CN112396254A - Destination prediction method, destination prediction device, destination prediction medium, and electronic device - Google Patents
Destination prediction method, destination prediction device, destination prediction medium, and electronic device Download PDFInfo
- Publication number
- CN112396254A CN112396254A CN202011459957.4A CN202011459957A CN112396254A CN 112396254 A CN112396254 A CN 112396254A CN 202011459957 A CN202011459957 A CN 202011459957A CN 112396254 A CN112396254 A CN 112396254A
- Authority
- CN
- China
- Prior art keywords
- information
- vehicle
- target
- historical
- destination
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000012549 training Methods 0.000 claims description 102
- 230000006399 behavior Effects 0.000 claims description 64
- 239000013598 vector Substances 0.000 claims description 24
- 238000012545 processing Methods 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 12
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 7
- 235000008694 Humulus lupulus Nutrition 0.000 claims description 3
- 230000010365 information processing Effects 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 17
- 238000004891 communication Methods 0.000 description 11
- 230000008569 process Effects 0.000 description 7
- 238000000638 solvent extraction Methods 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 4
- 230000005236 sound signal Effects 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
- G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman problem
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The present disclosure relates to a destination prediction method, apparatus, medium, and electronic device, the method comprising: acquiring target characteristic information related to a target vehicle, wherein the target characteristic information comprises current random characteristic information and target historical behavior track information of the target vehicle, and the current random characteristic information comprises current holiday information, current passing time information, current vehicle restriction information and current weather information; and inputting the target characteristic information into a destination prediction model to obtain the predicted destination information of the target vehicle output by the destination prediction model. By the technical scheme, the characteristic information of the destination prediction model during prediction is more comprehensive, so that the predicted destination information obtained by the destination prediction model is more accurate.
Description
Technical Field
The present disclosure relates to the field of computer technologies, and in particular, to a destination prediction method, apparatus, medium, and electronic device.
Background
The vehicle is a common vehicle in people's daily life, and with the development of vehicle technology, the degree of intelligence of the vehicle is higher and higher, for example, when a user prepares to go out, the destination of the user can be predicted, so that reference information is provided for the user. In the related art, the destination is usually predicted according to the historical track information of the vehicle, and the prediction accuracy is low, so that an accurate destination prediction result cannot be provided for a user.
Disclosure of Invention
The purpose of the present disclosure is to provide a destination prediction method, apparatus, medium, and electronic device, so as to improve the accuracy of a destination prediction result.
In order to achieve the above object, in a first aspect, the present disclosure provides a destination prediction method, the method including: acquiring target characteristic information related to a target vehicle, wherein the target characteristic information comprises current random characteristic information and target historical behavior track information of the target vehicle, and the current random characteristic information comprises current holiday information, current passing time information, current vehicle restriction information and current weather information; and inputting the target characteristic information into a destination prediction model to obtain the predicted destination information of the target vehicle output by the destination prediction model.
Optionally, the destination prediction model is trained by: acquiring original training data required by training the destination prediction model, wherein the original training data comprises historical behavior track information, historical destination information and historical random characteristic information related to the historical behavior track information of a plurality of vehicles, and the historical random characteristic information comprises historical holiday information, historical traffic time information, historical vehicle restriction information and historical weather information; and training a model through the original training data to obtain the destination prediction model.
Optionally, the training a model by the raw training data includes: preprocessing the original training data to obtain target training data; storing the target training data into a dynamic knowledge graph; and training the model through the target training data stored in the dynamic knowledge graph.
Optionally, the training the model by the target training data stored in the dynamic knowledge-graph includes: for each vehicle, according to the historical behavior track information of the vehicle, carrying out stroke division on the historical behavior track of the vehicle to obtain the historical stroke track information of the vehicle; respectively processing the historical travel track information of each vehicle by adopting a Skip-Gram model in a Word2vec model to obtain travel track characteristic vector information of the vehicles; and taking the travel track characteristic vector information of the vehicle and historical random characteristic information related to the historical behavior track information of the vehicle as the input of the model, taking the historical destination information of the vehicle as the target output of the model, and training the model.
Optionally, the performing, according to the historical behavior trajectory information of the vehicle, a trip division on the historical behavior trajectory of the vehicle includes: acquiring a plurality of distance segments which are passed by the vehicle from historical starting point information to historical destination information; determining a time length threshold value corresponding to each route segment; and for each route segment, under the condition that the time length of the vehicle passing through the route segment is greater than a corresponding time length threshold value, dividing the historical behavior track of the vehicle in the route segment.
Optionally, the determining a duration threshold corresponding to each of the hops includes: for each of the segments, determining a preset time period according to the time when the vehicle passes through the segment; and determining the time length threshold value according to the time length required by other vehicles to pass through the distance in the preset time period.
Optionally, the destination prediction model includes a plurality of SRU units, where the SRU units perform calculation by using an SDZ algorithm, and the generation of the memory unit state and the generation of the hidden layer state in the SRU units both use the SDZ algorithm.
In a second aspect, the present disclosure provides a destination prediction apparatus, the apparatus comprising: the system comprises a target characteristic information acquisition module, a target characteristic information acquisition module and a target characteristic information processing module, wherein the target characteristic information acquisition module is used for acquiring target characteristic information related to a target vehicle, the target characteristic information comprises current random characteristic information and target historical behavior track information of the target vehicle, and the current random characteristic information comprises current holiday information, current traffic time information, current vehicle restriction information and current weather information; and the input module is used for inputting the target characteristic information into a destination prediction model to obtain the predicted destination information of the target vehicle output by the destination prediction model.
