Behavior analysis method and device based on multi-dimensional combined position behavior trajectory chain

Abstract

The application provides a behavior analysis method and device based on a multi-dimensional combined position behavior trajectory chain, which comprises the following steps: performing MR positioning on a user by using a positioning mode combining accurate positioning and pattern matching; identifying indoor and outdoor, building layer identification and motion state identification of the user based on the MR positioning result; training a position positioning model by using an integrated learning technology and predicting the position of the user; constructing a spatial large database based on the position of the user, and predicting and analyzing the behavior of the user based on the spatial large database. The behavior analysis method and device based on the multi-dimensional combined position behavior trajectory chain can construct a trinity position positioning model based on the position behavior of the user by using the multi-dimensional combined positioning and integrated learning mode, and can construct a spatial large database based on the position of the user to predict and analyze the behavior of the user, thereby ensuring the discrimination of the authenticity and rationality of the behavior of the user.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, in particular to a behavior analysis method and device based on multi-dimensional combined position behavior trajectory chain. BACKGROUND

[0002] At present, the existing positioning methods mainly include APP (application program) positioning method, TA (timing advance) combined with AOA (angle of arrival) positioning method, triangular positioning method and fingerprint library-based positioning method. Among them, the APP positioning method mainly extracts the latitude and longitude reported by the user by analyzing the S1-U interface signaling to obtain the user's position information. The TA combined with AOA positioning method mainly estimates the distance between the base station and the UE according to the TA value in the MR (measurement report), and then obtains the position information of the terminal according to the angle information of AOA. The triangular positioning method mainly combines MR field strength information and network element parameter information, uses a triangle or polygon formed by the main service cell and two or more strongest adjacent areas to calculate the center point, and performs field strength weighted offset to obtain the positioning result.

[0003] The existing technical solutions have the technical problems of low positioning accuracy, large positioning error, and inaccurate analysis results of user behavior. SUMMARY

[0004] The present application provides a behavior analysis method and device based on multi-dimensional combined position behavior trajectory chain to solve the technical problem of inaccurate analysis results of user behavior.

[0005] In a first aspect, the present application provides a behavior analysis method based on multi-dimensional combined position behavior trajectory chain, comprising:

[0006] Precise positioning and pattern matching combined positioning method is used for user network measurement report (MR) positioning;

[0007] Based on the MR positioning result, the user is identified indoors and outdoors, the building is identified in layers, and the motion state is identified;

[0008] An integrated learning technology is used to train the position positioning model and predict the user's position;

[0009] A spatial database based on user position is constructed, and the user behavior is predicted and analyzed based on the spatial database.

[0010] In one embodiment, the integrated learning technology is used to train the position positioning model and predict the user's position, comprising:

[0011] Extracting user static position features, user motion trajectory features, network environment features, user scene features and user geographical environment features;

[0012] construct a location positioning model based on the user static location feature, the user motion trajectory feature, the network environment feature, the user scene feature and the user geographic environment feature, a preset weight corresponding to the user static location feature, a preset weight corresponding to the user motion trajectory feature, a preset weight corresponding to the network environment feature, a preset weight corresponding to the user scene feature and a preset weight corresponding to the user geographic environment feature, and train the location positioning model and predict the user location.

[0013] In one embodiment, the smaller the prediction range, the greater the preset weight corresponding to the network environment feature.

[0014] In one embodiment, the greater the location relevance, the greater the preset weight corresponding to the user static location feature.

[0015] In one embodiment, the weights corresponding to the user static location feature, the user motion trajectory feature, the network environment feature, the user scene feature and the user geographic environment feature decrease in turn.

[0016] In one embodiment, the weights corresponding to the user static location feature, the user motion trajectory feature, the network environment feature, the user scene feature and the user geographic environment feature are 40%, 35%, 12.5%, 10% and 2.5% in turn.

[0017] In a second aspect, the present application provides a behavior analysis device based on a multi-dimensional combined location behavior trajectory chain, comprising:

[0018] A positioning module is configured to perform MR positioning on a user by using a positioning method combining accurate positioning and pattern matching.

[0019] An identification module is configured to identify the indoor and outdoor of the user, the floor layer of the building and the motion state of the user based on the MR positioning result.

[0020] A prediction module is configured to train a location positioning model by using an integrated learning technology and predict the location of the user.

[0021] An analysis module is configured to construct a spatial large database based on the location of the user, and predict and analyze the behavior of the user based on the spatial large database.

[0022] In a third aspect, the present application provides an electronic device comprising a processor and a memory storing a computer program, wherein the processor implements the behavior analysis method based on a multi-dimensional combined location behavior trajectory chain according to the first aspect when executing the computer program.

[0023] In a fourth aspect, the present application provides a processor-readable storage medium, which stores a computer program for causing a processor to execute the behavior analysis method based on the multi-dimensional combined location behavior trajectory chain according to the first aspect.

[0024] In a fifth aspect, the present application further provides a computer program product, which comprises a computer program for implementing the behavior analysis method based on the multi-dimensional combined location behavior trajectory chain according to the first aspect when executed by a processor.

