CN108417276A - It is a kind of to look after intelligent monitoring method in real time towards health endowment post house - Google Patents
It is a kind of to look after intelligent monitoring method in real time towards health endowment post house Download PDFInfo
- Publication number
- CN108417276A CN108417276A CN201810190824.8A CN201810190824A CN108417276A CN 108417276 A CN108417276 A CN 108417276A CN 201810190824 A CN201810190824 A CN 201810190824A CN 108417276 A CN108417276 A CN 108417276A
- Authority
- CN
- China
- Prior art keywords
- data
- doctor
- room
- strong
- time
- 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
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H80/00—ICT specially adapted for facilitating communication between medical practitioners or patients, e.g. for collaborative diagnosis, therapy or health monitoring
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/20—Software design
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Medical Informatics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Epidemiology (AREA)
- Pathology (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The present invention propose it is a kind of looking after intelligent monitoring method in real time towards health endowment post house, can realize and support strong object Life cycle to curing and look after.First, digital sensor and the Intellisense terminal of strong objects perimeter are supported come the foster strong object itself of comprehensive perception doctor and its current state of local environment by being deployed in doctor, and collected data is saved in the database in high in the clouds；Then, the multidimensional mass data that strong object is supported for curing, the movement track for supporting strong object to doctor with the method for time series forecasting are predicted, and the life pattern of the foster strong object of doctor is constructed with the method for cluster；Finally, according to curing the comparison supported the matching of strong object movement track and support strong object life pattern to curing, abnormal judgement and real-time early warning are carried out.The present invention can be used for the fields such as medical treatment, endowment, to cure the service of looking after supported strong object and provide Life cycle, the phenomenon that helping to improve " data silo, service are isolated ".
Description
Technical field
The present invention relates to medical treatment, endowment fields to curing the method supported strong object and look after intellectual monitoring in real time, in conjunction with big data
Analytical technology allows health endowment to tend to interconnected, Thingsization, mobile, intelligence and integrates each side's Service Source and carried to old man
For safety, comprehensive, Life cycle health endowment service, specifically, propose a kind of towards the real-time of health endowment post house
Look after intelligent monitoring method.
Background technology
Health endowment service trade is in generation informations skills such as cloud computing, big data, mobile Internet, Internet of Things and business networkings
The new situation and new feature of " transboundary merging " are showed under the promotion of art, how to be provided for the elderly to high efficiency smart safe, complete
Face, Life cycle durative services become most challenging one of problem.
To push the development of healthy old-age care, national successively issue《The programs for the elderly development of " 13 " country and endowment
System Construction is planned》With《Wisdom health endowment industry development action plan (2017-2020)》, clearly propose to utilize a new generation
Information technology realizes that individual, family, community, mechanism are effectively docked and distributed rationally with health medical treatment resource.Business networking (IoS：
Internet of Services) be by cross-system, cross-cutting, trans-regional, across a network magnanimity Heterogeneous service by polymerization with
The complex services network of collaboration and formation is to solve health endowment service trade global networking, professional socialization, transboundary syncretization
There is cross-domain type, the platform core key technology of outstanding problems such as " dissipate, unrest, difference, lack, resentment ", theory and technology body under background
System's research is one of the key task of national " 13 " modern service industry scientific and technical innovation ad hoc planning.As health is supported by government
Old increasing is helped, and more and more mechanisms have put into healthy endowment industry.
At present in terms of intelligent health monitoring service, the giants such as IBM, NEC have launched towards house, community endowment clothes
Business based on the wisdom of Internet of Things and artificial intelligence technology support parents solution, the country occur many towards field of digital home
Health supervision product, but in the research and development of the core key parts such as wearable physiological characteristic monitoring, short-distance wireless communication still
There are technical bottleneck and rely on import, therefore, transboundary resource, power-assisted are realized " collaboration of multimachine structure " and " more for active mating fusion more
Institution Services " are problems instantly most in the urgent need to address.
For the service mode of current health endowment transboundary, either ubiquitous formula house community, multimachine structure doctor supports strong combine
Or the health control and service that IoT is merged with IoS are all undoubtedly Internet of Things, internet and business networking technology integrated application
Typical scene.In order to enhance strong object is supported to curing and its local environment is abnormal the monitoring of emergency situations and pre- under various scenes
Alert ability, the present invention combination big data analysis technology propose a kind of to look after intellectual monitoring side towards the real-time of health endowment post house
Method, boosting Yi Yang mechanisms realize fine-grained management, raising efficiency.
Invention content
Intelligent monitoring method being looked after in real time towards health endowment post house the object of the present invention is to provide a kind of, feature exists
It, can be with real-time early warning to take corresponding measure when being abnormal situation in strong object real-time monitoring can be supported to curing.
