Information prediction method and device, equipment, storage medium and computer program product
By obtaining and predicting the location information of the target object, using feature extraction and prediction models, the problem of inability to accurately predict passenger motion trajectories is solved, and the accuracy and real-time prediction are improved, helping tourism sites improve service capabilities.
Patent Information
- Application Number
- CN202510020798.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-02
- Publication Date
- 2025-05-27
AI Technical Summary
At present, it is impossible to accurately predict the movement trajectory of tourists, resulting in insufficient service capabilities of tourist destinations.
By obtaining the position information of the target object, using a preset feature extraction model for feature extraction, obtaining the target feature sequence, and predicting the target feature sequence through the trained target prediction model to obtain the motion trajectory of the target object.
It improves the accuracy of the prediction results, reduces the amount of data analyzed, and ensures the real-time and effectiveness of the calculations, helping tourist sites improve their reception capabilities for tourists.
Smart Images
Figure CN120045868A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of big data applications, and particularly to an information prediction method, apparatus, device, storage medium, and computer program product. Background Art
[0002] With the development of society, the current material life of the people has entered a relatively affluent stage, the consumption level has continued to increase, and the quality of life has been steadily improved. In recent years, the national tourism demand has increased rapidly, thus accelerating the development of the domestic tourism industry. However, the increasing number of tourists has severely tested the reception capacity of tourist destinations. How to predict the change of the tourist movement trend to improve the reception capacity of tourist destinations for tourists has become a technical problem that urgently needs to be solved.
[0003] Content of the Application
[0004] To solve the above technical problems, the present application expects to provide an information prediction method, apparatus, device, storage medium, and computer program product, which solves the problem that the current tourist movement trajectory cannot be accurately predicted to ensure the service capacity of the destination. By specifically analyzing and predicting the target population, the amount of data to be analyzed is reduced, the accuracy of the prediction result is improved, and the real-time performance and effectiveness of the calculation are ensured.
[0005] The technical solution of the present application is realized as follows:
[0006] The present application provides an information prediction method, and the method includes:
[0007] Obtain the location information of the target object;
[0008] Adopt a preset feature extraction model to extract features from the location information to obtain a target feature sequence; wherein, the feature length of each feature in the target feature sequence is the same;
[0009] Predict the target feature sequence through a trained target prediction model to obtain the movement trajectory of the target object.
[0010] In the above solution, the obtaining the location information of the target object includes:
[0011] Obtain the planned destination information of the target object within a future time period;
[0012] Obtain the location distribution information of the target object within a historical time period; wherein, the location information at least includes the planned destination information and the location distribution information.
[0013] In the above solution, the planned destination information includes time information and position coordinate information, and the position distribution information includes the residence duration information of the area where the device is located determined according to the preset area information and the position coordinate information of the area where the device is located.
[0014] In the above solution, the structure of the preset feature extraction model at least includes an encoder-decoder model.
[0015] In the encoder-decoder model, the encoder includes at least two layers of BiGRU units, and the decoder includes at least one layer of BiGRU units.
[0016] In the above solution, the method further includes:
[0017] Obtaining a dataset of samples to be trained;
[0018] Inputting the sample data in the dataset of samples to be trained into the preset feature extraction model to obtain a set of feature sequences to be trained;
[0019] Training the model to be trained with the set of feature sequences to be trained to obtain the trained target prediction model.
[0020] In the above solution, the step of training the model to be trained with the set of feature sequences to be trained to obtain the trained target prediction model includes:
[0021] Obtaining a feature sequence to be input from the set of feature sequences to be trained;
[0022] Inputting the feature sequence to be input into the feature learning layer of the model to be trained to obtain a learning feature result;
[0023] Inputting the learning feature result into the attention layer of the model to be trained to obtain a weight coefficient corresponding to the learning feature result;
[0024] Performing weighted analysis processing on the learning feature result and the corresponding weight coefficient through the output layer of the model to be trained to obtain a prediction result corresponding to the feature sequence to be input;
[0025] Calculating the actual result corresponding to the feature sequence to be input and the corresponding prediction result by using a preset loss function to obtain a loss value corresponding to the feature sequence to be input;
[0026] If the loss value is greater than a preset threshold, performing backpropagation in the model to be trained by using the loss value to adjust the parameters of the model to be trained to obtain a first trained model;
[0027] After updating the model to be trained to the first training model, perform the step of obtaining the feature sequence to be input from the set of feature sequences to be trained until the target prediction model corresponding to the loss value less than or equal to the preset threshold is obtained;
[0028] If the loss value is less than or equal to the preset threshold, determine that the target prediction model is the model to be trained.
