WO2020042580A1 - 人员分组方法、装置、电子装置及存储介质 - Google Patents

人员分组方法、装置、电子装置及存储介质 Download PDF

Info

Publication number
WO2020042580A1
WO2020042580A1 PCT/CN2019/077227 CN2019077227W WO2020042580A1 WO 2020042580 A1 WO2020042580 A1 WO 2020042580A1 CN 2019077227 W CN2019077227 W CN 2019077227W WO 2020042580 A1 WO2020042580 A1 WO 2020042580A1
Authority
WO
WIPO (PCT)
Prior art keywords
node
performance
ratio
current
preset condition
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.)
Ceased
Application number
PCT/CN2019/077227
Other languages
English (en)
French (fr)
Inventor
邓悦
金戈
徐亮
肖京
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.)
Ping An Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co Ltd
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 Ping An Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Publication of WO2020042580A1 publication Critical patent/WO2020042580A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06311—Scheduling, planning or task assignment for a person or group
    • G06Q10/063112—Skill-based matching of a person or a group to a task

Definitions

  • the present application relates to a grouping method, and particularly to a person grouping method, a person grouping device, an electronic device, and a storage medium.
  • a preferred embodiment of the present application provides a method for grouping people, including: obtaining multiple sets of sample data, each set of sample data including multiple condition attributes and corresponding decision attributes, wherein the decision attributes are performance of the sample data; Training a decision tree model based on the sample data, wherein the decision tree model includes a plurality of leaf nodes, each leaf node representing a group; each leaf node in the decision tree model is used as a current node, and all Achievement rate and achievement coverage rate of the current node, determine whether each of the current nodes meets a preset condition according to the achievement rate and the achievement coverage rate; when all the current nodes belonging to the same parent node do not meet the When the conditions are preset, it is determined whether the parent node is a root node; when the parent node is not a root node, each of the parent nodes is used as a new current node to trigger the calculation of the performance rate and the performance of all current nodes.
  • a step of coverage; when the parent node is a root node, calculating the performance ratio and the performance coverage of the root node Determine whether the root node meets a preset condition according to the performance ratio and the coverage ratio, and divide the root node into a performance group when the root node meets the preset condition, and when the root node does not meet Ignore the root node when preset conditions; for each of the current nodes, perform the following steps: use the current node as the target node, and divide the target node when the target node meets the preset conditions To the excellent performance group; when the target node does not meet the preset condition, determine whether other nodes in the same layer as the target node meet the preset condition; When another node in the same layer meets the preset condition, the target node is divided into a poor performance group; when all other nodes in the same layer that belong to the same parent node do not meet the preset condition, the target node is ignored .
  • a preferred embodiment of the present application further provides a personnel grouping device, including: an obtaining module for obtaining multiple sets of sample data, each set of sample data including multiple condition attributes and corresponding decision attributes, and the decision attributes are all The performance of the sample data is described; a training module is used to train a decision tree model according to the sample data, wherein the decision tree model includes multiple leaf nodes, each leaf node represents a group; an execution module is used to calculate all Achievement rate and achievement coverage rate of the current node, determine whether each of the current nodes meets a preset condition according to the achievement rate and the achievement coverage rate; when all the current nodes belonging to the same parent node do not meet the When the conditions are preset, the execution module determines whether the parent node is a root node; when the parent node is not a root node, the execution module uses each of the parent nodes as a new current node to trigger the calculation Steps for the achievement ratio and achievement coverage of all current nodes; when the parent node is the root node, the execution module Calculate the merit
  • a preferred embodiment of the present application further provides an electronic device, including a processor and a memory.
  • the memory stores at least one computer-readable instruction
  • the processor is configured to execute the computer-readable instruction to implement The grouping method described above.
  • a preferred embodiment of the present application further provides a non-volatile readable storage medium. At least one computer-readable instruction is stored on the non-volatile readable storage medium. When the computer-readable instruction is executed by a processor, Implement the personnel grouping method described previously.
  • the information of the poor performers is added to the decision tree modeling to make the grouping targets more clear and the data execution effect is better, to ensure that the finally excellent performant group is a group with a certain percentage of outstanding performers; moreover, The distinguishing characteristics of high-performance personnel are identified through the conditional attributes of high-performance groupings to build user portraits.
  • FIG. 1 is a flowchart of a personnel grouping method provided by a preferred embodiment of the present application.
  • FIG. 2 is a schematic diagram of a decision tree trained by the person grouping method of FIG. 1.
  • FIG. 2 is a schematic diagram of a decision tree trained by the person grouping method of FIG. 1.
  • FIG. 3 is a schematic structural diagram of a personnel grouping device provided by a preferred embodiment of the present application.
  • FIG. 4 is a schematic structural diagram of an electronic device according to a preferred embodiment of the present application.
  • FIG. 1 is a flowchart of a personnel grouping method provided by a preferred embodiment of the present application.
  • the personnel grouping method is applied in an electronic device 1. According to different requirements, the order of the steps of the personnel grouping method may be changed, and some steps may be omitted or combined.
  • the method for grouping people includes the following steps:
  • step S11 multiple sets of sample data are obtained, and each set of sample data includes multiple condition attributes and corresponding decision attributes, and the decision attributes are performance of the sample data.
  • the sample data needs to include data of a higher performing person (ie, a high performing person) and a poor performing person (ie, a low performing person), and the sample data may be stored in the electronic device.
  • the electronic device may also connect to an external sample library through a network to further obtain sample data stored in the sample library.
  • the electronic device may also collect and establish the sample library by means of big data.
  • condition attributes may be behavior trajectories (such as business trips), app activity, business expansion, consumption, interests, hobbies, participation in training, attendance rate, and so on.
  • the condition attributes include business trips, business expansions, training participations, and attendance rates.
  • Table 1 The sample data is shown in Table 1.
  • step S12 a decision tree model is trained according to the sample data, wherein the decision tree model includes a plurality of leaf nodes, and each leaf node represents a group.
  • the decision tree algorithm belongs to a supervised learning classification algorithm, and the decision tree model represents a mapping relationship between object attributes and object values.
  • the decision tree model has N layers (N is a natural number, N> 2), and the electronic device sets each node (including a root node, an internal node, and a leaf node) in the decision tree model according to the condition attribute. ).
  • the first layer of the decision tree model is a root node
  • the second layer of the decision tree model represents a plurality of nodes obtained by segmenting the root node with a first-level condition attribute.
  • the third layer of the decision tree model is a root node.
  • the layer represents multiple nodes obtained by segmenting the nodes in the previous layer with the second-level condition attributes, and so on.
  • training the decision tree model specifically includes:
  • step S121 the sample data is used as a training set to calculate the information gain of each condition attribute.
  • Step S122 selecting the condition attribute with the largest information gain as the root node of the decision tree model to segment the sample data to obtain the next-level node; where the larger the information gain of a condition attribute is, the condition attribute is selected.
  • the more information provided for classification the more conducive it is to determinism, and the more beneficial it is to classify the sample data. As shown in FIG. 2, if the condition attribute for which the maximum information gain is calculated is “travel situation”, the “travel situation” condition attribute is selected as the root node to segment the sample data.
  • Step S123 Recalculate the information gain of each condition attribute using the sample data contained in each node as a training set.
  • step S124 the condition attribute with the largest information gain is selected to divide the node to obtain the next-level node.
  • C 0
  • the condition attribute that calculates the maximum information gain is "business development situation”
  • the nodes are segmented to obtain the next-level nodes.
  • the condition attribute with the largest information gain is “participation in training”
  • the node is segmented according to the condition of “participation in training” And get the next node.
  • the conditional attributes used to segment multiple nodes on the same layer are usually different. The nodes formed after each segmentation have higher data purity than the nodes in the previous layer.
  • step S125 steps S123 and S124 are performed recursively until the division is stopped.
  • each leaf node represents a group, and each group contains a fixed ratio between the number of high-performing people and the number of low-performing people. Different from each other.
  • the ratio of the leaf node "Group 1" may be 1: 8.
  • a leaf node can include all high-performing people, or all poor-performing people.
  • Each set of sample data can only be divided into one of the leaf nodes, that is, each set of sample data cannot belong to two or more leaf nodes at the same time.
  • the next segmentation is stopped when the number of samples contained in each node of the current layer of the decision tree is less than a preset number.
  • the information gain Gain (S, A) of each condition attribute can be calculated as follows:
  • Gain (S, A) represents the information gain of the conditional attribute A on the training set S
  • Entropy (S) represents the information entropy of the training set S
  • Entropy (S, A) represents the information entropy of the A attribute.
  • the training set has a total of 16 sets of sample data, 11 sets with excellent performance, and 5 sets with poor performance.
  • the information gain calculation of the condition attribute of "travel situation" is taken as an example for description.
  • the information gain calculation process of other condition attributes is the same.
  • each leaf node in the decision tree model is used as a current node, and a performance ratio and a performance coverage ratio of all current nodes are calculated, and whether each of the current nodes is determined according to the performance ratio and the performance coverage ratio. Meet the preset conditions.
  • the target node When all other nodes in the same layer that belong to the same parent node do not meet the preset condition, the target node is ignored. That is, the target node is neither divided into the high performance group nor the target node into the poor performance group.
  • the performance ratio of each node is the ratio between the number of performance personnel at the node and the total number of personnel to be grouped at the node, and the performance ratio of the node can be determined by the number of performance personnel and performance at the node. The ratio between the number of poor people is calculated.
  • the performance coverage of each node is the ratio between the number of performance personnel of the node and the total number of performance personnel of the entire decision tree model. When the performance ratio and performance coverage of the node reach a performance ratio threshold and a performance coverage threshold, respectively, it is determined that the node meets the preset condition.