The object of the present invention is achieved like this：
A kind of to look after intelligent monitoring method in real time towards health endowment post house, feature is that this method includes walking in detail below
Suddenly：
Step 1：Acquisition doctor, which supports the initial data of strong object and stores, arrives cloud database；It specifically includes：
Strong time location of the object in the room of post house is supported by digitized sensor equipment and Intellisense terminal acquisition doctor
Data and room environment data are simultaneously saved in the database in high in the clouds.
Step 2：The data that the foster strong object of the doctor needed is obtained from cloud database are saved in local data base and carry out
Pretreatment；It specifically includes：
Step A：Choose post house inner room number, inner room room number, action sensor, action location data, that is, time series and when
Between stamp field database is being locally created；
Step B：Corresponding doctor is obtained from cloud database according to the field in local data base supports the original of strong object
Data simultaneously filter out invalid empty data；
Step C：Time data format conversion is judged into invalid data again at TimeSeriesRDD formats and deletion.
Since DataFrame formats have lacked compared to TimeSeriesRDD formats the format of a key, so passing through generation
DataFrame formats are added the row of a key by code, are then converted into TimeSeriesRDD formats.It can be simply by
The withColumn methods of DataFrame data formats are quickly converted.
Step 3：The movement track for curing foster strong object is predicted with the method for time series forecasting, and with cluster
Method construct the life pattern that doctor supports strong object；It specifically includes：
Step A：Prediction doctor supports the movement track of strong object
The action location data usage time series prediction technique that strong object is supported to curing is supported strong object according to doctor and is constantly updated
Historical data carry out modeling and more new model, then predict doctor and support the strong next movement track of object.
A1 load () method that the SQLContext objects in SparkSql provide) is used to obtain inner room from local data base
Room number and time stamp data are simultaneously converted into dataframe data formats；
A2 time series) is initialized as TimeSeriesRDD formats and selected HoltWinters models；
A3 HoltWinters.fitModel () method that spark-timeseries is provided) is utilized to create, training
HoltWinters models；
A4 the interval for) predicting the foster strong following 3 minutes movement tracks of object of doctor and data has been arranged to 5 seconds；
A5 36 predicted values) are preserved, and return to a1 after waiting for 3 minutes, cycle executes；
The present invention uses time series algorithm, and increases income computing engines in conjunction with Spark, realizes that doctor supports the action rail of strong object
The prediction of mark；
Time series forecasting is divided into short-term, mid-term, long-term forecast, and has had much for the prediction of different range
Corresponding prediction algorithm, such as：Simple chronological average method, the method for weighted moving average, weighting chronological average method, exponential smoothing
Method, seasonal trend predicted method etc..The algorithm of each time prediction has corresponding applicable scene, it is contemplated that doctor supports strong object
Life track seasonal trend feature is presented, so having selected Holt-Winters algorithms；
According to actual demand, a preprocessed data will be read within every 3 minutes, and re-create and train Holt-Winters moulds
Type, and predict that doctor supports the next 3 minutes movement tracks of strong object, then the movement track of prediction is added in specified data library
In table, movement track matching stage is supplied to execute matching operation.
Step B：Cluster doctor supports the life pattern of strong object
Doctor is supported strong object one day, and activity each local in the room of post house is defined as life pattern with rest rule, adopts
It doctor is supported into strong activity time of the object in inner room room is carried out with the increase income K-Means algorithms of computing engines running optimizatin of Spark
Cluster supports activity time range and corresponding threshold value of the strong object in each room to cluster out doctor, obtains doctor and support strong object
Life pattern；
B1) optimize K-Means clustering algorithms, algorithm optimization flow is as follows：
(1) data are read and are converted into RDD data formats；
(2) Map operations are executed to be formatted data and vectorization；
(3) distance for calculating each fragment data to the centers Canopy obtains Local C anopy central points；
(4) merge and generate Canopy central points；
(5) K-Means initialization operations are carried out, then carries out Map operations and executes K-Means Local Clusterings；
(6) result of Local Clustering is subjected to merger, calculates global cluster node, update central point；
(7) (5) and (6) are repeated until central point no longer changes；
B2 an object sc) is created by SparkContext, then uses textFile () method of sc to read doctor and supports
The inner room room number and time stamp data of strong object are simultaneously converted into RDD data formats；
B3 it) is operated using map and the data of the foster strong object of doctor is classified according to inner room room number, for further cluster
The data of [timestamp, inner room room number] Format Type are provided；
B4 it is) that the foster time data progress Canopy for being good for each inner room room of object of doctor is slightly clustered, obtains every inner room room
Between time data Temporal Clustering, and be saved in local data base；
B5) the Temporal Clustering to cluster out at the beginning of obtaining each Temporal Clustering and the end time, generates each room
Active period, and be saved in local data base；
B6) to the Temporal Clustering generated in b3, secondary cluster is carried out using the K-Means clustering algorithms of optimization, is obtained thinner
The cluster result of change clusters out smaller Temporal Clustering, and is saved in local data base；
B7 the division that the period) is carried out to the Temporal Clustering of secondary cluster, calculates the spacing value between each period, right
Spacing value obtains the rest threshold value in this room into row median value；
B8) active period and rest threshold value that doctor is supported to strong one day each room of object, is converted into dataframe numbers
According to format, it is saved in local data base.