[0025] The behavior analysis method and device based on the multi-dimensional combined location behavior trajectory chain provided by the present application construct a three-in-one location positioning model based on user location behavior through multi-dimensional combined positioning and integrated learning, and construct a spatial database based on user location to predict and analyze user behavior, thereby ensuring the discrimination of the authenticity and rationality of user behavior. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0027] Figure 1 is a flowchart of the behavior analysis method based on the multi-dimensional combined location behavior trajectory chain provided by the present application;

[0028] Figure 2 is a flowchart of the indoor and outdoor MR distinguishing algorithm provided by the present application;

[0029] Figure 3 is a structural diagram of the behavior analysis device based on the multi-dimensional combined location behavior trajectory chain provided by the present application;

[0030] Figure 4 is a structural diagram of the electronic device provided by the present application. DETAILED DESCRIPTION

[0031] At present, the existing positioning methods mainly include APP (application) positioning method, TA (timing advance) combined with AOA (angle of arrival) positioning method, triangular positioning method and fingerprint library-based positioning method. Among them, the APP positioning method mainly extracts the latitude and longitude reported by the user by analyzing the S1-U interface signaling to obtain the user location information. The TA combined with AOA positioning method mainly estimates the distance between the base station and the terminal / user equipment (UE) according to the TA value in the MR (measurement report), and then obtains the position information of the terminal according to the angle information of AOA. The triangular positioning method mainly combines the MR field strength information and the network element parameter information, uses the triangle or polygon formed by the main service cell and two or more strongest adjacent areas to calculate the center point, and performs field strength weighted offset to obtain the positioning result.

[0032] The APP positioning method needs to be calibrated in the outdoor environment to accurately position, and the latitude and longitude of most APPs are encrypted and cannot be directly analyzed, so the number of available sample points is limited.

[0033] The positioning accuracy of the TA combined with AOA positioning method is obviously affected by the environment, and the positioning is more accurate in open areas, but the positioning accuracy is poor in areas with many high buildings.

[0034] In the triangular positioning method, the proportion of the phenomenon that the adjacent area information of the MR data in the existing network is incomplete is large, so the three-point positioning method has poor implementability, low accuracy, large positioning error, and is obviously affected by the station spacing.

[0035] In summary, the above-mentioned methods have low accuracy or cannot be positioned in some scenarios (such as the identification of moving state on high-speed rail), and the above-mentioned positioning methods are relatively single and not well combined with other technologies.

[0036] The present application is mainly used to solve the technical problems of low accuracy, large positioning error, single positioning method, inability to combine with other technologies to analyze the behavior of the user, and inaccurate analysis result of the behavior of the user of the traditional positioning method, therefore, the present application constructs a trinity position positioning model based on the user location behavior by means of multi-dimensional combined positioning and integrated learning, and constructs a spatial database based on the user location to predict and analyze the behavior of the user, so as to realize the discrimination of the authenticity and rationality of the behavior of the user.

[0037] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme in the present application will be described clearly and completely in combination with the drawings in the present application. Obviously, the described embodiments are part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor belong to the scope of protection of the present application.

[0038] Figure 1 is a flowchart of the behavior analysis method based on multi-dimensional combined location behavior trajectory chain provided by the present application, as shown in Figure 1 The present application provides a behavior analysis method based on multi-dimensional combined location behavior trajectory chain, comprising:

[0039] Step 101, using a positioning method combining precise positioning and pattern matching to perform MR positioning for users.

[0040] Among them, the positioning method combining precise positioning and pattern matching performs precise positioning for identifiable users, and performs KNN algorithm matching through pattern matching positioning for unidentifiable users, part of the users are precisely matched, part of the users are optimally matched, and finally the MR positioning of the whole network is realized.

[0041] The precise positioning method in this embodiment includes: base station triangulation based on propagation model, location positioning based on TA combined with AOA, fingerprint database positioning of propagation model, location positioning based on OTT data, triangulation positioning based on WiFi network, location positioning associated with home broadband information, and fitting positioning based on road test data.

[0042] 1. Base station triangulation based on propagation model

[0043] Implementation principle: the radio signal strength decreases with the increase of the propagation distance, and the radio propagation distance and the signal strength have strong correlation. On the basis of a large number of practices, it can be concluded that the received signal strength conforms to the log-normal (log-normal) distribution model, and the propagation distance can be calculated through the attenuation of the signal in the propagation process.

[0044] Calculation method: the measured point receives cell signals from three different directions, and then converts them into the distance from the measured target to the base station according to the transmission loss model of the wireless signal, and then takes the three cells as the centers and the distances from them to the measured point as the radii to make circles, and the intersection of the three circles is the user position.

[0045] 2. Location positioning based on TA combined with AOA

[0046] Implementation principle: in the TDD_LTE MR record, the Tadv value and the AOA value are reported on each MR point, and the current MR position can be determined through the Tadv value and the AOA.

[0047] Calculation method: step one, the distance information from the current MR reporting point to the base station can be obtained through the Tadv field.

[0048] The calculation method is: 1 Ts time advance distance = 300000000 / (15000*2048)) / 2 = 4.89 meters; when D = 0: S = 0 ~ 4.89*(D+0.5) random value; when D > 0: S = 4.89*(D-0.5) ~ 4.89*(D+0.5) random value. Wherein, S: represents the distance from the antenna to the MR point, unit: meter, D: for direct Tadv report value. L = Sqrt(Power(S, 2) - Power(H, 2)), wherein L: represents the distance from the MR point to the cell, unit: meter, H is the cell antenna height, Sqrt() is the square root function.

