CN110334167A - Positional shift method for early warning and device based on neural network track - Google Patents
Positional shift method for early warning and device based on neural network track Download PDFInfo
- Publication number
- CN110334167A CN110334167A CN201910559453.0A CN201910559453A CN110334167A CN 110334167 A CN110334167 A CN 110334167A CN 201910559453 A CN201910559453 A CN 201910559453A CN 110334167 A CN110334167 A CN 110334167A
- Authority
- CN
- China
- Prior art keywords
- tracing point
- positional shift
- arrived
- characterizes
- neural network
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 36
- 239000000203 mixture Substances 0.000 claims abstract description 14
- 238000004364 calculation method Methods 0.000 claims description 36
- 238000012512 characterization method Methods 0.000 claims description 26
- 238000010276 construction Methods 0.000 claims description 14
- 239000000284 extract Substances 0.000 claims description 7
- 235000013399 edible fruits Nutrition 0.000 claims description 2
- 230000008569 process Effects 0.000 description 9
- 239000011159 matrix material Substances 0.000 description 8
- 238000012544 monitoring process Methods 0.000 description 3
- 241000208340 Araliaceae Species 0.000 description 2
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 2
- 235000003140 Panax quinquefolius Nutrition 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 235000008434 ginseng Nutrition 0.000 description 2
- 238000005498 polishing Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- 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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06N3/045—Combinations of networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/029—Location-based management or tracking services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Remote Sensing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of positional shift method for early warning and device based on neural network track.The positional shift method for early warning based on neural network track includes: acquisition historical trajectory data；Position sequence is extracted from historical trajectory data, and is based on position sequence, constructs prediction model；When collect it is N number of arrived position when, arrived position and prediction model based on N number of, carry out trajectory predictions, wherein N >=2；When collecting N+1 and arrived position, the errors between the actual path and trajectory predictions result that N+1 arrived position composition are calculated, whether error in judgement meets preset error condition, if it is, carrying out positional shift alarm.Scheme provided by the invention being capable of early warning positional shift promptly and accurately.
Description
Technical field
The present invention relates to field of computer technology, in particular to the positional shift method for early warning based on neural network track and
Device.
Background technique
Currently, location information can be provided for guardian by positioning method, to prevent children or cognition problematic old
People wanders away.For example, put a locator with old man or children, guardian passes through the monitoring that is mounted on mobile phone or computer
Software can see the position of locator in real time.After old man wanders away, guardian sees where lower mobile phone just can know that old man.This is fixed
The mode of position, is only available to guardian's location information, if guardian does not see locator data for a long time, it may be found that walking
After mistake is exactly for a long time time, even if man memory power can only be relied in addition, guardian sees locator data, artificially
Locator data and historical track are compared, to judge whether the user of locator occurs positional shift.Artificial judgement
Positional shift often has that accuracy and timeliness are poor.
Summary of the invention
The embodiment of the invention provides a kind of positional shift method for early warning and device based on neural network track, Neng Gouji
When accurate early warning positional shift.
Positional shift method for early warning based on neural network track, comprising:
Acquire historical trajectory data；
Position sequence is extracted from the historical trajectory data, and is based on the position sequence, constructs prediction model；
Further include:
When collect it is N number of arrived position when, based on it is described it is N number of arrived position and the prediction model, carry out track
It predicts, wherein N >=2；
When collecting N+1 and arrived position, calculates actual paths that N+1 arrived position composition and track is pre-
The error between result is surveyed, judges whether the error meets preset error condition, if it is, carrying out positional shift report
It is alert.
It is preferably, described to extract position sequence from the historical trajectory data, comprising:
Using following first calculation formula, the region calculated in the historical trajectory data between every two tracing point is consistent
Property weight；
First calculation formula:
Wherein, coh (p, q) characterizes the region consistency weight between p-th of tracing point and q-th of tracing point；Dist (p,
Q) actual range between p-th of tracing point and q-th of tracing point is characterized；Duration (p, q) characterization from p-th of tracing point to
The time required to q-th of tracing point；δ characterizes constant；P and q is the positive integer not less than 1, and p ≠ q；
Tracing point by the region consistency weight not less than preset region threshold is combined into regional ensemble；
According to preset inquiry radius, search is located at the tracing point in the inquiry radius in the regional ensemble；
It extracts and is stopped a little from the tracing point in the inquiry radius；
Described stop a little is clustered；
According to cluster result, position sequence is determined.
Preferably, described to be based on the position sequence, construct prediction model, comprising:
Using Skip-gram algorithm, the location point in the position sequence is trained to corresponding feature vector；
By in the corresponding feature vector input length memory models of the location point, the ginseng of the length memory models is determined
Number.
Preferably, the mistake calculated between the actual path and trajectory predictions result that N+1 arrived position composition
Difference, comprising:
Using following second calculation formula, the N+1 positions that arrived in position and the trajectory predictions result are calculated
The first deviation between point；
Second calculation formula:
Wherein, dist (m, n) characterizes first deviation；M.lat and m.lon characterizes N+1 respectively arrived position
The dimension and longitude set；N.lat and n.lon characterizes the dimension and longitude of the location point in the trajectory predictions result；
It is accumulative location point of the deviation more than the error threshold in the preset error condition continuously occur.
It is preferably, described to judge whether the error meets preset error condition, comprising:
Judge whether calculated first deviation is more than error threshold in preset error condition, if it is, sentencing
Whether disconnected accumulated result is not less than the count threshold in the error condition, if it is, executing the progress positional shift report
It is alert.