  • the threshold of excellence rate For example, setting the threshold of excellence rate to 40% and the threshold of excellence coverage rate to 20%. If it is calculated that the performance ratio of the node “Group 8" is 80% and the performance coverage ratio is 20%, then the node “Group 8" meets the preset conditions and is classified as a high performance group. For the node “Group 7", if the excellent coverage rate is 30% and the excellent coverage rate is 25%, the node “Group 7" does not meet the preset conditions, but because the node “Group 7" belongs to the same parent node as the same parent node The layer node “Group 8" has preset conditions, so the node “Group 7” is divided into poor performance groups. Conversely, if the node “Group 7" and the node “Group 8" do not meet the preset conditions, the node “Group 7" and the node “Group 8" are ignored.
  • the performance ratio threshold and the performance coverage threshold are stored in the electronic device in advance.
  • the performance ratio threshold and the performance coverage threshold may be adjusted according to actual needs.
  • step S14 each time when all the current nodes belonging to the same parent node do not meet the preset condition, it is determined whether the parent node is a root node.
  • Step S15 when the parent node is not a root node, using each of the parent nodes as a new current node, triggering the steps of calculating a performance ratio and a performance coverage ratio of all current nodes; when the parent node is a root Node, calculate the merit ratio and merit coverage of the root node, and determine whether the root node meets the preset conditions according to the merit ratio and the merit coverage, and when the root node meets the preset conditions, The root node is divided into a merit group, and the root node is ignored when the root node does not meet a preset condition.
  • the root node is judged.
  • the size and quality of the performance group and the performance group can be controlled.
  • the final performance group can include people with different characteristics.
  • the outstanding characteristics of this part of the outstanding people are: not traveling, strong business ability, and good attendance.
  • Group 8 is divided into the high-performance group, the condition attributes involved in this high-performance group include: business trip ⁇ 2 times, training participation ⁇ 1 time, and attendance rate ⁇ 80.
  • the outstanding characteristics of this group of outstanding performers are: frequent business trips, willingness to learn, and better attendance.
  • FIG. 3 is a schematic structural diagram of a personnel grouping device 300 provided by a preferred embodiment of the present application.
  • the personnel grouping device 300 runs in an electronic device.
  • the personnel grouping device 300 may include a plurality of functional modules composed of program code segments.
  • the program code of each program segment of the personnel grouping device 300 may be stored in a memory of the electronic device and executed by the at least one processor to implement a personnel grouping function.
  • the personnel grouping device 300 may be divided into a plurality of functional modules according to functions performed by the personnel grouping device 300.
  • the personnel grouping device 300 includes: an acquisition module 301, a training module 302, and an execution module 303.
  • the module referred to in the present application refers to a series of computer-readable instruction segments that can be executed by at least one processor and can perform fixed functions, which are stored in a memory. In this embodiment, functions of each module will be described in detail in subsequent embodiments.
  • the obtaining module 301 is configured to obtain multiple sets of sample data, and each set of sample data includes multiple condition attributes and corresponding decision attributes, and the decision attributes are performance of the sample data.
  • the sample data needs to include data of a higher performing person (ie, a high performing person) and a poor performing person (ie, a low performing person), and the sample data may be stored in the electronic device.
  • the electronic device may also connect to an external sample library through a network to further obtain sample data stored in the sample library.
  • the electronic device may also collect and establish the sample library by means of big data.
  • condition attributes may be behavior trajectories (such as business trips), app activity, business expansion, consumption, interests, hobbies, participation in training, attendance rate, and so on.
  • the condition attributes include business trips, business expansions, participation in training, and attendance rates, and the sample data is shown in Table 1 above.
  • the training module 302 is configured to train a decision tree model according to the sample data, wherein the decision tree model includes a plurality of leaf nodes, and each leaf node represents a group.
  • the decision tree algorithm belongs to a supervised learning classification algorithm, and the decision tree model represents a mapping relationship between object attributes and object values.
  • the decision tree model has N layers (N is a natural number, and N> 2), and the training module 302 sets each node (including a root node, an internal node, and a leaf) in the decision tree model according to the condition attribute. Node).
  • the first layer of the decision tree model is a root node
  • the second layer of the decision tree model represents a plurality of nodes obtained by segmenting the root node with a first-level condition attribute.
  • the third layer of the decision tree model is a root node.
  • the layer represents multiple nodes obtained by segmenting the nodes in the previous layer with the second-level condition attributes, and so on.
  • the training module 302 uses the sample data as a training set to calculate the information gain of each condition attribute, and selects the condition attribute with the largest information gain as the root node of the decision tree model to Segment the sample data to get the next level of nodes.
  • the condition attribute for which the maximum information gain is calculated is “travel situation”
  • the “travel situation” condition attribute is selected as the root node to segment the sample data.
  • the training module 302 further uses the sample data contained in each node as a training set to recalculate the information gain of each condition attribute, and selects the condition attribute with the largest information gain to segment the node to obtain the next-level node.
  • C 0
  • the condition attribute that calculates the maximum information gain is "business development situation”
  • the nodes are segmented to obtain the next-level nodes.
  • condition attribute with the largest information gain is “participation in training”
  • the node is segmented according to the condition of “participation in training” condition And get the next node.
  • conditional attributes used to segment multiple nodes on the same layer are usually different.
  • the nodes formed after each segmentation have higher data purity than the nodes in the previous layer.
  • the training module 302 further recursively performs the calculation steps of the information gain of each condition attribute and the step of selecting the condition attribute with the largest information gain to divide the node to obtain the next-level node until the segmentation stops.
  • each leaf node represents a group, and each group contains a fixed ratio between the number of high-performing people and the number of low-performing people. Different from each other. For example, the ratio of the leaf node "Group 1" may be 1: 8.
  • a leaf node can include all high-performing people, or all poor-performing people.
  • Each set of sample data can only be divided into one of the leaf nodes, that is, each set of sample data cannot belong to two or more leaf nodes at the same time.
  • the next segmentation is stopped when the number of samples contained in each node of the current layer of the decision tree is less than a preset number.
  • the information gain Gain (S, A) of each condition attribute can be calculated as follows:
  • Gain (S, A) represents the information gain of the conditional attribute A on the training set S
  • Entropy (S) represents the information entropy of the training set S
  • Entropy (S, A) represents the information entropy of the A attribute.
  • the training set has a total of 16 sets of sample data, 11 sets with excellent performance, and 5 sets with poor performance.
  • the information gain calculation of the condition attribute of "travel situation" is taken as an example for description.
  • the execution module 303 is configured to use each leaf node in the decision tree model as a current node, calculate a performance ratio and a performance coverage ratio of all current nodes, and determine each location based on the performance ratio and the performance coverage ratio. Describes whether the current node meets the preset conditions.
  • the execution module 303 is configured to:
  • the target node When all other nodes in the same layer that belong to the same parent node do not meet the preset condition, the target node is ignored. That is, the target node is neither divided into the high performance group nor the target node into the poor performance group.
  • the performance ratio of each node is the ratio between the number of performance personnel at the node and the total number of personnel to be grouped at the node, and the performance ratio of the node can be determined by the number of performance personnel and performance at the node. The ratio between the number of poor people is calculated.
  • the performance coverage of each node is the ratio between the number of performance personnel of the node and the total number of performance personnel of the entire decision tree model. When the performance ratio and performance coverage of the node reach a performance ratio threshold and a performance coverage threshold, respectively, it is determined that the node meets the preset condition.
  • the threshold of excellence rate For example, setting the threshold of excellence rate to 40% and the threshold of excellence coverage rate to 20%. If it is calculated that the performance ratio of the node “Group 8" is 80% and the performance coverage ratio is 20%, then the node “Group 8" meets the preset conditions and is classified as a high performance group. For the node “Group 7", if the excellent coverage rate is 30% and the excellent coverage rate is 25%, the node “Group 7" does not meet the preset conditions, but because the node “Group 7" belongs to the same parent node as the same parent node The layer node “Group 8" has preset conditions, so the node “Group 7” is divided into poor performance groups. Conversely, if the node “Group 7" and the node “Group 8" do not meet the preset conditions, the node “Group 7" and the node “Group 8" are ignored.
  • the performance ratio threshold and the performance coverage threshold are stored in the electronic device in advance.
  • the performance ratio threshold and the performance coverage threshold may be adjusted according to actual needs.
  • the execution module 303 is further configured to determine whether the parent node is a root node every time when all the current nodes belonging to the same parent node do not meet the preset condition.
  • the execution module 303 uses each of the parent nodes as a new current node, and triggers the steps of calculating a performance ratio and a performance coverage ratio of all current nodes; when the parent node When the root node is the root node, the performance ratio and the performance coverage ratio of the root node are calculated, and whether the root node meets a preset condition is determined according to the performance ratio and the performance coverage ratio.
  • the root node meets the preset condition Divide the root node into a superior group, and ignore the root node when the root node does not meet a preset condition.
  • the root node is judged.
  • the size and quality of the performance group and the performance group can be controlled.
  • the final performance group can include people with different characteristics.
  • the outstanding characteristics of this part of the outstanding people are: not traveling, strong business ability, and good attendance.
  • Group 8 is divided into the high-performance group, the condition attributes involved in this high-performance group include: business trip ⁇ 2 times, training participation ⁇ 1 time, and attendance rate ⁇ 80.
  • the outstanding characteristics of this group of outstanding performers are: frequent business trips, willingness to learn, and better attendance.
  • FIG. 4 is a schematic structural diagram of an electronic device 1 that implements the personnel grouping method in a preferred embodiment of the present application.
  • the electronic device 1 includes a memory 101, a processor 102, and computer-readable instructions 103, such as a personnel grouping program, stored in the memory 101 and executable on the processor 102.
  • step S11 multiple sets of sample data are obtained, and each set of sample data includes multiple condition attributes and corresponding decision attributes, and the decision attributes are performance of the sample data.
  • the sample data needs to include data of a higher performing person (ie, a high performing person) and a poor performing person (ie, a low performing person), and the sample data may be stored in the electronic device.
  • the electronic device may also connect to an external sample library through a network to further obtain sample data stored in the sample library.
  • the electronic device may also collect and establish the sample library by means of big data.
  • condition attributes may be behavior trajectories (such as business trips), app activity, business expansion, consumption, interests, hobbies, participation in training, attendance rate, and so on.