[0049] Step two, determine the MR point longitude and latitude, because S takes a random value, the spherical distance is directly ignored as a plane distance here.

[0050] X(2) = X(1) + LsinAOA

[0051] Y(2) = Y(1) + LcosAOA

[0052] Wherein, X(2) is the longitude of the user reported MR point, Y(2) is the latitude of the user reported MR point, X(1) is the cell longitude, Y(1) is the cell latitude, and AOA is the antenna angle of arrival.

[0053] 3, fingerprint database positioning of propagation model

[0054] The implementation principle is: based on high-precision map, using 3D ray propagation model to calculate the signal strength in each grid in the three-dimensional space of the area to be planned, and finally digitizing these grid information containing the field strength of each cell to form feature vector values. The sample data set of these feature vector values constitutes a fingerprint database, and the MR data is accurately positioned by using the fingerprint database algorithm.

[0055] The calculation method is: the level strength of the main service cell A measured by a certain MR point is a, and the level strengths of the adjacent cells B and C are b and c respectively, forming an MR cell field strength vector (a, b, c), and a feature vector formed by a certain fingerprint position is (Aa, Bb, Cc) (wherein Aa indicates that the level strength of the A cell at this position is a, then the Euclidean distance d of the two vectors satisfies the shortest match, that is, d = (a-A a ) 2 +(b-B b ) 2 +(c-C c ) 2 ] 1 / 2 ).

[0056] 4, location positioning based on OTT data

[0057] Implementation principle: Many APP applications will obtain the user's location information, and the user will use HTTP protocol to interact with the server during the location positioning process, and the user's accurate location will be intercepted through core network signaling packet capture.

[0058] 5. WiFi network-based triangulation positioning

[0059] Implementation principle: When the device is in the WiFi network, the collected data that can identify the AP (access point) can be sent to the location server, the server retrieves the geographic location of each AP, and combines the strength of each signal to calculate the geographic location of the device and return it to the user device. Using the three-point positioning technology, the user's location is realized.

[0060] Calculation method: The test point receives AP signals from three different directions, and then converts them into the distance from the test target to the AP according to the transmission loss model of wireless signals. Then take the three APs as the center and their respective distances to the test point as the radius to make a circle. The intersection of the three circles is the user's location.

[0061] 6. Associated with home broadband information for location positioning

[0062] Implementation principle: Using home broadband information to locate the home broadband users in the building, more accurately positioning the user's floor, which is conducive to the accurate positioning of wireless network problems.

[0063] 7. Fitting positioning based on road test data

[0064] Implementation principle: Based on the collected road test and sweep frequency data, analyze the trend of the field strength information of each cell, combine the direction of the test line to form a characteristic curve, and fit the MR sampling point through these characteristic curves to complete the positioning of the MR sampling data.

[0065] Calculation method: Extract the level strength of the associated main service cell and the level strength of the surrounding neighbor cells from the MR sampling point data as the identification characteristics of the MR sampling point, and then find the characteristic curve of the cell in the model library. Through fitting, the specific location of the sampling point is located.

[0066] Step 102, based on the MR positioning result, respectively identify the indoor and outdoor users, building layer identification and motion state identification.

[0067] 1. Indoor and outdoor user identification

[0068] The method for identifying the outdoor attributes of the user is based on the MR data in the critical area after preliminary positioning, and through the user's cell residence, position change, moving speed, main service cell change rate, RSRP level change, service attribute and other multiple dimensions, the indoor and outdoor users are distinguished.Figure 2 is the flowchart of the indoor and outdoor MR distinguishing algorithm provided by the present application, as shown in Figure 2 The reported MRO (original measurement report) is counted in units of MME UE S1AP ID (access point identification code), and the specific algorithm for MR data of the same MME UE S1AP ID is as follows:

[0069] a. According to the switching characteristics of the main service cell (according to the change of the cell where the user is located at several time points before and after), the indoor and outdoor users are distinguished.

[0070] b. According to the total cell data characteristics (according to the difference in the number of adjacent areas received indoors and outdoors), the indoor and outdoor users are distinguished.

[0071] c. According to the indoor and outdoor level intensity characteristics (according to the difference in the level intensity of the service cell and the adjacent area received indoors and outdoors), the indoor and outdoor cells are distinguished.

[0072] According to the analysis of the indoor and outdoor coverage environment characteristics and the comparison and analysis of the MR data characteristics, the indoor and outdoor users are identified from the aspects of "indoor and outdoor main service cell switching with obvious mobility characteristics", "the number of cells contained in MR", "the strength and difference of each pilot signal", etc.

[0073] 2. Building layer identification

[0074] For the users of the building, on the basis of completing the user feature classification, the corresponding positioning technology is adapted according to the data required by the positioning technology and the applicable scene, the full data positioning is completed, and the building layer user identification is performed.