Preferably, it extracts and is stopped a little in the tracing point from the inquiry radius, comprising:
Using following calculation formula groups, calculating is stopped a little；
Wherein, sp.lat characterizes the dimension stopped a little；Sp.lon characterizes the longitude stopped a little；R.lat characterization is located at described
Inquire the dimension of the tracing point in radius；R.lon characterization is located at the longitude of the tracing point in the inquiry radius；K characterization is located at
The total number of tracing point in the inquiry radius.
Preferably, the above-mentioned positional shift method for early warning based on neural network track further comprises:
Determine at least one target position；
When judging that error is unsatisfactory for preset error condition, calculating collected N+1 arrived position and every
The second deviation between one target position；
When calculated all second deviations are more than preset deviation threshold, then positional shift alarm is carried out.
Positional shift prior-warning device based on neural network track, comprising: acquisition unit, model construction unit, track are pre-
Survey unit and offset prewarning unit, wherein
The acquisition unit, for acquiring historical trajectory data；
The model construction unit, for extracting position sequence from the historical trajectory data that the acquisition unit acquires,
And it is based on the position sequence, construct prediction model；
The trajectory predictions unit, for when collect it is N number of arrived position when, based on it is described it is N number of arrived position and
The prediction model of the model construction building unit carries out trajectory predictions, wherein N >=2；
The offset prewarning unit arrived position for when collecting N+1 and arrived position, calculating N+1
The error between trajectory predictions result that the actual path of composition and the trajectory predictions unit predict, judges that the error is
It is no to meet preset error condition, if it is, carrying out positional shift alarm.
Preferably, the model construction unit, is used for:
Using following first calculation formula, the region calculated in the historical trajectory data between every two tracing point is consistent
Property weight；
First calculation formula:
Wherein, coh (p, q) characterizes the region consistency weight between p-th of tracing point and q-th of tracing point；Dist (p,
Q) actual range between p-th of tracing point and q-th of tracing point is characterized；Duration (p, q) characterization from p-th of tracing point to
The time required to q-th of tracing point；δ characterizes constant；P and q is the positive integer not less than 1, and p ≠ q；
Tracing point by the region consistency weight not less than preset region threshold is combined into regional ensemble；
According to preset inquiry radius, search is located at the tracing point in the inquiry radius in the regional ensemble；
Using following calculation formula groups, calculating is stopped a little；
Wherein, sp.lat characterizes the dimension stopped a little；Sp.lon characterizes the longitude stopped a little；R.lat characterization is located at described
Inquire the dimension of the tracing point in radius；R.lon characterization is located at the longitude of the tracing point in the inquiry radius；K characterization is located at
The total number of tracing point in the inquiry radius；
Described stop a little is clustered；
According to cluster result, position sequence is determined.
Preferably, the model construction unit, for using Skip-gram algorithm, by the position in the position sequence
Point is trained to corresponding feature vector；Described in determining in the corresponding feature vector input length memory models of the location point
The parameter of length memory models.
Preferably, the offset prewarning unit, is used for:
Using following second calculation formula, the N+1 positions that arrived in position and the trajectory predictions result are calculated
The first deviation between point；
Second calculation formula:
Wherein, dist (m, n) characterizes first deviation；M.lat and m.lon characterizes N+1 respectively arrived position
The dimension and longitude set；N.lat and n.lon characterizes the dimension and longitude of the location point in the trajectory predictions result；
It is accumulative location point of the deviation more than the error threshold in the preset error condition continuously occur；
Judge whether calculated first deviation is more than error threshold in preset error condition, if it is, sentencing
Whether disconnected accumulated result is not less than the count threshold in the error condition, if it is, executing the progress positional shift report
It is alert.
Preferably, the offset prewarning unit, is further used for:
Determine at least one target position；
When judging that error is unsatisfactory for preset error condition, calculating collected N+1 arrived position and every
The second deviation between one target position；
When calculated all second deviations are more than preset deviation threshold, then positional shift alarm is carried out.
The embodiment of the invention provides a kind of positional shift method for early warning and device based on neural network track, this is based on
The positional shift method for early warning of neural network track, comprising: acquisition historical trajectory data；Position is extracted from historical trajectory data
Sequence, and it is based on position sequence, construct prediction model；When collect it is N number of arrived position when, based on it is N number of arrived position and
The prediction model carries out trajectory predictions, wherein N >=2；When collecting N+1 and arrived position, calculates N+1 and has arrived
Whether the error between the actual path formed up to position and trajectory predictions result, error in judgement meet preset error condition,
If it is, carrying out positional shift alarm.Location point can be monitored in real time, it being capable of early warning positional shift promptly and accurately.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is the present invention
Some embodiments for those of ordinary skill in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 is the process of the positional shift method for early warning provided by one embodiment of the present invention based on neural network track
Figure；
Fig. 2 is the process for the positional shift method for early warning based on neural network track that another embodiment of the present invention provides
Figure；
Fig. 3 is the structural schematic diagram of LSTM unit composed structure provided by one embodiment of the present invention；
Fig. 4 is the anti-Early-warning Model structure chart of wandering away based on trajectory predictions that another embodiment of the present invention provides；
Fig. 5 is that the structure of the positional shift prior-warning device provided by one embodiment of the present invention based on neural network track is shown
It is intended to.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments, based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
As shown in Figure 1, the embodiment of the invention provides a kind of positional shift method for early warning based on neural network track, it should
Method may comprise steps of:
Step 101: acquisition historical trajectory data；
Step 102: extracting position sequence from historical trajectory data, and be based on position sequence, construct prediction model；
Step 103: when collect it is N number of arrived position when, arrived position and prediction model based on N number of, carry out track
It predicts, wherein N >=2；
Step 104: when collecting N+1 and arrived position, calculating the N+1 actual paths that arrived position composition
With the error between trajectory predictions result；
Step 105: whether error in judgement meets preset error condition, if so, thening follow the steps 106；Otherwise, it executes
Step 107；
Step 106: carrying out positional shift alarm, and terminate current process；
Step 107: ignoring this judging result.