  • the condition attributes include business trips, business expansions, participation in training, and attendance rates, and the sample data is shown in Table 1 above.
  • step S12 a decision tree model is trained according to the sample data, wherein the decision tree model includes a plurality of leaf nodes, and each leaf node represents a group.
  • the decision tree algorithm belongs to a supervised learning classification algorithm, and the decision tree model represents a mapping relationship between object attributes and object values.
  • the decision tree model has N layers (N is a natural number, N> 2), and the electronic device sets each node (including a root node, an internal node, and a leaf node) in the decision tree model according to the condition attribute. ).
  • the first layer of the decision tree model is a root node
  • the second layer of the decision tree model represents a plurality of nodes obtained by segmenting the root node with a first-level condition attribute.
  • the third layer of the decision tree model is a root node.
  • the layer represents multiple nodes obtained by segmenting the nodes in the previous layer with the second-level condition attributes, and so on.
  • training the decision tree model specifically includes:
  • Step S121 Use the sample data as a training set to calculate an information gain for each condition attribute.
  • Step S122 selecting the condition attribute with the largest information gain as the root node of the decision tree model to segment the sample data to obtain the next-level node; where the larger the information gain of a condition attribute is, the condition attribute is selected.
  • the more information provided for classification the more conducive it is to determinism, and the more beneficial it is to classify the sample data. As shown in FIG. 2, if the condition attribute for which the maximum information gain is calculated is “travel situation”, the “travel situation” condition attribute is selected as the root node to segment the sample data.
  • Step S123 Recalculate the information gain of each condition attribute using the sample data contained in each node as a training set.
  • step S124 the condition attribute with the largest information gain is selected to divide the node to obtain the next-level node.
  • C 0
  • the condition attribute that calculates the maximum information gain is "business development situation”
  • the nodes are segmented to obtain the next-level nodes.
  • the condition attribute with the largest information gain is “participation in training”
  • the node is segmented according to the condition of “participation in training” condition And get the next node.
  • the conditional attributes used to segment multiple nodes on the same layer are usually different. The nodes formed after each segmentation have higher data purity than the nodes in the previous layer.
  • each leaf node represents a group.
  • Each group contains a fixed ratio between the number of high-performing people and the number of low-performing people.
  • the ratio between the number of high-performing people and the number of low-performing people contained in multiple groups can be Different from each other.
  • the ratio of the leaf node "Group 1" may be 1: 8.
  • a leaf node can include all high-performing people, or all poor-performing people.
  • Each set of sample data can only be divided into one of the leaf nodes, that is, each set of sample data cannot belong to two or more leaf nodes at the same time.
  • the next segmentation is stopped when the number of samples contained in each node of the current layer of the decision tree is less than a preset number.
  • the information gain Gain (S, A) of each condition attribute can be calculated as follows:
  • Gain (S, A) represents the information gain of the conditional attribute A on the training set S
  • Entropy (S) represents the information entropy of the training set S
  • Entropy (S, A) represents the information entropy of the A attribute.
  • the training set has a total of 16 sets of sample data, 11 sets with excellent performance, and 5 sets with poor performance.
  • the information gain calculation of the condition attribute of "travel situation" is taken as an example for description.
  • the information gain calculation process of other condition attributes is the same.
  • each leaf node in the decision tree model is used as a current node, and a performance ratio and a performance coverage ratio of all current nodes are calculated, and whether each of the current nodes is determined according to the performance ratio and the performance coverage ratio. Meet the preset conditions.
  • the target node When all other nodes in the same layer that belong to the same parent node do not meet the preset condition, the target node is ignored. That is, the target node is neither divided into the high performance group nor the target node into the poor performance group.
  • the performance ratio of each node is the ratio between the number of performance personnel at the node and the total number of personnel to be grouped at the node, and the performance ratio of the node can be determined by the number of performance personnel and performance at the node. The ratio between the number of poor people is calculated.
  • the performance coverage of each node is the ratio between the number of performance personnel of the node and the total number of performance personnel of the entire decision tree model. When the performance ratio and performance coverage of the node reach a performance ratio threshold and a performance coverage threshold, respectively, it is determined that the node meets the preset condition.
  • the threshold of excellence rate For example, setting the threshold of excellence rate to 40% and the threshold of excellence coverage rate to 20%. If it is calculated that the performance ratio of the node “Group 8" is 80% and the performance coverage ratio is 20%, then the node “Group 8" meets the preset conditions and is classified as a high performance group. For the node “Group 7", if the excellent coverage rate is 30% and the excellent coverage rate is 25%, the node “Group 7" does not meet the preset conditions, but because the node “Group 7" belongs to the same parent node as the same parent node The layer node “Group 8" has preset conditions, so the node “Group 7” is divided into poor performance groups. Conversely, if the node “Group 7" and the node “Group 8" do not meet the preset conditions, the node “Group 7" and the node “Group 8" are ignored.
  • the performance ratio threshold and the performance coverage threshold are stored in the electronic device in advance.
  • the performance ratio threshold and the performance coverage threshold may be adjusted according to actual needs.
  • step S14 each time when all the current nodes belonging to the same parent node do not meet the preset condition, it is determined whether the parent node is a root node.
  • Step S15 when the parent node is not a root node, using each of the parent nodes as a new current node, triggering the steps of calculating a performance ratio and a performance coverage ratio of all current nodes; when the parent node is a root Node, calculate the merit ratio and merit coverage of the root node, and determine whether the root node meets the preset conditions according to the merit ratio and the merit coverage, and when the root node meets the preset conditions, The root node is divided into a merit group, and the root node is ignored when the root node does not meet a preset condition.
  • the root node is judged.
  • the size and quality of the performance group and the performance group can be controlled.
  • the final performance group can include people with different characteristics.
  • the outstanding characteristics of this part of the outstanding people are: not traveling, strong business ability, and good attendance.
  • Group 8 is divided into the high-performance group, the condition attributes involved in this high-performance group include: business trip ⁇ 2 times, training participation ⁇ 1 time, and attendance rate ⁇ 80.
  • the outstanding characteristics of this group of outstanding performers are: frequent business trips, willingness to learn, and better attendance.
  • the processor 102 executes the computer-readable instructions 103
  • the functions of the modules / units in the embodiment of the personnel grouping device described above are implemented, for example, units 301-303 in FIG.
  • the information of the poor performers is added to the decision tree modeling to make the grouping targets more clear and the data execution effect is better, to ensure that the finally excellent performant group is a group with a certain percentage of outstanding performers;
  • the distinguishing characteristics of high-performance personnel are identified through the conditional attributes of high-performance groupings to build user portraits.
  • the computer-readable instructions 103 may be divided into one or more modules / units, and the one or more modules / units are stored in the memory 101 and executed by the processor 102, To complete this application.
  • the one or more modules / units may be a series of computer-readable instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer-readable instructions 103 in the electronic device 1.
  • the computer-readable instructions 103 may be divided into an acquisition module 301, a training module 302, and an execution module 303 in FIG.
  • the electronic device 1 may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server.
  • a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server.
  • the schematic diagram is only an example of the electronic device 1 and does not constitute a limitation on the electronic device 1.
  • the schematic diagram may include more or fewer components than shown in the figure, or some components may be combined, or different Components, for example, the electronic device 1 may further include an input-output device, a network access device, a bus, and the like.
  • the so-called processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASICs), Ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor may be a microprocessor, or the processor 30 may be any conventional processor, etc.
  • the processor 102 is a control center of the electronic device 1, and uses various interfaces and lines to connect the entire electronic device 1. Various parts.
  • the memory 101 may be configured to store the computer-readable instructions 103 and / or modules / units, and the processor 102 may execute or execute the computer-readable instructions and / or modules / units stored in the memory 101, and Recalling the data stored in the memory 101 to implement various functions of the electronic device 1.
  • the memory 101 may mainly include a storage program area and a storage data area, where the storage program area may store an operating system, an application program required for at least one function (such as a sound playback function, an image playback function, etc.), etc .; the storage data area may Data (such as audio data, phonebook, etc.) created according to the use of the electronic device 1 are stored.
  • the memory 101 may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, an internal memory, a plug-in hard disk, a Smart Memory Card (SMC), and a Secure Digital (SD).
  • a non-volatile memory such as a hard disk, an internal memory, a plug-in hard disk, a Smart Memory Card (SMC), and a Secure Digital (SD).
  • SSD Secure Digital
  • flash memory card Flash card
  • flash memory device at least one disk storage device, flash memory device, or other volatile solid-state storage device.
  • the integrated module / unit of the electronic device 1 When the integrated module / unit of the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile readable storage medium. Based on this understanding, this application implements all or part of the processes in the methods of the above embodiments, and can also be completed by computer-readable instructions instructing related hardware.
  • the computer-readable instructions can be stored in a non-volatile memory. In the read storage medium, when the computer-readable instructions are executed by a processor, the steps of the foregoing method embodiments can be implemented.
  • the computer-readable instructions include computer-readable instruction codes, and the computer-readable instruction codes may be in a source code form, an object code form, an executable file, or some intermediate form.
  • the non-volatile readable medium may include: any entity or device capable of carrying the computer-readable instruction code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), electric carrier signals, telecommunication signals, and software distribution media.
  • ROM Read-Only Memory
  • RAM Random Access Memory
  • electric carrier signals telecommunication signals
  • telecommunication signals and software distribution media.
  • the content contained in the non-volatile readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdictions. For example, in some jurisdictions, according to legislation and patent practices, non- Volatile readable media does not include electrical carrier signals and telecommunication signals.
  • each functional unit in each embodiment of the present application may be integrated in the same processing unit, or each unit may exist separately physically, or two or more units may be integrated in the same unit.
  • the above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional modules.