[0075] According to the positioning of the 3D high-precision map building, the building layer index presentation is realized, and the layering algorithm includes:

[0076] a. The performance of RRU (radio remote unit) and the data of signaling and the test data of each layer are used to realize the building layer presentation.

[0077] b. The indoor building set planning drawing and simulation data are used to realize the layer identification.

[0078] The positioning algorithm includes Bluetooth indoor positioning, WiFi positioning, ZigBee indoor positioning, ultra-wideband positioning, etc.

[0079] The specific algorithm is as follows:

[0080] Bluetooth indoor positioning technology is to maintain the network as a multi-user based network connection mode by installing several Bluetooth access points in the room, and to ensure that the Bluetooth access point is always the master of the piconet, and then to measure the signal strength to triangulate the new blind node. The biggest advantage of Bluetooth indoor positioning technology is small device size, short distance, low power consumption, and easy integration into mobile devices such as mobile phones. As long as the Bluetooth function of the device is turned on, it can be positioned. Bluetooth transmission is not affected by the line of sight, but for complex space environment, the stability of Bluetooth system is slightly poor, the noise signal interference is large, and the price of Bluetooth device and equipment is relatively expensive. Bluetooth indoor positioning is mainly used for small-scale positioning of people, such as single-story hall or store.

[0081] WiFi positioning technology has two kinds, one is to compare the signal strength of mobile devices and three wireless network access points through differential algorithm, to accurately triangulate people and vehicles. The other is to record a large amount of signal strength of the determined position point in advance, and to determine the position by comparing the signal strength of the new device with the database with a large amount of data ("fingerprint" positioning). WiFi positioning can realize complex large-scale positioning, monitoring and tracking tasks in a wide range of application fields, and the total accuracy is relatively high, but the accuracy of indoor positioning can only reach about 2 meters, which cannot achieve accurate positioning. Due to the popularity of WiFi routers and mobile terminals, the positioning system can share the network with other customers, the hardware cost is very low, and the positioning system of WiFi can reduce the possibility of RF interference. WiFi positioning is suitable for positioning and navigation of people or vehicles, and can be used in medical institutions, theme parks, factories, shopping malls and other occasions requiring positioning and navigation.

[0082] ZigBee indoor positioning technology: ZigBee indoor positioning technology forms a network between a number of blind nodes to be positioned and a reference node with a known position and a gateway, and each small blind node communicates with each other to realize all positioning. ZigBee is a new short-range, low-rate wireless network technology. These sensors only need a little energy to pass data from one node to another in a relay manner through radio waves, and ZigBee is a low-power and low-cost communication system with very high efficiency. But the signal transmission of ZigBee is greatly affected by multipath effect and movement, and the positioning accuracy depends on the channel physical quality, signal source density, environment and algorithm accuracy, resulting in high cost of positioning software and great improvement space. ZigBee indoor positioning has been adopted by many large factories and workshops as a personnel on-duty management system.

[0083] Ultra-wideband positioning technology is a brand new technology which is quite different from traditional communication positioning technology. It uses anchor nodes and bridge nodes with known locations to communicate with new blind nodes, and uses triangulation or "fingerprint" positioning method to determine the location. Ultra-wideband communication does not use the carrier in the traditional communication system, but transmits data by sending and receiving extremely narrow pulses with nanoseconds or less, so it has a bandwidth of GHz. Because of the advantages of strong penetration, good anti-multipath effect, high security, low system complexity, and the ability to provide accurate positioning accuracy, the prospect of ultra-wideband positioning technology is quite broad. However, the newly added blind nodes also need active communication, which makes the power consumption high, and the layout is also needed in advance, so the cost cannot be reduced. Ultra-wideband indoor positioning can be used for indoor precise positioning and navigation in various fields, including people and large objects, such as car garage parking navigation, mine personnel positioning, valuable goods storage, etc.

[0084] 3. Motion state recognition

[0085] Firstly, by associating signaling data with MR and other data, user information, network occupation information, and the wireless environment of the terminal are integrated together, and time dimension and space dimension are analyzed to recognize the characteristics of user's static and dynamic state, behavior trajectory, and the scene.

[0086] Secondly, after associating the sampling points with signaling, engineering parameters and other multi-source data, the clustering analysis is performed by using k-means algorithm and other algorithms to mine the potential correlation of the sampling point data, and to identify the behavior characteristics of the current unknown user, so as to continuously enrich the user behavior characteristic label library in the form of self-learning.

[0087] Finally, according to the characteristics of high-speed movement, the distance and time difference between the two sampling points of the same user are calculated to obtain the moving speed of the user, and combined with the data of the cells along the high-speed (high-speed rail) line, it is judged whether the user is a high-speed (high-speed rail) user.

[0088] The algorithms for realizing motion state recognition include: barrier feature judgment method, motion state recognition, neighbor area feature recognition, Kmeans clustering hierarchical analysis algorithm, least squares method, K-nearest neighbor analysis, etc., as follows:

[0089] Barrier feature judgment method: the collected MR parameters and engineering parameters are substituted into the extended HATA model formula to calculate the theoretical loss value of the user, and compared with the difference between the actual cell transmission power and the UE receiving and transmitting power, to determine whether there is a barrier between the user and the base station. Through the comprehensive calculation of multiple MRs of the user, the proportion of MR points blocked by the user is used to preliminarily judge the indoor and outdoor attributes of the user.