Life pattern analysis not only needs the time data to every room to cluster, it is also necessary to the every of every room
A Temporal Clustering is clustered again, executes the most of the time that law-analysing is occupied on the number and time of cluster, so using
The K-Means clustering algorithms of optimization are slightly clustered by Canopy, are then clustered to K-Means using the result slightly clustered
Algorithm is initialized, to improve the execution efficiency of life pattern analysis.
Step 4：Matching doctor supports the movement track of strong object and compares the life pattern that doctor supports strong object, and strong object is supported to curing
Carry out abnormal judgement and early warning.It specifically includes：
Step A：Match movement track
A1 specified doctor) is obtained from cloud database to support nearest 3 minutes real time datas of strong object and be saved in local
Database obtains whole real time datas if inadequate 3 minutes；
A2 real-time time sequence and the time series progress dimension-reduction treatment of prediction of strong object) are supported to curing, acquisition doctor supports strong
The inner room room number of object；
A3 two sequences) are subjected to dynamic time warping algorithm and carry out similitude matching, calculate the accumulation of two sequences
Distance；
A4) Cumulative Distance and threshold value are judged；If two sequences are similar enough, the Cumulative Distance meeting calculated
Stablize in a range, is compared according to a large amount of, threshold value is finally arranged to 5；
A5) when Cumulative Distance be more than threshold value, just progress step B compare life pattern, terminate if being not above threshold value,
Wait for matched within next 30 seconds；
Step B：Compare life pattern
B1) real time data in strong object room is supported from acquisition doctor in cloud database and be saved in local data base；
B2) doctor is obtained from local data base support the life pattern for being good for object in the room；
B3) judge that current doctor supports the behavior of strong object whether in the life pattern in the room, if the life in the room
In pattern, and within the activity time, then activity time threshold value is set, otherwise, time of having a rest threshold value is set；If not in the room
In life pattern, then be arranged not time threshold；
B4 then) opening timing task calculates the respective behavior data time preserved in current time and local data base
Time difference t1 carry out abnormity early warning if no early warning is crossed and t1 is more than threshold value if early warning is crossed with regard to stopping task.
Step C：Environmental abnormality early warning
Indoor environment abnormity early warning is mainly temperature pre-warning, is as follows：
C1 doctor) is got from cloud database first and supports the room data of strong object and corresponding threshold value；
C2) judge whether temperature moves in circles more than threshold value if it is less than c1 is returned to；
C3) if it is greater than threshold value, and there is no early warning, with regard to carrying out environmental abnormality early warning.
Beneficial effects of the present invention
The present invention carries out big data analysis and personalized modeling to curing foster strong object daily life.The present invention can solve pair
Doctor supports the problem of strong object real-time monitoring and abnormity early warning, is provided for the foster strong object of doctor safe, comprehensive, Life cycle strong
Health endowment service.
Description of the drawings
Fig. 1 is the method for the present invention flow chart.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, right below in conjunction with drawings and examples
The present invention is further elaborated.It should be appreciated that specific embodiment described herein is used only for explaining the present invention, and
It is not used in the limitation present invention.
The present invention server environment be：
Server configures：4 cores, 64G memories, 480G (SSD) (Ali's cloud)
Operating system：Ubuntu 16.04x64
Database：MySQL 5.7.17
Web server：Tomcat 8.0.46
Hadoop：2.7.3
Spark：2.1.1
It is realized under 2017 development environments of IntelliJ IDEA, specific implementation is achieved by the steps of：
Step 1：Acquisition doctor, which supports the initial data of strong object and stores, arrives cloud database
Step A：By digitized sensor equipment and Intellisense terminal acquisition doctor support strong object in the room of post house when
Between position data and room environment data and be saved in the database in high in the clouds.
Step 2：The data that the foster strong object of the doctor needed is obtained from cloud database are saved in local data base and carry out
Pretreatment
Before carrying out data analysis, untreated initial data is disorderly and unsystematic, can not carry out data analysis, former
Beginning data format is as shown in table 1：
1 sensing equipment raw data mode of table
According to the data that movement track prediction and life pattern analysis need, post house inner room number, inner room room number, row are chosen
Database is being locally created in dynamic sensor, action location data (hereinafter referred to as time series) and these fields of timestamp, chooses
Field it is as shown in table 2：
The data field that table 2 is chosen
| Post house inner room number | Inner room room number | Action sensor | Action location data | Timestamp |
The realization logic of data acquisition is simultaneously uncomplicated, and main is exactly to obtain specified doctor from cloud database to support strong object
Data are simultaneously stored in local data base, but the capacity in view of being locally stored when the foster strong object number of doctor increases to a certain amount of is very
It is easy to reach the upper limit, so local file can be deleted with preprocessing process timing, be updated according to polling tasks timing, specifically
Realization step be：
(1) when opening the monitoring for supporting strong object to curing, a timed task is will produce, poll per minute is primary.