Wherein, historical trajectory data is obtained based on communication base station.The historical trajectory data may include active user
Movement track data, can also include the movement track data of other users relevant to active user.For example, for monitoring
Old man's positional shift, the historical trajectory data refer to the movement track of old man and the action rail of other households relevant to old man
Mark.Such as the positional shift of child, the historical trajectory data refer to child movement track and it is relevant to child its
The movement track of his household.Historical trajectory data is hour from base station cell report of user GPS track data time granularity,
Time interval i.e. between sample is 1 hour.Missing data carries out the polishing of data with interpolation, for example missing data is assigned
Give arbitrary value such as 0.
The position sequence refers to that key position point refers to, User Activity range and corresponding stop point, critical positions, family
The characteristic informations such as membership location are enclosed in front yard.
In the embodiment shown in fig. 1, by acquiring historical trajectory data；Position-order is extracted from historical trajectory data
Column, and it is based on position sequence, construct prediction model；When collect it is N number of arrived position when, based on N number of position and pre- of arrived
Model is surveyed, carries out trajectory predictions, wherein N >=2；When collecting N+1 and arrived position, calculating N+1 arrived position
Error between the actual path and trajectory predictions result of composition, judges whether the error meets preset error condition, such as
Fruit is then to carry out positional shift alarm.Location point can be monitored in real time, it being capable of early warning positional shift promptly and accurately.
The above-mentioned specific embodiment that position sequence is extracted from historical trajectory data can include: calculated using following first
Formula calculates the region consistency weight in historical trajectory data between every two tracing point；
First calculation formula:
Wherein, coh (p, q) characterizes the region consistency weight between p-th of tracing point and q-th of tracing point；Dist (p,
Q) actual range between p-th of tracing point and q-th of tracing point is characterized；Duration (p, q) characterization from p-th of tracing point to
The time required to q-th of tracing point；δ characterizes constant；P and q is the positive integer not less than 1, and p ≠ q；
Tracing point by region consistency weight not less than preset region threshold is combined into regional ensemble；According to preset
Radius is inquired, search is located at the tracing point in inquiry radius in regional ensemble；It extracts and stays from the tracing point in inquiry radius
Foot point；It is a little clustered to stopping；According to cluster result, position sequence is determined.
In addition, above-mentioned be based on position sequence, the specific embodiment of prediction model is constructed can include: use Skip-gram
Location point in position sequence is trained to corresponding feature vector by algorithm；The corresponding feature vector of location point is inputted into length
In memory models, the parameter of length memory models is determined.
Above-mentioned calculating N+1 arrived the specific of the error between the actual path and trajectory predictions result of position composition
Embodiment can include: utilize following second calculation formula, calculate the N+1 positions that arrived in position and trajectory predictions result
Set the first deviation between a little；
Second calculation formula:
Wherein, dist (m, n) characterizes first deviation；M.lat and m.lon characterizes N+1 respectively arrived position
The dimension and longitude set；N.lat and n.lon characterizes the dimension and longitude of the location point in the trajectory predictions result；
It is accumulative location point of the deviation more than the error threshold in the preset error condition continuously occur.
Whether above-mentioned error in judgement meets the specific embodiment of preset error condition can include: judges calculated
Whether one deviation is more than error threshold in preset error condition, if it is, judging whether accumulated result is not less than institute
The count threshold in error condition is stated, carries out positional shift alarm if it is, executing.
In an alternative embodiment of the invention, the specific embodiment party stopped a little is extracted in the above-mentioned tracing point from inquiry radius
Formula can include:
Using following calculation formula groups, calculating is stopped a little；
Wherein, sp.lat characterizes the dimension stopped a little；Sp.lon characterizes the longitude stopped a little；R.lat characterization is located at described
Inquire the dimension of the tracing point in radius；R.lon characterization is located at the longitude of the tracing point in the inquiry radius；K characterization is located at
The total number of tracing point in the inquiry radius.
In an alternative embodiment of the invention, the above-mentioned positional shift method for early warning based on neural network track can be wrapped further
It includes: determining at least one target position；When judging that error is unsatisfactory for preset error condition, collected N+1 is calculated
A the second deviation that arrived between position and each described target position；When calculated all second deviations are more than
When preset deviation threshold, then positional shift alarm is carried out.
To monitor old man's positional shift, for preventing old man to wander away, position of the expansion explanation based on neural network track
Deviate method for early warning.As shown in Fig. 2, it is specific to be somebody's turn to do the positional shift method for early warning based on neural network track can include:
Step 201: acquisition historical trajectory data；
The historical trajectory data includes the historical track of old man's historical trajectory data and other households relevant to old man
Data.The historical trajectory data is hour from base station cell report of user GPS track data time granularity, i.e., before sample
Time interval be 1 hour.Missing data carries out the polishing of data with interpolation, and interpolation can be with interpolation any number such as 0.
Historical trajectory data is standardized, and GPS data is converted into location point and is marked.
