CN113886669B - Self-adaptive clustering method for electricity user portraits - Google Patents

Self-adaptive clustering method for electricity user portraits Download PDF

Info

Publication number
CN113886669B
CN113886669B CN202111248953.6A CN202111248953A CN113886669B CN 113886669 B CN113886669 B CN 113886669B CN 202111248953 A CN202111248953 A CN 202111248953A CN 113886669 B CN113886669 B CN 113886669B
Authority
CN
China
Prior art keywords
cluster
data
clustering
class
fusion
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
Application number
CN202111248953.6A
Other languages
Chinese (zh)
Other versions
CN113886669A (en
Inventor
李涛
李凌
吕雪涛
汪波
何年容
王星
贾冰蕾
李成
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
HUBEI ELECTRIC POWER Co JINGZHOU POWER SUPPLY Co
State Grid Corp of China SGCC
Original Assignee
HUBEI ELECTRIC POWER Co JINGZHOU POWER SUPPLY Co
State Grid Corp of China SGCC
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by HUBEI ELECTRIC POWER Co JINGZHOU POWER SUPPLY Co, State Grid Corp of China SGCC filed Critical HUBEI ELECTRIC POWER Co JINGZHOU POWER SUPPLY Co
Priority to CN202111248953.6A priority Critical patent/CN113886669B/en
Publication of CN113886669A publication Critical patent/CN113886669A/en
Application granted granted Critical
Publication of CN113886669B publication Critical patent/CN113886669B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/906—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/903—Querying
    • G06F16/9035—Filtering based on additional data, e.g. user or group profiles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • 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
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computational Linguistics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to an adaptive clustering method for electric power user portraits, belonging to the technical field of data mining. The method comprises the steps of realizing feature extraction by adopting an automatic encoder principle, and reducing the dimension of high-dimension data by using a proper square loss function to obtain low-dimension information with higher information density; performing cluster analysis by adopting K-means algorithm operation, and obtaining initial cluster category in low dimensionality by the low dimensionality information; adopting a unimodal statistical test as a basic algorithm of fusion to carry out category fusion; feature extraction, cluster analysis and class fusion optimization are integrated, a cluster mode is constructed, a single peak statistical test value among classes is calculated after initial cluster classes are obtained, and the inter-class fusion is carried out according to the value; the method has the advantages that the proper number of class clusters is obtained under the condition that the number of the classes is not known in advance, and the clustering performance is effectively improved. The method solves the problems that the prior art replaces one cluster parameter with other parameters, has poor and satisfactory effect on constructing a cluster mode by large-scale high-dimension data, and the cluster performance is unsatisfactory.