(2) after getting the information of the foster strong object of doctor, first determine whether locally whether existing doctor supports the pre- place for being good for object
The data of reason start to the behavior of current all behavioral datas to determine if it does not, being ready for initialization and acquiring the foster strong object of doctor
The data of position data, acquisition are stored in local, and filename is identified by curing foster strong object ID.
(3) if it find that local existing doctor supports the preprocessed data of strong object, that just from preprocessed data last
The time of data starts to obtain, and the data of acquisition are stored in local, and filename is identified by curing foster strong object ID.
(4) when initial data acquisition completion, pre-treatment step will be entered, in order to save local disk size, when pre-
After processing terminates, the data file of initial acquisition will be deleted.
Data prediction carried out after changing data format, and each minute poll is primary, so data
Data volume to be treated and little is pre-processed, and since the data of data source are cleaner, needs pretreated step simultaneously
Seldom, specific implementation step is：
(1) trigger data pre-processes link from raw data acquisition, first determines whether that data whether there is hashed field, if
In the presence of then deleting this data, then carry out lower a data detection；If just carrying out time format there is no hashed field to turn
It changes, directly carries out lower a data detection.
(2) it if completing cycle, preserves in the data to file that pretreatment is completed, and deletes doctor and support strong object original number
According to file.
Step 3：The movement track for curing foster strong object is predicted with the method for time series forecasting, and with cluster
Method construct the life pattern that doctor supports strong object
A) prediction doctor supports the movement track of strong object
Doctor, which supports location data of the strong object in health endowment post house inner room, the spy of regularity, periodicity and real-time
Sign, regularity show as foster strong object the walking about between room of doctor and a rule state are presented；It is foster strong periodically to show as doctor
The daily movement track of object is roughly the same, meets the foster strong object status of doctor to form one to time series modeling
Model；Real-time shows as doctor and supports strong object once deviateing movement track usually, real time data will and time series forecasting
Data occur larger difference.Therefore it can be predicted with the method for time series forecasting curing foster strong object movement track.
The present invention is combined time series algorithm and Spark computing engines of increasing income, and is supported the action of strong object to curing and is positioned number
According to usage time series prediction technique, the historical data that strong object is constantly updated is supported according to doctor and carries out modeling and more new model, so
After predict doctor and support the strong next movement track of object.The specific steps are：
A1 load () method that the SQLContext objects in SparkSql provide) is used to obtain inner room from local data base
Room number and time stamp data are simultaneously converted into dataframe data formats；
A2 time series) is initialized as TimeSeriesRDD formats and selected HoltWinters models；
A3 HoltWinters.fitModel () method that spark-timeseries is provided) is utilized to create, training
HoltWinters models；
A4 the interval for) predicting the foster strong following 3 minutes movement tracks of object of doctor and data has been arranged to 5 seconds；
A5 36 predicted values) are preserved, and return to a1 after waiting for 3 minutes, cycle executes.
According to actual demand, a preprocessed data will be read within every 3 minutes, and re-create and train Holt-Winters moulds
Type, and predict that doctor supports the next 3 minutes movement tracks of strong object, then the movement track of prediction is added in specified data library
In table, movement track matching stage is supplied to execute matching operation.
B) cluster doctor supports the life pattern of strong object
By curing the location data supported strong object and walked about in the room of post house, the movement track for realizing the foster strong object of doctor is pre-
It surveys, but only predicts to judge that the exception that doctor supports strong object behavior has limitation by movement track, strong object hair is supported for curing
The unexpected situations such as raw tumble can not reach higher discrimination.Therefore the present invention carries out the data for curing foster strong object again
Analysis, the movable and rest rule that will cure foster one day each place in the room of post house of strong object are defined as life pattern, use
Doctor is supported strong activity time of the object in inner room room and gathers by the increase income K-Means algorithms of computing engines running optimizatin of Spark
Class supports activity time range and corresponding threshold value of the strong object in each room to cluster out doctor, obtains the life that doctor supports strong object
Pattern living, the specific steps are：
B1 an object sc) is created by SparkContext, then uses textFile () method of sc to read doctor and supports
The inner room room number and time stamp data of strong object are simultaneously converted into RDD data formats；
B2 it) is operated using map and the data of the foster strong object of doctor is classified according to inner room room number, for further cluster
The data of [timestamp, inner room room number] Format Type are provided, such as [20170801102000,0] indicate August in 2017 1
10 points of 0 second 20 minutes No. 0 rooms；
B3 it is) that the foster time data progress Canopy for being good for each inner room room of object of doctor is slightly clustered, obtains every inner room room
Between time data Temporal Clustering, and be saved in local data base；
B4) the Temporal Clustering to cluster out at the beginning of obtaining each Temporal Clustering and the end time, generates each room
Active period, and be saved in local data base；
B5) to the Temporal Clustering generated in b3, secondary cluster is carried out by K-Means clustering algorithms, acquisition more refines poly-
Class is as a result, cluster out smaller Temporal Clustering, and be saved in local data base；
B6 the division that the period) is carried out to the Temporal Clustering of secondary cluster, calculates the spacing value between each period, right
Spacing value obtains the rest threshold value in this room into row median value；
B7) active period and rest threshold value that doctor is supported to strong one day each room of object, is converted into dataframe numbers
According to format, it is saved in local data base.