Step 202: calculating the region consistency weight in historical trajectory data between every two tracing point；
The step can utilize following first calculation formula, calculate in the historical trajectory data between every two tracing point
Region consistency weight；
First calculation formula:
Wherein, coh (p, q) characterizes the region consistency weight between p-th of tracing point and q-th of tracing point；Dist (p,
Q) actual range between p-th of tracing point and q-th of tracing point is characterized；Duration (p, q) characterization from p-th of tracing point to
The time required to q-th of tracing point；δ characterizes constant；P and q is the positive integer not less than 1, and p ≠ q.
Step 203: the tracing point by region consistency weight not less than preset region threshold is combined into regional ensemble；
Step 204: according to preset inquiry radius, search is located at the tracing point in inquiry radius in regional ensemble；
Step 205: extracting and stopped a little from the tracing point in inquiry radius；
The step extracts the specific embodiment stopped a little, and using following calculation formula groups, calculating is stopped a little；
Wherein, sp.lat characterizes the dimension stopped a little；Sp.lon characterizes the longitude stopped a little；R.lat characterization is located at described
Inquire the dimension of the tracing point in radius；R.lon characterization is located at the longitude of the tracing point in the inquiry radius；K characterization is located at
The total number of tracing point in the inquiry radius.
Step 206: a little being clustered to stopping；
The process of step cluster is mainly, to stop as tracing point, to re-execute the steps 202 to step 205.
Step 207: according to cluster result, determining position sequence；
Step 208: using Skip-gram algorithm, the location point in position sequence is trained to corresponding feature vector；
The step is mainly using following
Trajs indicates that the track of critical positions Sequence composition is gathered, f expression location point, and model parameter θ, F (C | f) it is to work as f
The probability that track context c occurs when appearance, C (f) indicate the set for the track context that f occurs.We use formula above
θ is found to optimize formula.
Formalization processing is carried out to θ using the extension softmax of logistic regression, conditional probability is made to be converted into following public affairs
Formula:
Wherein vc, vfIt is the column vector of track context c and location point p respectively, dimension d, C are in all track contexts
The set that position is constituted.The feature vector expression of each critical positions is obtained so that conditional probability by the training of above-mentioned model
F (c | f) it maximizes.
Step 209: by the corresponding feature vector input length memory models of location point, determining the ginseng of length memory models
Number；
The step detailed process is mainly that the location point of obtained critical positions sequence is defined as XA, XAFor n-dimensional vector:
XA=(x1 x2 … xn), XAAs the input of long Memory Neural Networks (LSTM) in short-term, learn temporal aspect, which remembers in short-term
The neuron for recalling neural network (LSTM) is memory unit structure, as shown in Figure 3.If solving the output that t moment hides layer unit
Lt, should first calculate memory unit ct, following formula:
ct=it*tanh(WcXAt+ucLt-1+bc)+ftct-1
Wherein, XAtCharacterize the location point of the critical positions sequence of t moment, wc、ucAnd bcCharacterize respectively the first coefficient matrix,
Second coefficient matrix and the first offset parameter, Lt-1、ct-1It is the information of output and the memory of previous moment hidden unit, wherein
itAnd ftIt is the input gate and forgetting door of t moment, the itAnd ftCorresponding formula is such as given a definition.
it=σ (wiqt+uiXA(t-1)+bi)
ft=σ (wfqt+ufXA(t-1)+bf)
wi、uiAnd biThird coefficient matrix, the 4th coefficient matrix and the second offset parameter, w are characterized respectivelyf、ufAnd bfRespectively
The 5th coefficient matrix, the 6th coefficient matrix and third offset parameter are characterized, σ characterizes a model constants, qtCharacterize the defeated of t moment
Enter.
Obtain the memory unit c of t momenttAfterwards, the out gate o of t moment memory unit can be acquiredt, and then acquire hidden unit
LtOutput, out gate otWith hidden unit LtCorresponding formula is as follows.
ot=σ (woqt+uoXA(t-1)+bo)
Lt=ottanh(ct)
wo、uoCharacterize the 7th coefficient matrix, the 8th coefficient matrix and the 4th offset parameter, last n-th of time respectively with o
The output of point is user trajectory output LA。
Step 210: determine at least one target position, when collect it is N number of arrived position when, arrived position based on N number of
It sets and prediction model, carries out trajectory predictions, wherein N >=2；
The target position is the position where other households relevant to old man.
Step 211: when collecting N+1 and arrived position, calculating the N+1 actual paths that arrived position composition
With the error between trajectory predictions result；
Using following second calculation formula, calculate location points that N+1 arrived in position and trajectory predictions result it
Between the first deviation；
Second calculation formula:
Wherein, dist (m, n) characterizes first deviation；M.lat and m.lon characterizes N+1 respectively arrived position
The dimension and longitude set；N.lat and n.lon characterizes the dimension and longitude of the location point in the trajectory predictions result；
Step 212: accumulative location point of the deviation more than the error threshold in the preset error condition continuously occur；
Step 213: judge whether calculated first deviation is more than error threshold in preset error condition, if
It is to then follow the steps 214；Otherwise, step 217 is executed；
Step 214: the count threshold whether accumulated result is not less than in the error condition is judged, if it is, executing
Step 215；Otherwise, step 216 is executed；
Step 215: carrying out positional shift alarm, and terminate current process；
Step 216: calculating collected N+1 the second deviation that arrived between position and each target position
Value；
Using following third calculation formula, calculate location points that N+1 arrived in position and trajectory predictions result it
Between the first deviation；
Third calculation formula:
Wherein, dist (m, n) characterizes first deviation；M.lat and m.lon characterizes N+1 respectively arrived position
The dimension and longitude set；G.lat and g.lon characterizes the dimension and longitude of the target position；
Step 217: when calculated all second deviations are more than preset deviation threshold, then carrying out positional shift report
It is alert.