[0029] In the above solution, the feature learning layer at least includes a double-layer stochastic matrix unit (SRU).
[0030] In the above solution, the preset loss function is where y i is the prediction result of the i-th sample in the feature sequence to be input, p i is the actual result of the i-th sample in the feature sequence to be input, and N is the number of samples included in the feature sequence to be input.
[0031] In the above solution, the method further includes:
[0032] Based on the movement trajectory, count the population at the corresponding location; and / or,
[0033] Based on the movement trajectory, perform information recommendation for the target object.
[0034] This application provides an information prediction device, which at least includes: an acquisition unit, an extraction unit, and a prediction unit; where:
[0035] The acquisition unit is used to acquire the location information of the target object;
[0036] The extraction unit is used to use a preset feature extraction model to extract features from the location information to obtain a target feature sequence; where the feature length of each feature in the target feature sequence is the same;
[0037] The prediction unit is used to predict the target feature sequence through a trained target prediction model to obtain the movement trajectory of the target object.
[0038] This application provides an information prediction device, which at least includes: a communication interface, a memory, a processor, and a communication bus; where:
[0039] The memory is used to store executable instructions;
[0040] The communication bus is used to realize the communication connection between the communication interface, the processor, and the memory;
[0041] The processor is configured to execute the information prediction program stored in the memory, and implement the steps of the information prediction method described in any one of the above.
[0042] This application provides a storage medium, on which an information prediction program is stored. When the processing program is executed, it is used to implement the steps of the information prediction method described in any one of the above.
[0043] This application provides a computer program product, including a computer program, which implements the steps of the information prediction method described in any one of the above when executed by a processor.
[0044] The embodiments of this application provide an information prediction method, device, equipment, storage medium and computer program product. By obtaining the position information of a target object, using a preset feature extraction model to extract features from the position information to obtain a target feature sequence, and predicting the target feature sequence through a trained target prediction model to obtain the movement trajectory of the target object. In this way, through the target prediction model, the preset feature extraction model is used to perform feature extraction processing on the unknown information to obtain the target feature sequence for prediction processing, and the possible future movement trajectory of the target object is obtained. Thus, the future trajectory of the target object is predicted through the target prediction model, solving the problem that the current movement trajectory of passengers cannot be accurately predicted to ensure the service capacity of the destination. By specifically analyzing and predicting the target population, the amount of data to be analyzed is reduced, the accuracy of the prediction result is improved, and the real-time performance and effectiveness of the calculation are ensured. Description of the Drawings
[0045] Figure 1 It is a schematic flowchart of an information prediction method provided by an embodiment of this application;
[0046] Figure 2 It is a schematic implementation flowchart of an application embodiment provided by an embodiment of this application;
[0047] Figure 3 It is a schematic structural diagram of a BiGRU provided by an embodiment of this application;
[0048] Figure 4 It is a schematic structural diagram of a GRU provided by an embodiment of this application;
[0049] Figure 5 It is a schematic structural diagram of an encoder-decoder architecture provided by an embodiment of this application;
[0050] Figure 6 It is a schematic structural diagram of an SRU model provided by an embodiment of this application;
[0051] Figure 7Schematic diagram of the structure of an attention mechanism model provided by an embodiment of the present application;
[0052] Figure 8 Schematic diagram of the structure of a final prediction model provided by an embodiment of the present application;
[0053] Figure 9 Schematic diagram of the structure of an information prediction device provided by an embodiment of the present application;
[0054] Figure 10 Schematic diagram of the structure of an information prediction device provided by an embodiment of the present application. Detailed implementation manners
[0055] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application.
[0056] An embodiment of the present application provides an information prediction method. Referring to Figure 1 as shown, the method is applied to an information prediction device, and the method includes the following steps:
[0057] Step 101: Obtain the position information of the target object.
[0058] In the embodiment of the present application, the target object is a user who needs to perform trajectory prediction, such as a tourist. The information prediction device can, for example, obtain relevant information about the positions of the target object in the past period and the future period from corresponding ticketing systems, map systems, etc. after obtaining the corresponding authorization, so as to obtain the position information.
[0059] Step 102: Use a preset feature extraction model to extract features from the position information to obtain a target feature sequence.