Description

Self-adaptive clustering method for electricity user portraits
Technical Field
The invention relates to an adaptive clustering method for electric power user portraits, belonging to the technical field of data mining.
Background
According to the basic attribute, electricity consumption behavior, payment behavior and demand behavior of the power users, feature classification and grading are carried out, typical features are extracted from each type, a threshold value is given to the tags, individual and group portraits of the power users are carried out according to the final tags and the service demand scene, the customer behavior is predicted, the electricity consumption is accurately estimated, the power supply loss is reduced, the service satisfaction is improved, and the electric energy is saved, so that the method is a vital work of power enterprises today. The individual and group portraits of the power users are carried out, the data is firstly divided into clusters of similar data points in a large amount of unlabeled data, but in practice, the number of clusters is not known, the construction of a cluster mode is very troublesome, the existing K-based representation framework, the feature extraction framework and the density-based framework are used for constructing the cluster mode, more complex parameters are exchanged by using a cluster number parameter which is easier to understand, and the detected cluster number is controlled by the cluster parameter which is easier to understand to a great extent; such a method of replacing one cluster parameter with another parameter is for a large high-dimensional data set consisting of: for modern data clusters composed of images, videos and texts, the constructed clustering mode is bad and satisfactory, and the clustering performance is difficult to be satisfied.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides an electric power user portrait self-adaptive clustering method, which adopts an automatic encoder principle to realize feature extraction, uses a proper loss function to reduce the dimension of high dimension data and obtain low dimension information with higher information density; performing cluster analysis by adopting K-means algorithm operation, and obtaining initial cluster category in low dimensionality by the low dimensionality information; adopting a unimodal statistical test as a basic algorithm of fusion to carry out category fusion; the feature extraction, the cluster analysis and the class fusion optimization are integrated, a cluster mode is constructed, after an initial cluster class is obtained, a single peak statistical test value among the classes is calculated, and fusion among the classes is carried out according to the value; the method has the advantages that the proper number of class clusters is obtained under the condition that the number of the classes is not known in advance, and the clustering performance is effectively improved.
The invention realizes the aim through the following technical scheme:
a self-adaptive clustering method for electric power user portraits is characterized in that: the method is realized by the following steps:
A. extracting features;
B. clustering analysis;
C. category fusion;
the feature extraction realizes data diversification, including time sequence data and category data, and the data does not need to be standardized; firstly, carrying out feature extraction on input data through a coding network, then carrying out initial clustering through a K-means algorithm, then carrying out category fusion through a unimodal statistical test value matrix, carrying out unified optimization through a unified loss function, repeating the steps of feature extraction, cluster analysis and category fusion until the processes are stable, and finally outputting a labelAnd the number of clusters K.
The feature extraction is achieved by the sub-steps of:
a.1 Automatic encoder preparation: the automatic encoder is divided into an encoding part enc (·) and a decoding part dec (·) and inputs the encoding part enc (·) as X;
a.2 Automatic encoder training: the input and output of the network are identical, i.e., x=dec (enc (X)), training minimizes the Loss function value;
a.3 Feature extraction, obtaining feature data using the encoded portion of the trained automatic encoder: enc (X).
The cluster analysis is realized by the following substeps:
b.1 Inputting the original input feature (X) into the encoding section to obtain a low-dimensional feature vector enc (X);
b.2 Clustering enc (X) obtained in the step 2.1 by using a K-means algorithm to obtain an original category which is recorded asWherein i represents a cluster tag;
b.3 Updating the class center to be the actual vector closest to the class center obtained by the K-means algorithm to obtain the clustering class data label.
The category fusion is realized through the following substeps:
c.1 Every two kinds of clustering class data labels are projected onto the connecting line of the centers of the two kinds of classes;
c.2 Calculating unimodal statistical test values of Dip values and Dip-p-value values for data of every two category centers, and obtaining a value by using the Dip-p-value values: number of class clusters the symmetric Matrix of unimodal statistical test values of the number of class clusters;
c.3 For the maximum value in the unimodal statistical test matrix, if the maximum value is greater than the threshold, fusing the two classes, and updating the matrix until fusion and merging are not performed.
A self-adaptive clustering method for electric power user portraits is characterized in that: the automatic encoder network training process uses a gradient descent method to find the minimum value of the formula (1):