Landscapes

  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Educational Administration (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种人员分组方法、装置、电子装置及存储介质，所述方法包括：获取多组样本数据；训练决策树模型；以决策树模型每一叶子节点作为当前节点，计算所有当前节点的绩优率及绩优覆盖率，确定每一当前节点是否符合预设条件；当属于同一父节点的所有当前节点均不符合所述预设条件时，判断所述父节点是否为根节点；当所述父节点为根节点时，计算所述根节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定所述根节点是否符合预设条件，当所述根节点符合预设条件时将所述根节点划分至绩优分组，当所述根节点不符合预设条件时忽略所述根节点。该方法通过绩优分组的条件属性识别绩优人员的显著特征，从而构建用户画像。

Description

人员分组方法、装置、电子装置及存储介质
本申请要求于2018年08月27日提交中国专利局，申请号为201810982127.6发明名称为“人员分组方法、装置、电子装置及计算机可读存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及分组方法，具体涉及一种人员分组方法、人员分组装置、电子装置以及存储介质。
背景技术
企业通常拥有海量的人员基础数据和行为数据。企业通常期望搭建用户画像，对绩优人员的显著特征进行识别。现有技术中，通常采用无监督学习的聚类算法对无序的对象进行分群、归类，已达到更好地理解研究对象的目的。然而，利用该类方法建立用户画像存在缺陷，如，分群目标不明确，分群效果不可控，分群结果不可解释，且各分群的重要特征无法识别等。
发明内容
鉴于以上内容，有必要提出一种人员分组方法、人员分组装置、电子装置以及存储介质，能够解决以上问题。
本申请一较佳实施方式提供一种人员分组方法，包括：获取多组样本数据，每一组样本数据包括多个条件属性以及对应的决策属性，所述决策属性为所述样本数据的绩效；根据所述样本数据训练一决策树模型，其中，所述决策树模型包括多个叶子节点，每一叶子节点代表一个分组；以所述决策树模型中的每一叶子节点作为当前节点，计算所有当前节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件；当属于同一父节点的所有所述当前节点均不符合所述预设条件时，判断所述父节点是否为根节点；当所述父节点不是根节点时，以每一个所述父节点作为新的当前节点，触发所述计算所有当前节点的绩优率以及绩优覆盖率的步骤；当所述父节点为根节点时，计算所述根节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定所述根节点是否符合预设条件，当所述根节点符合预设条件时将所述根节点划分至绩优分组，当所述根节点不符合预设条件时忽略所述根节点；针对每一个所述当前节点，执行以下步骤：将所述当前节点作为目标节点，当所述目标节点符合所述预设条件时，将所述目标节点划分至所述绩优分组；当所述目标节点不符合所述预设条件时，判断与所述目标节点属于同一父节点的其它同层节点是否符合所述预设条件；当属于同一父节点的至少一其它同层节点符合所述预设条件时，将所述目标节点划分至绩差分组；当属于同一父节点的所有其它同层节点均不符合所述预设条件时，忽略所述目标节点。
本申请一较佳实施方式还提供一种人员分组装置，包括：获取模块，用于获取多组样本数据，每一组样本数据包括多个条件属性以及对应的决策属性，所述决策属性为所述样本数据的绩效；训练模块，用于根据所述样本数据训练一决策树模型，其中，所述决策树模型包括多个叶子节点，每一叶子节点代表一个分组；执行模块，用于计算所有当前节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件；当属于同一父节点的所有所述当前节点均不符合所述预设条件时，所述执行模块判断所述父节点是否为根节点；当所述父节点不是根节点时，所述执行模块以每一个所述父节点作为新的当前节点，触发所述计算所有当前节点的绩优率以及绩优覆盖率的步骤；当所述父节点为根节点时，所述执行模块计算所述根节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定所述根节点是否符合预设条件，当所述根节点符合预设条件时将所述根节点划分至绩优分组，当所述根节点不符合预设条件时忽略所述根节点；针对每一个所述当前节点，所述执行模块用于：将所述当前节点作为目标节点，当所述目标节点符合所述预设条件时，将所述目标节点划分至所述绩优分组；当所述目标节点不符合所述预设条件时，判断与所述目标节点属于同一父节点的其它同层节点是否符合所述预设条件；当属于同一父节点的至少一其它同层节点符合所述预设条件时，将所述目标节点划分至绩差分组；当属于同一父节点的所有其它同层节点均不符合所述预设条件时，忽略所述目标节点。
本申请一较佳实施方式还提供一种电子装置，包括处理器和存储器，所述存储器中存储有至少一个计算机可读指令，所述处理器用于执行所述计算机可读指令以实现如前所述的人员分组方法。
本申请一较佳实施方式还提供一种非易失性可读存储介质，所述非易失性可读存储介质上存储至少一个计算机可读指令，所述计算机可读指令被处理器执行时实现如前所述的人员分组方法。
本申请实施例在决策树建模中加入绩差人员的信息，使得分组目标更明确，且数据的执行效果更好，保证最终获得的绩优分组为绩优人数比例达到一定程度的群体；而且，可通过绩优分组的条件属性识别绩优人员的显著特征，从而构建用户画像。
附图说明
图1是本申请一较佳实施例提供的人员分组方法的流程图。
图2是图1的人员分组方法所使训练的决策树的示意图。
图3是本申请一较佳实施例提供的人员分组装置的结构示意图。
图4为本申请一较佳实施例提供的电子装置的结构示意图。
具体实施方式
图1是本申请一较佳实施例提供的人员分组方法的流程图。所述人员分组方法应用于一电子装置1中。根据不同需求，所述人员分组方法的步骤顺序可以改变，某些步骤可以省略或合并。所述人员分组方法包括以下步骤：
步骤S11，获取多组样本数据，每一组样本数据包括多个条件属性以及对 应的决策属性，所述决策属性为所述样本数据的绩效。
其中，所述样本数据需同时包括绩效较优的人员(即，绩优人员)和绩效较差的人员(即，绩差人员)的数据，所述样本数据可存储于所述电子装置中。在另一实施方式中，所述电子装置还可以通过接入网络连接之一外部样本库，进而来获取所述样本库存储的样本数据。在其它实施方式中，所述电子装置还可以通过大数据方式来收集并建立所述样本库。
在本实施方式中，所述条件属性可为行为轨迹(如，出差情况)、app活跃情况、业务扩展情况、消费情况、兴趣爱好、参加培训情况、考勤率等。以所述条件属性包括出差情况、业务扩展情况、参加培训情况以及考勤率为例进行说明，所述样本数据如表1所示。
表1样本数据
Figure PCTCN2019077227-appb-000001
步骤S12，根据所述样本数据训练一决策树模型，其中，所述决策树模型包括多个叶子节点中，每一叶子节点代表一个分组。
其中，决策树算法属于监督学习分类算法，而决策树模型代表对象属性与对象值之间的一种映射关系。其中，所述决策树模型具有N层(N为自然数，N>2)，所述电子装置根据所述条件属性设置所述决策树模型中的每个节点(包括根节点、内部节点和叶子节点)的值。所述决策树模型的第一层为根节点，所述决策树模型的第二层代表以第一级条件属性对所述根节点进行分割得到的多个节点，所述决策树模型的第三层代表以第二级条件属性对上一层节点进行分割得到的多个节点，等等。如图2所示，所述决策树模型包括四层，即，N＝4。
在本实施方式中，训练所述决策树模型具体包括：
步骤S121，以所述样本数据作为训练集计算每一条件属性的信息增益(information gain)。
步骤S122，选择信息增益最大的条件属性作为所述决策树模型的根节点以分割所述样本数据以得到下一层节点；其中，某一条件属性的信息增益越大，说明选择所述条件属性对分类提供的信息越多，越有利于确定性，越有利于将所述样本数据进行分类。如图2所示，若计算信息增益最大的条件属性为“出差情况”，则选择“出差情况”条件属性作为根节点以分割所述样本数据。
步骤S123，以每一节点包含的样本数据作为训练集重新计算每一条件属性的信息增益。
步骤S124，选择信息增益最大的条件属性分割所述节点以得到下一层节点。如图2所示，对于包含每月出差情况为0次(C＝0)的样本的节点，若计算信息增益最大的条件属性为“业务拓展情况”，则根据“业务拓展情况”条件属性对所述节点进行分割而获得下一层节点。对于包含每月出差情况为≧2次(C≧2)的样本的节点，若计算信息增益最大的条件属性为“参加培训情况”，则根据“参加培训情况”条件属性对所述节点进行分割而获得下一层节点。实际训练过程中，对位于同一层的多个节点进行分割所采用的条件属性通常不同。每一次分割后形成的节点比上一层节点的数据纯度更高。
步骤S125，递归执行步骤S123以及S124，直至分割停止。此时，每一叶节点代表一个分组，每一分组包含的绩优人员数量与绩差人员数量之间具有固定的比值，而多个分组包含的绩优人员数量与绩差人员数量之间的比值可以互不相同。如，叶子节点“分组1”的所述比值可为1:8。当然，某一叶节点包含的可均为绩优人员，也可以均为绩差人员。其中，每一组样本数据只能被划分至其中一个叶子节点中，即，每一组样本数据不能同属于两个及两个以上的叶子节点。
在本实例中，在所述决策树的层数达到一预设层数(如：4层)时停止下一次分割。在另一实施例中，在所述决策树的当前层的每一节点包含的样本数量小于一预设数量时停止下一次分割。
其中，每一条件属性的信息增益Gain(S,A)可通过如下方式计算：
Gain(S,A)＝Entropy(S)-Entropy(S,A)
其中，Gain(S,A)表示A条件属性在训练集S上的信息增益，Entropy(S)表示训练集S的信息熵，Entropy(S,A)表示A属性的信息熵。
例如，如表1所示，在步骤S121中，所述训练集共有样本数据16组，绩优的有11组，绩差的有5组。以“出差情况”条件属性的信息增益计算为例进行说明，其它条件属性的信息增益计算过程同理。对于C＝0的情况，共有样本数据5组，绩优的有4组，绩差的有1组；对于C＝1的情况，共有样本数据4组，绩优的有2组，绩差的有2组；对于C≧2的情况，共有样本数据7组，绩优的有5组，绩差的有2组。因此，“出差情况”条件属性的信息增益的计算方式如下：
Figure PCTCN2019077227-appb-000002
Figure PCTCN2019077227-appb-000003
Figure PCTCN2019077227-appb-000004
步骤S13，以所述决策树模型中的每一叶子节点作为当前节点，计算所有当前节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件。
针对每一个所述当前节点，执行以下步骤：
将所述当前节点作为目标节点，当所述目标节点符合所述预设条件时，将所述目标节点划分至所述绩优分组；
当所述目标节点不符合所述预设条件时，判断与所述目标节点属于同一父节点的其它同层节点是否符合所述预设条件；
当属于同一父节点的至少一其它同层节点符合所述预设条件时，将所述目标节点划分至绩差分组；
当属于同一父节点的所有其它同层节点均不符合所述预设条件时，忽略所述目标节点。即，既不将所述目标节点划分至所述绩优分组，也不将所述目标节点划分至所述绩差分组。
其中，每一节点的绩优率为所述节点的绩优人员数量与所述节点的待分组人员的总数量之间的比值，所述节点的绩优率可通过所述节点包含的绩优人员数量与绩差人员数量之间的比值进行计算。每一节点的绩优覆盖率为所述节点的绩优人员数量与整个决策树模型的绩优人员的总数量之间的比值。当所述节点的绩优率以及绩优覆盖率分别达到一绩优率阈值与一绩优覆盖率阈值时，确定所述节点符合所述预设条件。反之，当所述节点的绩优率未达 到所述绩优率阈值，或是，所述节点的绩优覆盖率未达到所述绩优覆盖率阈值，又或是，所述节点的绩优率以及绩优覆盖率均未达到所述绩优率阈值与所述绩优覆盖率阈值时，确定所述节点不符合所述预设条件。