[0060] Wherein, the feature lengths of each feature in the target feature sequence are the same.
[0061] In the embodiment of the present application, the preset feature extraction model is a model that has been determined in advance and is used to perform unified processing on feature extraction. By using the preset feature extraction model to extract features from the position information of the target object, it is realized to convert relevant position features of different lengths into position features of the same length, thereby obtaining the target feature sequence.
[0062] Step 103: Use the trained target prediction model to predict the target feature sequence to obtain the motion trajectory of the target object.
[0063] In the embodiments of the present application, the target prediction model can be stored in the storage unit of the information prediction device, or can be obtained by the information prediction device from the server device that provides the target prediction model. Specifically, it can be determined according to the actual situation and will not be specifically limited here. The information prediction device obtains the trained target prediction model, and then uses the target prediction model to perform prediction processing on the target feature sequence. In this way, the target prediction model can output the motion trajectory corresponding to the target object. Thus, the future feasible motion trajectory of the target object is predicted. After obtaining the motion trajectory of the target object in this way, the motion trajectory can be displayed to realize route planning for the target object; or, after obtaining the motion trajectory of the target object, position statistics can be performed based on the motion trajectories of a large number of different objects obtained through statistics to obtain the possible future crowd flow information corresponding to each position, and then corresponding control processing can be indicated according to the crowd flow information; or, the motion trajectory of the target object can also be used to provide targeted information recommendations for the target object to ensure the usage experience effect of the target object.
[0064] Based on the foregoing embodiments, in other embodiments of the present application, step 101 of obtaining the position information of the target object can be implemented through the following steps:
[0065] Obtain the planned destination information of the target object within a future time period;
[0066] Obtain the position distribution information of the target object within a historical time period; where the position information at least includes the planned destination information and the position distribution information.
[0067] In the embodiments of the present application, the future time period can be a preset time period. For example, it can be the travel time period specified by the user. The planned destination information of the target object within the future time period can be input into the information prediction device after relevant operations of the user, or can be determined by the information prediction device after ticket reservation according to the ticket reservation system used by the user. The position distribution information of the target object within the historical time period can be input into the information prediction device by the user, or can be obtained by the target object from the corresponding statistical system, such as a map system.
[0068] Based on the foregoing embodiments, in other embodiments of the present application, the planned destination information includes time information and position coordinate information, and the position distribution information includes the residence duration information of the area where the target object is located determined according to the preset area information and the position coordinate information of the area where the target object is located.
[0069] In the embodiments of the present application, the time information included in the planned destination information is the time corresponding to when arriving at the planned destination in the future. The location coordinate information can be represented by the earth location coordinates of the destination. The planned destination information includes the relevant information of one or more destinations. The location distribution information includes the relevant information of one or more historical locations. The residence duration information of the area where one is located and the location coordinate information of the area where one is located are determined according to the preset area information. For example, it can be divided by city, and the residence duration in a certain area in the city, such as a scenic area, and the location coordinate information of the scenic area, etc.
[0070] Based on the foregoing embodiments, in other embodiments of the present application, the structure of the preset feature extraction model at least includes an encoder-decoder model.
[0071] Based on the foregoing embodiments, in other embodiments of the present application, the encoder included in the encoder-decoder model at least includes two layers of BiGRU units, and the decoder includes at least one layer of BiGRU units.
[0072] Based on the foregoing embodiments, in other embodiments of the present application, the method further includes the following steps:
[0073] Obtain a dataset of samples to be trained;
[0074] Input the sample data in the dataset of samples to be trained into the preset feature extraction model to obtain a set of feature sequences to be trained;
[0075] Use the set of feature sequences to be trained to train the model to be trained to obtain a trained target prediction model.
[0076] In the embodiments of the present application, the dataset of samples to be trained is a dataset with complete trajectory changes collected in advance. The sample data in the dataset of samples to be trained is subjected to feature extraction processing using the preset feature extraction model to obtain a corresponding set of feature sequences to be trained, and then the model to be trained is trained based on the set of feature sequences to be trained until a target prediction model is obtained.