[0090] Motion recognition: through the user occupies each base station sampling point number and the proportion of the situation, the latitude and longitude of the base station to calculate the distance, determine the user's general activity range, preliminary judgment of the user's stability. Statistics of user occupation base station type, the longest site in the user occupation macro station is the main occupation base station of the user;

[0091] Through the base station latitude and longitude, based on the distance between any two points on the earth formula to calculate the distance from each base station to the main occupation base station Max(d)<500m.

[0092] Adjacent area characteristics: through the base station reported adjacent area, MR measurement adjacent area has room for more than a certain proportion of users as indoor users. Indoor user adjacent area is less, and the adjacent area signal is weak.

[0093] Kmeans clustering hierarchical analysis algorithm: through the MRO sampling in the same position (the MRO information of the same position at different heights is different), the clustering analysis of the sampling data is obtained.

[0094] Implementation method:

[0095] (1), the size of the MRO data set after processing is n, let I represent the number of iterations, the initial value of I is equal to 1, and k initial cluster centers Z j (I), j = 1, 2, 3,..., k, Z j (I) represents the jth cluster center in the Ith iteration.

[0096] (2), calculate the distance D(x i ,Z j (I)) of each data object and the cluster center, i = 1, 2,..., n, j = 1, 2, 3...k, x i represents the ith sample point, if it satisfies: D(x i ,Z k (I)) = min{D(x i ,Z j (I)), I = 1, 2, 3,..., n}, then x i belongs to C k , Z k (I) represents the kth cluster center in the Ith iteration, C k represents the kth cluster with Z k (I) as the cluster center.

[0097] (3), calculate the new k cluster centers Z j (I+1), j = 1, 2, 3,..., k, Z j(I+1)j represents the jth cluster center in the (I+1)th iteration: that is, the arithmetic mean of the latitude of each element of the cluster center is taken.

[0098] (4) If Z j (I+1)j = Z j (I), j = 1, 2, 3, …, k, then I = I+1, return (2) otherwise the algorithm ends.

[0099] Least squares method: a fast gridding ray tracing method based on graph theory, network flow theory and analytic geometry, which converts each calculation region into a large-scale connected graph, and uses the shortest path algorithm in the graph to calculate the path ray information combined with the knowledge of analytic geometry.

[0100] Least squares method physical model establishment fingerprint library algorithm:

[0101] Calculation formula:

[0102]

[0103]

[0104] Where L represents the path ray length, a represents the signal strength, b represents the size of the grid, d represents the increased flow, c represents the capacity, n represents the number of paths, c path represents the path size, and f(alpha, beta) represents the flow.

[0105] Introduce the loss function: err = 1 / 2(L-Ax) 2 , when min(err), x = (A T A) -1 A T L.

[0106] Find all outdoor grids adjacent to the building where grid i is located, and calculate the attenuation L ik of each outdoor grid k using the outdoor model.

[0107] L ik = min{L k +η0+η1(d ik -2.5)}

[0108] Where L k represents the path ray length of the kth grid, η0 represents the level strength, η1 represents the signal gain, and d ik represents the difference between the ith grid and the kth grid.

[0109] K-Nearest Neighbor Analysis: KNN is used to match the top k most similar grids. For the longitude and latitude of these k grids, different weights are given to obtain the final longitude and latitude of the test point. The weights are obtained from the level distribution of each cell corresponding to the sample points in the grid.

[0110] Assumption: Each sample point takes n cells, and MR also takes n cells. The level of each sample point tested to the cell obeys Gaussian distribution:

[0111]

[0112] Where μ is the mathematical expectation of the cell level distribution, and δ is the standard deviation of the cell level distribution.

[0113] Since each grid corresponds to n cells, the matching probability of MR and grid (Cj) is:

[0114]

[0115] Where C j is the label of the jth most similar grid, and n is the number of cells corresponding to Cj.

[0116] The longitude and latitude weights of each grid are:

[0117]

[0118] Where k is the number of grids; the matching probability of the k grids is normalized as the weight of the grid longitude and latitude, and i＝1 k weight i = 1.

[0119] The longitude and latitude of MR after weighting are:

[0120]

[0121]

[0122] Where lonti is the longitude, lati is the dimension, C i is the label of the ith most similar grid, i = 1, 2, …, k.

[0123] Step 103, using ensemble learning technology to train the location positioning model and predict the user location.

[0124] Specifically, the location positioning model is trained using ensemble learning technology, and the user location is predicted, including:

[0125] extracting user static position features, user motion trajectory features, network environment features, user scene features and user geographic environment features;

[0126] Based on the user static position features, the user motion trajectory features, the network environment features, the user scene features and the user geographic environment features, the user static position features corresponding to the preset weight, the user motion trajectory features corresponding to the preset weight, the network environment features corresponding to the preset weight, the user scene features corresponding to the preset weight and the user geographic environment features corresponding to the preset weight, a position positioning model is constructed and trained, and the user position is predicted.

[0127] Among them, the model construction and training include the following steps: on the basis of fusion processing multi-dimensional data, a position positioning model covering "feature recognition + position positioning + model training" is constructed, and the model is trained by applying integrated learning technology to continuously improve the accuracy of the positioning model, and the MR data is accurately positioned.