例如，设置所述绩优率阈值为40％，所述绩优覆盖率阈值为20％。若计算出节点“分组8”的绩优率为80％，绩优覆盖率为20％，则节点“分组8”符合预设条件，被划分至绩优分组。对于节点“分组7”，若绩优覆盖率为30％，绩优覆盖率为25％，则节点“分组7”不符合预设条件，但是由于与所述节点“分组7”属于同一父节点的同层节点“分组8”预设条件，因此节点“分组7”被划分至绩差分组。反之，若节点“分组7”与节点“分组8”均不符合预设条件，则忽略节点“分组7”以及节点“分组8”。
在本实施方式中，所述绩优率阈值与所述绩优覆盖率阈值预先存储于所述电子装置中。所述绩优率阈值与所述绩优覆盖率阈值可根据实际需要进行调整。
步骤S14，每次当属于同一父节点的所有所述当前节点均不符合所述预设条件时，判断所述父节点是否为根节点。
步骤S15，当所述父节点不是根节点时，以每一个所述父节点作为新的当前节点，触发所述计算所有当前节点的绩优率以及绩优覆盖率的步骤；当所述父节点为根节点时，计算所述根节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定所述根节点是否符合预设条件，当所述根节点符合预设条件时将所述根节点划分至绩优分组，当所述根节点不符合预设条件时忽略所述根节点。
根据上述例子，若节点“分组7”与节点“分组8”均不符合预设条件，则需要考虑父节点是否为根节点。由于节点“分组7”与节点“分组8”的父节点并非根节点，则继续判断所述父节点的所述绩优率以及所述绩优覆盖率是否符合预设条件。可以理解，当所有的内部节点和叶子节点均不符合预设条件时，才对所述根节点进行判断。
其中，通过调整所述绩优率阈值与所述绩优覆盖率阈值，可控制所述绩优分组与所述绩差分组的大小和质量。
其中，最终得到的绩优分组可包括具有不同特征的绩优人群。如，若“分组4”被划分至所述绩优分组，则该部分绩优人群涉及的条件属性包括：出差情况＝0次、业务拓展情况≧1次，考勤率≧80。也就是说，该部分绩优人群的显著特征为：不出差、业务能力强、出勤情况较好。
若“分组8”被划分至所述绩优分组，则该部分绩优人群涉及的条件属性包括：出差情况为≧2次、参加培训情况≧1次，考勤率≧80。也就是说，该部分绩优人群的显著特征为：经常出差、乐于学习、出勤情况较好。
图3为本申请一较佳实施方式提供的人员分组装置300的结构示意图。在一些实施例中，所述人员分组装置300运行于电子装置中。所述人员分组装置300可以包括多个由程序代码段所组成的功能模块。所述人员分组装置 300的各个程序段的程序代码可以存储于电子装置的存储器中，并由所述至少一个处理器所执行，以实现人员分组功能。
本实施例中，所述人员分组装置300根据其所执行的功能，可以被划分为多个功能模块。如图3所示，所述人员分组装置300包括：获取模块301、训练模块302、执行模块303。本申请所称的模块是指一种能够被至少一个处理器所执行并且能够完成固定功能的一系列计算机可读指令段，其存储在存储器中。在本实施例中，关于各模块的功能将在后续的实施例中详述。
所述获取模块301用于获取多组样本数据，每一组样本数据包括多个条件属性以及对应的决策属性，所述决策属性为所述样本数据的绩效。
其中，所述样本数据需同时包括绩效较优的人员(即，绩优人员)和绩效较差的人员(即，绩差人员)的数据，所述样本数据可存储于所述电子装置中。在另一实施方式中，所述电子装置还可以通过接入网络连接之一外部样本库，进而来获取所述样本库存储的样本数据。在其它实施方式中，所述电子装置还可以通过大数据方式来收集并建立所述样本库。
在本实施方式中，所述条件属性可为行为轨迹(如，出差情况)、app活跃情况、业务扩展情况、消费情况、兴趣爱好、参加培训情况、考勤率等。以所述条件属性包括出差情况、业务扩展情况、参加培训情况以及考勤率为例进行说明，所述样本数据如上表1所示。
所述训练模块302用于根据所述样本数据训练一决策树模型，其中，所述决策树模型包括多个叶子节点，每一叶子节点代表一个分组。
其中，决策树算法属于监督学习分类算法，而决策树模型代表对象属性与对象值之间的一种映射关系。其中，所述决策树模型具有N层(N为自然数，N>2)，所述训练模块302根据所述条件属性设置所述决策树模型中的每个节点(包括根节点、内部节点和叶子节点)的值。所述决策树模型的第一层为根节点，所述决策树模型的第二层代表以第一级条件属性对所述根节点进行分割得到的多个节点，所述决策树模型的第三层代表以第二级条件属性对上一层节点进行分割得到的多个节点，等等。如图2所示，所述决策树模型包括四层，即，N＝4。
在本实施方式中，所述训练模块302以所述样本数据作为训练集计算每一条件属性的信息增益(information gain)，并选择信息增益最大的条件属性作为所述决策树模型的根节点以分割所述样本数据以得到下一层节点。其中，某一条件属性的信息增益越大，说明选择所述条件属性对分类提供的信息越多，越有利于确定性，越有利于将所述样本数据进行分类。如图2所示，若计算信息增益最大的条件属性为“出差情况”，则选择“出差情况”条件属性作为根节点以分割所述样本数据。
所述训练模块302进一步以每一节点包含的样本数据作为训练集重新计算每一条件属性的信息增益，并选择信息增益最大的条件属性分割所述节点以得到下一层节点。如图2所示，对于包含每月出差情况为0次(C＝0)的样本的节点，若计算信息增益最大的条件属性为“业务拓展情况”，则根据“业务拓展情况”条件属性对所述节点进行分割而获得下一层节点。对于包含每月出差 情况为≧2次(C≧2)的样本的节点，若计算信息增益最大的条件属性为“参加培训情况”，则根据“参加培训情况”条件属性对所述节点进行分割而获得下一层节点。实际训练过程中，对位于同一层的多个节点进行分割所采用的条件属性通常不同。每一次分割后形成的节点比上一层节点的数据纯度更高。
所述训练模块302进一步递归执行每一条件属性的信息增益的计算步骤以及选择信息增益最大的条件属性分割所述节点以得到下一层节点的步骤，直至分割停止。此时，每一叶节点代表一个分组，每一分组包含的绩优人员数量与绩差人员数量之间具有固定的比值，而多个分组包含的绩优人员数量与绩差人员数量之间的比值可以互不相同。如，叶子节点“分组1”的所述比值可为1:8。当然，某一叶节点包含的可均为绩优人员，也可以均为绩差人员。其中，每一组样本数据只能被划分至其中一个叶子节点中，即，每一组样本数据不能同属于两个及两个以上的叶子节点。
在本实例中，在所述决策树的层数达到一预设层数(如：4层)时停止下一次分割。在另一实施例中，在所述决策树的当前层的每一节点包含的样本数量小于一预设数量时停止下一次分割。
其中，每一条件属性的信息增益Gain(S,A)可通过如下方式计算：
Gain(S,A)＝Entropy(S)-Entropy(S,A)
其中，Gain(S,A)表示A条件属性在训练集S上的信息增益，Entropy(S)表示训练集S的信息熵，Entropy(S,A)表示A属性的信息熵。
例如，如表1所示，在所述训练模块302选择根节点的过程中，所述训练集共有样本数据16组，绩优的有11组，绩差的有5组。以“出差情况”条件属性的信息增益计算为例进行说明，其它条件属性的信息增益计算过程同理。对于C＝0的情况，共有样本数据5组，绩优的有4组，绩差的有1组；对于C＝1的情况，共有样本数据4组，绩优的有2组，绩差的有2组；对于C≧2的情况，共有样本数据7组，绩优的有5组，绩差的有2组。因此，“出差情况”条件属性的信息增益的计算方式如下：
Figure PCTCN2019077227-appb-000005
Figure PCTCN2019077227-appb-000006
Figure PCTCN2019077227-appb-000007
所述执行模块303用于以所述决策树模型中的每一叶子节点作为当前节点，计算所有当前节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件。
针对每一个所述当前节点，所述执行模块303用于：
将所述当前节点作为目标节点，当所述目标节点符合所述预设条件时，将所述目标节点划分至所述绩优分组；
当所述目标节点不符合所述预设条件时，判断与所述目标节点属于同一父节点的其它同层节点是否符合所述预设条件；
当属于同一父节点的至少一其它同层节点符合所述预设条件时，将所述目标节点划分至绩差分组；
当属于同一父节点的所有其它同层节点均不符合所述预设条件时，忽略所述目标节点。即，既不将所述目标节点划分至所述绩优分组，也不将所述目标节点划分至所述绩差分组。
其中，每一节点的绩优率为所述节点的绩优人员数量与所述节点的待分组人员的总数量之间的比值，所述节点的绩优率可通过所述节点包含的绩优人员数量与绩差人员数量之间的比值进行计算。每一节点的绩优覆盖率为所述节点的绩优人员数量与整个决策树模型的绩优人员的总数量之间的比值。当所述节点的绩优率以及绩优覆盖率分别达到一绩优率阈值与一绩优覆盖率阈值时，确定所述节点符合所述预设条件。反之，当所述节点的绩优率未达到所述绩优率阈值，或是，所述节点的绩优覆盖率未达到所述绩优覆盖率阈值，又或是，所述节点的绩优率以及绩优覆盖率均未达到所述绩优率阈值与所述绩优覆盖率阈值时，确定所述节点不符合所述预设条件。
例如，设置所述绩优率阈值为40％，所述绩优覆盖率阈值为20％。若计算出节点“分组8”的绩优率为80％，绩优覆盖率为20％，则节点“分组8”符合预设条件，被划分至绩优分组。对于节点“分组7”，若绩优覆盖率为30％，绩优覆盖率为25％，则节点“分组7”不符合预设条件，但是由于与所述节点“分组7”属于同一父节点的同层节点“分组8”预设条件，因此节点“分组7”被划分至绩差分组。反之，若节点“分组7”与节点“分组8”均不符合预设条件，则忽略节点“分组7”以及节点“分组8”。
在本实施方式中，所述绩优率阈值与所述绩优覆盖率阈值预先存储于所述电子装置中。所述绩优率阈值与所述绩优覆盖率阈值可根据实际需要进行调整。
所述执行模块303还用于每次当属于同一父节点的所有所述当前节点均不符合所述预设条件时，判断所述父节点是否为根节点。
当所述父节点不是根节点时，所述执行模块303以每一个所述父节点作为新的当前节点，触发所述计算所有当前节点的绩优率以及绩优覆盖率的步骤；当所述父节点为根节点时，计算所述根节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定所述根节点是否符合预设条件，当所述根节点符合预设条件时将所述根节点划分至绩优分组，当所述根节点不符合预设条件时忽略所述根节。
根据上述例子，若节点“分组7”与节点“分组8”均不符合预设条件，则需要考虑父节点是否为根节点。由于节点“分组7”与节点“分组8”的父节点并非根节点，则继续判断所述父节点的所述绩优率以及所述绩优覆盖率是否符合预设条件。可以理解，当所有的内部节点和叶子节点均不符合预设条件时， 才对所述根节点进行判断。
其中，通过调整所述绩优率阈值与所述绩优覆盖率阈值，可控制所述绩优分组与所述绩差分组的大小和质量。
其中，最终得到的绩优分组可包括具有不同特征的绩优人群。如，若“分组4”被划分至所述绩优分组，则该部分绩优人群涉及的条件属性包括：出差情况＝0次、业务拓展情况≧1次，考勤率≧80。也就是说，该部分绩优人群的显著特征为：不出差、业务能力强、出勤情况较好。
若“分组8”被划分至所述绩优分组，则该部分绩优人群涉及的条件属性包括：出差情况为≧2次、参加培训情况≧1次，考勤率≧80。也就是说，该部分绩优人群的显著特征为：经常出差、乐于学习、出勤情况较好。
如图4所示，图4是本申请一较佳实施方式中实现所述人员分组方法的电子装置1的结构示意图。所述电子装置1包括存储器101、处理器102以及存储于所述存储器101中并可在所述处理器102上运行的计算机可读指令103，例如人员分组程序。
所述处理器102执行所述计算机可读指令103时实现上述实施例中人员分组方法的步骤：
步骤S11，获取多组样本数据，每一组样本数据包括多个条件属性以及对应的决策属性，所述决策属性为所述样本数据的绩效。
其中，所述样本数据需同时包括绩效较优的人员(即，绩优人员)和绩效较差的人员(即，绩差人员)的数据，所述样本数据可存储于所述电子装置中。在另一实施方式中，所述电子装置还可以通过接入网络连接之一外部样本库，进而来获取所述样本库存储的样本数据。在其它实施方式中，所述电子装置还可以通过大数据方式来收集并建立所述样本库。
在本实施方式中，所述条件属性可为行为轨迹(如，出差情况)、app活跃情况、业务扩展情况、消费情况、兴趣爱好、参加培训情况、考勤率等。以所述条件属性包括出差情况、业务扩展情况、参加培训情况以及考勤率为例进行说明，所述样本数据如上表1所示。
步骤S12，根据所述样本数据训练一决策树模型，其中，所述决策树模型包括多个叶子节点，每一叶子节点代表一个分组。