[0077] Based on the foregoing embodiments, in other embodiments of the present application, the step of using the set of feature sequences to be trained to train the model to be trained to obtain a trained target prediction model can be implemented through the following steps:
[0078] Obtain an input feature sequence from the set of feature sequences to be trained;
[0079] Input the input feature sequence into the feature learning layer of the model to be trained to obtain a learning feature result;
[0080] Input the learning feature result into the attention layer of the model to be trained to obtain a weight coefficient corresponding to the learning feature result;
[0081] Perform weighted analysis processing on the learning feature results and corresponding weight coefficients through the output layer of the model to be trained, and obtain the prediction results corresponding to the feature sequence to be input;
[0082] Use a preset loss function to calculate the actual results and corresponding prediction results corresponding to the feature sequence to be input, and obtain the loss value corresponding to the feature sequence to be input;
[0083] If the loss value is greater than the preset threshold, perform backpropagation of the loss value in the model to be trained, adjust the parameters of the model to be trained, and obtain the first trained model;
[0084] After updating the model to be trained to the first trained model, execute the step of obtaining the feature sequence to be input from the set of feature sequences to be trained until the target prediction model corresponding to the loss value less than or equal to the preset threshold is obtained;
[0085] If the loss value is greater than or equal to the preset threshold, determine the target prediction model as the model to be trained.
[0086] In the embodiments of the present application, a model training method is adopted to train the model to be trained based on the set of feature sequences to be trained, and obtain the target prediction model when the loss value is less than or equal to the preset threshold.
[0087] Based on the foregoing embodiments, in other embodiments of the present application, the feature learning layer includes at least a double-layer stochastic matrix unit SRU.
[0088] Based on the foregoing embodiments, in other embodiments of the present application, the preset loss function is where y i is the prediction result of the i-th sample in the feature sequence to be input, p i is the actual result of the i-th sample in the feature sequence to be input, and N is the number of samples included in the feature sequence to be input.
[0089] Based on the foregoing embodiments, in other embodiments of the present application, the method further includes the following steps:
[0090] Based on the movement trajectory, count the population at the corresponding position; and / or,
[0091] Based on the movement trajectory, perform information recommendation for the target object.
[0092] In the embodiments of the present application, after obtaining the movement trajectories of a large number of target objects, based on the large number of movement trajectories, count the population at the same position to obtain the corresponding population, realize the prediction of the flow of people for subsequent management, or, after determining the movement trajectories of the target objects, according to the movement trajectories, perform targeted recommendations for the target objects on food, sightseeing information, etc., and realize information recommendation.
[0093] Based on the foregoing embodiments, an implementation system for implementing an information prediction method provided by an embodiment of the present application includes: a data collection module, a feature extraction module, and a trajectory prediction module. When the system is applied to predict the trajectory of scenic area tourists, the corresponding implementation process can refer to Figure 2 as shown. Among them, Figure 2 the data collection module at least includes two processes of collection and screening. First, collect the scenic area ticket sales data within a period of time. On this basis, screen to obtain the information of tourists arriving at the scenic area. Finally, combine the operator data to obtain the movement data of the current tourists within a period of time; due to the characteristics that the lengths of the movement trajectory sequences of different tourists in different scenic areas are different, the feature extraction module can extract the collected sequence data of different lengths to obtain feature sequences of the same length; the trajectory prediction module needs to perform supervised training on the feature sequences to achieve tourist trajectory prediction.
[0094] Specifically, the data collection module mainly includes the collection of scenic area ticket sales data and the collection of location signaling sequence data. The corresponding specific implementation process can include:
[0095] 1. Obtain the scenic area on-site ticket sales data ticket = {t1, t2,..., tn}. Here, 30 days of the scenic area is used as the time range for extracting ticket sales data, that is, determine the ticket booking information within 30 days.
[0096] 2. Based on the tourist information included in ticket, obtain the location signaling sequence data loc = {l1, l2,..., ln} before purchasing the ticket, where n represents the default stay time of tourists at the tourist destination, which is also set to 30 days here.
[0097] 3. Combine the urban area information to screen the data loc, and eliminate the data with insignificant movement characteristics. For example, the tourists' positions hovering in a small area or remaining unchanged within a certain time range due to a certain activity are determined as invalid data and eliminated or merged.
[0098] It should be noted that ticket contains all the tourist ticket purchase information within the specified time period. Its main function is to identify the tourist ticket purchase time and the location information of the scenic area. Among them, ti represents the set of all tourist ticket purchase records on the i-th day. The information included in each point in ti can be shown in Table 1.