Optionally, the destination prediction model is obtained by training a training device of the destination prediction model, where the training device of the destination prediction model includes: a training data obtaining module, configured to obtain original training data required for training the destination prediction model, where the original training data includes historical behavior trajectory information, historical destination information, and historical random feature information related to the historical behavior trajectory information of each of the multiple vehicles, and the historical random feature information includes historical holiday information, historical transit time information, historical vehicle restriction information, and historical weather information; and the training module is used for training a model through the original training data to obtain the destination prediction model.
Optionally, the training module comprises: the preprocessing submodule is used for preprocessing the original training data to obtain target training data; the storage submodule is used for storing the target training data into a dynamic knowledge graph; and the training submodule is used for training the model through the target training data stored in the dynamic knowledge graph.
Optionally, the training submodule includes: the trip dividing submodule is used for carrying out trip dividing on the historical behavior track of the vehicle according to the historical behavior track information of the vehicle so as to obtain the historical trip track information of the vehicle; the processing submodule is used for respectively processing the historical travel track information of each vehicle by adopting a Skip-Gram model in a Word2vec model to obtain travel track characteristic vector information of the vehicles; and the model training submodule is used for taking the travel track characteristic vector information of the vehicle and historical random characteristic information related to the historical behavior track information of the vehicle as the input of the model, taking the historical destination information of the vehicle as the target output of the model and training the model.
Optionally, the run-length division submodule includes: the acquisition submodule is used for acquiring a plurality of stretches which are passed by the vehicle from the historical starting point information to the historical destination information; the determining submodule is used for determining a time length threshold value corresponding to each route segment; and the partitioning submodule is used for partitioning the historical behavior track of the vehicle in each route segment under the condition that the time length of the vehicle passing through the route segment is greater than the corresponding time length threshold value.
Optionally, the determining sub-module is configured to: for each of the segments, determining a preset time period according to the time when the vehicle passes through the segment; and determining the time length threshold value according to the time length required by other vehicles to pass through the distance in the preset time period.
In a third aspect, the present disclosure provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method provided by the first aspect of the present disclosure.
In a fourth aspect, the present disclosure provides an electronic device comprising: a memory having a computer program stored thereon; a processor for executing the computer program in the memory to implement the steps of the method provided by the first aspect of the present disclosure.
According to the technical scheme, the target characteristic information related to the target vehicle is obtained, the target characteristic information is input into the destination prediction model, and the predicted destination information of the target vehicle output by the destination prediction model is obtained. The target characteristic information may include current random characteristic information, and since the driving track and the possible destination information of the target vehicle are influenced in many ways, the current random characteristic information is synthesized in the present disclosure when the destination of the target vehicle is predicted, and the current random characteristic information may include current holiday information, current transit time information, current vehicle restriction information, and current weather information. Because the characteristic information of the destination prediction model during prediction is more comprehensive, the predicted destination information obtained through the destination prediction model is more accurate and closer to the actual destination of the user of the target vehicle, and the user requirements are better met.
Additional features and advantages of the disclosure will be set forth in the detailed description which follows.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure and are incorporated in and constitute a part of this specification, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure without limiting the disclosure. In the drawings:
fig. 1 is a flow chart illustrating a method of destination prediction according to an exemplary embodiment.
FIG. 2 is a flow diagram illustrating a method of training a destination predictive model in accordance with an exemplary embodiment.
Fig. 3 is a flow chart illustrating a method of training a destination predictive model according to another exemplary embodiment.
FIG. 4 is a flow diagram illustrating a method of training a model with target training data stored in a dynamic knowledge graph in accordance with an exemplary embodiment.
FIG. 5 is a flow chart illustrating a method of trip partitioning a historical behavior trace of a vehicle according to an exemplary embodiment.
FIG. 6 is a diagram illustrating processing of historical trip trajectory information by a Skip-Gram model in a Word2vec model in accordance with an exemplary embodiment.
Fig. 7 is a schematic diagram illustrating an SRU unit according to an exemplary embodiment.
Fig. 8 is a schematic diagram illustrating a structure of a destination prediction model according to an example embodiment.
FIG. 9 is a schematic diagram illustrating a method of training a model according to an exemplary embodiment.
Fig. 10 is a block diagram illustrating a destination prediction apparatus according to an example embodiment.
FIG. 11 is a block diagram illustrating an electronic device in accordance with an example embodiment.
FIG. 12 is a block diagram illustrating an electronic device in accordance with an example embodiment.
Detailed Description
The following detailed description of specific embodiments of the present disclosure is provided in connection with the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating the present disclosure, are given by way of illustration and explanation only, not limitation.
Fig. 1 is a flow chart illustrating a destination prediction method according to an exemplary embodiment, which may be applied to an electronic device having a processing capability, such as a terminal or a server. As shown in fig. 1, the method may include S101 and S102.
In S101, target feature information related to the target vehicle is acquired.
The target vehicle may be any vehicle that requires destination prediction. The target characteristic information may include current random characteristic information, target historical behavior trace information of the target vehicle. Wherein the target historical behavior trace information of the target vehicle may include travel trace information of the target vehicle over a historical period. The current random characteristic information may include current holiday information, current transit time information, current vehicle restriction information, current weather information.