其中，决策树算法属于监督学习分类算法，而决策树模型代表对象属性与对象值之间的一种映射关系。其中，所述决策树模型具有N层(N为自然数，N>2)，所述电子装置根据所述条件属性设置所述决策树模型中的每个节点(包括根节点、内部节点和叶子节点)的值。所述决策树模型的第一层为根节点，所述决策树模型的第二层代表以第一级条件属性对所述根节点进行分割得到的多个节点，所述决策树模型的第三层代表以第二级条件属性对上一层节点进行分割得到的多个节点，等等。如图2所示，所述决策树模型包括四层，即，N＝4。
在本实施方式中，训练所述决策树模型具体包括：
步骤S121，以所述样本数据作为训练集计算每一条件属性的信息增益 (information gain)。
步骤S122，选择信息增益最大的条件属性作为所述决策树模型的根节点以分割所述样本数据以得到下一层节点；其中，某一条件属性的信息增益越大，说明选择所述条件属性对分类提供的信息越多，越有利于确定性，越有利于将所述样本数据进行分类。如图2所示，若计算信息增益最大的条件属性为“出差情况”，则选择“出差情况”条件属性作为根节点以分割所述样本数据。
步骤S123，以每一节点包含的样本数据作为训练集重新计算每一条件属性的信息增益。
步骤S124，选择信息增益最大的条件属性分割所述节点以得到下一层节点。如图2所示，对于包含每月出差情况为0次(C＝0)的样本的节点，若计算信息增益最大的条件属性为“业务拓展情况”，则根据“业务拓展情况”条件属性对所述节点进行分割而获得下一层节点。对于包含每月出差情况为≧2次(C≧2)的样本的节点，若计算信息增益最大的条件属性为“参加培训情况”，则根据“参加培训情况”条件属性对所述节点进行分割而获得下一层节点。实际训练过程中，对位于同一层的多个节点进行分割所采用的条件属性通常不同。每一次分割后形成的节点比上一层节点的数据纯度更高。
步骤S125，递归执行步骤S123以及S124，直至分割停止。此时，每一叶节点代表一个分组，每一分组包含的绩优人员数量与绩差人员数量之间具有固定的比值，而多个分组包含的绩优人员数量与绩差人员数量之间的比值可以互不相同。如，叶子节点“分组1”的所述比值可为1:8。当然，某一叶节点包含的可均为绩优人员，也可以均为绩差人员。其中，每一组样本数据只能被划分至其中一个叶子节点中，即，每一组样本数据不能同属于两个及两个以上的叶子节点。
在本实例中，在所述决策树的层数达到一预设层数(如：4层)时停止下一次分割。在另一实施例中，在所述决策树的当前层的每一节点包含的样本数量小于一预设数量时停止下一次分割。
其中，每一条件属性的信息增益Gain(S,A)可通过如下方式计算：
Gain(S,A)＝Entropy(S)-Entropy(S,A)
其中，Gain(S,A)表示A条件属性在训练集S上的信息增益，Entropy(S)表示训练集S的信息熵，Entropy(S,A)表示A属性的信息熵。
例如，如表1所示，在步骤S121中，所述训练集共有样本数据16组，绩优的有11组，绩差的有5组。以“出差情况”条件属性的信息增益计算为例进行说明，其它条件属性的信息增益计算过程同理。对于C＝0的情况，共有样本数据5组，绩优的有4组，绩差的有1组；对于C＝1的情况，共有样本数据4组，绩优的有2组，绩差的有2组；对于C≧2的情况，共有样本数据7组，绩优的有5组，绩差的有2组。因此，“出差情况”条件属性的信息增益的计算方式如下：
Figure PCTCN2019077227-appb-000008
Figure PCTCN2019077227-appb-000009
Figure PCTCN2019077227-appb-000010
步骤S13，以所述决策树模型中的每一叶子节点作为当前节点，计算所有当前节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件。
针对每一个所述当前节点，执行以下步骤：
将所述当前节点作为目标节点，当所述目标节点符合所述预设条件时，将所述目标节点划分至所述绩优分组；
当所述目标节点不符合所述预设条件时，判断与所述目标节点属于同一父节点的其它同层节点是否符合所述预设条件；
当属于同一父节点的至少一其它同层节点符合所述预设条件时，将所述目标节点划分至绩差分组；
当属于同一父节点的所有其它同层节点均不符合所述预设条件时，忽略所述目标节点。即，既不将所述目标节点划分至所述绩优分组，也不将所述目标节点划分至所述绩差分组。
其中，每一节点的绩优率为所述节点的绩优人员数量与所述节点的待分组人员的总数量之间的比值，所述节点的绩优率可通过所述节点包含的绩优人员数量与绩差人员数量之间的比值进行计算。每一节点的绩优覆盖率为所述节点的绩优人员数量与整个决策树模型的绩优人员的总数量之间的比值。当所述节点的绩优率以及绩优覆盖率分别达到一绩优率阈值与一绩优覆盖率阈值时，确定所述节点符合所述预设条件。反之，当所述节点的绩优率未达到所述绩优率阈值，或是，所述节点的绩优覆盖率未达到所述绩优覆盖率阈值，又或是，所述节点的绩优率以及绩优覆盖率均未达到所述绩优率阈值与所述绩优覆盖率阈值时，确定所述节点不符合所述预设条件。
例如，设置所述绩优率阈值为40％，所述绩优覆盖率阈值为20％。若计算出节点“分组8”的绩优率为80％，绩优覆盖率为20％，则节点“分组8”符合预设条件，被划分至绩优分组。对于节点“分组7”，若绩优覆盖率为30％，绩优覆盖率为25％，则节点“分组7”不符合预设条件，但是由于与所述节点“分组7”属于同一父节点的同层节点“分组8”预设条件，因此节点“分组7”被划分至绩差分组。反之，若节点“分组7”与节点“分组8”均不符合预设条件，则忽略节点“分组7”以及节点“分组8”。
在本实施方式中，所述绩优率阈值与所述绩优覆盖率阈值预先存储于所述电子装置中。所述绩优率阈值与所述绩优覆盖率阈值可根据实际需要进行 调整。
步骤S14，每次当属于同一父节点的所有所述当前节点均不符合所述预设条件时，判断所述父节点是否为根节点。
步骤S15，当所述父节点不是根节点时，以每一个所述父节点作为新的当前节点，触发所述计算所有当前节点的绩优率以及绩优覆盖率的步骤；当所述父节点为根节点时，计算所述根节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定所述根节点是否符合预设条件，当所述根节点符合预设条件时将所述根节点划分至绩优分组，当所述根节点不符合预设条件时忽略所述根节点。
根据上述例子，若节点“分组7”与节点“分组8”均不符合预设条件，则需要考虑父节点是否为根节点。由于节点“分组7”与节点“分组8”的父节点并非根节点，则继续判断所述父节点的所述绩优率以及所述绩优覆盖率是否符合预设条件。可以理解，当所有的内部节点和叶子节点均不符合预设条件时，才对所述根节点进行判断。
其中，通过调整所述绩优率阈值与所述绩优覆盖率阈值，可控制所述绩优分组与所述绩差分组的大小和质量。
其中，最终得到的绩优分组可包括具有不同特征的绩优人群。如，若“分组4”被划分至所述绩优分组，则该部分绩优人群涉及的条件属性包括：出差情况＝0次、业务拓展情况≧1次，考勤率≧80。也就是说，该部分绩优人群的显著特征为：不出差、业务能力强、出勤情况较好。
若“分组8”被划分至所述绩优分组，则该部分绩优人群涉及的条件属性包括：出差情况为≧2次、参加培训情况≧1次，考勤率≧80。也就是说，该部分绩优人群的显著特征为：经常出差、乐于学习、出勤情况较好。
或者，所述处理器102执行所述计算机可读指令103时实现上述人员分组装置实施例中各模块/单元的功能，例如图3中的单元301-303。
本申请实施例在决策树建模中加入绩差人员的信息，使得分组目标更明确，且数据的执行效果更好，保证最终获得的绩优分组为绩优人数比例达到一定程度的群体；而且，可通过绩优分组的条件属性识别绩优人员的显著特征，从而构建用户画像。
示例性的，所述计算机可读指令103可以被分割成一个或多个模块/单元，所述一个或者多个模块/单元被存储在所述存储器101中，并由所述处理器102执行，以完成本申请。所述一个或多个模块/单元可以是能够完成特定功能的一系列计算机可读指令段，该指令段用于描述所述计算机可读指令103在所述电子装置1中的执行过程。例如，所述计算机可读指令103可以被分割成图3中的获取模块301、训练模块302、执行模块303。
所述电子装置1可以是桌上型计算机、笔记本、掌上电脑及云端服务器等计算设备。本领域技术人员可以理解，所述示意图仅仅是电子装置1的示例，并不构成对电子装置1的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件，例如所述电子装置1还可以包括输入输出设备、网络接入设备、总线等。
所称处理器102可以是中央处理单元(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器30也可以是任何常规的处理器等，所述处理器102是所述电子装置1的控制中心，利用各种接口和线路连接整个电子装置1的各个部分。
所述存储器101可用于存储所述计算机可读指令103和/或模块/单元，所述处理器102通过运行或执行存储在所述存储器101内的计算机可读指令和/或模块/单元，以及调用存储在存储器101内的数据，实现所述电子装置1的各种功能。所述存储器101可主要包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需的应用程序(比如声音播放功能、图像播放功能等)等；存储数据区可存储根据电子装置1的使用所创建的数据(比如音频数据、电话本等)等。此外，存储器101可以包括高速随机存取存储器，还可以包括非易失性存储器，例如硬盘、内存、插接式硬盘，智能存储卡(Smart Media Card,SMC)，安全数字(Secure Digital,SD)卡，闪存卡(Flash Card)、至少一个磁盘存储器件、闪存器件、或其他易失性固态存储器件。
所述电子装置1集成的模块/单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个非易失性可读取存储介质中。基于这样的理解，本申请实现上述实施例方法中的全部或部分流程，也可以通过计算机可读指令来指令相关的硬件来完成，所述的计算机可读指令可存储于一非易失性可读存储介质中，该计算机可读指令在被处理器执行时，可实现上述各个方法实施例的步骤。其中，所述计算机可读指令包括计算机可读指令代码，所述计算机可读指令代码可以为源代码形式、对象代码形式、可执行文件或某些中间形式等。所述非易失性可读介质可以包括：能够携带所述计算机可读指令代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、电载波信号、电信信号以及软件分发介质等。需要说明的是，所述非易失性可读介质包含的内容可以根据司法管辖区内立法和专利实践的要求进行适当的增减，例如在某些司法管辖区，根据立法和专利实践，非易失性可读介质不包括电载波信号和电信信号。
在本申请所提供的几个实施例中，应该理解到，所揭露的电子装置和方法，可以通过其它的方式实现。例如，以上所描述的电子装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式。
另外，在本申请各个实施例中的各功能单元可以集成在相同处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在相同单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用硬件加软件 功能模块的形式实现。
对于本领域技术人员而言，显然本申请不限于上述示范性实施例的细节，而且在不背离本申请的精神或基本特征的情况下，能够以其他的具体形式实现本申请。因此，无论从哪一点来看，均应将实施例看作是示范性的，而且是非限制性的，本申请的范围由所附权利要求而不是上述说明限定，因此旨在将落在权利要求的等同要件的含义和范围内的所有变化涵括在本申请内。不应将权利要求中的任何附图标记视为限制所涉及的权利要求。此外，显然“包括”一词不排除其他单元或步骤，单数不排除复数。电子装置权利要求中陈述的多个单元或电子装置也可以由同一个单元或电子装置通过软件或者硬件来实现。第一，第二等词语用来表示名称，而并不表示任何特定的顺序。
最后应说明的是，以上实施例仅用以说明本申请的技术方案而非限制，尽管参照较佳实施例对本申请进行了详细说明，本领域的普通技术人员应当理解，可以对本申请的技术方案进行修改或等同替换，而不脱离本申请技术方案的精神和范围。