By predicting several groups of the elderly's movement tracks, the results showed that method prediction provided in an embodiment of the present invention is quasi-
True rate is up to 0.7% or more.
The i.e. above-mentioned positional shift method for early warning based on neural network track can be summarized as, as shown in Fig. 2, the input of model
For important position sequence file, the data in the critical positions sequential file include each user n time point (cycle T, often
A acquisition time granularity) location track data composition.The output of model is the location track of T+1 time point user, model pair
User is modeled with long Memory Neural Networks (LSTM) in short-term, excavates the periodic feature of user location track.To real time position
Track and predicted position trajectory processing judge whether to trigger early warning.Eabedding layers of the Landmark location point structure to input
It builds position vector, and the position vector of building is exported and gives the length constructed according to historical trajectory data Memory Neural Networks in short-term
(LSTM), to obtain monitoring result.
As shown in figure 5, the embodiment of the present invention provides a kind of positional shift prior-warning device based on neural network track, the base
Positional shift prior-warning device in neural network track includes: acquisition unit 501, model construction unit 502, trajectory predictions unit
503 and offset prewarning unit 504, wherein
Acquisition unit 501, for acquiring historical trajectory data；
Model construction unit 502, for extracting position sequence from the historical trajectory data that acquisition unit 501 acquires, and
Based on position sequence, prediction model is constructed；
Trajectory predictions unit 503, for when collect it is N number of arrived position when, arrived position and model structure based on N number of
The prediction model of the building of unit 502 is built, carries out trajectory predictions, wherein N >=2；
Prewarning unit 504 is deviated, arrived set of locations for when collecting N+1 and arrived position, calculating N+1
At the trajectory predictions result that predicts of actual path and trajectory predictions unit 503 between error, whether error in judgement meet
Preset error condition, if it is, carrying out positional shift alarm.
In an alternative embodiment of the invention, model construction unit 502, for utilizing following first calculation formula, calculating is gone through
Region consistency weight in history track data between every two tracing point；
First calculation formula:
Wherein, coh (p, q) characterizes the region consistency weight between p-th of tracing point and q-th of tracing point；Dist (p,
Q) actual range between p-th of tracing point and q-th of tracing point is characterized；Duration (p, q) characterization from p-th of tracing point to
The time required to q-th of tracing point；δ characterizes constant；P and q is the positive integer not less than 1, and p ≠ q；
Tracing point by region consistency weight not less than preset region threshold is combined into regional ensemble；
According to preset inquiry radius, search is located at the tracing point in the inquiry radius in the regional ensemble；
Using following calculation formula groups, calculating is stopped a little；
Wherein, sp.lat characterizes the dimension stopped a little；Sp.lon characterizes the longitude stopped a little；R.lat characterization is located at described
Inquire the dimension of the tracing point in radius；R.lon characterization is located at the longitude of the tracing point in the inquiry radius；K characterization is located at
The total number of tracing point in the inquiry radius；
It is a little clustered to stopping；According to cluster result, position sequence is determined.
In an alternative embodiment of the invention, model construction unit 502, for using Skip-gram algorithm, by position sequence
In location point be trained to corresponding feature vector；By in the corresponding feature vector input length memory models of location point, determine
The parameter of length memory models.
In an alternative embodiment of the invention, prewarning unit 504 is deviated, for utilizing following second calculation formula, calculates N
The first deviation between+1 location point that arrived in position and the trajectory predictions result；
Second calculation formula:
Wherein, dist (m, n) characterizes first deviation；M.lat and m.lon characterizes N+1 respectively arrived position
The dimension and longitude set；N.lat and n.lon characterizes the dimension and longitude of the location point in the trajectory predictions result；
It is accumulative location point of the deviation more than the error threshold in the preset error condition continuously occur；Judgement calculates
The first deviation whether be more than error threshold in preset error condition, if it is, judging whether accumulated result not small
Count threshold in the error condition, if it is, executing the progress positional shift alarm.
In an alternative embodiment of the invention, prewarning unit 504 is deviated, is further used for determining at least one target position；
When judging that error is unsatisfactory for preset error condition, calculate collected N+1 arrived position and each described in
The second deviation between target position；When calculated all second deviations are more than preset deviation threshold, then carry out
Positional shift alarm.
In conclusion more than the present invention each embodiment at least has the following beneficial effects:
1, in embodiments of the present invention, by acquiring historical trajectory data；Position sequence is extracted from historical trajectory data,
And it is based on position sequence, construct prediction model；When collect it is N number of arrived position when, based on N number of position and described pre- of arrived
Model is surveyed, carries out trajectory predictions, wherein N >=2；When collecting N+1 and arrived position, calculating N+1 arrived position
Whether the error between the actual path and trajectory predictions result of composition, error in judgement meet preset error condition, if so,
Then carry out positional shift alarm.Location point can be monitored in real time, it being capable of early warning positional shift promptly and accurately.
2, in embodiments of the present invention, by calculating the region consistency in historical trajectory data between every two tracing point
Weight；Tracing point by region consistency weight not less than preset region threshold is combined into regional ensemble；It is looked into according to preset
Radius is ask, search is located at the tracing point in inquiry radius in regional ensemble；It extracts and stops from the tracing point in inquiry radius
Point；It is a little clustered to stopping；According to cluster result, position sequence is determined, reduce the interference of insignificant location point, guaranteeing
While forecasting accuracy, operation burden is reduced.