[0128] Specifically, during positioning, on the one hand, the highest precision method such as OTT positioning is selected for positioning, and on the other hand, the integrated learning method is used to position and predict based on user static features, user trajectory features, scene features, geographic environment features and basic position signaling data. The prediction effect is evaluated by calculating the spatial distance between the high-precision positioning point and the predicted positioning point, and a large amount of training is performed to achieve the best effect, so that in the case where the high-precision positioning method is not applicable, the trained positioning model can also be used to predict the accurate position information.

[0129] Specifically, the user position prediction using integrated learning technology includes the following steps:

[0130] 1. Feature extraction:

[0131] User static features: position code, cell code, base station dimension, base station longitude, base station address, cell name, cell type (1: macro cell, 2: micro cell, 3: power division or coupling base station source, 4: RRU), antenna hanging height, electronic tilt angle, mechanical tilt angle, antenna direction angle, coverage type (1: outdoor, 2: indoor) antenna latitude, antenna longitude, base station type (1: 2G, 2: 3G, 3: 4G).

[0132] User motion trajectory features: select the position points within the last 5 minutes or the last 5 position points (take the way with more position points) as the historical position point set, wherein the position of each point, the total trajectory of the historical position point set, the instantaneous speed at each position point in the historical position point set, the motion direction at each position point in the historical position point set, the time spent at each point in the historical position point set, the total displacement distance of the historical position point set, the total displacement speed of the historical position point set, and the total displacement direction of the historical position point set.

[0133] Network environment features: 2G network, 3G network, 4G network, 5G network, and WiFi signal.

[0134] User scenario features: base station area type (city, county, and rural area), base station area name, base station city, base station county, nearby traffic landmarks (city road, national road, highway, subway, airport, and port), and nearby building.

[0135] User geographical environment features: climate (sunny, rainy, foggy, and snowy), land (grassland, forest, desert, wetland, and hard ground), river, lake, and mountain.

[0136] 2. The weight design of each feature adopts the following principles:

[0137] 1) The smaller the prediction range, the greater the feature weight, for example, 5G network feature weight > 4G network weight > 3G network weight > 2G network weight.

[0138] 2) The higher the position relevance, the greater the weight, for example, cell location weight > cell type weight > cell name weight.

[0139] 3) According to experience, the weight proportions of user static position features, user motion trajectory features, user network environment features, user scenario features, and user geographical environment features are 4:3.5:1.25:1:0.25.

[0140] 3. Model training and verification test: based on the preprocessed feature data, a random forest model is constructed and trained and verified.

[0141] For example, a random forest model is constructed using 300 CRT trees with a depth of 20;

[0142] The prediction result matches 80% within a range of 50 m, and 90% within a range of 100 m.

[0143] Step 104, constructing a spatial big database based on user location, and predicting and analyzing user behavior based on the spatial big database.

[0144] The spatial big data includes the following parts:

[0145] Geographic data: refers to the data directly or indirectly associated with a certain place on earth, including natural geographic data and socio-economic data; content: land cover type data, geomorphology data, soil data, hydrology data, vegetation data, residential data, river data, administrative boundary and socio-economic data, etc.; characteristics: large data volume, more regular, slow change.

[0146] Trajectory data: refers to the user activity data obtained by GNSS and other measurement methods and network check-in methods, which can be used to reflect the user's location and social preferences; content: individual trajectory data, group trajectory data, vehicle trajectory data, etc.; characteristics: large data volume, information fragmentation, low accuracy, semi-structured.

[0147] Spatial media data: includes location-based digital text, graphics, images, video images and other media data, mainly from mobile social network microblog and other new Internet applications.

[0148] The main functions of spatial big data are as follows:

[0149] City operation service: city planning, disease control, intelligent transportation, energy saving and emission reduction, environmental protection, emergency response;

[0150] Personal life service: social communication, personalized information push, driving safety, intelligent driving;

[0151] Enterprise economic service: enterprise scheduling, store site selection, advertisement push, location marketing;

[0152] Spatial big data analysis includes data processing:

[0153] Data processing is mainly to use big data tools such as HADOOP to realize data storage, processing, extraction and filtering; use ARCGIS and other presentation tools to realize spatial analysis, information display, condition definition and data production.

[0154] Spatial big data has multiple uses - from analyzing location data that was previously unavailable to gain more accurate insights, to analyzing data in motion to seize opportunities that were previously missed. Spatial big data platform enables your organization to solve complex problems that were previously impossible to solve.

[0155] Specifically, the business idea of spatial big data analysis is: data-information-knowledge-intelligence-solution.