[0099] Table 1
[0100] Serial number Field Description 1 imsi Passenger imsi identifier 2 time Ticket purchase time 3 longitude Scenic spot longitude 4 latitude Scenic spot latitude
[0101] The data loc is a set of location signaling sequences. li represents all the location signaling sequences of all the extracted passenger information within the past i days. The specific information included in each point in li can be shown in Table 2.
[0102] Table 2
[0103] Serial number Field name Description 1 imsi imsi identifier 2 timestamp Timestamp 3 longitude Longitude 4 latitude Latitude
[0104] When filtering the loc data, it is possible to traverse and search, sequentially traverse the passenger data in chronological order, and use the region search method to determine whether the current data point is within the urban area. In some application scenarios, it can be implemented through the geo plugin of the database. At the same time, when traversing, judge the distance between the current point and the previous point. If the current distance from the previous point is less than 500 meters, mark the current point and continue to traverse backward until the first point greater than 500m is found, and then filter all the marked points, so as to reduce the impact of the passenger's short-distance movement on the feature extraction effect of the feature extraction module and reduce the deviation generated when predicting the trajectory.
[0105] The main function of the feature extraction module is to use an encoder-decoder to generate feature sequences of the same length from location signaling sequences of different lengths. Specifically, the encoder-decoder can use a bidirectional gated recurrent unit (BiGRU) as a component of the encoder-decoder to extract the bidirectional features of the input sequence for data support for trajectory prediction. Correspondingly, the local structure of the BiGRU can refer to Figure 3 as shown, where xi is the input location signaling sequence, GRU is the gated recurrent unit, and the specific structure can refer to Figure 4 as shown, and the specific calculation process can be expressed by referring to the following calculation formula: z t =σ(W z ·[h t-1 ,x t ), r t =σ(W r ·[h t-1 ,x t ), where, where, z t is the update gate, r t is the reset gate, W i is the corresponding weight vector, h (t-1) represents the hidden state at time t-1, x t is the input sequence at time t, and * is the element-by-element multiplication of vectors.
[0106] The encoder-decoder architecture can calculate input sequences of different lengths and generate feature sequences of the same length through the decoder. The specific encoder-decoder architecture design can refer toFigure 5 As shown in Figure 5 , the encoder stage includes two layers of BiGRU units to extract the bidirectional features of the input position signaling data sequence for encoding, and a Convolutional Neural Networks (CNN) to obtain the local sensitive feature encoding of the input data sequence using convolutional characteristics. The complete feature encoding C of the entire input data sequence is obtained by concatenating all the outputs using a weight vector. The decoder includes one layer of BiGRU units to generate the output feature vector based on the intermediate expression vector. In this way, during the process of feature extraction, the trajectory prediction stage is a supervised learning, so it is necessary to save the feature information of the current trajectory sequence in a timely manner.
[0107] In this way, the collected data includes the determined trajectory information and the possible trajectory information within a certain period in the future. The feature sequence after real feature extraction also contains the feature information of the trajectory and the trajectory information within a certain period in the future. In this way, the supervised training of the trajectory prediction model can be better realized.
[0108] The main function of the trajectory prediction module is to construct a trajectory prediction model based on the SRU model based on the feature sequence output by the feature extraction module, so as to realize the prediction of the future movement trajectory of passengers. In the embodiment of the present application, the output of the prediction is the movement trajectory sequence of passengers within a certain period in the future. By obtaining all the passenger itinerary trajectory data related to the tourist destination in the city in advance according to relevant data information, and performing corresponding feature extraction on different trajectory data sequences to obtain the historical trajectory sequence and the prediction trajectory sequence, the purpose of supervised training is achieved. Among them, the SRU model is a recurrent neural network model, which puts most of the operations into parallel processing, can quickly and effectively learn the complete feature dependencies between sequences, and thus better realize real-time trajectory prediction.
[0109] Correspondingly, a model structure of the SRU model included in the trajectory prediction module applied in the present application can be referred to Figure 6 As shown in Figure 6 , the corresponding calculation process can be represented by the following formula: f t =σ(W f x t +b f ), r t =σ(W r x t +b r ), h t =r t ⊙g(c t )+(1 - r t )⊙x, where W i and b i are the weight matrix and bias matrix corresponding to the gate logic respectively, f t and rt are the forget gate and reset gate respectively, ⊙ is the matrix point-by-point calculation, x t is the input at time t, c t The unit status.