Life pattern analysis not only needs the time data to every room to cluster, it is also necessary to the every of every room
A Temporal Clustering is clustered again, executes the most of the time that law-analysing is occupied on the number and time of cluster, so using
The K-Means clustering algorithms of optimization are slightly clustered by Canopy, are then clustered to K-Means using the result slightly clustered
Algorithm is initialized, and to improve the execution efficiency of life pattern analysis, algorithm optimization flow is as follows：
(1) data are read and are converted into RDD data formats.
(2) Map operations are executed to be formatted data and vectorization.
(3) distance for calculating each fragment data to the centers Canopy obtains Local C anopy central points.
(4) merge and generate Canopy central points.
(5) K-Means initialization operations are carried out, then carries out Map operations and executes K-Means Local Clusterings.
(6) result of Local Clustering is subjected to merger, calculates global cluster node, update central point.
(7) (5) and (6) are repeated until central point no longer changes.
The life pattern that strong object is supported for curing is defined as the active period and time of having a rest section in each room.
Data format for clustering the foster strong object life pattern of doctor is as shown in table 3：
The data format of 3 life pattern of table
| Timestamp | Post house inner room number |
| Timestamp | Post house inner room number |
| Timestamp | Post house inner room number |
| …… | …… |
The life pattern that doctor supports strong object need not carry out predicting new data as movement track for every 3 minutes, but daily
Historical data to curing foster strong object is once clustered, and is then shown that doctor supports one day life pattern of strong object, will be clustered out
Life pattern be saved in database by date.It will be different in view of curing the daily life pattern of foster strong object, still
Life pattern weekly on the same day is roughly the same, thus will doctor support the life pattern of strong object respectively with Monday to Sunday into
Row preserves.
Step 4：Matching doctor supports the movement track of strong object and compares the life pattern that doctor supports strong object, and strong object is supported to curing
Carry out abnormal judgement and early warning
Step A：Match movement track
Due to supporting the requirement of real-time height of strong object monitoring to curing, so abnormity early warning mechanism obtains a data in 30 seconds,
3 minutes real time datas are obtained every time to be judged, obtained if without 3 minutes real time datas data as much as possible into
Row judges.The specific steps are：
A1 specified doctor) is obtained from cloud database to support nearest 3 minutes real time datas of strong object and be saved in local
Database obtains real time data as much as possible if inadequate 3 minutes；
A2 real-time time sequence and the time series progress dimension-reduction treatment of prediction of strong object) are supported to curing, acquisition doctor supports strong
The inner room room number of object；
A3 two sequences) are subjected to dynamic time warping algorithm and carry out similitude matching, calculate the accumulation of two sequences
Distance；
The principle of dynamic time warping algorithm：
Two time serieses X and Y are defined first, and X and Y respectively refers to the real-time time sequence that generation doctor supports strong object in the present invention
With the time series of prediction, the length of X and Y are respectively n and m, X=[x1, x2, x3..., xn], Y=[y1, y2, y3..., ym],
The distance between sequence of points and point calculation formula are Euclidean distance：
(1) matrix of a n × m is constructed first, and (i, j) element of matrix is XiAnd YjThe Euclidean distance D of two points
(XI, Yj)。
(2) regular path W, W are definedk=(i, j), k are expressed as sequence X, the mapping of Y.Regular path is from W1Start to WkKnot
Beam；Wk-1Next point WkIt cannot be matched across some point, point alignment that can only be adjacent with oneself；Wk-1Next point
WkMust be as event dullness carries out.
(3) index rank may be reached by meeting the regular path of constraints, calculate the accumulation in all regular paths
Then distance selects the path of the minimum i.e. regular Least-cost of Cumulative Distance.
A4) Cumulative Distance and threshold value are judged.If similar enough, the Cumulative Distance calculated of two sequences
It can stablize in a range, be compared according to a large amount of, threshold value is finally arranged to 5；
A5) when Cumulative Distance be more than threshold value, just progress step B compare life pattern, terminate if being not above threshold value,
Wait for matched within next 30 seconds.
Step B：Compare life pattern
B1) real time data in strong object room is supported from acquisition doctor in cloud database and be saved in local data base；
B2) doctor is obtained from local data base support the life pattern for being good for object in the room；
B3) judge that current doctor supports the behavior of strong object whether in the life pattern in the room, if the life in the room
In pattern, and within the activity time, then activity time threshold value is set, otherwise, time of having a rest threshold value is set；If not in the room
In life pattern, then be arranged not time threshold.
B4 then) opening timing task calculates the respective behavior data time preserved in current time and local data base
Time difference t1, if early warning is crossed with regard to stopping task, if no early warning is crossed and t1Abnormity early warning is just carried out more than threshold value.
C) environmental abnormality early warning
It is mainly temperature pre-warning that doctor, which supports strong object indoor environment abnormity early warning, is as follows：
C1 doctor) is got from cloud database first and supports the room data of strong object and corresponding threshold value；
C2) judge whether temperature moves in circles more than threshold value if it is less than c1 is returned to；
C3) if it is greater than threshold value, and there is no early warning, with regard to carrying out environmental abnormality early warning.
Claims (5)
1. a kind of looking after intelligent monitoring method in real time towards health endowment post house, which is characterized in that this method includes following tool
Body step：
Step 1：Acquisition doctor, which supports the initial data of strong object and stores, arrives cloud database；
Step 2：Obtained from cloud database the doctor needed support strong object data be saved in local data base and carry out in advance from
Reason；
Step 3：With Time Series Forecasting Methods to curing the method for supporting the movement track of strong object and predicting, and using cluster
Construct the life pattern that doctor supports strong object；
Step 4：Matching doctor supports the movement track for being good for object and compares the life pattern that doctor supports strong object, is carried out to curing foster strong object
Abnormal judgement and early warning.
2. looking after intelligent monitoring method in real time as described in claim 1, which is characterized in that the step 1 is specially：Pass through number
Word sensor and Intellisense terminal acquisition doctor support strong time location data and room environment of the object in the room of post house
Data are simultaneously saved in the database in high in the clouds.
3. looking after intelligent monitoring method in real time as described in claim 1, which is characterized in that the step 2 specifically includes：
Step A：Choose post house inner room number, inner room room number, action sensor, action location data, that is, time series and timestamp
Database is being locally created in field；
Step B：The initial data that corresponding doctor supports strong object is obtained from cloud database according to the field in local data base
And filter out invalid empty data；
Step C：Time data format conversion is judged into invalid data again at TimeSeriesRDD formats and deletion.
4. looking after intelligent monitoring method in real time as described in claim 1, which is characterized in that the step 3 specifically includes：
Step A：Prediction doctor supports the movement track of strong object
A1 load () method that the SQLContext objects in SparkSql provide) is used to obtain inner room room from local data base
Between number and time stamp data and be converted into dataframe data formats；
A2 time series) is initialized as TimeSeriesRDD formats and selected HoltWinters models；
A3 HoltWinters.fitModel () method that spark-timeseries is provided) is utilized to create, training
HoltWinters models；
A4 the interval for) predicting the foster strong following 3 minutes movement tracks of object of doctor and data has been arranged to 5 seconds；
A5 36 predicted values) are preserved, and return to a1 after waiting for 3 minutes, cycle executes；
Step B：Cluster doctor supports the life pattern of strong object
B1) optimize K-Means clustering algorithms, algorithm optimization flow is as follows：
(1) data are read and are converted into RDD data formats；
(2) Map operations are executed to be formatted data and vectorization；
(3) distance for calculating each fragment data to the centers Canopy obtains Local C anopy central points；
(4) merge and generate Canopy central points；
(5) K-Means initialization operations are carried out, then carries out Map operations and executes K-Means Local Clusterings；
(6) result of Local Clustering is subjected to merger, calculates global cluster node, update central point；
(7) it repeats（5）With（6）Until central point no longer changes；
B2 an object sc) is created by SparkContext, then uses textFile () method of sc to read doctor foster strong
The inner room room number and time stamp data of object are simultaneously converted into RDD data formats；
B3 it) is operated using map and the data of the foster strong object of doctor is classified according to inner room room number, provided for further cluster
The data of [timestamp, inner room room number] Format Type；
B4 it is) that the foster time data progress Canopy for being good for each inner room room of object of doctor is slightly clustered, when obtaining every inner room room
Between data Temporal Clustering, and be saved in local data base；
B5) the Temporal Clustering to cluster out at the beginning of obtaining each Temporal Clustering and the end time, generates the work in each room
The dynamic period, and it is saved in local data base；
B6) to the Temporal Clustering generated in b3, secondary cluster is carried out using the K-Means clustering algorithms of optimization, what acquisition more refined
Cluster result clusters out smaller Temporal Clustering, and is saved in local data base；
B7 the division that the period) is carried out to the Temporal Clustering of secondary cluster, calculates the spacing value between each period, to interval
It is worth the rest threshold value that this room is obtained into row median value；
B8) active period and rest threshold value that doctor is supported to strong one day each room of object, is converted into dataframe data lattice
Formula is saved in local data base.
5. looking after intelligent monitoring method in real time as described in claim 1, which is characterized in that the step 4 specifically includes：
Step A：Match movement track
A1 specified doctor) is obtained from cloud database to support nearest 3 minutes real time datas of strong object and be saved in local data
Library obtains whole real time datas if inadequate 3 minutes；
A2 real-time time sequence and the time series progress dimension-reduction treatment of prediction that strong object) is supported to curing, obtain doctor and support strong object
Inner room room number；
A3 two sequences) are subjected to dynamic time warping algorithms and carry out similitude matchings, calculate the accumulations of two sequences away from
From；
A4) Cumulative Distance and threshold value are judged；If two sequences are similar enough, the Cumulative Distance calculated can be stablized
In a range, is compared according to a large amount of, threshold value is finally arranged to 5；
A5) when Cumulative Distance be more than threshold value, just progress step B compare life pattern, terminate if being not above threshold value, etc.
It waits for being matched for next 30 seconds；
Step B：Compare life pattern
B1) real time data in strong object room is supported from acquisition doctor in cloud database and be saved in local data base；
B2) doctor is obtained from local data base support the life pattern for being good for object in the room；
B3) judge that current doctor supports the behavior of strong object whether in the life pattern in the room, if the life pattern in the room
In, and within the activity time, then activity time threshold value is set, otherwise, time of having a rest threshold value is set；If not in the life in the room
In pattern, then be arranged not time threshold；
B4) opening timing task, then calculate current time and corresponding behavioral data time for being preserved in local data base when
Between it is poor, if early warning is crossed with regard to stopping task, if no early warning cross andAbnormity early warning is just carried out more than threshold value；
Step C：Environmental abnormality early warning
Environmental abnormality early warning is temperature pre-warning, is as follows：
C1 doctor) is got from cloud database first and supports the room data of strong object and corresponding threshold value；
C2) judge whether temperature moves in circles more than threshold value if it is less than c1 is returned to；
C3) if it is greater than threshold value, and there is no early warning, with regard to carrying out environmental abnormality early warning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810190824.8A CN108417276B (en) | 2018-03-08 | 2018-03-08 | Real-time care intelligent monitoring method for health care post house |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810190824.8A CN108417276B (en) | 2018-03-08 | 2018-03-08 | Real-time care intelligent monitoring method for health care post house |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108417276A true CN108417276A (en) | 2018-08-17 |
| CN108417276B CN108417276B (en) | 2022-02-11 |
Family
ID=63130527
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810190824.8A Active CN108417276B (en) | 2018-03-08 | 2018-03-08 | Real-time care intelligent monitoring method for health care post house |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108417276B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110716925A (en) * | 2019-09-05 | 2020-01-21 | 中电科大数据研究院有限公司 | Cross-border behavior recognition method based on trajectory analysis |
| CN111523570A (en) * | 2020-04-09 | 2020-08-11 | 辽宁百思特达半导体科技有限公司 | Smart city system based on community post house and control method thereof |
| CN113807388A (en) * | 2021-08-02 | 2021-12-17 | 滁州学院 | Behavior rule discovery method for solitary old people based on multivariate sensor data |
| CN116602664A (en) * | 2023-07-17 | 2023-08-18 | 青岛市胶州中心医院 | Comprehensive diagnosis and treatment nursing system for neurosurgery patients |
| CN117371339A (en) * | 2023-12-08 | 2024-01-09 | 西电济南变压器股份有限公司 | Transformer operation monitoring system based on Internet of things |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103927847A (en) * | 2014-04-18 | 2014-07-16 | 深圳市丰泰瑞达实业有限公司 | Intelligent nursing system for old people |
| CN104270714A (en) * | 2014-09-11 | 2015-01-07 | 华为技术有限公司 | Method and device for determining action trails of user |
| CN105069731A (en) * | 2015-09-15 | 2015-11-18 | 山东爱伯特智联信息技术有限公司 | Aged care service system and method with safety and health information acquisition |
| US20170182363A1 (en) * | 2015-12-28 | 2017-06-29 | Xiamen Cowell Industrial Ltd. | Exercise Health Management Method, Terminal and System |
| CN107018492A (en) * | 2017-04-11 | 2017-08-04 | 中科院微电子研究所昆山分所 | Position monitoring method, server and its system in a kind of garden |
-
2018
- 2018-03-08 CN CN201810190824.8A patent/CN108417276B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103927847A (en) * | 2014-04-18 | 2014-07-16 | 深圳市丰泰瑞达实业有限公司 | Intelligent nursing system for old people |