Claims (20)

  1. 一种人员分组方法，其特征在于，包括：
    获取多组样本数据，每一组样本数据包括多个条件属性以及对应的决策属性，所述决策属性为所述样本数据的绩效；
    根据所述样本数据训练一决策树模型，其中，所述决策树模型包括多个叶子节点，每一叶子节点代表一个分组；
    以所述决策树模型中的每一叶子节点作为当前节点，计算所有当前节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件；
    当属于同一父节点的所有所述当前节点均不符合所述预设条件时，判断所述父节点是否为根节点；
    当所述父节点不是根节点时，以每一个所述父节点作为新的当前节点，触发所述计算所有当前节点的绩优率以及绩优覆盖率的步骤；
    当所述父节点为根节点时，计算所述根节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定所述根节点是否符合预设条件，当所述根节点符合预设条件时将所述根节点划分至绩优分组，当所述根节点不符合预设条件时忽略所述根节点；
    针对每一个所述当前节点，执行以下步骤：
    将所述当前节点作为目标节点，当所述目标节点符合所述预设条件时，将所述目标节点划分至所述绩优分组；
    当所述目标节点不符合所述预设条件时，判断与所述目标节点属于同一父节点的其它同层节点是否符合所述预设条件；
    当属于同一父节点的至少一其它同层节点符合所述预设条件时，将所述目标节点划分至绩差分组；
    当属于同一父节点的所有其它同层节点均不符合所述预设条件时，忽略所述目标节点。
  2. 如权利要求1所述的人员分组方法，其特征在于，所述根据所述样本数据训练一决策树模型包括：
    以所述样本数据作为训练集计算每一条件属性的信息增益；
    选择信息增益最大的条件属性作为所述决策树模型的根节点以分割所述样本数据以得到下一层节点；
    以每一节点包含的样本数据作为训练集重新计算每一条件属性的信息增益；
    选择信息增益最大的条件属性分割所述节点以得到下一层节点；以及
    递归执行重新计算每一条件属性的信息增益的步骤以及选择信息增益最大的条件属性分割所述节点的步骤，直至分割停止。
  3. 如权利要求2所述的人员分组方法，其特征在于，在所述决策树的层数达到一预设层数时停止下一次分割。
  4. 如权利要求2所述的人员分组方法，其特征在于，在所述决策树的当前层的每一节点包含的样本数量小于一预设数量时停止下一次分割。
  5. 如权利要求1所述的人员分组方法，其特征在于，每一节点的绩优率为所述节点的绩优人员数量与所述节点的待分组人员的总数量之间的比值，每一节点的绩优覆盖率为所述节点的绩优人员数量与整个决策树模型的绩优人员的总数量之间的比值。
  6. 如权利要求5所述的人员分组方法，其特征在于，所述根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件包括：
    当所述当前节点的绩优率以及绩优覆盖率分别达到一绩优率阈值与一绩优覆盖率阈值时，确定所述当前节点符合所述预设条件；以及
    当所述当前节点的绩优率未达到所述绩优率阈值，或所述当前节点的绩优覆盖率未达到所述绩优覆盖率阈值，或所述当前节点的绩优率以及绩优覆盖率均未达到所述绩优率阈值与所述绩优覆盖率阈值时，确定所述当前节点不符合所述预设条件。
  7. 如权利要求1所述的人员分组方法，其特征在于，所述样本数据同时包括绩优人员和绩差人员的数据。
  8. 一种人员分组装置，其特征在于，包括：
    获取模块，用于获取多组样本数据，每一组样本数据包括多个条件属性以及对应的决策属性，所述决策属性为所述样本数据的绩效；
    训练模块，用于根据所述样本数据训练一决策树模型，其中，所述决策树模型包括多个叶子节点，每一叶子节点代表一个分组；
    执行模块，用于计算所有当前节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件；
    当属于同一父节点的所有所述当前节点均不符合所述预设条件时，所述执行模块判断所述父节点是否为根节点；
    当所述父节点不是根节点时，所述执行模块以每一个所述父节点作为新的当前节点，触发所述计算所有当前节点的绩优率以及绩优覆盖率的步骤；
    当所述父节点为根节点时，所述执行模块计算所述根节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定所述根节点是否符合预设条件，当所述根节点符合预设条件时将所述根节点划分至绩优分组，当所述根节点不符合预设条件时忽略所述根节点；
    针对每一个所述当前节点，所述执行模块用于：
    将所述当前节点作为目标节点，当所述目标节点符合所述预设条件时，将所述目标节点划分至所述绩优分组；
    当所述目标节点不符合所述预设条件时，判断与所述目标节点属于同一父节点的其它同层节点是否符合所述预设条件；
    当属于同一父节点的至少一其它同层节点符合所述预设条件时，将所述目标节点划分至绩差分组；
    当属于同一父节点的所有其它同层节点均不符合所述预设条件时，忽略所述目标节点。
  9. 一种电子装置，其特征在于，包括处理器和存储器，所述存储器中存储至少一个计算机可读指令，所述处理器执行所述计算机可读指令以实现以下步骤：
    获取多组样本数据，每一组样本数据包括多个条件属性以及对应的决策属性，所述决策属性为所述样本数据的绩效；
    根据所述样本数据训练一决策树模型，其中，所述决策树模型包括多个叶子节点，每一叶子节点代表一个分组；
    以所述决策树模型中的每一叶子节点作为当前节点，计算所有当前节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件；
    当属于同一父节点的所有所述当前节点均不符合所述预设条件时，判断所述父节点是否为根节点；
    当所述父节点不是根节点时，以每一个所述父节点作为新的当前节点，触发所述计算所有当前节点的绩优率以及绩优覆盖率的步骤；
    当所述父节点为根节点时，计算所述根节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定所述根节点是否符合预设条件，当所述根节点符合预设条件时将所述根节点划分至绩优分组，当所述根节点不符合预设条件时忽略所述根节点；
    针对每一个所述当前节点，执行以下步骤：
    将所述当前节点作为目标节点，当所述目标节点符合所述预设条件时，将所述目标节点划分至所述绩优分组；
    当所述目标节点不符合所述预设条件时，判断与所述目标节点属于同一父节点的其它同层节点是否符合所述预设条件；
    当属于同一父节点的至少一其它同层节点符合所述预设条件时，将所述目标节点划分至绩差分组；
    当属于同一父节点的所有其它同层节点均不符合所述预设条件时，忽略所述目标节点。
  10. 如权利要求9所述的电子装置，其特征在于，所述处理器在根据所述样本数据训练一决策树模型时，执行所述计算机可读指令以实现以下步骤：
    以所述样本数据作为训练集计算每一条件属性的信息增益；
    选择信息增益最大的条件属性作为所述决策树模型的根节点以分割所述样本数据以得到下一层节点；
    以每一节点包含的样本数据作为训练集重新计算每一条件属性的信息增益；
    选择信息增益最大的条件属性分割所述节点以得到下一层节点；以及
    递归执行重新计算每一条件属性的信息增益的步骤以及选择信息增益最大的条件属性分割所述节点的步骤，直至分割停止。
  11. 如权利要求10所述的电子装置，其特征在于，所述处理器执行所述计算机可读指令时还用以实现以下步骤：
    在所述决策树的层数达到一预设层数时停止下一次分割。
  12. 如权利要求10所述的电子装置，其特征在于，所述处理器执行所述计算机可读指令时还用以实现以下步骤：
    在所述决策树的当前层的每一节点包含的样本数量小于一预设数量时停止下一次分割。
  13. 如权利要求9所述的电子装置，其特征在于，每一节点的绩优率为所述节点的绩优人员数量与所述节点的待分组人员的总数量之间的比值，每一节点的绩优覆盖率为所述节点的绩优人员数量与整个决策树模型的绩优人员的总数量之间的比值。
  14. 如权利要求13所述的电子装置，其特征在于，所述根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件时，所述处理器执行所述计算机可读指令以实现以下步骤：
    当所述当前节点的绩优率以及绩优覆盖率分别达到一绩优率阈值与一绩优覆盖率阈值时，确定所述当前节点符合所述预设条件；以及
    当所述当前节点的绩优率未达到所述绩优率阈值，或所述当前节点的绩优覆盖率未达到所述绩优覆盖率阈值，或所述当前节点的绩优率以及绩优覆盖率均未达到所述绩优率阈值与所述绩优覆盖率阈值时，确定所述当前节点不符合所述预设条件。
  15. 一种非易失性可读存储介质，其特征在于，所述非易失性可读存储介质上存储至少一个计算机可读指令，所述计算机可读指令被处理器执行时实现以下步骤：
    获取多组样本数据，每一组样本数据包括多个条件属性以及对应的决策属性，所述决策属性为所述样本数据的绩效；
    根据所述样本数据训练一决策树模型，其中，所述决策树模型包括多个叶子节点，每一叶子节点代表一个分组；
    以所述决策树模型中的每一叶子节点作为当前节点，计算所有当前节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件；
    当属于同一父节点的所有所述当前节点均不符合所述预设条件时，判断所述父节点是否为根节点；
    当所述父节点不是根节点时，以每一个所述父节点作为新的当前节点，触发所述计算所有当前节点的绩优率以及绩优覆盖率的步骤；
    当所述父节点为根节点时，计算所述根节点的绩优率以及绩优覆盖率，根据所述绩优率以及所述绩优覆盖率确定所述根节点是否符合预设条件，当所述根节点符合预设条件时将所述根节点划分至绩优分组，当所述根节点不符合预设条件时忽略所述根节点；
    针对每一个所述当前节点，执行以下步骤：
    将所述当前节点作为目标节点，当所述目标节点符合所述预设条件时， 将所述目标节点划分至所述绩优分组；
    当所述目标节点不符合所述预设条件时，判断与所述目标节点属于同一父节点的其它同层节点是否符合所述预设条件；
    当属于同一父节点的至少一其它同层节点符合所述预设条件时，将所述目标节点划分至绩差分组；
    当属于同一父节点的所有其它同层节点均不符合所述预设条件时，忽略所述目标节点。
  16. 如权利要求15所述的存储介质，其特征在于，在根据所述样本数据训练一决策树模型时，所述计算机可读指令被所述处理器执行以实现以下步骤：
    以所述样本数据作为训练集计算每一条件属性的信息增益；
    选择信息增益最大的条件属性作为所述决策树模型的根节点以分割所述样本数据以得到下一层节点；
    以每一节点包含的样本数据作为训练集重新计算每一条件属性的信息增益；
    选择信息增益最大的条件属性分割所述节点以得到下一层节点；以及
    递归执行重新计算每一条件属性的信息增益的步骤以及选择信息增益最大的条件属性分割所述节点的步骤，直至分割停止。
  17. 如权利要求16所述的存储介质，其特征在于，所述计算机可读指令被所述处理器执行时还用以实现以下步骤：
    在所述决策树的层数达到一预设层数时停止下一次分割。
  18. 如权利要求16所述的存储介质，其特征在于，所述计算机可读指令被所述处理器执行时还用以实现以下步骤：
    在所述决策树的当前层的每一节点包含的样本数量小于一预设数量时停止下一次分割。
  19. 如权利要求15所述的存储介质，其特征在于，每一节点的绩优率为所述节点的绩优人员数量与所述节点的待分组人员的总数量之间的比值，每一节点的绩优覆盖率为所述节点的绩优人员数量与整个决策树模型的绩优人员的总数量之间的比值。
  20. 如权利要求19所述的存储介质，其特征在于，所述根据所述绩优率以及所述绩优覆盖率确定每一个所述当前节点是否符合预设条件时，所述计算机可读指令被所述处理器执行以实现以下步骤：
    当所述当前节点的绩优率以及绩优覆盖率分别达到一绩优率阈值与一绩优覆盖率阈值时，确定所述当前节点符合所述预设条件；以及
    当所述当前节点的绩优率未达到所述绩优率阈值，或所述当前节点的绩优覆盖率未达到所述绩优覆盖率阈值，或所述当前节点的绩优率以及绩优覆盖率均未达到所述绩优率阈值与所述绩优覆盖率阈值时，确定所述当前节点不符合所述预设条件。
PCT/CN2019/077227 2018-08-27 2019-03-06 人员分组方法、装置、电子装置及存储介质 Ceased WO2020042580A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810982127.6 2018-08-27
CN201810982127.6A CN109214671B (zh) 2018-08-27 2018-08-27 人员分组方法、装置、电子装置及计算机可读存储介质