[0110] In this way, by using the double-layer SRU as a prediction model, the predicted output trajectory sequence of the SRU is fitted with the actual sequence, and the output weight of the model is continuously adjusted to improve the prediction accuracy. Correspondingly, when training the prediction model, the evaluation index that can be used can be evaluated using the following formula, which is recorded as: Where N is the number of specific samples, p i The actual value of the i-th sample predicted by the model, y i is the prediction result of the current i-th sample.
[0111] Furthermore, in order to increase the matching degree of the prediction results, the attention mechanism is used to fit the key feature sequence to improve the prediction accuracy. The model structure of the corresponding attention mechanism model can be referred to Figure 7 As shown, the calculation formula can be recorded as: Among them, α i is the weight, which indicates the attention level of the i-th input vector when a given query q is given, x i Output vector for SRU.
[0112] In summary, the structure of the final prediction model can be determined by referring to Figure 8 As shown in the figure, the corresponding implementation process of model training can be recorded as follows: first, the feature sequence of the feature extraction module is input into the prediction model, the feature sequence is learned by the double-layer SRU, the input weight information is obtained by the output attention layer, and the weight information is used to perform weighted averaging on the input to obtain the corresponding prediction result; then, the prediction result output by the model is fitted with the evaluation formula to continuously adjust the hyperparameters of the final prediction model through back propagation, until the model prediction result gradually tends to a stable state. After the accuracy verification is passed, the final prediction model with determined hyperparameters can be used to predict passenger trajectories.
[0113] In this way, the deep learning model is used to fit historical user characteristics, the prediction speed is faster, and it can be trained adaptively according to the characteristics of different cities. Furthermore, the feature extraction method based on the encoder-decoder can mine the feature information contained in sequences of different lengths, which can avoid the impact of data on model performance. Moreover, after predicting the travel trajectory of passengers, it can predict all tourist destinations in the city, and provide crowd flow information for tourist destinations based on the prediction results, which is conducive to crowd flow management, and can further recommend content information to specific groups of people.
[0114] The information prediction method provided by the embodiment of the present application obtains the position information of the target object, uses a preset feature extraction model to extract features from the position information to obtain a target feature sequence, and predicts the target feature sequence through a trained target prediction model to obtain the movement trajectory of the target object. In this way, through the target prediction model, the preset feature extraction model is used to perform feature extraction processing on the unknown information to obtain the target feature sequence for prediction processing, and the possible future movement trajectory of the target object is obtained. Thus, the future trajectory of the target object is predicted through the target prediction model, solving the problem that the current movement trajectory of passengers cannot be accurately predicted to ensure the service capacity of the destination. By specifically analyzing and predicting the target population, the amount of data analyzed is reduced, the accuracy of the prediction result is improved, and the real-time performance and effectiveness of the calculation are ensured.
[0115] Based on the foregoing embodiments, an embodiment of the present application provides an information prediction device, which can be applied to Figure 1 and the information prediction method provided by the corresponding embodiment. Referring to Figure 9 as shown, the information prediction device 2 may include: an acquisition unit 21, an extraction unit 22, and a prediction unit 23; where:
[0116] The acquisition unit 21 is configured to acquire the position information of the target object;
[0117] The extraction unit 22 is configured to use a preset feature extraction model to extract features from the position information to obtain a target feature sequence; wherein, the feature length of each feature in the target feature sequence is the same;
[0118] The prediction unit 23 is configured to predict the target feature sequence through a trained target prediction model to obtain the movement trajectory of the target object.
[0119] In other embodiments of the present application, the acquisition unit includes: a first acquisition module and a second acquisition module; where:
[0120] The first acquisition module is configured to acquire the planned destination information of the target object within a future time period;
[0121] The second acquisition module is configured to acquire the position distribution information of the target object within a historical time period; wherein, the position information includes at least the planned destination information and the position distribution information.
[0122] In other embodiments of the present application, the planned destination information includes time information and position coordinate information, and the position distribution information includes the residence duration information of the area where the target object is located determined according to the preset area information and the position coordinate information of the area where the target object is located.
[0123] In other embodiments of the present application, the structure of the preset feature extraction model includes at least an encoder-decoder model.
[0124] In other embodiments of the present application, the encoder included in the encoder-decoder model includes at least two layers of BiGRU units, and the decoder includes at least one layer of BiGRU units.