The behavior trajectory of a vehicle is affected in many ways, for example, on weekdays the destination of a user is usually the work location, while on saturday and statutory holidays the probability that the destination of a user is the work location is less. For another example, because a traffic department may take a current limiting measure for a specific road, traffic limiting measures may be different at different road sections and different times, and the driving track of a vehicle may also exhibit different characteristics under the influence of the traffic limiting measures. Furthermore, the travel trajectory of the vehicle is also affected by the transit time, for example during early peak hours and late peak hours, the travel trajectory of the vehicle may be different from other hours. In addition, the vehicle restriction information also affects the driving track of the vehicle, for example, many roads in a city restrict the passing of large vehicles, the driving track of the large vehicle may be different from that of a common car due to the restriction of the road passing rule, and the possibility that the destination of the large vehicle is a certain place in the city is small. In addition, in bad weather, under the influence of weather, the vehicle generally passes slowly, and the driving track of the vehicle may be different from that in normal times.
In summary, the driving track and the possible destination information of the target vehicle are influenced in many ways, and in the present disclosure, when the destination of the target vehicle is predicted, the current random feature information is integrated, so that the considered feature information is more comprehensive.
In S102, the target feature information is input to the destination prediction model, and the predicted destination information of the target vehicle output by the destination prediction model is obtained.
The destination prediction model may be a network model trained in advance for predicting the destination of the vehicle. The destination prediction model can predict the destination of the target vehicle according to the current random characteristic information, and the destination prediction model integrates the current random characteristic information during prediction, so that the considered characteristic information is more comprehensive, and the predicted destination information obtained by the destination prediction model is more accurate, is closer to the actual destination of the user of the target vehicle and better meets the requirements of the user.
According to the technical scheme, the target characteristic information related to the target vehicle is obtained, the target characteristic information is input into the destination prediction model, and the predicted destination information of the target vehicle output by the destination prediction model is obtained. The target characteristic information may include current random characteristic information, and since the driving track and the possible destination information of the target vehicle are influenced in many ways, the current random characteristic information is synthesized in the present disclosure when the destination of the target vehicle is predicted, and the current random characteristic information may include current holiday information, current transit time information, current vehicle restriction information, and current weather information. Because the characteristic information of the destination prediction model during prediction is more comprehensive, the predicted destination information obtained through the destination prediction model is more accurate and closer to the actual destination of the user of the target vehicle, and the user requirements are better met.
The following describes a method for training a destination prediction model in the present disclosure. Fig. 2 is a flowchart illustrating a method of training a destination prediction model according to an exemplary embodiment, and as shown in fig. 2, the method may include S201 and S202.
In S201, raw training data required for training the destination prediction model is acquired.
The raw training data may include historical behavior trace information, historical destination information, and historical random feature information associated with the historical behavior trace information for each of the plurality of vehicles. The plurality of vehicles may be any of a plurality of vehicles selected in advance, and may include, for example, the target vehicle described above, historical behavior trajectory information of the vehicle, that is, travel trajectory information of the vehicle in a historical period, and historical destination information of the vehicle, that is, information of a destination to which the vehicle has arrived in the historical period. The historical random characteristic information may include historical holiday information, historical transit time information, historical vehicle restriction information, and historical weather information.
Illustratively, for example, the historical behavior trace information of the vehicle 1 includes that the vehicle 1 passes through the intersection R in the historical period, and the historical random feature information related to the historical behavior trace information may include holiday information, transit time information, vehicle restriction information, weather information when the vehicle 1 passes through the intersection R.
In S202, the model is trained by the raw training data to obtain a destination prediction model.
Since the driving track of the vehicle and the possible destination information are influenced by various factors, the accuracy and the robustness of the trained destination prediction model can be improved by using the historical random characteristic information related to the historical behavior track information as training data in the disclosure.
Fig. 3 is a flowchart illustrating a method for training a destination prediction model according to another exemplary embodiment, and as shown in fig. 3, the method may include S301 to S304, and the above-mentioned S202 may include S302 to S304.
In S301(201), raw training data necessary for training the destination prediction model is acquired.
The embodiment of step S301 may refer to S201.
In S302, the original training data is preprocessed to obtain target training data.
For example, the operation of preprocessing the raw training data may include at least one of: the data format is unified, missing data is supplemented, isolated data which cannot be associated with other data is deleted, and data is deduplicated. The original training data is preprocessed, so that the integrity and the accuracy of the training data can be improved.
In S303, the target training data is stored into the dynamic knowledge-graph.
In the related art, it is common to store the historical behavior trace information of the vehicle in a relational database, and to associate the data by a table-to-table linking operation. However, the vehicle track dynamically changes with time, and the behavior track information of the vehicle is data with time attribute and with time sequence, for example, the vehicle historical behavior track information includes license plate number, owner name, passing time, passing place. The relational database is adopted to store the time-sequence characteristic information, when the complete historical behavior track information of the vehicle is obtained, different data need to be obtained from a plurality of tables respectively, the operation is complex, the speed is slow when the data is called, and the whole behavior track of the vehicle is difficult to accurately restore.
According to the method and the device, the target training data are stored in the dynamic knowledge graph, the dynamic knowledge graph is a graph formed by nodes and relations with time attributes, pull-through and modeling of different types of feature data can be achieved, the flexibility is high, the dynamic knowledge graph has self-optimization and self-adaption capabilities, and historical behavior tracks and historical random feature information of the vehicles with time sequence can be recorded more accurately.
In S304, the model is trained by the target training data stored in the dynamic knowledge map.