representing the loss of the automatic encoder, wherein B represents a small input batch, X represents input data, i.e. the output data desired by the automatic encoder, enc (·) represents data encoded by the encoding network, dec (·) represents data encoded by the decoding network, and->Representing the square of the euclidean distance.
A self-adaptive clustering method for electric power user portraits is characterized in that: the method comprises the steps of firstly clustering the number of very overestimated class clusters through a common K-means algorithm, solving the center of the class clusters, then constructing a matrix of unimodal statistical test values for the current class clusters, then carrying out class cluster fusion according to the matrix of unimodal statistical test values, and finally outputting the number K of the class clusters and specific labels。
A self-adaptive clustering method for electric power user portraits is characterized in that: the cluster center is obtained according to a formula (2):
wherein the method comprises the steps ofRepresenting class cluster label +.>Is a cluster-like center of (2). />Representing the cluster-like center obtained using the K-means algorithm.
The point value of the calculated unimodal statistical test value is obtained by the formula (3):
wherein the method comprises the steps ofIs one-dimensional data.
A self-adaptive clustering method for electric power user portraits is characterized in that: the loss evaluation of the clusters is obtained using equation (4), and the whole process is optimized by minimizing equation (6):
representing the loss of clusters, wherein->Is allocated to->Std (·) represents standard deviation and mean (·) represents mean. />Representing the Euclidean distance of the cluster-like center, +.>Obtained from equation (5):
wherein the method comprises the steps ofIndicating loss of automatic encoder->Representing the loss of clustering.
A self-adaptive clustering method for electric power user portraits is characterized in that: the cluster center after cluster fusion is obtained through a formula (7):
and the merging is performed by continuously updating the matrix of the unimodal statistical test values.
Compared with the prior art, the invention has the beneficial effects that:
the self-adaptive clustering method for the electric power user portraits is characterized in that a square loss function is used for feature extraction, high-dimensional data are subjected to dimension reduction, a K-means algorithm with a wider application range is used for cluster analysis to operate, a single-peak statistical test is used as a basic algorithm for fusion in a category fusion stage, low-dimensional information with higher information density is obtained, and initial clustering categories in low dimensions are obtained from the low-dimensional information; carrying out class fusion by adopting a unimodal statistical test, integrating characteristic extraction, cluster analysis and class fusion into a whole according to the characteristics of a unimodal function, constructing a cluster mode, calculating a unimodal statistical test value among classes after obtaining an initial cluster class, and carrying out fusion among classes according to the value; the method has the advantages that the proper number of class clusters is obtained under the condition that the number of the classes is not known in advance, and the clustering performance is effectively improved. The method solves the problems that the prior art replaces one cluster parameter with other parameters, has poor and satisfactory effect on constructing a cluster mode by large-scale high-dimension data, and the cluster performance is unsatisfactory.
Drawings
FIG. 1 is a schematic workflow diagram of feature extraction of the present invention;
FIG. 2 is a schematic workflow diagram of cluster analysis of the present invention;
FIG. 3 is a schematic of the overall workflow of the present invention.
Detailed Description
Embodiments of the adaptive clustering method for power consumer profiles are described in further detail below (see fig. 1-3) in conjunction with the accompanying drawings:
it should be understood that the examples described herein are for the purpose of illustration and explanation only and are not intended to limit the present invention.
An adaptive clustering method for the pictures of electric power users is characterized in that: the method is realized by the following steps:
step A, extracting features;
step B, cluster analysis;
step C, category fusion;
the feature extraction realizes data diversification, including time sequence data and category data, and the data does not need to be standardized; firstly, carrying out feature extraction on input data through a coding network, then carrying out initial clustering through a K-means algorithm, then carrying out category fusion through a unimodal statistical test value matrix, carrying out unified optimization through a unified loss function, repeating the steps of feature extraction, cluster analysis and category fusion until the processes are stable, and finally outputting a labelAnd the number of clusters K.
And step A, extracting the characteristics, wherein the specific implementation comprises the following substeps:
a.1 A) of: automatic encoder preparation: the automatic encoder is divided into an encoding part enc (·) and a decoding part dec (·) and inputs the encoding part enc (·) as X; the data is diversified, time sequence data and category data are contained, and standardization of the data is not needed; network preparation: dividing the automatic Encoder network into an encoding portion (Encoder) and a decoding portion (Decoder);
a.2: training an automatic coder: the input and output of the network are identical, i.e., x=dec (enc (X)), training minimizes the Loss function value; training a network: the input of the network is the output of the network (see fig. 1, 2 and table one), the sample passes through eight full-connection layers, six layers of LeakyReLu layers, the extracted feature enc (X) is output from the coding part of the network, and the training process uses the gradient descent method to find the minimum value for the formula (1):