[0125] In other embodiments of the present application, the information prediction device further includes: a training unit; wherein:
[0126] The obtaining unit is further configured to obtain a dataset of samples to be trained;
[0127] The extraction unit is further configured to input the sample data in the dataset of samples to be trained into a preset feature extraction model to obtain a set of feature sequences to be trained;
[0128] The training unit is configured to perform model training on the model to be trained by using the set of feature sequences to be trained to obtain a trained target prediction model.
[0129] In other embodiments of the present application, the training unit is specifically configured to implement the following steps:
[0130] Obtain a feature sequence to be input from the set of feature sequences to be trained;
[0131] Input the feature sequence to be input into the feature learning layer of the model to be trained to obtain a learning feature result;
[0132] Input the learning feature result into the attention layer of the model to be trained to obtain a weight coefficient corresponding to the learning feature result;
[0133] Perform weighted analysis processing on the learning feature result and the corresponding weight coefficient through the output layer of the model to be trained to obtain a prediction result corresponding to the feature sequence to be input;
[0134] Calculate the actual result corresponding to the feature sequence to be input and the corresponding prediction result by using a preset loss function to obtain a loss value corresponding to the feature sequence to be input;
[0135] If the loss value is greater than a preset threshold, perform backpropagation in the model to be trained by using the loss value to adjust the parameters of the model to be trained to obtain a first trained model;
[0136] After updating the model to be trained to the first trained model, execute the step of obtaining the feature sequence to be input from the set of feature sequences to be trained until the target prediction model corresponding to when the loss value is less than or equal to the preset threshold is obtained;
[0137] If the loss value is less than or equal to the preset threshold, determine that the target prediction model is the model to be trained.
[0138] In other embodiments of the present application, the feature learning layer includes at least two layers of stochastic random unit (SRU).
[0139] In other embodiments of the present application, the preset loss function is where y i is the prediction result of the i-th sample in the to-be-input feature sequence, p i is the actual result of the i-th sample in the to-be-input feature sequence, and N is the number of samples included in the to-be-input feature sequence.
[0140] In other embodiments of the present application, the information prediction device further includes: a statistical unit, and / or a recommendation unit; where:
[0141] The statistical unit is configured to count the population quantity at the corresponding position based on the movement trajectory; and / or,
[0142] The recommendation unit is configured to perform information recommendation for the target object based on the movement trajectory.
[0143] It should be noted that the process of information interaction between units and modules in this embodiment can refer to the description in other embodiments, and will not be elaborated here.
[0144] The information prediction device provided by the embodiment of the present application obtains the position information of the target object, uses a preset feature extraction model to extract features from the position information to obtain a target feature sequence, and predicts the target feature sequence through a trained target prediction model to obtain the movement trajectory of the target object. In this way, through the target prediction model, the preset feature extraction model is used to perform feature extraction processing on the unknown information to obtain the target feature sequence for prediction processing, so as to obtain the possible future movement trajectory of the target object. In this way, the future trajectory of the target object is predicted through the target prediction model, which solves the problem that the current movement trajectory of passengers cannot be accurately predicted to ensure the service capacity of the destination. By specifically analyzing and predicting the target population, the amount of data to be analyzed is reduced, the accuracy of the prediction result is improved, and the real-time performance and effectiveness of the calculation are ensured.
[0145] Based on the foregoing embodiments, an embodiment of the present application provides an information prediction device, which can be applied to Figure 1 and the information prediction method provided by the corresponding embodiment. Referring to Figure 10 as shown, the information prediction device 3 may include: a communication interface 31, a memory 32, a processor 33, and a communication bus 34; where:
[0146] The memory 32 is used to store executable information;
[0147] The communication bus 34 is used to implement communication connections between the communication interface 31, the processor 33, and the memory 32;
[0148] The processor 33 is configured to execute the information prediction program stored in the memory 32 to implement as Figure 1The implementation process in the information prediction method provided by the corresponding embodiments will not be elaborated here.
[0149] Based on the foregoing embodiments, an embodiment of the present application provides a computer-readable storage medium, simply referred to as a storage medium. The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the implementation process in the information prediction method provided by the corresponding embodiments Figure 1 The implementation process in the information prediction method provided by the corresponding embodiments will not be elaborated here.
[0150] Based on the foregoing embodiments, an embodiment of the present application further provides a computer program product, including a computer program, which can be executed by the processor 33 of the information prediction device 3 to complete any of the foregoing method steps.
[0151] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.
[0152] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.
[0153] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.