By the scheme, the target training data are stored in the dynamic knowledge graph, required data are obtained from the dynamic knowledge graph when the model is trained, data do not need to be obtained from a plurality of tables respectively in the related technology, and the data calling speed can be improved. And moreover, the dynamic knowledge graph is adopted to store target training data, so that the time sequence characteristics of the vehicle behavior track information are better matched.
FIG. 4 is a flow diagram illustrating a method of training a model with target training data stored in a dynamic knowledge graph in accordance with an exemplary embodiment. As shown in fig. 4, an exemplary implementation of S304 may include S401 to S403.
In S401, for each vehicle, the historical behavior trajectory of the vehicle is divided into routes according to the historical behavior trajectory information of the vehicle, so as to obtain the historical route trajectory information of the vehicle.
The method comprises the steps of carrying out stroke division on a historical behavior track of a vehicle, namely dividing a running track of the vehicle into a plurality of stroke sections.
An exemplary embodiment of this step S401 may be as shown in fig. 5, including S4011 to S4013.
In S4011, a plurality of hops that the vehicle passes in the middle from the history start point information to the history destination information are acquired.
Illustratively, three preset position points S1, S2 and S3 are included among the history start point information and the history destination information, and the preset position points may be preset, for example, preset position points are set according to road intersections, each intersection being a preset position point. Accordingly, the plurality of roadbed segments may include a leg from the historical origin to S1, a leg from S1 to S2, a leg from S2 to S3, and a leg from S3 to the historical destination.
In S4012, a respective duration threshold value for each link is determined.
Alternatively, the step S4012 may include: for each segment, determining a preset time period according to the time when the vehicle passes through the segment; the duration threshold is determined based on the duration of time required for other vehicles to traverse the segment within a preset time period.
For example, for the stretch from S1 to S2, the time at which the vehicle passes through the stretch may be the time at which the vehicle passes the S1 location point, and for example, a time period of a preset length of time before the time may be the preset time period. For example, the time when the vehicle passes through the road segment is 12:00, and the preset time period can be 11: 55-12: 00. For example, the average value of the time lengths required by other vehicles to pass through the distance in the preset time period may be used as a time length threshold corresponding to the distance, and the time lengths of the vehicles to pass through the distance may be compared with the time length threshold, so as to determine whether the historical behavior track of the vehicle needs to be divided in the distance.
Due to the influence of holidays, traffic restriction measures, rush hour periods and other factors, the time required for the vehicle to pass through the same route segment at different times is often different, for example, the time required for the vehicle to pass through the route segment at an early peak is relatively long, and the time required for the vehicle to pass through the route segment at other time periods is relatively short, so that the accuracy is not high enough if the uniform threshold value is taken as a standard to judge whether the vehicle travel needs to be divided at the route segment. In the disclosure, the preset time period is determined according to the time when the vehicle passes through the segment, and the time threshold is determined according to the time required by other vehicles to pass through the segment in the preset time period, so that the determined time threshold is adaptive to the passing time of the vehicle.
In S4013, for each of the stretches, when the time period during which the vehicle passes through the stretch is greater than the corresponding time period threshold, the historical behavior trajectory of the vehicle is divided at the stretch.
If the length of time that the vehicle has traveled through the stretch is greater than the corresponding length of time threshold, indicating that the vehicle may have been stationary for a period of time within the stretch, the historical behavior trace of the vehicle is partitioned for the stretch. For example, for the distance from S1 to S2, the time period for the vehicle to pass through the distance is 20 minutes, the time period threshold corresponding to the distance is 10 minutes, which may indicate that the time for the vehicle to pass through the distance is long, and the vehicle may stay for a certain period of time, then the historical behavior track of the vehicle is divided in the distance, and the historical travel track information of the vehicle obtained by performing the travel division may include a plurality of track points.
In S402, the Skip-Gram model in the Word2vec model is adopted to process the historical travel track information of each vehicle respectively to obtain the travel track characteristic vector information of the vehicle.
FIG. 6 is a diagram illustrating processing of historical trip trajectory information by a Skip-Gram model in a Word2vec model in accordance with an exemplary embodiment.
As shown in FIG. 6, { x1,x2,x3,...,xk,...,xvThe track points expressed by the method represent the coding mode of one-hot representation (one-hot), wherein R represents the coding mode of one-hot representation (one-hot) of the track points expressed by the methodvIs the set of all track points, v represents the number of the sets, k represents the kth track point; w is formed as RV×NA vector matrix of size V N, the transverse vector h of each dimension representing a locus point, is formed by{x1,x2,x3,...,xk,...,xvMultiplying the vector matrix W can convert the track points coded by the one-hot representation (one-hot) into a vector representation h ═ h1,...,hi,...,hN-1,hN}，h∈RNIs N. According to the principle of the Skip-Gram model, any track point in the track is adopted to represent the adjacent track point of the track point, namely, the more similar the vectors of the adjacent track points in the vehicle track are, the W' is equal to RN×VThe vector transformation matrix representing the current track point and the adjacent track point has the dimension size of N multiplied by V, and the adjacent track point y is obtained by converting the result obtained by multiplying the track point vector h and the vector matrix W' through softmax1,y2,ycAnd C represents the C-th adjacent track point. The Skip-Gram model can refer to the related art for specific ways of processing the historical travel track information.