It should be noted that, in this document, such as first and second etc relational terms are used merely to an entity
Or operation is distinguished with another entity or operation, is existed without necessarily requiring or implying between these entities or operation
Any actual relationship or order.Moreover, the terms "include", "comprise" or its any other variant be intended to it is non-
It is exclusive to include, so that the process, method, article or equipment for including a series of elements not only includes those elements,
It but also including other elements that are not explicitly listed, or further include solid by this process, method, article or equipment
Some elements.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including
There is also other identical factors in the process, method, article or equipment of the element.
Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above method embodiment can pass through
The relevant hardware of program instruction is completed, and program above-mentioned can store in computer-readable storage medium, the program
When being executed, step including the steps of the foregoing method embodiments is executed；And storage medium above-mentioned includes: ROM, RAM, magnetic disk or light
In the various media that can store program code such as disk.
Finally, it should be noted that the foregoing is merely presently preferred embodiments of the present invention, it is merely to illustrate skill of the invention
Art scheme, is not intended to limit the scope of the present invention.Any modification for being made all within the spirits and principles of the present invention,
Equivalent replacement, improvement etc., are included within the scope of protection of the present invention.
Claims (11)
1. the positional shift method for early warning based on neural network track characterized by comprising
Acquire historical trajectory data；
Position sequence is extracted from the historical trajectory data, and is based on the position sequence, constructs prediction model；
Further include:
When collect it is N number of arrived position when, based on it is described it is N number of arrived position and the prediction model, carry out trajectory predictions,
Wherein N >=2；
When collecting N+1 and arrived position, the N+1 actual paths and trajectory predictions knot that arrived position composition are calculated
Error between fruit, judges whether the error meets preset error condition, if it is, carrying out positional shift alarm.
2. the positional shift method for early warning according to claim 1 based on neural network track, which is characterized in that it is described from
Position sequence is extracted in the historical trajectory data, comprising:
Using following first calculation formula, the region consistency power in the historical trajectory data between every two tracing point is calculated
Value；
First calculation formula:
Wherein, coh (p, q) characterizes the region consistency weight between p-th of tracing point and q-th of tracing point；Dist (p, q) table
Levy the actual range between p-th of tracing point and q-th of tracing point；Duration (p, q) is characterized from p-th of tracing point to q
The time required to a tracing point；δ characterizes constant；P and q is the positive integer not less than 1, and p ≠ q；
Tracing point by the region consistency weight not less than preset region threshold is combined into regional ensemble；
According to preset inquiry radius, search is located at the tracing point in the inquiry radius in the regional ensemble；
It extracts and is stopped a little from the tracing point in the inquiry radius；
Described stop a little is clustered；
According to cluster result, position sequence is determined.
3. the positional shift method for early warning according to claim 1 based on neural network track, which is characterized in that the base
In the position sequence, prediction model is constructed, comprising:
Using Skip-gram algorithm, the location point in the position sequence is trained to corresponding feature vector；
By in the corresponding feature vector input length memory models of the location point, the parameter of the length memory models is determined.
4. the positional shift method for early warning according to claim 1 based on neural network track, which is characterized in that the meter
Calculate the error between the actual path and trajectory predictions result that N+1 arrived position composition, comprising:
Using following second calculation formula, calculate location points that N+1 arrived in position and the trajectory predictions result it
Between the first deviation；
Second calculation formula:
Wherein, dist (m, n) characterizes first deviation；M.lat and m.lon characterizes N+1 respectively arrived position
Dimension and longitude；N.lat and n.lon characterizes the dimension and longitude of the location point in the trajectory predictions result；
It is accumulative location point of the deviation more than the error threshold in the preset error condition continuously occur.
5. the positional shift method for early warning according to claim 4 based on neural network track, which is characterized in that described to sentence
Whether the error of breaking meets preset error condition, comprising:
Judge whether calculated first deviation is more than error threshold in preset error condition, if it is, judgement is tired
Whether meter result is not less than the count threshold in the error condition, if it is, executing the progress positional shift alarm.
6. the positional shift method for early warning according to claim 2 based on neural network track, which is characterized in that it is described from
It extracts and is stopped a little in tracing point in the inquiry radius, comprising:
Using following calculation formula groups, calculating is stopped a little；
Wherein, sp.lat characterizes the dimension stopped a little；Sp.lon characterizes the longitude stopped a little；R.lat characterization is located at the inquiry
The dimension of tracing point in radius；R.lon characterization is located at the longitude of the tracing point in the inquiry radius；K characterization is located at described
Inquire the total number of the tracing point in radius.
7. the positional shift method for early warning according to any one of claims 1 to 5 based on neural network track, feature exist
In further comprising:
Determine at least one target position；
When judging that error is unsatisfactory for preset error condition, calculating collected N+1 arrived position and each
The second deviation between the target position；
When calculated all second deviations are more than preset deviation threshold, then positional shift alarm is carried out.
8. the positional shift prior-warning device based on neural network track characterized by comprising acquisition unit, model construction list
Member, trajectory predictions unit and offset prewarning unit, wherein
The acquisition unit, for acquiring historical trajectory data；
The model construction unit, for extracting position sequence, and base from the historical trajectory data that the acquisition unit acquires
In the position sequence, prediction model is constructed；
The trajectory predictions unit, for when collect it is N number of arrived position when, based on N number of position and described of arrived
The prediction model of model construction building unit carries out trajectory predictions, wherein N >=2；
The offset prewarning unit arrived position composition for when collecting N+1 and arrived position, calculating N+1
Actual path and the trajectory predictions result that predicts of the trajectory predictions unit between error, judge whether the error full
The preset error condition of foot, if it is, carrying out positional shift alarm.