[0154] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions for implementing the steps specified in one process or a plurality of processes and / or blocks Figure 1 one process or a plurality of processes and / or blocks Figure 1 steps for the functions specified in one block or a plurality of blocks.
[0155] As described above, only the preferred embodiments of the present application are given, and they are not intended to limit the protection scope of the present application.
Claims
1. An information prediction method, characterized in that: The method comprises: Get the location information of the target object; Using a preset feature extraction model, extracting features from the position information to obtain a target feature sequence; wherein the feature length of each feature in the target feature sequence is the same; The target feature sequence is predicted using a trained target prediction model to obtain the motion trajectory of the target object.
2. The method according to claim 1, characterized in that The step of obtaining the location information of the target object includes: Acquire the planned destination information of the target object in a future time period; Acquire location distribution information of the target object within a historical time period; wherein the location information at least includes the planned destination information and the location distribution information.
3. The method according to claim 2, characterized in that The planned destination information includes time information and location coordinate information, and the location distribution information includes stay duration information of the area determined according to the preset area information and location coordinate information of the area.
4. The method according to claim 1, characterized in that The structure of the preset feature extraction model at least includes an encoder-decoder model.
5. The method according to claim 4, characterized in that The encoder included in the encoder-decoder model includes at least two layers of BiGRU units, and the decoder includes at least one layer of BiGRU units.
6. The method according to claim 1, characterized in that The method further comprises: Obtain the sample data set to be trained; Inputting the sample data in the sample data set to be trained into the preset feature extraction model to obtain a feature sequence set to be trained; The model to be trained is trained using the feature sequence set to be trained to obtain the trained target prediction model.
7. The method according to claim 6, characterized in that The method of using the to-be-trained feature sequence set to perform model training on the to-be-trained model to obtain the trained target prediction model comprises: Acquire a feature sequence to be input from the feature sequence set to be trained; Inputting the feature sequence to be input into the feature learning layer of the model to be trained to obtain a learning feature result; Inputting the learning feature result into the attention layer of the model to be trained to obtain a weight coefficient corresponding to the learning feature result; Performing weighted analysis on the learning feature results and corresponding weight coefficients through the output layer of the model to be trained to obtain a prediction result corresponding to the feature sequence to be input; Using a preset loss function to calculate the actual result corresponding to the feature sequence to be input and the corresponding predicted result, to obtain a loss value corresponding to the feature sequence to be input; If the loss value is greater than a preset threshold, back-propagating the loss value in the model to be trained, adjusting the parameters of the model to be trained, and obtaining a first training model; After the model to be trained is updated to the first training model, the step of obtaining a feature sequence to be input from the feature sequence set to be trained is executed until the target prediction model corresponding to the loss value being less than or equal to the preset threshold is obtained; If the loss value is less than or equal to the preset threshold, the target prediction model is determined to be the model to be trained.
8. The method according to claim 7, characterized in that The feature learning layer includes at least a double layer of random matrix units SRU.
9. The method according to claim 7, characterized in that: The preset loss function is: Among them, y i is the prediction result of the i-th sample in the feature sequence to be input, p i is the actual result of the i-th sample in the feature sequence to be input, and N is the number of samples included in the feature sequence to be input.
10. The method according to claim 1, characterized in that The method further comprises: Based on the movement trajectory, counting the population at the corresponding location; and / or, Based on the motion trajectory, information is recommended for the target object.
11. An information prediction device, characterized in that: The device at least comprises: an acquisition unit, an extraction unit and a prediction unit; wherein: The acquisition unit is used to acquire the location information of the target object; The extraction unit is used to extract features from the position information using a preset feature extraction model to obtain a target feature sequence; wherein the feature length of each feature in the target feature sequence is the same; The prediction unit is used to predict the target feature sequence through a trained target prediction model to obtain the motion trajectory of the target object.
12. An information prediction device, characterized in that: The device at least includes: a communication interface, a memory, a processor and a communication bus; wherein: The memory is used to store executable instructions; The communication bus is used to realize the communication connection between the communication interface, the processor and the memory; The processor is used to execute the information prediction program stored in the memory to implement the steps of the information prediction method according to any one of claims 1 to 10.
13. A storage medium, characterized in that: The storage medium stores an information prediction program, which is used to implement the steps of the information prediction method according to any one of claims 1 to 10 when executed.
14. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the computer program implements the steps of the information prediction method according to any one of claims 1 to 10.