Publications (1)

Publication Number Publication Date
WO2020042580A1 true WO2020042580A1 (zh) 2020-03-05

Family

ID=64989355

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/077227 Ceased WO2020042580A1 (zh) 2018-08-27 2019-03-06 人员分组方法、装置、电子装置及存储介质

Country Status (2)

Country Link
CN (1) CN109214671B (zh)
WO (1) WO2020042580A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112633680A (zh) * 2020-12-21 2021-04-09 深圳前海微众银行股份有限公司 风险事件的策略确定方法、装置、设备及可读存储介质
CN113486007A (zh) * 2021-06-29 2021-10-08 天九共享网络科技集团有限公司 信息采集方法、系统、电子设备及存储介质

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109214671B (zh) * 2018-08-27 2022-03-01 平安科技（深圳）有限公司 人员分组方法、装置、电子装置及计算机可读存储介质
CN111898700A (zh) * 2020-08-13 2020-11-06 北京科技大学 一种基于决策树算法和环境传感参数的人员在室识别方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102214213A (zh) * 2011-05-31 2011-10-12 中国科学院计算技术研究所 一种采用决策树的数据分类方法和系统
CN105354588A (zh) * 2015-09-28 2016-02-24 北京邮电大学 一种构造决策树的方法
CN105894177A (zh) * 2016-03-25 2016-08-24 国家电网公司 一种基于决策树算法的电力设备运行风险分析评估方法
CN109214671A (zh) * 2018-08-27 2019-01-15 平安科技（深圳）有限公司 人员分组方法、装置、电子装置及计算机可读存储介质

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101877086A (zh) * 2009-12-15 2010-11-03 中国科学院声学研究所 一种适用于业务运营支撑系统的客户管理方法
CN101916402A (zh) * 2010-08-03 2010-12-15 中南大学 一种铁路编组调车方法
CN102254044A (zh) * 2011-05-10 2011-11-23 洪涛 一种在线社交网络的人脉系数测量方法
CN104636989A (zh) * 2015-02-11 2015-05-20 广东电网有限责任公司中山供电局 电力系统监控告警信息处理方法和系统
CN107292186B (zh) * 2016-03-31 2021-01-12 阿里巴巴集团控股有限公司 一种基于随机森林的模型训练方法和装置
CN106611284A (zh) * 2016-06-20 2017-05-03 四川用联信息技术有限公司 一种哈夫曼物料采购决策算法
CN107679994A (zh) * 2017-07-26 2018-02-09 平安科技（深圳）有限公司 核保决策树的创建方法、装置、计算机设备及存储介质
CN108416368A (zh) * 2018-02-08 2018-08-17 北京三快在线科技有限公司 样本特征重要性的确定方法及装置，电子设备

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102214213A (zh) * 2011-05-31 2011-10-12 中国科学院计算技术研究所 一种采用决策树的数据分类方法和系统
CN105354588A (zh) * 2015-09-28 2016-02-24 北京邮电大学 一种构造决策树的方法
CN105894177A (zh) * 2016-03-25 2016-08-24 国家电网公司 一种基于决策树算法的电力设备运行风险分析评估方法
CN109214671A (zh) * 2018-08-27 2019-01-15 平安科技（深圳）有限公司 人员分组方法、装置、电子装置及计算机可读存储介质

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112633680A (zh) * 2020-12-21 2021-04-09 深圳前海微众银行股份有限公司 风险事件的策略确定方法、装置、设备及可读存储介质
CN112633680B (zh) * 2020-12-21 2024-04-16 深圳前海微众银行股份有限公司 风险事件的策略确定方法、装置、设备及可读存储介质
CN113486007A (zh) * 2021-06-29 2021-10-08 天九共享网络科技集团有限公司 信息采集方法、系统、电子设备及存储介质

Also Published As

Publication number Publication date
CN109214671B (zh) 2022-03-01
CN109214671A (zh) 2019-01-15

Similar Documents

Publication Publication Date Title
CN104281882B (zh) 基于用户特征的预测社交网络信息流行度的方法及系统
WO2020042579A1 (zh) 分组归纳方法、装置、电子装置及存储介质
KR102874907B1 (ko) 인공지능 기반 사람 이미지 파일링 방법, 기기 및 저장매체
WO2020119053A1 (zh) 一种图片聚类方法、装置、存储介质及终端设备
WO2021184727A1 (zh) 数据异常检测方法、装置、电子设备及存储介质
WO2021082697A1 (zh) 基于批量告警事件的定位方法、装置、电子设备及介质
WO2020042580A1 (zh) 人员分组方法、装置、电子装置及存储介质
CN111553215A (zh) 人员关联方法及其装置、图卷积网络训练方法及其装置
CN115086880B (zh) 一种出行特征识别方法、装置、设备及存储介质
CN103678670A (zh) 一种微博热词与热点话题挖掘系统及方法
CN108520471B (zh) 重叠社区发现方法、装置、设备及存储介质
WO2021047021A1 (zh) 信息挖掘方法、装置、设备和存储介质
CN110825868A (zh) 一种基于话题热度的文本推送方法、终端设备及存储介质
US9754015B2 (en) Feature rich view of an entity subgraph
WO2023035535A1 (zh) 语义分割网络的训练方法、装置、设备及存储介质
TW202123026A (zh) 資料歸檔方法、裝置、電腦裝置及存儲介質
CN115035017A (zh) 细胞密度分群方法、装置、电子设备及存储介质
CN117473315A (zh) 一种基于多层感知机的图分类模型构建方法和图分类方法
CN112069048A (zh) 日志处理方法、装置及存储介质
CN113378902B (zh) 一种基于优化视频特征的视频抄袭检测方法
CN114723652A (zh) 细胞密度确定方法、装置、电子设备及存储介质
CN105468669A (zh) 一种融合用户关系的自适应微博话题追踪方法
CN115909450A (zh) 一种数据识别方法及装置
CN116049889A (zh) 基于前缀树的差分隐私轨迹发布方法及装置
CN114390550B (zh) 一种网络类型识别的方法、相关装置、设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19855964

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19855964

Country of ref document: EP

Kind code of ref document: A1