9. the positional shift prior-warning device according to claim 8 based on neural network track, which is characterized in that the mould
Type construction unit, is used for:
Using following first calculation formula, the region consistency power in the historical trajectory data between every two tracing point is calculated
Value；
First calculation formula:
Wherein, coh (p, q) characterizes the region consistency weight between p-th of tracing point and q-th of tracing point；Dist (p, q) table
Levy the actual range between p-th of tracing point and q-th of tracing point；Duration (p, q) is characterized from p-th of tracing point to q
The time required to a tracing point；δ characterizes constant；P and q is the positive integer not less than 1, and p ≠ q；
Tracing point by the region consistency weight not less than preset region threshold is combined into regional ensemble；
According to preset inquiry radius, search is located at the tracing point in the inquiry radius in the regional ensemble；
Using following calculation formula groups, calculating is stopped a little；
Wherein, sp.lat characterizes the dimension stopped a little；Sp.lon characterizes the longitude stopped a little；R.lat characterization is located at the inquiry
The dimension of tracing point in radius；R.lon characterization is located at the longitude of the tracing point in the inquiry radius；K characterization is located at described
Inquire the total number of the tracing point in radius；
Described stop a little is clustered；
According to cluster result, position sequence is determined.
10. the positional shift prior-warning device according to claim 8 based on neural network track, which is characterized in that
Location point in the position sequence is trained to correspondence for using Skip-gram algorithm by the model construction unit
Feature vector；By in the corresponding feature vector input length memory models of the location point, the length memory models are determined
Parameter；
And/or
The offset prewarning unit, is used for:
Using following second calculation formula, calculate location points that N+1 arrived in position and the trajectory predictions result it
Between the first deviation；
Second calculation formula:
Wherein, dist (m, n) characterizes first deviation；M.lat and m.lon characterizes N+1 respectively arrived position
Dimension and longitude；N.lat and n.lon characterizes the dimension and longitude of the location point in the trajectory predictions result；
It is accumulative location point of the deviation more than the error threshold in the preset error condition continuously occur；
Judge whether calculated first deviation is more than error threshold in preset error condition, if it is, judgement is tired
Whether meter result is not less than the count threshold in the error condition, if it is, executing the progress positional shift alarm.
11. according to any positional shift prior-warning device based on neural network track of claim 8 to 10, feature exists
In the offset prewarning unit is further used for:
Determine at least one target position；
When judging that error is unsatisfactory for preset error condition, calculating collected N+1 arrived position and each
The second deviation between the target position；
When calculated all second deviations are more than preset deviation threshold, then positional shift alarm is carried out.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910559453.0A CN110334167B (en) | 2019-06-26 | 2019-06-26 | Position deviation early warning method and device based on neural network track |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910559453.0A CN110334167B (en) | 2019-06-26 | 2019-06-26 | Position deviation early warning method and device based on neural network track |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110334167A true CN110334167A (en) | 2019-10-15 |
| CN110334167B CN110334167B (en) | 2023-06-09 |
Family
ID=68142736
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910559453.0A Active CN110334167B (en) | 2019-06-26 | 2019-06-26 | Position deviation early warning method and device based on neural network track |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110334167B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110795522A (en) * | 2019-11-06 | 2020-02-14 | 中国人民解放军战略支援部队信息工程大学 | Method and device for predicting track position of mobile user |
| CN111028471A (en) * | 2019-12-16 | 2020-04-17 | 山东恒云信息科技有限公司 | Method for preventing tourists from approaching high-risk area by using neural network technology |
| CN111238492A (en) * | 2020-01-14 | 2020-06-05 | 深圳猛犸电动科技有限公司 | Fixed point drift correction method and device, terminal equipment and storage medium |
| CN111260082A (en) * | 2020-03-08 | 2020-06-09 | 北京瀚科科技集团有限公司 | Spatial object motion trajectory curve prediction method based on neural network |
| CN112814500A (en) * | 2020-12-31 | 2021-05-18 | 南京海索信息科技有限公司 | Intelligent lock real-time monitoring system and method |
| CN113537323A (en) * | 2021-07-02 | 2021-10-22 | 香港理工大学深圳研究院 | Indoor track error evaluation method based on LSTM neural network |
| CN113689660A (en) * | 2020-05-19 | 2021-11-23 | 上海惠芽信息技术有限公司 | Safety early warning method of wearable device and wearable device |
| WO2023019840A1 (en) * | 2021-08-18 | 2023-02-23 | 上海闻泰电子科技有限公司 | Wireless positioning method and apparatus, electronic device, and storage medium |
| CN115985097A (en) * | 2022-12-29 | 2023-04-18 | 浪潮通信信息系统有限公司 | High-speed user operation track validity judgment method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764560A (en) * | 2018-05-22 | 2018-11-06 | 电子科技大学 | Aircraft scene trajectory predictions method based on shot and long term Memory Neural Networks |
| WO2019000887A1 (en) * | 2017-06-26 | 2019-01-03 | 北京三快在线科技有限公司 | Method and device for recommending information |