The Skip-Gram model calculates the similarity between the input vector and the output vector by sampling the sequence of the context environment, and is normalized by the level softmax. The discrete track points are subjected to vector characterization by using a Skip-Gram model, if the discrete track points are adjacent position points, the vectors are similar, and the processing of the travel track information is realized.
In S403, the travel route feature vector information of the vehicle and the historical random feature information related to the historical behavior trajectory information of the vehicle are input as the model, the historical destination information of the vehicle is output as the target of the model, and the model is trained.
In the present disclosure, a plurality of SRU (Simple loop Unit) units are included, a ground prediction model of an SRU Unit may be calculated by using an SDZ (singular-driver Zoneout) algorithm between packages, and the generation of the memory Unit state and the generation of the hidden layer state in the SRU Unit both use the SDZ algorithm.
Because the behavior tracks of the vehicle have the sequence and the sequential behaviors have the interdependence, the recurrent neural network has very good effect in processing the sequence of the time sequence and can effectively capture the interdependence between the sequential behaviors. Therefore, the sequence and dependency of the sequential behaviors of the behavior sequence can be well captured by adopting the SRU unit, the problems of gradient loss and explosion of the traditional recurrent neural network are solved, and parallel computation can be realized.
Fig. 7 is a schematic structural diagram of an SRU unit according to an exemplary embodiment, and as shown in fig. 7, the calculation process and relationship between the memory unit state and the hidden layer state of the SRU unit are as follows:
ft＝σ(Wf*xt+bf)
rt＝σ(Wr*xt+br)
ct＝ftΘct-1+(1-ft)Θxt
ht＝rtΘg(ct)+(1-rt)Θxt
wherein the forgetting gate outputs ftDepending on the current input trajectory x at time tt，WfWeight ratio representing forgetting gate, bfIndicating the deviation of the forgotten door at time t by aligning with the track xtThe linear change result of (2) is output of the forgetting gate obtained by the sigma function; reset gate rtAccording to the current input track xtBy inputting the trajectory x for time ttIs linearly varied and then the output of the reset gate, W, is obtained through the sigma functionrRepresents the weight proportion of the reset gate, brRepresents the offset of the reset gate at time t; memory cell state c at time t of simple cycle unit (SRU)tIs dependent on the output f of the forgetting gatetInput trajectory xtAnd memory cell state c at time t-1t-1Θ denotes the inner product between vectors and forgets the output f of the thresholdtAgain depending on the data x at that timetSo that the generation of the state of the memory cell depends only on the input x at the current momentt. Hidden layer state htThe output state of the current time t is represented, wherein g represents an activation function; the SRU unit eliminates and relieves the dependency relationship between the statesThe timing dependence of the various thresholds is such thattAnd rtThe calculation of the method can be processed in parallel, and the efficiency of model training and prediction is improved.
SDZ (Surrisal-drive Zoneout) is a method for self-adapting Zoneout probability, solves the problem that Zoneout specifies fixed probability, and improves the robustness of a neural network to disturbance in a hidden state in the training process. The unit activation in the SDZ algorithm is to randomly replace the activation of the previous step with a certain probability, and partial parameters are updated in the process of updating the weight, so that the training data can be improved to a certain degree.
Fig. 8 is a schematic diagram illustrating a structure of a destination prediction model according to an example embodiment. As shown in fig. 8, X1, X2, and X3 are model input data, and fig. 8 only illustrates these three data, and the number of model input data in actual application is not limited to this. And generating the memory unit state in the SRU unit layer and the hidden layer state by adopting an SDZ algorithm. The memory state of the same layer of the SRU unit comprises two parts: memory cell state c generated by SRUtAnd memory cell state at previous time
With simultaneous use of variable ZtTo control the pair of memory cell states and the new memory cell state
The influence of (c). Meanwhile, the SRU units adopt the SDZ algorithm to calculate, so that the relation between layers of the hidden layer state is enhanced, the problem of long track dependence in destination prediction can be solved by adding the SDZ algorithm, the accuracy of the model is ensured, and the model calculation process can be shown in the following formula:
where Θ represents the inner product between vectors, and the state of the memory cell at time t
Is memory cell state c at time ttAnd memory cell state at previous time
By using the variable ZtControl is carried out to obtain the hidden state at the time t
By using the variable ZtControlling hidden state h at time ttAnd memory cell state at previous time
And (4) proportional composition.
According to the method, historical random characteristic information related to historical behavior track information is fused as input of a model, such as historical holiday information, historical transit time information, historical vehicle restriction information, historical weather information and the like, historical destination information of a vehicle is output as a target of the model, the model is trained, random factors influencing vehicle track information are considered, and accuracy, robustness and operation efficiency of the model are improved.
FIG. 9 is a schematic diagram illustrating a model training process according to an exemplary embodiment, wherein the training process is described in detail above and is not repeated here.
Based on the same inventive concept, the present disclosure also provides a destination prediction apparatus, and fig. 10 is a block diagram of a destination prediction apparatus according to an exemplary embodiment, as shown in fig. 10, the apparatus 1000 may include:
a target characteristic information obtaining module 1001, configured to obtain target characteristic information related to a target vehicle, where the target characteristic information includes current random characteristic information and target historical behavior trajectory information of the target vehicle, and the current random characteristic information includes current holiday information, current transit time information, current vehicle restriction information, and current weather information;
the input module 1002 is configured to input the target feature information into a destination prediction model, so as to obtain predicted destination information of the target vehicle output by the destination prediction model.