[0156] Key technologies:

[0157] Sensor network - location sensing technology, including satellite, WiFi, Bluetooth positioning technology, to obtain massive spatial location data of people or things;

[0158] Holographic location map - grid multi-dimensional location information number, including people, things, events and so on, to achieve the real sense of multi-dimensional data in space correlation;

[0159] Spatial big data warehouse - unstructured data, semi-structured data, structured data parallel storage and management, to establish a heterogeneous database, adapt to the complexity, diversity and chaos of spatial big data;

[0160] Data security - multi-level application security and design, each layer of security is designed for a specific purpose, and can be used to provide authorization rules;

[0161] Spatial big data mining - using neural network mining algorithm, to ensure the high parallelism, self-adaptation and self-organization characteristics of spatial big data analysis and calculation, with good prediction ability and control ability, supporting blind analysis of big data, obtaining different knowledge learning, or for business decision, or for enterprise management or for expansion development, or for improving user experience;

[0162] Spatial big data visualization - full perspective perception data, restore the truth, control the spatial distribution of things and correlation, predict future trends;

[0163] Spatial big data analysis and presentation: basic data report, customer flow line image, customer loss analysis, shop union suggestion, etc.

[0164] The application constructs a three-in-one location positioning model based on user location behavior by multi-dimensional combination positioning and integrated learning, and constructs a spatial big database based on user location to predict and analyze user behavior, so as to realize the discrimination of the authenticity and rationality of user behavior (such as location verification, whether there is fraud or not, etc.). In addition, the application uses the method of constructing a model by using big data to control and prompt the risk of banks, insurance companies and the like.

[0165] The multi-dimensional combined location behavior trajectory chain-based behavior analysis device provided by the application is described below. The multi-dimensional combined location behavior trajectory chain-based behavior analysis device described below can be correspondingly referred to the multi-dimensional combined location behavior trajectory chain-based behavior analysis method described above.

[0166] Figure 3 The multi-dimensional combined location behavior trajectory chain-based behavior analysis device provided by the application is described below. The multi-dimensional combined location behavior trajectory chain-based behavior analysis device described below can be correspondingly referred to the multi-dimensional combined location behavior trajectory chain-based behavior analysis method described above. Figure 3 As shown in FIG. 1, the application provides a multi-dimensional combined location behavior trajectory chain-based behavior analysis device, which comprises a positioning module 301, an identification module 302, a prediction module 303 and an analysis module 304, wherein:

[0167] The positioning module 301 is configured to perform network-wide measurement report (MR) positioning on the user by using a positioning method combining accurate positioning and pattern matching; the identification module 302 is configured to identify the indoor and outdoor state of the user, the floor layer of the building and the motion state of the user based on the MR positioning result; the prediction module 303 is configured to train a location positioning model by using an ensemble learning technique and predict the location of the user; and the analysis module 304 is configured to construct a spatial database based on the location of the user and perform prediction and analysis on the behavior of the user based on the spatial database.

[0168] In one embodiment, the training of the location positioning model by using the ensemble learning technique and the prediction of the location of the user include:

[0169] extracting the static location feature of the user, the motion trajectory feature of the user, the network environment feature, the scene feature in which the user is located and the geographic environment feature in which the user is located;

[0170] constructing and training a location positioning model and predicting the location of the user based on the static location feature of the user, the motion trajectory feature of the user, the network environment feature, the scene feature in which the user is located and the geographic environment feature in which the user is located, the preset weight corresponding to the static location feature of the user, the preset weight corresponding to the motion trajectory feature of the user, the preset weight corresponding to the network environment feature, the preset weight corresponding to the scene feature in which the user is located and the preset weight corresponding to the geographic environment feature in which the user is located.

[0171] In one embodiment, the smaller the prediction range of the network environment feature, the greater the preset weight corresponding to the network environment feature.

[0172] In one embodiment, the higher the location relevance of the static location feature of the user, the greater the preset weight corresponding to the static location feature of the user.

[0173] In one embodiment, the weight proportions corresponding to the static location feature of the user, the motion trajectory feature of the user, the network environment feature, the scene feature in which the user is located and the geographic environment feature in which the user is located decrease in turn.

[0174] In one embodiment, the weight proportions corresponding to the static location feature of the user, the motion trajectory feature of the user, the network environment feature, the scene feature in which the user is located and the geographic environment feature in which the user is located are 40%, 35%, 12.5%, 10% and 2.5% in turn.

[0175] It should be noted that the behavior analysis device based on the multi-dimensional combined location behavior trajectory chain provided in the embodiment of the present application can realize all the method steps realized by the method embodiment and achieve the same technical effects, and the same parts and beneficial effects in the embodiment as the method embodiment will not be described in detail.

[0176] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call a computer program in the memory 430 to execute steps of a behavior analysis method based on a multi-dimensional combined position behavior trajectory chain, such as including:

[0177] A location method combining precise positioning and pattern matching is used to perform full-network measurement report MR positioning of users; based on the MR positioning results, users are identified as indoors or outdoors, as well as as they are identified as building layers and as moving objects; an ensemble learning technique is used to train the location model and predict user locations; a large spatial database based on user locations is constructed, and user behavior is predicted and analyzed based on the large spatial database.

[0178] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0179] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, which, when executed by a computer, enable the computer to perform the steps of the behavior analysis method based on multi-dimensional combined position behavior trajectory chain provided by the above methods, for example including:

[0180] The user is positioned by using a positioning method combining precise positioning and pattern matching, and a measurement report (MR) positioning is performed; indoor and outdoor identification, building layer identification and motion state identification are performed based on the MR positioning result; an integrated learning technology is used to train a position positioning model and predict the user position; a spatial database based on the user position is constructed, and user behavior prediction and analysis are performed based on the spatial database.

