terminal group, a multi-layer perceptron is used to construct a priority scoring model, and the priority scoring model is used to determine the master mobile terminal in each mobile terminal group; When a mobile terminal group needs to initiate a paging request, the master mobile terminal of the group first initiates paging on behalf of the group. If the paging fails, other mobile terminals in the group are ranked according to their priority scores, and then try to initiate paging in turn until the paging is successful or all mobile terminals are exhausted. When determining the paging order of each mobile terminal group, a scheduling algorithm based on a graph is adopted, each mobile terminal group is abstracted as a node of the graph, and the potential interference relationship between the mobile terminal groups is abstracted as directed edges between the nodes. The graph is divided by a minimum cut algorithm to obtain a number of subgraphs that do not interfere with each other. The mobile terminal groups in the subgraphs are paged in parallel, and the subgraphs are paged serially. Track the location changes and dynamic changes of signal strength of mobile terminals in real time. When it is detected that the location change of a mobile terminal exceeds the preset location change threshold or the signal strength change exceeds the preset signal strength change threshold, trigger incremental update and only adjust the affected mobile terminal grouping to avoid global regrouping. During the paging process, a reinforcement learning framework is used to continuously optimize the priority scoring model and grouping strategy of each mobile terminal. The success rate of each paging is used as a feedback reward. The model parameters are updated through the policy gradient algorithm, and the weight parameters in the priority scoring model are adjusted, so that the system can adaptively adjust the paging strategy based on historical experience.

2. The method according to claim 1, characterized in that Acquire basic information of multiple mobile terminals, the basic information including geographic location information, signal strength data, and movement trajectory records; use the basic information as input parameters, and use a clustering algorithm to group the multiple mobile terminals to obtain a number of mobile terminal groups, including: Obtaining geographic location information, signal strength data, and movement trajectory records of at least two mobile terminals; Preprocess the geographic location information and use the Kalman filter algorithm to remove outliers and noise data to obtain cleaned location data; Normalize the signal strength data, and use the maximum-minimum normalization method to uniformly map the signal strength data of different mobile terminals to the interval [0, 1] to obtain normalized strength data; The mobile trajectory records are compressed, and the Douglas-Peucker algorithm is used to extract key position points including stop points and turning points to obtain compressed trajectory data; The cleaned position data, the normalized intensity data, and the compressed trajectory data are constructed into a feature vector; The feature vector is input into the K-means clustering algorithm, and the similarity between the mobile terminals is measured by calculating the Euclidean distance between the mobile terminals to obtain the grouping result of the mobile terminals; For each mobile terminal group, density clustering is performed on the location data in the mobile terminal group to obtain a typical location distribution of the mobile terminal group; For each mobile terminal group, statistically analyzing the signal strength data in the mobile terminal group to obtain the signal strength distribution of the mobile terminal group; For each mobile terminal group, frequent sequence mining is performed on the trajectory data in the mobile terminal group to obtain the typical movement trajectory pattern of the mobile terminal group; The location distribution, signal strength distribution and movement trajectory pattern characteristic attributes of each mobile terminal group are integrated to construct a mobile terminal group portrait.

3. The method according to claim 1, characterized in that For each mobile terminal group, a multi-layer perceptron is used to construct a priority scoring model; the priority scoring model is used to determine the master mobile terminal in each mobile terminal group, including: Obtain signal strength, geographic location and movement trajectory information of multiple mobile terminals; For each moving trajectory of the mobile terminal, two trajectories are sampled to obtain a sampling point sequence of equal length; Calculate the Euclidean distance matrix between two sampling point sequences; Find the path with the minimum cost in the distance matrix and take the path with the minimum cost as the optimal match between the two trajectories; Calculate the similarity of two trajectories based on the best match; For each mobile terminal, a priority scoring model is used to calculate the comprehensive priority score of each mobile terminal based on signal strength, geographic location and similarity of movement trajectory. The weights in the priority scoring model are preset according to actual needs. Determine the mobile terminal with the highest comprehensive priority score in each mobile terminal group as the candidate master control mobile terminal of the group; If there are multiple candidate master mobile terminals with the same comprehensive priority scores, the signal strengths of the multiple candidate master mobile terminals are compared, and the candidate master mobile terminal with the highest signal strength is determined as the master mobile terminal of the group; The master mobile terminal information of each mobile terminal group is output, and the master mobile terminal information includes the mobile terminal identification, signal strength, geographical location coordinates and the mobile terminal group to which it belongs.

4. The method according to claim 1, characterized in that: When a mobile terminal group needs to initiate a paging request, the master mobile terminal of the group first initiates paging on behalf of the group. If the paging fails, other mobile terminals in the group are ranked by priority scores and try to initiate paging in turn until the paging is successful or all mobile terminals are exhausted, including: A1. Obtain pre-configured mobile terminal grouping information and determine the master mobile terminal and the standby mobile terminal in each group; A2. When a mobile terminal group needs to initiate a paging request, determine whether the master mobile terminal in the group is available; If the master mobile terminal is available, the master mobile terminal initiates a paging request on behalf of the group; If the master mobile terminal is unavailable, obtaining preset priority information of the backup mobile terminals in the group; A3. Sort the standby mobile terminals in descending order according to their priorities to obtain a sorting result; A4. According to the sorting results, the standby mobile terminals with priorities higher than the preset threshold try to initiate paging requests and wait for responses. If a response is received from the target mobile terminal within the preset time threshold, the paging request is judged to be successful and the paging process ends; If no response is received from the target mobile terminal within the preset time threshold, the paging request is judged to have failed; A5. If there are any standby mobile terminals that have not attempted paging, select the next standby mobile terminal whose priority is higher than the preset threshold, and repeat steps A1-A4; A6. If paging is still unsuccessful after traversing all standby mobile terminals in the group, it is determined that the paging request of the mobile terminal group has failed, and the paging process ends.

5. The method according to claim 1, characterized in that: When determining the paging order of each mobile terminal group, a scheduling algorithm based on a graph is adopted, each mobile terminal group is abstracted as a node of the graph, and the potential interference relationship between the mobile terminal groups is abstracted as directed edges between the nodes. The graph is divided by a minimum cut algorithm to obtain a number of subgraphs that do not interfere with each other. The mobile terminal groups in the subgraphs are paged in parallel, and the subgraphs are paged serially, including: According to the mobile terminal grouping information, an initial interference graph is constructed, wherein each node in the initial interference graph represents a mobile terminal group, and the directed edges between the nodes represent potential interference relations between the groups; The maximum flow minimum cut algorithm is used to divide the initial interference graph into several subgraphs that do not interfere with each other. The minimum cut indicates that the interference between subgraphs is minimal, ensuring that the parallel paging within the subgraph will not interfere with each other. For each subgraph obtained by division, all mobile terminal groups in it are added to the same parallel paging queue. The groups in the parallel paging queue can perform paging operations at the same time without causing interference; The topological order of the subgraphs is obtained through the depth-first search algorithm. The topological order represents the dependency relationship between the subgraphs. The subgraphs that complete the paging first are ranked first, and the subgraphs that complete the paging later are ranked last. According to the topological order, determine the serial paging scheme between subgraphs to ensure that the paging between subgraphs will not affect each other; In the parallel paging queue of each subgraph, a polling scheduling algorithm is used to perform paging operations on each mobile terminal group in turn. The polling scheduling adopts a cyclic manner to allocate paging time slices to each group in turn until the queue is empty; In the serial paging process, if all the groups in the current subgraph have completed paging, then according to the topological order relationship, jump to the next subgraph and start its parallel paging queue; Parallel paging and serial paging are continuously executed in a recursive manner until the paging queues of all subgraphs are empty, completing the entire paging scheduling process.

6. The method according to claim 1, characterized in that Track the location changes and dynamic changes of signal strength of mobile terminals in real time. When it is detected that the location change of a mobile terminal exceeds the preset location change threshold or the signal strength change exceeds the preset signal strength change threshold, trigger incremental update and adjust only the affected groups to avoid global regrouping, including: Obtain real-time location information and signal strength information of mobile terminals; Compare the real-time location information of the mobile terminal with a preset location threshold, and compare the signal strength information with a preset signal strength threshold; If the change of the real-time location information of the mobile terminal exceeds a preset location threshold or the change of the signal strength information exceeds a preset signal strength threshold, an incremental update is triggered; Determine the affected group based on the current location and signal strength of the mobile terminal using a pre-established location-signal strength-group mapping table; Using a K-means clustering algorithm, taking the location and signal strength of the mobile terminal as features, re-clustering the affected groups to obtain new grouping results; Compare the new grouping results with the original global grouping results one by one. If the number of mobile terminals or the location center of the group changes significantly, it is determined that the group needs to be adjusted; According to the grouping that needs to be adjusted, the mapping relationship between the mobile terminal and the base station is updated, and each group is mapped to the base station with the best signal strength to obtain an updated global grouping solution; The updated global grouping scheme is packaged into an incremental update instruction and sent to the corresponding mobile terminal and base station through the mobile network to notify them to update the local grouping cache to complete the incremental update.

7. The method according to claim 3, characterized in that During the paging process, the reinforcement learning framework is used to continuously optimize the priority scoring model and grouping strategy of each mobile terminal. The success rate of each paging is used as a feedback reward. The model parameters are updated through the policy gradient algorithm, and the weight parameters in the priority scoring model are adjusted, so that the system can adaptively adjust the paging strategy based on historical experience, including: Obtain relevant data during the paging process, including the location information, moving speed and power attributes of the mobile terminal, and historical paging success rate; Clean and normalize the relevant data and extract key features as the input of the model; Through training, the mobile terminal priority evaluation model can predict the priority score of the mobile terminal according to its attributes; According to the priority scores of mobile terminals, the K-means clustering algorithm is used to group the mobile terminals, and mobile terminals with similar priority scores are divided into the same group; In each paging process, a specific mobile terminal group is selected for paging according to a preset strategy, wherein the preset strategy includes randomly selecting a mobile terminal group with a preset probability and selecting a mobile terminal group with the highest priority score with a preset probability; The success rate of each paging is recorded as the reward feedback for reinforcement learning, and the policy gradient algorithm is used to update the parameters of the mobile terminal priority scoring model; Through continuous data collection and reinforcement learning, the mobile terminal priority scoring model and paging strategy are continuously optimized to adapt to changes in actual paging scenarios and improve the overall paging success rate.

Images