Optionally, the destination prediction model is obtained by training a training device of the destination prediction model, where the training device of the destination prediction model includes: a training data obtaining module, configured to obtain original training data required for training the destination prediction model, where the original training data includes historical behavior trajectory information, historical destination information, and historical random feature information related to the historical behavior trajectory information of each of the multiple vehicles, and the historical random feature information includes historical holiday information, historical transit time information, historical vehicle restriction information, and historical weather information; and the training module is used for training a model through the original training data to obtain the destination prediction model.
Optionally, the training module comprises: the preprocessing submodule is used for preprocessing the original training data to obtain target training data; the storage submodule is used for storing the target training data into a dynamic knowledge graph; and the training submodule is used for training the model through the target training data stored in the dynamic knowledge graph.
Optionally, the training submodule includes: the trip dividing submodule is used for carrying out trip dividing on the historical behavior track of the vehicle according to the historical behavior track information of the vehicle so as to obtain the historical trip track information of the vehicle; the processing submodule is used for respectively processing the historical travel track information of each vehicle by adopting a Skip-Gram model in a Word2vec model to obtain travel track characteristic vector information of the vehicles; and the model training submodule is used for taking the travel track characteristic vector information of the vehicle and historical random characteristic information related to the historical behavior track information of the vehicle as the input of the model, taking the historical destination information of the vehicle as the target output of the model and training the model.
Optionally, the run-length division submodule includes: the acquisition submodule is used for acquiring a plurality of stretches which are passed by the vehicle from the historical starting point information to the historical destination information; the determining submodule is used for determining a time length threshold value corresponding to each route segment; and the partitioning submodule is used for partitioning the historical behavior track of the vehicle in each route segment under the condition that the time length of the vehicle passing through the route segment is greater than the corresponding time length threshold value.
Optionally, the determining sub-module is configured to: for each of the segments, determining a preset time period according to the time when the vehicle passes through the segment; and determining the time length threshold value according to the time length required by other vehicles to pass through the distance in the preset time period.
With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.
Fig. 11 is a block diagram illustrating an electronic device 700 in accordance with an example embodiment. As shown in fig. 11, the electronic device 700 may include: a processor 701 and a memory 702. The electronic device 700 may also include one or more of a multimedia component 703, an input/output (I/O) interface 704, and a communication component 705.
The processor 701 is configured to control the overall operation of the electronic device 700, so as to complete all or part of the steps in the destination prediction method. The memory 702 is used to store various types of data to support operation at the electronic device 700, such as instructions for any application or method operating on the electronic device 700 and application-related data, such as contact data, transmitted and received messages, pictures, audio, video, and the like. The Memory 702 may be implemented by any type of volatile or non-volatile Memory device or combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic Memory, flash Memory, magnetic disk, or optical disk. The multimedia components 703 may include screen and audio components. Wherein the screen may be, for example, a touch screen and the audio component is used for outputting and/or inputting audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may further be stored in the memory 702 or transmitted through the communication component 705. The audio assembly also includes at least one speaker for outputting audio signals. The I/O interface 704 provides an interface between the processor 701 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 705 is used for wired or wireless communication between the electronic device 700 and other devices. Wireless Communication, such as Wi-Fi, bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IOT, eMTC, or other 5G, etc., or a combination of one or more of them, which is not limited herein. The corresponding communication component 705 may thus include: Wi-Fi module, Bluetooth module, NFC module, etc.
In an exemplary embodiment, the electronic Device 700 may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components for performing the destination prediction method described above.
In another exemplary embodiment, a computer readable storage medium comprising program instructions which, when executed by a processor, implement the steps of the destination prediction method described above is also provided. For example, the computer readable storage medium may be the memory 702 described above including program instructions executable by the processor 701 of the electronic device 700 to perform the destination prediction method described above.
Fig. 12 is a block diagram illustrating an electronic device 1900 according to an example embodiment. For example, the electronic device 1900 may be provided as a server. Referring to fig. 12, an electronic device 1900 includes a processor 1922, which may be one or more in number, and a memory 1932 for storing computer programs executable by the processor 1922. The computer program stored in memory 1932 may include one or more modules that each correspond to a set of instructions. Further, the processor 1922 may be configured to execute the computer program to perform the destination prediction method described above.
Additionally, electronic device 1900 may also include a power component 1926 and a communication component 1950, the power component 1926 may be configured to perform power management of the electronic device 1900, and the communication component 1950 may be configured to enable communication, e.g., wired or wireless communication, of the electronic device 1900. In addition, the electronic device 1900 may also include input/output (I/O) interfaces 1958. The electronic device 1900 may operate based on an operating system, such as Windows Server, stored in memory 1932TM，Mac OS XTM，UnixTM，LinuxTMAnd so on.
In another exemplary embodiment, a computer readable storage medium comprising program instructions which, when executed by a processor, implement the steps of the destination prediction method described above is also provided. For example, the computer readable storage medium may be the memory 1932 described above that includes program instructions executable by the processor 1922 of the electronic device 1900 to perform the destination prediction method described above.
In another exemplary embodiment, a computer program product is also provided, which comprises a computer program executable by a programmable apparatus, the computer program having code portions for performing the above-mentioned destination prediction method when executed by the programmable apparatus.
The preferred embodiments of the present disclosure are described in detail with reference to the accompanying drawings, however, the present disclosure is not limited to the specific details of the above embodiments, and various simple modifications may be made to the technical solution of the present disclosure within the technical idea of the present disclosure, and these simple modifications all belong to the protection scope of the present disclosure.