[0181] In another aspect, the embodiments of the present application further provide a processor-readable storage medium, which stores a computer program for causing a processor to perform the steps of the method provided by the above-mentioned embodiments, for example, including:

[0182] The user is positioned by using a positioning method combining precise positioning and pattern matching, and a measurement report (MR) positioning is performed; indoor and outdoor identification, building layer identification and motion state identification are performed based on the MR positioning result; an integrated learning technology is used to train a position positioning model and predict the user position; a spatial database based on the user position is constructed, and user behavior prediction and analysis are performed based on the spatial database.

[0183] The processor-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to a magnetic storage (such as a floppy disk, a hard disk, a magnetic tape, a magneto-optical disk (MO), etc.), an optical storage (such as a CD, a DVD, a BD, a HVD, etc.), and a semiconductor memory (such as a ROM, an EPROM, an EEPROM, a non-volatile memory (NAND FLASH), a solid-state disk (SSD)), etc.

[0184] The device embodiments described above are only schematic, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments. Those skilled in the art can understand and implement without creative labor.

[0185] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0186] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A behavior analysis method based on multi-dimensional combined position behavior trajectory chain, characterized in that, The method comprises the following steps: A positioning mode combining accurate positioning and pattern matching is used to perform a whole-network measurement report (MR) positioning on a user; Based on the MR positioning result, indoor-outdoor identification, building layer identification and motion state identification are performed on the user; An integrated learning technology is used to train a position positioning model and predict the user position; A spatial large database based on the user position is constructed, and user behavior prediction and analysis are performed based on the spatial large database; The integrated learning technology is used to train the position positioning model and predict the user position, which comprises the following steps: User static position features, user motion trajectory features, network environment features, user scene features and user geographic environment features are extracted; A position positioning model is constructed based on the user static position features, the user motion trajectory features, the network environment features, the user scene features, the user geographic environment features, preset weights corresponding to the user static position features, preset weights corresponding to the user motion trajectory features, preset weights corresponding to the network environment features, preset weights corresponding to the user scene features and preset weights corresponding to the user geographic environment features, and the position positioning model is trained and used to predict the user position. 2.The behavior analysis method based on multi-dimensional combination position behavior trajectory chain according to claim 1, wherein, The smaller the prediction range of the network environment features, the greater the preset weight corresponding to the network environment features. 3.The behavior analysis method based on multi-dimensional combination position behavior trajectory chain according to claim 1, wherein, The higher the position correlation of the user static position features, the greater the preset weight corresponding to the user static position features. 4.The method of claim 1, wherein, The weights corresponding to the user static position features, the user motion trajectory features, the network environment features, the user scene features and the user geographic environment features decrease in turn.

5. The behavior analysis method based on multi-dimensional combination position behavior trajectory chain according to claim 4, characterized in that, The weights corresponding to the user static position features, the user motion trajectory features, the network environment features, the user scene features and the user geographic environment features are 40%, 35%, 12.5%, 10% and 2.5% in turn.

6. A behavior analysis device based on a multi-dimensional combined position behavior trajectory chain, characterized by The method comprises the following steps: A positioning module is configured to perform a whole-network measurement report (MR) positioning on a user by using a positioning mode combining accurate positioning and pattern matching; An identification module is configured to perform indoor-outdoor identification, building layer identification and motion state identification on the user based on the MR positioning result; A prediction module is configured to train a position positioning model by using an integrated learning technology and predict the user position; An analysis module is configured to construct a spatial large database based on the user position and perform user behavior prediction and analysis based on the spatial large database; The integrated learning technology is used to train the position positioning model and predict the user position, which comprises the following steps: User static position features, user motion trajectory features, network environment features, user scene features and user geographic environment features are extracted; A position positioning model is constructed based on the user static position features, the user motion trajectory features, the network environment features, the user scene features, the user geographic environment features, preset weights corresponding to the user static position features, preset weights corresponding to the user motion trajectory features, preset weights corresponding to the network environment features, preset weights corresponding to the user scene features and preset weights corresponding to the user geographic environment features, and the position positioning model is trained and used to predict the user position. The position positioning model is constructed and trained based on the user static position feature, the user motion trajectory feature, the network environment feature, the user scene feature, the user geographic environment feature, a preset weight corresponding to the user static position feature, a preset weight corresponding to the user motion trajectory feature, a preset weight corresponding to the network environment feature, a preset weight corresponding to the user scene feature, and a preset weight corresponding to the user geographic environment feature, and the user position is predicted.

7. An electronic device comprising a processor and a memory having a computer program stored therein, characterized in that The processor implements the behavior analysis method based on the multi-dimensional combined position behavior trajectory chain according to any one of claims 1 to 5 when the computer program is executed.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the behavior analysis method based on the multi-dimensional combined position behavior trajectory chain according to any one of claims 1 to 5.

9. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the behavior analysis method based on the multi-dimensional combined position behavior trajectory chain according to any one of claims 1 to 5. The computer program is executed by the processor to implement the behavior analysis method based on the multi-dimensional combined position behavior trajectory chain according to any one of claims 1 to 5.

Images