| CN109410496A (en) * | 2018-10-25 | 2019-03-01 | 北京交通大学 | Attack early warning method, apparatus and electronic equipment |
-
2019
- 2019-06-26 CN CN201910559453.0A patent/CN110334167B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019000887A1 (en) * | 2017-06-26 | 2019-01-03 | 北京三快在线科技有限公司 | Method and device for recommending information |
| CN108764560A (en) * | 2018-05-22 | 2018-11-06 | 电子科技大学 | Aircraft scene trajectory predictions method based on shot and long term Memory Neural Networks |
| CN109410496A (en) * | 2018-10-25 | 2019-03-01 | 北京交通大学 | Attack early warning method, apparatus and electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 章梦杰等: "多维空间中基于模式的移动对象轨迹预测", 《计算机系统应用》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110795522B (en) * | 2019-11-06 | 2023-01-06 | 中国人民解放军战略支援部队信息工程大学 | Method and device for predicting track position of mobile user |
| CN110795522A (en) * | 2019-11-06 | 2020-02-14 | 中国人民解放军战略支援部队信息工程大学 | Method and device for predicting track position of mobile user |
| CN111028471B (en) * | 2019-12-16 | 2021-04-06 | 山东恒云信息科技有限公司 | Method for preventing tourists from approaching high-risk area by using neural network technology |
| CN111028471A (en) * | 2019-12-16 | 2020-04-17 | 山东恒云信息科技有限公司 | Method for preventing tourists from approaching high-risk area by using neural network technology |
| CN111238492B (en) * | 2020-01-14 | 2021-12-17 | 深圳猛犸电动科技有限公司 | Fixed point drift correction method and device, terminal equipment and storage medium |
| CN111238492A (en) * | 2020-01-14 | 2020-06-05 | 深圳猛犸电动科技有限公司 | Fixed point drift correction method and device, terminal equipment and storage medium |
| CN111260082A (en) * | 2020-03-08 | 2020-06-09 | 北京瀚科科技集团有限公司 | Spatial object motion trajectory curve prediction method based on neural network |
| CN111260082B (en) * | 2020-03-08 | 2023-11-07 | 北京瀚科科技集团有限公司 | Spatial object motion trail prediction model construction method based on neural network |
| CN113689660A (en) * | 2020-05-19 | 2021-11-23 | 上海惠芽信息技术有限公司 | Safety early warning method of wearable device and wearable device |
| CN112814500A (en) * | 2020-12-31 | 2021-05-18 | 南京海索信息科技有限公司 | Intelligent lock real-time monitoring system and method |
| CN112814500B (en) * | 2020-12-31 | 2022-03-18 | 南京海索信息科技有限公司 | Intelligent lock real-time monitoring system and method |
| CN113537323A (en) * | 2021-07-02 | 2021-10-22 | 香港理工大学深圳研究院 | Indoor track error evaluation method based on LSTM neural network |
| CN113537323B (en) * | 2021-07-02 | 2023-11-07 | 香港理工大学深圳研究院 | Indoor track error assessment method based on LSTM neural network |
| WO2023019840A1 (en) * | 2021-08-18 | 2023-02-23 | 上海闻泰电子科技有限公司 | Wireless positioning method and apparatus, electronic device, and storage medium |
| CN115985097A (en) * | 2022-12-29 | 2023-04-18 | 浪潮通信信息系统有限公司 | High-speed user operation track validity judgment method |
| CN115985097B (en) * | 2022-12-29 | 2024-04-26 | 浪潮通信信息系统有限公司 | High-speed user running track effectiveness judging method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110334167B (en) | 2023-06-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110334167A (en) | Positional shift method for early warning and device based on neural network track | |
| Lin et al. | Road traffic speed prediction: A probabilistic model fusing multi-source data | |
| Tung et al. | Goal-based trajectory analysis for unusual behaviour detection in intelligent surveillance | |
| CN110619430B (en) | Space-time attention mechanism method for traffic prediction | |
| CN106384120B (en) | A kind of resident's activity pattern method for digging and device based on mobile phone location data | |
| CN104408913A (en) | Traffic flow three parameter real time prediction method taking regard of space-time correlation | |
| Hu et al. | Integrated determination of network origin–destination trip matrix and heterogeneous sensor selection and location strategy | |
| Yang et al. | When and where: Predicting human movements based on social spatial-temporal events | |
| Fortino et al. | Exploiting marked temporal point processes for predicting activities of daily living | |
| US11519734B2 (en) | Facility management based on cumulative device location data | |
| CN115688952A (en) | Big data-based campus congestion area prediction method | |
| Wang et al. | The Indoor Predictability of Human Mobility: Estimating Mobility With Smart Home Sensors | |
| De Paola et al. | Detection of points of interest in a smart campus | |
| CN115938104A (en) | Dynamic short-time road network traffic state prediction model and prediction method | |
| CN110110690B (en) | Target pedestrian tracking method, device, equipment and storage medium | |
| Lu | An adaptive system for predicting freeway travel times | |
| Boytsov | Context reasoning, context prediction and proactive adaptation in pervasive computing systems | |
| Li et al. | Exploiting unlabeled RSSI fingerprints in multi-building and multi-floor indoor localization through deep semi-supervised learning based on mean teacher | |
| Dai et al. | Trajectory outlier detection based on dbscan and velocity entropy | |
| Sarasvathi | Reliability analysis of an IoT-based air pollution monitoring system using machine learning algorithm-BDBN | |
| Alshaer | Mobile object-tracking approach using a combination of fuzzy logic and neural networks | |
| Deepan et al. | Air quality index prediction using seasonal autoregressive integrated moving average transductive long short‐term memory | |
| CN113889282A (en) | Individual mobile intervention infectious disease prevention and control method and system | |
| Wongsathan | The logistic growth regression model with the genetic algorithm for predicting the third wave of the COVID-19 epidemic in Thailand | |
| CN113032688A (en) | Method for predicting access position of social network user at given future time |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |