CN116484307B - Cloud computing-based intelligent ring remote control method - Google Patents
Cloud computing-based intelligent ring remote control method Download PDFInfo
- Publication number
- CN116484307B CN116484307B CN202310735937.2A CN202310735937A CN116484307B CN 116484307 B CN116484307 B CN 116484307B CN 202310735937 A CN202310735937 A CN 202310735937A CN 116484307 B CN116484307 B CN 116484307B
- Authority
- CN
- China
- Prior art keywords
- data
- monitoring data
- monitoring
- data points
- abnormal
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 63
- 238000012544 monitoring process Methods 0.000 claims abstract description 206
- 230000002159 abnormal effect Effects 0.000 claims abstract description 120
- 238000001514 detection method Methods 0.000 claims abstract description 90
- 230000005856 abnormality Effects 0.000 claims abstract description 55
- 238000012216 screening Methods 0.000 claims abstract description 7
- 238000005457 optimization Methods 0.000 claims description 50
- 238000012360 testing method Methods 0.000 claims description 16
- 238000003064 k means clustering Methods 0.000 claims description 12
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 5
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 description 7
- 238000005259 measurement Methods 0.000 description 6
- 230000001133 acceleration Effects 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 239000004973 liquid crystal related substance Substances 0.000 description 4
- 238000004138 cluster model Methods 0.000 description 3
- 230000007423 decrease Effects 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 230000002650 habitual effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention relates to the technical field of intelligent ring remote control, in particular to a cloud computing-based intelligent ring remote control method. The method comprises the following steps: obtaining each target cluster and an initial abnormal data detection model; acquiring monitoring data of a user to be monitored in a current time period acquired by an intelligent ring; acquiring data points corresponding to all the monitoring data based on the monitoring data of the user to be monitored and an initial abnormal data detection model in the current time period, and determining a first abnormal degree of all the monitoring data according to the relative positions of the data points corresponding to all the monitoring data and all the target clusters; and obtaining a target abnormal data detection model according to the distribution condition and the first abnormality degree of the data points in the neighborhood of the data points corresponding to each monitoring data, and further screening the abnormal monitoring data. The invention improves the detection precision of the abnormal data.
Description
Technical Field
The invention relates to the technical field of intelligent ring remote control, in particular to a cloud computing-based intelligent ring remote control method.
Background
Remote control of the intelligent ring includes gesture recognition, touch operation, voice control and other custom shortcuts. In the process of remote control through gesture recognition, the intelligent finger ring sends data acquired by the built-in sensor to the cloud server, key feature extraction is carried out on the cloud server, different gestures are distinguished through different features, and corresponding operation is further carried out. However, a great amount of human body behavior noise exists in the collected monitoring data, so that noise and abnormality detection needs to be carried out on the collected data, and the reliability of remote control is improved.
The existing abnormality detection process is as follows: and extracting data features of the windows from the time sequence data through sliding windows, clustering the extracted features through a K-means clustering algorithm, calculating the distance from the feature value corresponding to each sliding window to the nearest clustering center, and taking the distance as a measure of the abnormal degree of the data points. In the process of intelligent ring remote control, carrying out anomaly detection on monitoring data acquired in real time, and carrying out anomaly detection on data points by online K-means clustering, wherein the process comprises the following steps: (1) Firstly, training an initial clustering model through initial data points; (2) When a new data point is acquired, firstly extracting characteristics, and then judging whether the data point is abnormal or not through a current clustering model; (3) And updating the clustering model based on all data after the number of the newly added data points reaches a number threshold, and carrying out subsequent abnormal detection on the data points by using the updated clustering model. In the existing incremental clustering process, the update of a clustering model is triggered by setting a data point quantity threshold value for data points newly added in real time, new clustering is carried out on all historical data points in the updating process to obtain a clustering result, along with the increase of monitoring data, the positions of different monitoring data characteristic values in clusters in the clustering model change, namely, the change of one data point in the clustering model in the updating process of the model, and along with the updating of the model, the detection precision of abnormal data is reduced.
Disclosure of Invention
In order to solve the problem of low detection precision in the prior art when abnormal detection is carried out on intelligent ring monitoring data, the invention aims to provide a cloud computing-based intelligent ring remote control method, which adopts the following technical scheme:
the invention provides a cloud computing-based intelligent ring remote control method, which comprises the following steps:
acquiring initial data points based on test data of the intelligent ring in a test stage, clustering the initial data points to acquire target cluster and initial abnormal data detection models; acquiring monitoring data of a user to be monitored in a current time period acquired by an intelligent ring;
acquiring data points corresponding to all the monitoring data based on the monitoring data of the user to be monitored and the initial abnormal data detection model in the current time period, and determining a first abnormal degree of all the monitoring data according to the relative positions of the data points corresponding to all the monitoring data and all the target clusters;
obtaining a distance optimization factor corresponding to each monitoring data according to the distribution condition of the data points in the neighborhood of the data point corresponding to each monitoring data and the corresponding first abnormality degree; and clustering data points corresponding to all the monitoring data based on the distance optimization factors and the relative distances between the data points corresponding to the monitoring data to obtain a target abnormal data detection model, and screening abnormal monitoring data based on the target abnormal data detection model.
Preferably, the determining the first abnormality degree of each monitoring data according to the relative positions of the data points corresponding to each monitoring data and each target cluster includes:
for any monitored data: and acquiring the minimum value of Euclidean distance between the data point corresponding to the monitoring data and the central points of all the target clusters, and taking the normalization result of the minimum value of Euclidean distance as the first abnormality degree of the monitoring data.
Preferably, the obtaining a distance optimization factor corresponding to each monitoring data according to the distribution condition of the data points in the neighborhood of the data point corresponding to each monitoring data and the corresponding first abnormality degree includes:
for any monitored data:
counting the number of data points in the reverse K neighbor of the data point corresponding to the monitoring data, and recording the number as the first number of the data point corresponding to the monitoring data; recording the ratio of the first quantity to K as a first duty cycle; wherein K is a preset first value;
respectively counting the number of data points in the reverse K neighbors of each data point in the reverse K neighbors of the data point corresponding to the monitoring data, and taking the number of data points as a second number corresponding to each data point in the reverse K neighbors of the data point corresponding to the monitoring data; the ratio of the second quantity to K is recorded as a second duty ratio corresponding to each data point in reverse K neighbors of the data point corresponding to the monitoring data;
and obtaining a distance optimization factor corresponding to the monitoring data according to the difference condition between the first duty ratio and the second duty ratio and the first abnormality degree of the monitoring data.
Preferably, according to the difference between the first duty ratio and the second duty ratio and the first abnormality degree of the monitored data, obtaining a distance optimization factor corresponding to the monitored data includes:
calculating an average value of second duty ratios corresponding to all data points in reverse K neighbors of the data points corresponding to the monitoring data; recording the absolute value of the difference between the first duty ratio and the average value as a first characteristic index;
and obtaining a distance optimization factor corresponding to the monitoring data according to the first characteristic index and the first abnormality degree of the monitoring data, wherein the first characteristic index and the first abnormality degree are in positive correlation with the distance optimization factor.
Preferably, the clustering the data points corresponding to all the monitored data based on the distance optimization factor and the relative distance between the data points corresponding to the monitored data to obtain the target abnormal data detection model includes:
determining the product of the distance between the data point corresponding to each monitoring data and the clustering center and the corresponding distance optimizing factor as the target distance between the data point corresponding to each monitoring data and the clustering center; and clustering data points corresponding to all the monitoring data by adopting a K-means clustering algorithm based on the target distance to obtain a target abnormal data detection model.
Preferably, the clustering the initial data points to obtain each target cluster and an initial abnormal data detection model includes:
clustering all initial data points by adopting a K-means clustering algorithm to obtain at least two clusters; counting the number of data points in each cluster, and sequencing all clusters according to the sequence from big to small based on the number of the data points in each cluster to obtain a cluster sequence;
judging whether the number of data points in a first cluster in the cluster sequence meets a preset condition, and if so, taking the first cluster in the cluster sequence as a target cluster; if not, judging whether the sum of the number of data points in the first cluster and the number of data points in the second cluster in the cluster sequence meets a preset condition, and the like until the sum of the number of data points meets the preset condition, and taking the corresponding cluster as a target cluster;
and taking the clustering model obtained after the initial data point clustering is completed as an initial abnormal data detection model.
Preferably, screening the anomaly monitoring data based on the target anomaly data detection model includes:
according to the distribution characteristics of the data points in the target abnormal data detection model, calculating the abnormal degree of each monitoring data again, and taking the calculated abnormal degree as a second abnormal degree of each monitoring data;
and respectively judging whether the second abnormality degree of each monitoring data is larger than a preset abnormality degree threshold value, and if so, taking the corresponding monitoring data as abnormality monitoring data.
The invention has at least the following beneficial effects:
1. according to the method, initial data points are obtained based on test data of the intelligent ring in a test stage, an initial abnormal data detection model is obtained by clustering the initial data points, then the distance of each monitoring data in the clustering process of the next time is optimized by measuring the distance of the data points of the monitoring data of a user to be monitored in the initial abnormal data detection model in the current time period, and then the initial abnormal data detection model is updated to obtain a target abnormal data detection model, namely the abnormal data detection model is updated.
2. According to the method, when the distance between the data points corresponding to the monitoring data is optimized, the first abnormality degree of each monitoring data is determined according to the relative position between the data points corresponding to each monitoring data and the target cluster, the distance between the data points in the clustering process is optimized by integrating the position of the data points corresponding to each monitoring data in the data space and the local spatial distribution condition, namely, the data points in the deviating state in the initial abnormal data detection model are integrally judged through the distribution characteristics of the data points in the neighborhood range and the first abnormality degree, the distance optimization factor corresponding to each monitoring data is further determined, the distance in the next clustering is optimized based on the distance optimization factor, so that abnormal data cannot be misjudged as normal data in the next clustering process, and the detection precision of the abnormal data is improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a cloud computing-based intelligent ring remote control method according to an embodiment of the present invention.
Detailed Description
In order to further explain the technical means and effects adopted by the invention to achieve the preset aim, the following describes in detail a cloud computing-based intelligent ring remote control method according to the invention with reference to the attached drawings and the preferred embodiment.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The invention provides a cloud computing-based intelligent ring remote control method.
Cloud computing-based intelligent ring remote control method embodiment:
the specific scene aimed at by this embodiment is: the intelligent ring is worn for a user to be monitored, the built-in sensor of the intelligent ring is utilized to collect monitoring data of the user to be monitored in real time, the collected monitoring data are sent to the cloud server, key feature extraction is carried out by the cloud server, different gestures are distinguished through different features, corresponding operation is further carried out, but a large amount of human body behavior noise exists in the monitoring data collected by the intelligent ring, the reliability of remote control of the subsequent intelligent ring is reduced due to the existence of noise and abnormal data, therefore, in the remote control of the intelligent ring, abnormal detection needs to be carried out on the monitoring data collected by the intelligent ring, in the process of detecting real-time abnormal data based on local abnormal factors of clustering in the cloud computing platform, the collected monitoring data are optimized for the distance of data points in an optimization model in the clustering process through abnormal degree change of the data points in a historical model, the abnormal data detection accuracy of the model is guaranteed, and the reliability of remote control of the intelligent ring is further improved.
The embodiment provides a cloud computing-based intelligent ring remote control method, as shown in fig. 1, which comprises the following steps:
step S1, obtaining initial data points based on test data of the intelligent ring in a test stage, clustering the initial data points, and obtaining target cluster groups and an initial abnormal data detection model; and acquiring monitoring data of a user to be monitored in the current time period acquired by the intelligent ring.
Firstly, wearing an intelligent ring for a user to be monitored, and collecting monitoring data of the user to be monitored by utilizing a built-in sensor of the intelligent ring, wherein the monitoring data comprise acceleration and angular velocity; in the embodiment, the monitoring data are collected every 1 second, so that every 1 second corresponds to one acceleration and one angular velocity, and in a specific application, an implementer can set the collection frequency of the detection data according to specific conditions; the monitoring data of the user to be monitored in the current time period is monitored abnormally, the current time period is a set formed by historical time with a time interval smaller than or equal to the preset time length from the current time, the preset time length in the embodiment is one week, so that the current time period in the embodiment is the last week, and in specific application, an implementer can set according to specific conditions. So far, the monitoring data of each acquisition moment of the user to be monitored in the current time period are obtained.
In the process of carrying out anomaly detection on monitoring data of a user to be monitored, which are collected by the intelligent ring, the anomaly data caused by the conditions of rapid movement or waving of the hand and the like need to be screened out, and the anomaly data are removed. In the process of performing incremental optimization on the existing cluster-based local anomaly factor detection (namely CBLOF) model, because cluster classification in the CBLOF is changed due to change of data space, when anomaly data points are increased, the original anomaly degree of the anomaly data points in the CBLOF model is reduced due to increase of the anomaly data points, and therefore the cluster classification in the process of optimizing the cluster model needs to be adjusted according to the anomaly degree of the incremental data points in the process of incremental clustering.
According to the embodiment, test data of each acquisition moment of the intelligent ring in a test stage are acquired, each test data comprises an acceleration and an angular velocity, namely, a two-dimensional data point can be acquired based on each test data, the two-dimensional data point is subjected to characteristic extraction of a sliding window, the extracted characteristic is the average acceleration and the average angular velocity in a window with the length of L=11, namely, a characteristic value corresponding to each test data is acquired, the characteristic value corresponding to each test data is respectively put into a data space, a plurality of data points are acquired, and the data points acquired at the moment are recorded as initial data points. In a specific application, the window length may be adjusted based on the sensor sampling frequency of the intelligent ring.
And clustering all the initial data points by adopting K-means clustering to obtain a plurality of clustering clusters, and taking a clustering model obtained after the initial data point clustering is completed as an initial abnormal data detection model. In this embodiment, the value of K is set to 20 when K-means is clustered, and in a specific application, an implementer can set according to a specific situation.
After a clustering result is obtained, counting the number of data points in each cluster, and based on the number of the data points in each cluster, sequencing all the clusters in a sequence from large to small to obtain a cluster sequence, judging whether the number of the data points in a first cluster in the cluster sequence meets a preset condition, and if so, taking the first cluster in the cluster sequence as a target cluster; if not, judging whether the sum of the number of the data points in the first cluster and the number of the data points in the second cluster in the cluster sequence meets a preset condition, and the like until the sum of the number of the data points meets the preset condition, and taking the corresponding cluster as a target cluster. The preset condition in this embodiment is that the sum of the number of data points is greater than 80% of the total number of all data points in the initial abnormal data detection model. Therefore, in this embodiment, when the number of data points in the first cluster in the cluster sequence is greater than 80% of the total number of all data points in the initial abnormal data detection model, the first cluster in the cluster sequence is taken as the target cluster; when the number of data points in the first cluster in the cluster sequence is smaller than or equal to 80% of the total number of all data points in the initial abnormal data detection model, if the sum value of the number of data points in the first cluster in the cluster sequence and the number of data points in the second cluster is larger than 80% of the total number of all data points in the initial abnormal data detection model, the first cluster and the second cluster in the cluster sequence are used as target clusters; if the sum of the number of data points in the first cluster and the number of data points in the second cluster in the cluster sequence is less than or equal to 80% of the total number of all data points in the initial abnormal data detection model, judging the size relation between the sum of the number of data points in the first cluster, the number of data points in the second cluster and the number of data points in the third cluster in the cluster sequence and 80% of the total number of all data points in the initial abnormal data detection model, and so on until the sum of the number of data points meets a preset condition, and taking the corresponding cluster as a target cluster when the preset condition is met. The target cluster is a large cluster with a large number of data points, and the clusters except the target cluster are small clusters with a small number of data points. In a specific application, the practitioner may set the preset conditions according to the specific situation.
So far, the clustering model in the CBLOF for real-time abnormal data detection is obtained through the initial data points, namely the initial abnormal data detection model is obtained, and the target cluster in the initial abnormal data detection model is obtained.
Step S2, obtaining data points corresponding to all the monitoring data based on the monitoring data of the user to be monitored and the initial abnormal data detection model in the current time period, and determining a first abnormal degree of all the monitoring data according to the relative positions of the data points corresponding to all the monitoring data and all the target clusters.
After the initial abnormal data detection model is obtained, the abnormal detection of the monitoring data collected by the intelligent ring can be started, and in the process of abnormal detection of the monitoring data collected by the intelligent ring, the collected data volume is continuously increased, so that after the new data increment of the monitoring data reaches a threshold value, namely, after the monitoring data of each collection moment in the current time period is obtained, the initial abnormal data detection model is updated. In the updating process, the importance degree of the data points in the initial abnormal data detection model in the model updating process is judged by combining the importance degree of the data points in the data space, and the distance measurement is carried out on the data points, so that the abnormal detection precision of the data points is ensured not to be reduced due to the increase of the data points in the model updating process.
After the initial abnormal data detection model is obtained, the monitoring data of each acquisition moment of the user to be monitored in the current time period are obtained through the feature extraction process in the step S1, the two-dimensional values corresponding to the monitoring data in each current time period are put into the initial abnormal data detection model to obtain data points corresponding to the monitoring data, the center point of the nearest target cluster is selected through the position of each data point, and the nearest distance between the data point and the center point of the target cluster is used as a measurement basis of the abnormality degree of each monitoring data. When the euclidean distance between the data point corresponding to the monitoring data and the larger cluster is longer, the more likely that the corresponding monitoring data is abnormal data, namely the greater the corresponding abnormality degree, so the first abnormality degree of each monitoring data is determined by combining the euclidean distance between the data point corresponding to each monitoring data and the central point of the target cluster.
Specifically, for any monitored data: and acquiring the minimum value of Euclidean distance between the data point corresponding to the monitoring data and the central points of all the target clusters, and taking the normalization result of the minimum value of Euclidean distance as the first abnormality degree of the monitoring data. The specific calculation formula of the first abnormality degree of the ith monitoring data is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,first degree of abnormality for the ith monitored data, +.>For the minimum value of Euclidean distance between the data point corresponding to the ith monitoring data and the central points of all target clusters, +.>For detecting the minimum value of Euclidean distance between every two data points in the model of the initial abnormal data, +.>The maximum value of Euclidean distance between every two data points in the model is detected for the initial abnormal data.
Introducing the maximum value and the minimum value of the Euclidean distance between every two data points in the initial abnormal data detection model into a calculation formula of the first abnormal degree to perform linear normalization processing on the minimum value of the Euclidean distance between the data point corresponding to the ith monitoring data and the central points of all target cluster; when the minimum value of Euclidean distance between the data point corresponding to the ith monitoring data and the central points of all target clusters is smaller, the less likely the ith monitoring data is abnormal data, namely the smaller the corresponding first abnormality degree is; when the minimum value of Euclidean distance between the data point corresponding to the ith monitoring data and the central points of all target clusters is larger, the more likely the ith monitoring data is abnormal data, namely the greater the corresponding first abnormality degree is.
The first abnormality degree of each monitoring data in the current time period is obtained by combining the initial abnormality data detection model.
Step S3, obtaining a distance optimization factor corresponding to each monitoring data according to the distribution condition of the data points in the neighborhood of the data point corresponding to each monitoring data and the corresponding first abnormality degree; and clustering data points corresponding to all the monitoring data based on the distance optimization factors and the relative distances between the data points corresponding to the monitoring data to obtain a target abnormal data detection model, and screening abnormal monitoring data based on the target abnormal data detection model.
After the first degree of abnormality of each monitoring data in the current time period is obtained, the embodiment performs approximate optimization of the neighbor data points on the numerical value according to the first degree of abnormality of each monitoring data, after the optimization, the data points are put into a newly added data point set, wherein the newly added data point set is formed by the data points corresponding to the monitoring data collected in the current time period, and after the number of the data points in the newly added data point set reaches a number threshold, namely, after the monitoring data of each collection time in the current time period is obtained, the data points and the data points in the initial abnormal data detection model are put together to perform optimization of the CBLOF model. However, in this process, abnormal data points may continuously enter the model, which may cause more and more abnormal data points, and then some abnormal data points enter a large cluster during the K-means clustering process, which causes the accuracy of the subsequently optimized model for abnormal data point detection to be reduced.
And the reference value in the next model optimization process is measured by the distance between the data points in the current newly added data point set (namely the set formed by the data points corresponding to the monitoring data acquired in the current time period) in the last model. The data points in the local area in the next measurement process are judged through the distance measurement of the data points in the previous model in the model updating process.
For a newly added data point set, namely a set formed by data points corresponding to all monitoring data in the current time period, the first abnormality degree determined by the position of the newly added data point set in the last CBLOF clustering model (namely an initial abnormal data detection model) is used as a distance optimization basis of the data points in a clustering iteration process, because the center point of a clustering cluster can deviate due to the newly added data points in the new clustering process, and in order to ensure the detection precision of abnormal data points, the local distribution situation of the data points needs to be measured. Therefore, in this embodiment, the distribution condition of the data points in the neighborhood of the data point corresponding to each monitored data in the current time period and the corresponding first abnormality degree are combined to determine the distance optimization factor corresponding to each monitored data, so as to perform the next clustering, and obtain a new abnormal data detection model.
Specifically, for any monitored data: counting the number of data points in the reverse K neighbor of the data point corresponding to the monitoring data, and recording the number as the first number of the data point corresponding to the monitoring data; recording the ratio of the first quantity to K as a first duty cycle; wherein K is a preset first value; respectively counting the number of data points in the reverse K neighbors of each data point in the reverse K neighbors of the data point corresponding to the monitoring data, and taking the number of data points as a second number corresponding to each data point in the reverse K neighbors of the data point corresponding to the monitoring data; the ratio of the second quantity to K is recorded as a second duty ratio corresponding to each data point in reverse K neighbors of the data point corresponding to the monitoring data; calculating an average value of second duty ratios corresponding to all data points in reverse K neighbors of the data points corresponding to the monitoring data; recording the absolute value of the difference between the first duty ratio and the average value as a first characteristic index; and obtaining a distance optimization factor corresponding to the monitoring data according to the first characteristic index and the first abnormality degree of the monitoring data, wherein the first characteristic index and the first abnormality degree are in positive correlation with the distance optimization factor. The positive correlation relationship indicates that the dependent variable increases along with the increase of the independent variable, the dependent variable decreases along with the decrease of the independent variable, and the specific relationship can be multiplication relationship, addition relationship, idempotent of an exponential function and is determined by practical application; as a specific embodiment, a specific calculation formula of the distance optimization factor corresponding to the monitoring data is given. The specific calculation formula of the distance optimization factor corresponding to the ith monitoring data is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,for the distance optimization factor corresponding to the ith monitoring data, K is a preset first value, < ->First degree of abnormality for the ith monitored data, +.>The number of data points in the reverse K-nearest neighbor of the data point corresponding to the ith monitored data, +.>The number of data points in the reverse K neighbor of the q-th data point in the reverse K neighbor of the data point corresponding to the ith monitoring data, +.>And e is a natural constant, and I is an absolute value sign for the total number of the monitoring data acquired in the current time period.
In this embodiment, the first value is preset to be 20, and in a specific application, the practitioner may set the first value according to a specific situation.
Representing a first number of data points corresponding to the ith monitoring data, namely the number of data points in a reverse K neighbor of the data point corresponding to the ith monitoring data; />Representing a first duty ratio, wherein the first duty ratio is used for representing the number duty ratio of data points in the reverse nearest neighbor of the data point corresponding to the ith monitoring data; />Representing a second number of data points corresponding to the q-th data point in the reverse K neighbor of the data point corresponding to the i-th monitoring data, namely the number of data points in the reverse K neighbor of the q-th data point in the reverse K neighbor of the data point corresponding to the i-th monitoring data; />Representing a second duty ratio corresponding to the q-th data point in the reverse K neighbor of the data point corresponding to the i-th monitoring data, namely, the number duty ratio of the data points in the reverse K neighbor of the q-th data point in the reverse K neighbor of the data point corresponding to the i-th monitoring data; />Representing the average value of the second duty ratios corresponding to all data points in the reverse K neighbor of the data point corresponding to the ith monitoring data; />The difference between the first duty cycle and the mean value of the second duty cycle corresponding to all data points in the reverse K neighbor of the data point corresponding to the ith monitored data is characterized for the first characteristic index. In the embodiment, the distribution condition of the data points corresponding to the ith monitoring data and the average reverse nearest neighbor data points of the surrounding data points is judged, the corresponding abnormality degree of the monitoring data corresponding to the ith monitoring data in the last clustering process is combined, the distance optimization factor corresponding to the ith monitoring data is determined, and the ∈10 is introduced into the calculation formula of the distance optimization factor>Is for->And (5) carrying out normalization processing. When the difference between the number of data points in the reverse K-nearest neighbor of the data point corresponding to the i-th monitoring data and the average number of data points in the reverse K-nearest neighbor of the data point in the reverse K-nearest neighbor thereof is greater and the first degree of abnormality of the i-th monitoring data is greater,the data point corresponding to the ith monitored data is deviated from the local density distribution, and the distance measurement of the data point in the next clustering process is larger, namely the distance optimization factor corresponding to the ith monitored data is larger.
By adopting the method, the distance optimization factor corresponding to each monitoring data in the current time period can be obtained and used for updating the initial abnormal data detection model subsequently.
According to the method, the distance of the data point in the clustering process is optimized through the distance and the local space distribution condition of the data point corresponding to the monitoring data, the data point is fused through the local density of the data point, namely, for the data point in the deviating state in the current clustering model, the integral judgment is carried out through the number of the data points in the neighbor range, the data point fusion of the local area is carried out on the data point through the data point state of the neighbor and the reverse nearest neighbor, and the number of surrounding data points is reduced, so that the data point cannot become a big cluster in the next clustering model optimizing process, namely, cannot be misjudged as normal data, and the abnormal detection precision of the monitoring data is improved.
After the distance optimization factor is obtained, the distance between the data points in the clustering iterative process is optimized through the distance optimization factor. Specifically, determining the product of the distance between the data point corresponding to each monitoring data and the clustering center and the corresponding distance optimizing factor as the target distance between the data point corresponding to each monitoring data and the clustering center; and clustering data points corresponding to all the monitoring data by adopting a K-means clustering algorithm based on the target distance to obtain a clustering result. According to the embodiment, the distance optimization factor is used for optimizing the objective function of the K-means clustering in the initial abnormal data detection model, the optimized objective function is obtained, and the clustering result is obtained by clustering data points corresponding to all monitoring data according to the optimized objective function. The K-means clustering algorithm is the prior art and will not be described in detail here. The optimized objective function specifically comprises the following steps:
wherein F is an optimized objective function,distance optimization factor corresponding to ith monitoring data, < ->For the data point corresponding to the ith monitored data, < +.>For the center point of the jth cluster, < +.>And (3) taking the Euclidean distance between the data point corresponding to the ith monitoring data and the center point of the jth cluster, wherein M is the number of the monitoring data in the current time period, and min () is a minimum function.
According to the embodiment, the distance in the clustering process is optimized based on the distance optimization factor corresponding to each monitoring data, an optimized clustering model and a new clustering result are obtained, and the clustering model obtained after the clustering is completed is recorded as a target abnormal data detection model corresponding to the current time period.
According to the distribution characteristics of the data points in the target abnormal data detection model, calculating the abnormal degree of each monitoring data again, and taking the calculated abnormal degree as a second abnormal degree of each monitoring data; it should be noted that, when calculating the second degree of abnormality of the monitoring data, only all the monitoring data in the current time period are put into the target abnormal data detection model, and the second degree of abnormality of each monitoring data is calculated according to the relative positions of the data points corresponding to each monitoring data and the target cluster in the target abnormal data detection model, and the calculation method is the same as the calculation method of the first degree of abnormality in step S2, so that the embodiment will not be repeated. After the second abnormality degree of each monitoring data is obtained, judging whether the abnormality index of each monitoring data is larger than a preset abnormality degree threshold value or not respectively, and if so, taking the corresponding monitoring data as abnormality monitoring data; and if the monitoring data is smaller than or equal to the normal monitoring data, the corresponding monitoring data is used as the normal monitoring data. The preset abnormality degree threshold in this embodiment is 0.7, and in a specific application, the practitioner may set according to a specific situation.
In the embodiment, the distance measurement between each data point and the central point of the cluster is optimized through the distance optimization factor in the clustering process of the K-means, so that the influence of the newly added data points is smaller in the clustering process, the detection precision of the optimized cluster model in the abnormal detection of the monitoring data acquired by the intelligent ring is higher, the habitual abnormal data of personal remote control of a user is reserved, and the remote control can be realized more and more accurately along with the continuous optimization of the cluster model.
The method comprises the steps of carrying out distance optimization on data points through a distance optimization factor, carrying out incremental clustering model optimization on newly-added monitoring data, emptying a newly-added data set after abnormal monitoring data is obtained through an objective function in a K-means clustering algorithm in the optimized model, starting abnormal condition judgment of the newly-added monitoring data, and updating an abnormal data detection model by adopting the method provided by the embodiment, wherein the monitoring data in abnormal monitoring of the new monitoring data is the monitoring data collected by an intelligent ring in the next time period of the current time period. By adopting the method provided by the embodiment, the screening of the abnormal monitoring data is completed, and the habitual abnormal data of the personal remote control of the user to be monitored is reserved, so that the remote control of the intelligent ring is realized along with the continuous optimization of the clustering model.
According to the method, initial data points are obtained based on test data of the intelligent ring in a test stage, an initial abnormal data detection model is obtained by clustering the initial data points, then the distance of monitoring data of a user to be monitored in a current time period in the initial abnormal data detection model is optimized by measuring the distance of each monitoring data in the next clustering process, and then the initial abnormal data detection model is updated to obtain a target abnormal data detection model, namely the updating of the abnormal data detection model is completed. In the embodiment, when the distance between the data points corresponding to the monitoring data is optimized, the first abnormal degree of each monitoring data is determined according to the relative position between the data points corresponding to each monitoring data and the target cluster, the distance between the data points in the clustering process is optimized by integrating the position of the data points corresponding to each monitoring data in the data space and the local spatial distribution condition, namely, the data points in the deviating state in the initial abnormal data detection model are integrally judged through the distribution characteristics of the data points in the neighborhood range and the first abnormal degree, so that the distance optimization factor corresponding to each monitoring data is determined, the distance in the next clustering is optimized based on the distance optimization factor, the abnormal data is prevented from being misjudged as normal data in the next clustering process, and the detection precision of the abnormal data is improved.
Claims (6)
1. The cloud computing-based intelligent ring remote control method is characterized by comprising the following steps of:
acquiring initial data points based on test data of the intelligent ring in a test stage, clustering the initial data points to acquire target cluster and initial abnormal data detection models; acquiring monitoring data of a user to be monitored in a current time period acquired by an intelligent ring;
acquiring data points corresponding to all the monitoring data based on the monitoring data of the user to be monitored and the initial abnormal data detection model in the current time period, and determining a first abnormal degree of all the monitoring data according to the relative positions of the data points corresponding to all the monitoring data and all the target clusters;
obtaining a distance optimization factor corresponding to each monitoring data according to the distribution condition of the data points in the neighborhood of the data point corresponding to each monitoring data and the corresponding first abnormality degree; clustering data points corresponding to all monitoring data based on the distance optimization factors and the relative distances between the data points corresponding to the monitoring data to obtain a target abnormal data detection model, and screening abnormal monitoring data based on the target abnormal data detection model;
obtaining a distance optimization factor corresponding to each monitoring data according to the distribution condition of the data points in the neighborhood of the data point corresponding to each monitoring data and the corresponding first abnormality degree, including:
for any monitored data:
counting the number of data points in the reverse K neighbor of the data point corresponding to the monitoring data, and recording the number as the first number of the data point corresponding to the monitoring data; recording the ratio of the first quantity to K as a first duty cycle; wherein K is a preset first value;
respectively counting the number of data points in the reverse K neighbors of each data point in the reverse K neighbors of the data point corresponding to the monitoring data, and taking the number of data points as a second number corresponding to each data point in the reverse K neighbors of the data point corresponding to the monitoring data; the ratio of the second quantity to K is recorded as a second duty ratio corresponding to each data point in reverse K neighbors of the data point corresponding to the monitoring data;
and obtaining a distance optimization factor corresponding to the monitoring data according to the difference condition between the first duty ratio and the second duty ratio and the first abnormality degree of the monitoring data.
2. The cloud computing-based intelligent ring remote control method according to claim 1, wherein determining the first degree of abnormality of each monitoring data according to the relative positions of the data points corresponding to each monitoring data and each target cluster comprises:
for any monitored data: and acquiring the minimum value of Euclidean distance between the data point corresponding to the monitoring data and the central points of all the target clusters, and taking the normalization result of the minimum value of Euclidean distance as the first abnormality degree of the monitoring data.
3. The cloud computing-based intelligent ring remote control method according to claim 1, wherein obtaining the distance optimization factor corresponding to the monitored data according to the difference between the first duty ratio and the second duty ratio and the first abnormality degree of the monitored data comprises:
calculating an average value of second duty ratios corresponding to all data points in reverse K neighbors of the data points corresponding to the monitoring data; recording the absolute value of the difference between the first duty ratio and the average value as a first characteristic index;
and obtaining a distance optimization factor corresponding to the monitoring data according to the first characteristic index and the first abnormality degree of the monitoring data, wherein the first characteristic index and the first abnormality degree are in positive correlation with the distance optimization factor.
4. The cloud computing-based intelligent ring remote control method according to claim 1, wherein clustering all data points corresponding to the monitored data based on the distance optimization factor and the relative distance between the data points corresponding to the monitored data to obtain a target abnormal data detection model comprises:
determining the product of the distance between the data point corresponding to each monitoring data and the clustering center and the corresponding distance optimizing factor as the target distance between the data point corresponding to each monitoring data and the clustering center; and clustering data points corresponding to all the monitoring data by adopting a K-means clustering algorithm based on the target distance to obtain a target abnormal data detection model.
5. The cloud computing-based intelligent ring remote control method according to claim 1, wherein the clustering of the initial data points to obtain each target cluster and an initial abnormal data detection model comprises:
clustering all initial data points by adopting a K-means clustering algorithm to obtain at least two clusters; counting the number of data points in each cluster, and sequencing all clusters according to the sequence from big to small based on the number of the data points in each cluster to obtain a cluster sequence;
judging whether the number of data points in a first cluster in the cluster sequence meets a preset condition, and if so, taking the first cluster in the cluster sequence as a target cluster; if not, judging whether the sum of the number of data points in the first cluster and the number of data points in the second cluster in the cluster sequence meets a preset condition, and the like until the sum of the number of data points meets the preset condition, and taking the corresponding cluster as a target cluster;
and taking the clustering model obtained after the initial data point clustering is completed as an initial abnormal data detection model.
6. The cloud computing-based intelligent ring remote control method according to claim 1, wherein screening anomaly monitoring data based on the target anomaly data detection model comprises:
according to the distribution characteristics of the data points in the target abnormal data detection model, calculating the abnormal degree of each monitoring data again, and taking the calculated abnormal degree as a second abnormal degree of each monitoring data;
and respectively judging whether the second abnormality degree of each monitoring data is larger than a preset abnormality degree threshold value, and if so, taking the corresponding monitoring data as abnormality monitoring data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310735937.2A CN116484307B (en) | 2023-06-21 | 2023-06-21 | Cloud computing-based intelligent ring remote control method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310735937.2A CN116484307B (en) | 2023-06-21 | 2023-06-21 | Cloud computing-based intelligent ring remote control method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116484307A CN116484307A (en) | 2023-07-25 |
| CN116484307B true CN116484307B (en) | 2023-09-19 |
Family
ID=87214145
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310735937.2A Active CN116484307B (en) | 2023-06-21 | 2023-06-21 | Cloud computing-based intelligent ring remote control method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116484307B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116792370B (en) * | 2023-08-29 | 2023-11-21 | 德州华海石油机械股份有限公司 | TPB intelligent safety alarm method and system |
| CN116957421B (en) * | 2023-09-20 | 2024-01-05 | 山东济宁运河煤矿有限责任公司 | Washing and selecting production intelligent monitoring system based on artificial intelligence |
| CN117251749B (en) * | 2023-11-17 | 2024-02-27 | 陕西通信规划设计研究院有限公司 | Data processing method of Internet of things based on incremental analysis |
| CN117272216B (en) * | 2023-11-22 | 2024-02-09 | 中国建材检验认证集团湖南有限公司 | Data analysis method for automatic flow monitoring station and manual water gauge observation station |
| CN117688410B (en) * | 2024-02-02 | 2024-05-24 | 山东同利新材料有限公司 | Intelligent management method for production data of diethyl maleate |
| CN117807551B (en) * | 2024-02-29 | 2024-05-24 | 深圳市魔样科技股份有限公司 | Heart rate abnormality capturing method and system based on intelligent ring |
| CN117877736B (en) * | 2024-03-12 | 2024-05-24 | 深圳市魔样科技股份有限公司 | Intelligent ring abnormal health data early warning method based on machine learning |
| CN117972618A (en) * | 2024-04-01 | 2024-05-03 | 青岛航天半导体研究所有限公司 | Method and system for detecting secondary power failure of hybrid integrated circuit |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2019153309A (en) * | 2018-03-05 | 2019-09-12 | 株式会社アールティ | Image processing device, bin picking system, image processing method, and image processing program |
| CN113792754A (en) * | 2021-08-12 | 2021-12-14 | 国网江西省电力有限公司电力科学研究院 | Method for processing DGA (differential global alignment) online monitoring data of converter transformer by removing different elements and then repairing |
| CN115270986A (en) * | 2022-08-08 | 2022-11-01 | 贵州电网有限责任公司电力科学研究院 | Data anomaly detection method and device and computer equipment |
| CN116029842A (en) * | 2023-03-28 | 2023-04-28 | 北京环球医疗救援有限责任公司 | Cleaning and denoising method and system for medical insurance big data |
-
2023
- 2023-06-21 CN CN202310735937.2A patent/CN116484307B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2019153309A (en) * | 2018-03-05 | 2019-09-12 | 株式会社アールティ | Image processing device, bin picking system, image processing method, and image processing program |
| CN113792754A (en) * | 2021-08-12 | 2021-12-14 | 国网江西省电力有限公司电力科学研究院 | Method for processing DGA (differential global alignment) online monitoring data of converter transformer by removing different elements and then repairing |
| CN115270986A (en) * | 2022-08-08 | 2022-11-01 | 贵州电网有限责任公司电力科学研究院 | Data anomaly detection method and device and computer equipment |
| CN116029842A (en) * | 2023-03-28 | 2023-04-28 | 北京环球医疗救援有限责任公司 | Cleaning and denoising method and system for medical insurance big data |
Non-Patent Citations (1)
| Title |
|---|
| 基于异常检测的K-means改进算法研究;薛晨杰 等;《软件导刊》(第4期);第74-78页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116484307A (en) | 2023-07-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116484307B (en) | Cloud computing-based intelligent ring remote control method | |
| Wang et al. | A comparative study on human activity recognition using inertial sensors in a smartphone | |
| Candás et al. | An automatic data mining method to detect abnormal human behaviour using physical activity measurements | |
| EP3090684A1 (en) | Pedometer and method for analyzing motion data | |
| EP3090685A1 (en) | Pedometer and method for analyzing motion data | |
| WO2021128923A1 (en) | Exercise amount measurement method and apparatus, and electronic device | |
| CN102855634A (en) | Image detection method and image detection device | |
| CN116821809B (en) | Vital sign data acquisition system based on artificial intelligence | |
| CN112464738B (en) | Improved naive Bayes algorithm user behavior identification method based on mobile phone sensor | |
| CN116541732B (en) | Meteorological monitoring system based on ultrasonic data and optimization algorithm | |
| CN105243285A (en) | Big data health forecast system | |
| CN103892830A (en) | Emotion detection method and system based on human skin resistance changes | |
| CN112690761B (en) | Sleep state detection method, device, equipment and computer readable medium | |
| CN110123280A (en) | A kind of finger dexterity detection method based on the identification of intelligent mobile terminal operation behavior | |
| CN117315886B (en) | UWB radar-based method and device for detecting impending falling of personnel | |
| CN116028914B (en) | Intelligent finger ring identity authentication method and system | |
| CN115700595A (en) | Identity recognition method and device based on radio frequency fingerprint deep learning | |
| CN114903475A (en) | Health detector based on PCANet neural network and multi-sensor information fusion and method thereof | |
| KR20140020440A (en) | Activity diagnosis apparatus and method of the same | |
| CN114065335A (en) | Building energy consumption prediction method based on multi-scale convolution cyclic neural network | |
| Duffy et al. | An investigation into smartphone based weakly supervised activity recognition systems | |
| CN117807551B (en) | Heart rate abnormality capturing method and system based on intelligent ring | |
| CN117828282B (en) | Data efficient processing method based on adaptive filtering | |
| CN109907730A (en) | A kind of data processing method, device and storage medium and computer equipment | |
| Levasseur et al. | Time Series Representation for Real-World Applications of Deep Neural Networks |
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 | ||
| CP03 | Change of name, title or address | ||
| CP03 | Change of name, title or address |
Address after: 518000 1st, 2nd, 4th, 5th and 6th floors of Qiufen (10A) Building, Taihua wutong Industrial Park, Sanwei Community, Hangcheng Street, Bao'an District, Shenzhen, Guangdong Patentee after: Shenzhen Moyang Technology Co.,Ltd. Country or region after: China Address before: 518000 8th and 9th floors of Guizhu Building 15, Taihua wutong Industrial Park, Sanwei Community, Hangcheng Street, Bao'an District, Shenzhen, Guangdong Patentee before: SHENZHEN MOYOUNG TECHNOLOGY CO.,LTD. Country or region before: China |