representing the loss of the automatic encoder, wherein B represents a small input batch, X represents input data, i.e. output data intended by the automatic encoder, enc (·) represents data encoded by the encoding network, dec (·) represents data encoded by the decoding network, and->Representing the square of the euclidean distance.
Table one: decomposing network parameters
A.3 A) of: feature extraction, namely obtaining feature data by using an encoding part of a trained automatic encoder: enc (X) and finally obtaining characteristic data.
Step B, clustering analysis, (see fig. 2 and 3), the specific implementation includes the following sub-steps:
b.1 A) of: inputting the original data into an encoding part to obtain a feature vector enc (X); namely, inputting the original data (X) into an encoding part to obtain a low-dimensional feature vector enc (X);
b.2 A) of: clustering enc (X) obtained in the step 2.1 by using a K-means algorithm to obtain an original category which is recorded asWherein i represents a cluster tag;
substep B.2) for the encoded result, the result is first estimated by means of the usual K-means algorithmA cluster with very high cluster number is obtained, the cluster center is calculated, then a Matrix of a unimodal statistical test value is constructed for the current cluster, finally cluster fusion is carried out according to the Matrix of the unimodal statistical test value, and finally the cluster number K and a specific label are output。
The cluster center is obtained according to a formula (2):
wherein the method comprises the steps ofRepresenting the cluster label as +.>Is a cluster-like center of (2). Is a cluster-like center of (2). />Representing the cluster-like center obtained using the K-means algorithm.
B.3 A) of: updating the cluster center to be the actual vector closest to the cluster center obtained by the K-means algorithm, and obtaining the clustering type data label.
The point value for calculating the unimodal statistical test value is obtained by the formula (3)
Wherein the method comprises the steps ofIs one-dimensional data.
Representing the loss of clusters, < > in equation (4)>Is a class cluster label assigned to x, std (·) represents standard deviation mean (·) represents average. />Representing the Euclidean distance of the cluster-like center.
The loss evaluation of the clusters is obtained using equation (4) and the whole process is optimized by minimizing equation (6).
Euclidean distance of cluster-like centerObtained from equation (5):
wherein the method comprises the steps ofRepresenting loss from the encoder, < >>Representing the loss of clustering.
Step C, category fusion, (see fig. 1 and 3), the specific implementation includes the following sub-steps:
c.1 A) of: every two kinds of data are projected onto the connecting line of the centers of the two kinds of data; namely, each two kinds of clustering class data labels are projected onto the connecting line of the centers of the two kinds of classes;
c.2 A) of: calculating the values of the unimodal statistical test of the data in each two categories as a Dip value and a Dip-p-value, and obtaining a unimodal statistical test value symmetric Matrix with the size of the number of class clusters by using the Dip-p-value;
preferably, the cluster center after cluster fusion in step C.2) is obtained by the formula (7):
and is combined by continuously updating the Matrix of the unimodal statistical test values.
C.3 A) of: and for the maximum value in the unimodal statistical test value matrix, if the maximum value is greater than the threshold value, fusing the two classes, and updating the matrix at the same time until fusion and combination are not performed.
The method is characterized in that the method comprises the steps of drawing a power user, classifying and grading features according to the basic attribute, electricity consumption behavior, payment behavior and demand behavior of the power user, extracting typical features from each type, giving a label threshold value, and drawing individual and group images of the power user according to the final label and a business demand scene, predicting the behavior of a client, accurately predicting the electricity consumption, reducing the power supply loss, improving the service satisfaction degree and saving the electric energy, and is a vital work of an electric enterprise today. For individual and group portraits of power users, firstly, data is divided into clusters of similar data points in a large amount of unlabeled data, but in practice, the number of clusters is not known, and the problem that a cluster mode is very troublesome to construct exists. The invention provides a self-adaptive clustering method for electric power user portraits, which integrates feature extraction, cluster analysis and category fusion optimization, obtains proper number of clusters under the condition that the number of the clusters is not known in advance, and effectively improves the clustering performance.
The above description is merely a preferred embodiment of the present invention, and the above illustration is not to be construed as limiting the spirit of the present invention in any way, and any simple modification or variation of the above embodiments according to the technical spirit of the present invention, and equivalent embodiments that may be changed or modified to equivalent variations using the above disclosed technical spirit of the present invention, will still fall within the scope of the technical solutions of the present invention, without departing from the spirit and scope of the present invention.