It should be noted that, in the foregoing embodiments, various features described in the above embodiments may be combined in any suitable manner, and in order to avoid unnecessary repetition, various combinations that are possible in the present disclosure are not described again.
In addition, any combination of various embodiments of the present disclosure may be made, and the same should be considered as the disclosure of the present disclosure, as long as it does not depart from the spirit of the present disclosure.
Claims (10)
1. A method of destination prediction, the method comprising:
acquiring target characteristic information related to a target vehicle, wherein the target characteristic information comprises current random characteristic information and target historical behavior track information of the target vehicle, and the current random characteristic information comprises current holiday information, current passing time information, current vehicle restriction information and current weather information;
and inputting the target characteristic information into a destination prediction model to obtain the predicted destination information of the target vehicle output by the destination prediction model.
2. The method of claim 1, wherein the destination prediction model is trained by:
acquiring original training data required by training the destination prediction model, wherein the original training data comprises historical behavior track information, historical destination information and historical random characteristic information related to the historical behavior track information of a plurality of vehicles, and the historical random characteristic information comprises historical holiday information, historical traffic time information, historical vehicle restriction information and historical weather information;
and training a model through the original training data to obtain the destination prediction model.
3. The method of claim 2, wherein training the model with the raw training data comprises:
preprocessing the original training data to obtain target training data;
storing the target training data into a dynamic knowledge graph;
and training the model through the target training data stored in the dynamic knowledge graph.
4. The method of claim 3, wherein the training of the model with the target training data stored in the dynamic knowledge-graph comprises:
for each vehicle, according to the historical behavior track information of the vehicle, carrying out stroke division on the historical behavior track of the vehicle to obtain the historical stroke track information of the vehicle;
respectively processing the historical travel track information of each vehicle by adopting a Skip-Gram model in a Word2vec model to obtain travel track characteristic vector information of the vehicles;
and taking the travel track characteristic vector information of the vehicle and historical random characteristic information related to the historical behavior track information of the vehicle as the input of the model, taking the historical destination information of the vehicle as the target output of the model, and training the model.
5. The method of claim 4, wherein the step of performing trip division on the historical behavior track of the vehicle according to the historical behavior track information of the vehicle comprises:
acquiring a plurality of distance segments which are passed by the vehicle from historical starting point information to historical destination information;
determining a time length threshold value corresponding to each route segment;
and for each route segment, under the condition that the time length of the vehicle passing through the route segment is greater than a corresponding time length threshold value, dividing the historical behavior track of the vehicle in the route segment.
6. The method of claim 5, wherein said determining a respective duration threshold for each of said hops comprises:
for each of the segments, determining a preset time period according to the time when the vehicle passes through the segment; and determining the time length threshold value according to the time length required by other vehicles to pass through the distance in the preset time period.
7. The method of any of claims 1-6, wherein the destination prediction model comprises a plurality of SRU units, wherein the SRU units are computed using an SDZ algorithm, and wherein the generation of memory unit states and the generation of hidden layer states within the SRU units both use the SDZ algorithm.
8. A destination prediction apparatus, characterized in that the apparatus comprises:
the system comprises a target characteristic information acquisition module, a target characteristic information acquisition module and a target characteristic information processing module, wherein the target characteristic information acquisition module is used for acquiring target characteristic information related to a target vehicle, the target characteristic information comprises current random characteristic information and target historical behavior track information of the target vehicle, and the current random characteristic information comprises current holiday information, current traffic time information, current vehicle restriction information and current weather information;
and the input module is used for inputting the target characteristic information into a destination prediction model to obtain the predicted destination information of the target vehicle output by the destination prediction model.
9. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 7.
10. An electronic device, comprising:
a memory having a computer program stored thereon;
a processor for executing the computer program in the memory to carry out the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011459957.4A CN112396254A (en) | 2020-12-11 | 2020-12-11 | Destination prediction method, destination prediction device, destination prediction medium, and electronic device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011459957.4A CN112396254A (en) | 2020-12-11 | 2020-12-11 | Destination prediction method, destination prediction device, destination prediction medium, and electronic device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112396254A true CN112396254A (en) | 2021-02-23 |
Family
ID=74624795