| CN104270714A (en) * | 2014-09-11 | 2015-01-07 | 华为技术有限公司 | Method and device for determining action trails of user |
| CN105069731A (en) * | 2015-09-15 | 2015-11-18 | 山东爱伯特智联信息技术有限公司 | Aged care service system and method with safety and health information acquisition |
| US20170182363A1 (en) * | 2015-12-28 | 2017-06-29 | Xiamen Cowell Industrial Ltd. | Exercise Health Management Method, Terminal and System |
| CN107018492A (en) * | 2017-04-11 | 2017-08-04 | 中科院微电子研究所昆山分所 | Position monitoring method, server and its system in a kind of garden |
Non-Patent Citations (1)
| Title |
|---|
| 李佳婧: "养老设施适应性设计研究--从健康到护理", 《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 (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110716925A (en) * | 2019-09-05 | 2020-01-21 | 中电科大数据研究院有限公司 | Cross-border behavior recognition method based on trajectory analysis |
| CN110716925B (en) * | 2019-09-05 | 2023-08-04 | 中电科大数据研究院有限公司 | Cross-border behavior recognition method based on track analysis |
| CN111523570A (en) * | 2020-04-09 | 2020-08-11 | 辽宁百思特达半导体科技有限公司 | Smart city system based on community post house and control method thereof |
| CN113807388A (en) * | 2021-08-02 | 2021-12-17 | 滁州学院 | Behavior rule discovery method for solitary old people based on multivariate sensor data |
| CN113807388B (en) * | 2021-08-02 | 2022-10-28 | 滁州学院 | Behavior rule discovery method for solitary old people based on multivariate sensor data |
| CN116602664A (en) * | 2023-07-17 | 2023-08-18 | 青岛市胶州中心医院 | Comprehensive diagnosis and treatment nursing system for neurosurgery patients |
| CN116602664B (en) * | 2023-07-17 | 2023-09-22 | 青岛市胶州中心医院 | Comprehensive diagnosis and treatment nursing system for neurosurgery patients |
| CN117371339A (en) * | 2023-12-08 | 2024-01-09 | 西电济南变压器股份有限公司 | Transformer operation monitoring system based on Internet of things |
| CN117371339B (en) * | 2023-12-08 | 2024-03-26 | 西电济南变压器股份有限公司 | Transformer operation monitoring system based on Internet of things |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108417276B (en) | 2022-02-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108417276A (en) | It is a kind of to look after intelligent monitoring method in real time towards health endowment post house | |
| Wen et al. | Big data driven marine environment information forecasting: a time series prediction network | |
| Rawassizadeh et al. | Scalable daily human behavioral pattern mining from multivariate temporal data | |
| Vora et al. | FAAL: Fog computing-based patient monitoring system for ambient assisted living | |
| Tsai et al. | Enhanced parallel cat swarm optimization based on the Taguchi method | |
| Wang et al. | LDPA: A local data processing architecture in ambient assisted living communications | |
| Kovalchuk et al. | Distributed data-driven platform for urgent decision making in cardiological ambulance control | |
| Poornima et al. | A survey of predictive analytics using big data with data mining | |
| Fadlullah et al. | On delay-sensitive healthcare data analytics at the network edge based on deep learning | |
| Mittal et al. | Application of Bayesian Belief Networks for context extraction from wireless sensors data | |
| Kolomvatsos | A distributed, proactive intelligent scheme for securing quality in large scale data processing | |
| CN108717875A (en) | A kind of chronic disease intelligent management system based on big data | |
| Larcher et al. | Event-driven framework for detecting unusual patterns in AAL environments | |
| CN113207101B (en) | Information processing method based on 5G city component sensor and Internet of things cloud platform | |
| Upadhyay et al. | Periodic pattern mining from spatio-temporal database using novel global pollination artificial fish swarm optimizer-based clustering and modified FP tree | |
| Chen et al. | Periodic pattern mining based on GPS trajectories | |
| Wang et al. | A new method for discovering behavior patterns among animal movements | |
| CN111523570A (en) | Smart city system based on community post house and control method thereof | |
| Melnykova et al. | The Problem of Analysing the Relationships between Individual Characteristics of Individuals with COVID'19. | |
| Lee et al. | Incorporating big data and social sensors in a novel early warning system of dengue outbreaks | |
| Badinelli et al. | Integrating the internet of things and big data analytics into decision support models for healthcare management | |
| Huang et al. | A cloud-based accessible architecture for large-scale adl analysis services | |
| Khourdifi et al. | Hadoop and MapReduce technology as a solution for Wireless Body Area Networks in e-Health | |
| Li et al. | Spatial data analysis for intelligent buildings: Awareness of context and data uncertainty | |
| CN110909765A (en) | Pedestrian behavior mode classification method facing track big data |
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 |