Claims (6)

1. A self-adaptive clustering method for electric power user portraits is characterized in that: according to the basic attribute, electricity consumption behavior, payment behavior and demand behavior of the power users, carrying out feature classification and grading, extracting typical features from each type, giving a threshold value of the label, carrying out individual and group portraits of the power users according to the final label and combining service demand scenes, and predicting the behavior of the clients; the method is realized by the following steps:
step A, extracting features;
step B, cluster analysis;
step C, category fusion;
the feature extraction realizes data diversification, including time sequence data and category data, and the data does not need to be standardized; firstly, carrying out feature extraction on input data through a coding network, then carrying out initial clustering through a K-means algorithm, then carrying out category fusion through a unimodal statistical test value matrix, carrying out unified optimization through a unified loss function, repeating the steps of feature extraction, cluster analysis and category fusion until the processes are stable, and finally outputting a label C i And the number of clusters K;
step A, feature extraction, which is realized by the following substeps:
a.1 Automatic encoder preparation: the automatic encoder is divided into an encoding part enc (·) and a decoding part dec (·) and inputs the encoding part enc (·) as X;
a.2 Automatic encoder training: the input and output of the network are identical, i.e., x=dec (enc (X)), training minimizes the Loss function value;
a.3 Feature extraction, obtaining feature data using the encoded portion of the trained automatic encoder: enc (X);
step B, cluster analysis, which is realized by the following substeps:
b.1 Inputting the original input feature (X) into the encoding section to obtain a low-dimensional feature vector enc (X);
b.2 Step (ii) step (iii)Clustering enc (X) obtained in step 2.1 by using a K-means algorithm to obtain an original category marked as C i Wherein i represents a cluster tag;
b.3 Updating the class center to be the actual vector closest to the class center obtained by the K-means algorithm to obtain a clustering class data tag;
step C, category fusion-realized by the following substeps:
c.1 Every two kinds of clustering class data labels are projected onto the connecting line of the centers of the two kinds of classes;
c.2 Calculating unimodal statistical test values of Dip values and Dip-p-value values for data of every two category centers, and obtaining a value by using the Dip-p-value values: number of class clusters the symmetric Matrix of unimodal statistical test values of the number of class clusters;
c.3 For the maximum value in the unimodal statistical test matrix, if the maximum value is greater than the threshold, fusing the two classes, and updating the matrix until fusion and merging are not performed.
2. The adaptive clustering method for the electricity user portraits of claim 1, wherein the method comprises the following steps: the automatic encoder network training process uses a gradient descent method to minimize equation (1):
L rec representing the loss of the automatic encoder, wherein B represents a small input batch, X represents input data, i.e. the desired output data of the automatic encoder, enc (·) represents data encoded by the encoding network, dec (·) represents data after the data has passed the decoding network,representing the square of the euclidean distance.
3. The adaptive clustering method for electric power user portraits according to claim 1, characterized in thatThe method comprises the following steps: the method comprises the steps of firstly clustering the number of very overestimated class clusters through a common K-means algorithm, solving the center of the class clusters, then constructing a matrix of unimodal statistical test values for the current class clusters, then carrying out class cluster fusion according to the matrix of unimodal statistical test values, and finally outputting the number K of the class clusters and specific labels C i 。
4. A method for adaptive clustering of electricity consumer portraits according to claim 1 or claim 3, characterized in that: the cluster center is obtained according to the formula (2)
Wherein u is i A cluster center indicating a cluster label i, u i km Representing a cluster-like center obtained by using a K-means algorithm;
the point value for calculating the unimodal statistical test value is obtained by the formula (3)
Wherein the method comprises the steps ofIs one-dimensional data.
5. The adaptive clustering method for the electricity user portraits of claim 1, wherein the method comprises the following steps: the loss evaluation of the clusters is obtained using equation (4), and the whole process is optimized by minimizing equation (6):
L clu representing the loss of clustering, wherein c x Is assigned to xClass cluster tag, std (·) represents standard deviation, mean (·) represents average value, D c The Euclidean distance representing the cluster-like center is given by equation (5):
L＝L rec +L clu (6)
wherein L is rec Indicating loss of automatic encoder, L clu Representing the loss of clustering.
6. The adaptive clustering method of electricity consumer portraits of claim 1, wherein the adaptive clustering method of electricity consumer portraits is characterized in that: the cluster center after cluster fusion is obtained through a formula (7):
and the merging is performed by continuously updating the matrix of the unimodal statistical test values.
CN202111248953.6A 2021-10-26 2021-10-26 Self-adaptive clustering method for electricity user portraits Active CN113886669B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111248953.6A CN113886669B (en) 2021-10-26 2021-10-26 Self-adaptive clustering method for electricity user portraits

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111248953.6A CN113886669B (en) 2021-10-26 2021-10-26 Self-adaptive clustering method for electricity user portraits

Publications (2)

Publication Number Publication Date
CN113886669A CN113886669A (en) 2022-01-04
CN113886669B true CN113886669B (en) 2024-02-23

Family

ID=79014578

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111248953.6A Active CN113886669B (en) 2021-10-26 2021-10-26 Self-adaptive clustering method for electricity user portraits

Country Status (1)