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011459957.4A Pending CN112396254A (en) | 2020-12-11 | 2020-12-11 | Destination prediction method, destination prediction device, destination prediction medium, and electronic device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112396254A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113065074A (en) * | 2021-04-02 | 2021-07-02 | 大连理工大学 | Track destination prediction method based on knowledge graph and self-attention mechanism |
| CN113379304A (en) * | 2021-06-28 | 2021-09-10 | 北京百度网讯科技有限公司 | Method and device for determining electricity consumption |
| CN113821736A (en) * | 2021-09-01 | 2021-12-21 | 北京中交兴路信息科技有限公司 | Vehicle destination prediction method, device, equipment and medium based on prefix tree |
| CN115083147A (en) * | 2022-05-10 | 2022-09-20 | 广州汽车集团股份有限公司 | Destination prediction method and device and vehicle |
| CN115690163A (en) * | 2023-01-04 | 2023-02-03 | 中译文娱科技(青岛)有限公司 | Target tracking method, system and storage medium based on image content |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2009054130A1 (en) * | 2007-10-23 | 2009-04-30 | Panasonic Corporation | Destination display device and destination display method |
| CN109977283A (en) * | 2019-03-14 | 2019-07-05 | 中国人民大学 | A kind of the tourism recommended method and system of knowledge based map and user's footprint |
| CN111178628A (en) * | 2019-12-30 | 2020-05-19 | 沈阳民航东北凯亚有限公司 | Luggage arrival time prediction method and device |
| CN111369787A (en) * | 2018-12-26 | 2020-07-03 | 杭州海康威视系统技术有限公司 | Vehicle track prediction method and device and electronic equipment |
| CN112052339A (en) * | 2020-07-24 | 2020-12-08 | 广东水利电力职业技术学院（广东省水利电力技工学校） | Target track prediction method, system and storage medium based on knowledge graph |
-
2020
- 2020-12-11 CN CN202011459957.4A patent/CN112396254A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2009054130A1 (en) * | 2007-10-23 | 2009-04-30 | Panasonic Corporation | Destination display device and destination display method |
| CN111369787A (en) * | 2018-12-26 | 2020-07-03 | 杭州海康威视系统技术有限公司 | Vehicle track prediction method and device and electronic equipment |
| CN109977283A (en) * | 2019-03-14 | 2019-07-05 | 中国人民大学 | A kind of the tourism recommended method and system of knowledge based map and user's footprint |
| CN111178628A (en) * | 2019-12-30 | 2020-05-19 | 沈阳民航东北凯亚有限公司 | Luggage arrival time prediction method and device |
| CN112052339A (en) * | 2020-07-24 | 2020-12-08 | 广东水利电力职业技术学院（广东省水利电力技工学校） | Target track prediction method, system and storage medium based on knowledge graph |
Non-Patent Citations (1)
| Title |
|---|
| 张国兴: "具有长期依赖的出租车轨迹目的地预测", 《中国优秀硕士学位论文全文数据库（电子期刊）工程科技Ⅱ辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113065074A (en) * | 2021-04-02 | 2021-07-02 | 大连理工大学 | Track destination prediction method based on knowledge graph and self-attention mechanism |
| CN113379304A (en) * | 2021-06-28 | 2021-09-10 | 北京百度网讯科技有限公司 | Method and device for determining electricity consumption |
| CN113821736A (en) * | 2021-09-01 | 2021-12-21 | 北京中交兴路信息科技有限公司 | Vehicle destination prediction method, device, equipment and medium based on prefix tree |
| CN113821736B (en) * | 2021-09-01 | 2023-09-08 | 北京中交兴路信息科技有限公司 | Method, device, equipment and medium for predicting vehicle destination based on prefix tree |
| CN115083147A (en) * | 2022-05-10 | 2022-09-20 | 广州汽车集团股份有限公司 | Destination prediction method and device and vehicle |
| CN115690163A (en) * | 2023-01-04 | 2023-02-03 | 中译文娱科技(青岛)有限公司 | Target tracking method, system and storage medium based on image content |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112396254A (en) | Destination prediction method, destination prediction device, destination prediction medium, and electronic device | |
| Tekouabou et al. | Improving parking availability prediction in smart cities with IoT and ensemble-based model | |
| Tedjopurnomo et al. | A survey on modern deep neural network for traffic prediction: Trends, methods and challenges | |
| CN111862605B (en) | Road condition detection method and device, electronic equipment and readable storage medium | |
| CN113643532B (en) | Regional traffic prediction method and device | |
| Ip et al. | Vehicle trajectory prediction based on LSTM recurrent neural networks | |
| CN113159403B (en) | Intersection pedestrian track prediction method and device | |
| CN111862590A (en) | Road condition prediction method, road condition prediction device and storage medium | |
| CN114826947B (en) | Flow matrix recovery prediction method and system based on automatic encoder | |
| Wang et al. | A new traffic speed forecasting method based on bi-pattern recognition | |
| CN116698075B (en) | Road network data processing method and device, electronic equipment and storage medium | |
| Rajan et al. | Robustness generalizations of the shortest feasible path problem for electric vehicles | |
| Du et al. | Online stochastic routing incorporating real-time traffic information | |
| CN116258253A (en) | Vehicle OD prediction method based on Bayesian neural network | |
| CN115366919A (en) | Trajectory prediction method, system, electronic device and storage medium | |
| CN115660156A (en) | Method and system for predicting passenger flow congestion delay at traffic station | |
| CN113192315B (en) | Traffic flow distribution prediction method, prediction device and terminal equipment | |
| CN110942178B (en) | Charging pile recommendation method based on link prediction method of resource allocation index | |
| Panovski et al. | Public transportation prediction with convolutional neural networks | |
| Ramkumar et al. | Urban bus arrival time prediction using linear regression and kalman filter—a comparison | |
| Lam et al. | Flexible sliding windows for kernel regression based bus arrival time prediction | |
| CN113344239B (en) | Traffic congestion condition prediction method and system based on two-stage spectral clustering | |
| WO2022190827A1 (en) | Operation state estimation system and operation state estimation method | |
| CN117974407A (en) | Park traffic jam prediction method and system based on dynamic graph clustering element learning | |
| CN116109021B (en) | Travel time prediction method, device, equipment and medium based on multitask learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: No.27, 1st floor, building 14, Haidian Road, zhongjiancai District, Beijing 100096 Applicant after: Beijing PERCENT Technology Group Co.,Ltd. Address before: No.27, 1st floor, building 14, Haidian Road, zhongjiancai District, Beijing 100096 Applicant before: BEIJING BAIFENDIAN INFORMATION SCIENCE & TECHNOLOGY Co.,Ltd. |
|
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210223 |