Country Link
CN (1) CN113886669B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115375934B (en) * 2022-10-25 2023-04-11 北京鹰瞳科技发展股份有限公司 Method for training analysis of clustered models and related product
CN116312745B (en) * 2023-05-19 2023-08-08 山西智能大数据研究院有限公司 Intestinal flora super donor image information detection generation method

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107423769A (en) * 2017-08-03 2017-12-01 四川大学 Electric load curve adaptive clustering scheme based on morphological feature
RU2656708C1 (en) * 2017-06-29 2018-06-06 Самсунг Электроникс Ко., Лтд. Method for separating texts and illustrations in images of documents using a descriptor of document spectrum and two-level clustering
CN111259979A (en) * 2020-02-10 2020-06-09 大连理工大学 Deep semi-supervised image clustering method based on label self-adaptive strategy
CN112464004A (en) * 2020-11-26 2021-03-09 大连理工大学 Multi-view depth generation image clustering method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
RU2656708C1 (en) * 2017-06-29 2018-06-06 Самсунг Электроникс Ко., Лтд. Method for separating texts and illustrations in images of documents using a descriptor of document spectrum and two-level clustering
CN107423769A (en) * 2017-08-03 2017-12-01 四川大学 Electric load curve adaptive clustering scheme based on morphological feature
CN111259979A (en) * 2020-02-10 2020-06-09 大连理工大学 Deep semi-supervised image clustering method based on label self-adaptive strategy
CN112464004A (en) * 2020-11-26 2021-03-09 大连理工大学 Multi-view depth generation image clustering method

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
K-means改进算法在电力用户聚类辨识中的应用;李秋硕;王岩;孙宇军;肖勇;张朝鑫;;信息技术;20171025(第10期);全文 *
基于K-means聚类的物流园区用户画像分析;刘凯;;物流工程与管理;20200315(第03期);全文 *
基于改进快速密度峰值算法的电力负荷曲线聚类分析;陈俊艺;丁坚勇;田世明;卜凡鹏;朱炳翔;黄事成;周凯;;电力系统保护与控制;20181016(第20期);全文 *
电力数据及信息融合的电力客户行为分析;毛澍;张丽;;农村电气化;20200910(第09期);全文 *

Also Published As

Publication number Publication date
CN113886669A (en) 2022-01-04

Similar Documents

Publication Publication Date Title
CN113886669B (en) Self-adaptive clustering method for electricity user portraits
Kumar et al. Colorization transformer
CN111339433A (en) Information recommendation method and device based on artificial intelligence and electronic equipment
CN112819065B (en) Unsupervised pedestrian sample mining method and unsupervised pedestrian sample mining system based on multi-clustering information
CN103886623A (en) Image compression method and equipment, and system
CN108462888A (en) The intelligent association analysis method and system of user&#39;s TV and internet behavior
CN112668627A (en) Large-scale image online clustering system and method based on contrast learning
CN111292384B (en) Cross-domain diversity image generation method and system based on generative confrontation network
CN110751191A (en) Image classification method and system
CN112884758A (en) Defective insulator sample generation method and system based on style migration method
CN106803962B (en) 3D video depth map method for choosing frame inner forecast mode based on bayesian criterion
CN109460393A (en) Visualization system is repaired in a kind of preliminary examination based on big data in advance
CN107770813B (en) LTE uplink interference classification method based on PCA and two-dimensional skewness characteristics
CN111741313B (en) 3D-HEVC rapid CU segmentation method based on image entropy K-means clustering
CN117131449A (en) Data management-oriented anomaly identification method and system with propagation learning capability
Periasamy et al. A Common Palette Creation Algorithm for Multiple Images with Transparency Information
CN116561639A (en) Multi-mode data emotion analysis method for open source information
CN110782315B (en) Business system and method based on jewelry social electronic commerce platform
CN113298112B (en) Integrated data intelligent labeling method and system
CN112200275B (en) Artificial neural network quantification method and device
CN114708463A (en) Intelligent rock specimen image identification method
CN114241350A (en) Video coding test sequence determination method, related device and computer program product
CN113298893A (en) Artificial intelligence image processing method based on power dispatching
Li et al. Semi-siamese network for content-based video relevance prediction
CN111316326A (en) Image encoding method, apparatus and computer-readable storage medium

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