CN113177212B - Joint prediction method and device - Google Patents
Joint prediction method and device Download PDFInfo
- Publication number
- CN113177212B CN113177212B CN202110447583.2A CN202110447583A CN113177212B CN 113177212 B CN113177212 B CN 113177212B CN 202110447583 A CN202110447583 A CN 202110447583A CN 113177212 B CN113177212 B CN 113177212B
- Authority
- CN
- China
- Prior art keywords
- decision
- data source
- node
- tree structure
- source end
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Bioethics (AREA)
- General Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Computer Security & Cryptography (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the specification provides a joint prediction method and device. The joint prediction method applied to the data source end comprises the following steps: obtaining a first decision algorithm; the first decision algorithm comprises an attribute judgment method for the private data of the data source end; according to a first decision algorithm, carrying out attribute judgment on local private data to obtain local decision information; the local decision information is provided to the decision-making party.
Description
Technical Field
One or more embodiments of the present disclosure relate to electronic information technology, and more particularly, to a joint prediction method and apparatus.
Background
In the big data era, there are very many data islands. The data of each user is dispersed in different data sources. However, it is necessary to perform joint prediction for a user by using various data stored in different data sources so as to perform service control according to the result of the joint prediction.
At present, the method for performing joint prediction mainly includes that each data source end collects and sends data owned by each data source end to a decision-making party, and the decision-making party performs joint prediction according to the acquired data of each data source end. However, the joint prediction method can cause data leakage and is not high in safety. It is therefore desirable to provide a more secure joint prediction method.
Disclosure of Invention
One or more embodiments of the present specification describe joint prediction methods and apparatus that can improve the security of joint prediction.
According to a first aspect, there is provided a joint prediction method applied to a data source, including:
obtaining a first decision algorithm; the first decision algorithm comprises an attribute judgment method for the private data of the data source end;
according to a first decision algorithm, carrying out attribute judgment on local private data to obtain local decision information;
the local decision information is provided to the decision-making party.
Wherein the first decision algorithm is a first tree structure;
the first tree structure is generated by utilizing at least one node corresponding to the data source end and at least two edges connected to each node; wherein each node characterizes: judging the attribute of a private data of the data source end; different edges represent different attribute judgment results.
Wherein the obtaining of the local decision information includes:
for each node corresponding to the data source end in the first tree structure, executing:
obtaining an attribute judgment result according to the attribute judgment method represented by the node and local private data; and
selecting a decision edge from at least two edges of the node for connecting the next-level node according to the currently obtained attribute judgment result; the attribute judgment result of the decision edge representation is the same as the currently obtained attribute judgment result;
and obtaining local decision information according to the obtained decision edges.
Wherein, the obtaining the local decision information according to each obtained decision edge comprises:
sequentially adding the serial numbers of the decision edges into a branch set according to the sequence of the decision edges in the first tree structure from the root node to the leaf node;
determining the set of branches as the local decision information.
Wherein, the sequentially adding the serial numbers of the decision edges into the branch set comprises:
and judging whether at least two adjacent nodes corresponding to the data source end exist in the first tree structure, if so, adding the serial number of the decision edge connected with the last level node in the at least two adjacent nodes into the branch set according to the serial number of each decision edge connected with the at least two adjacent nodes.
Wherein the first tree structure comprises:
nodes corresponding to all data source ends participating in the joint prediction and at least two edges connected to each node;
alternatively, the first and second liquid crystal display panels may be,
and the nodes corresponding to the data source end and at least two edges connected on each node.
According to a second aspect, there is provided a joint prediction method applied to a decision-making party, including:
obtaining a second decision algorithm; the second decision algorithm comprises decision branches corresponding to the data source ends participating in the joint prediction and incidence relations among the decision branches; different decision branches represent different attribute judgment results;
acquiring local decision information provided by each data source end;
and obtaining a joint prediction result according to the second decision algorithm and the obtained local decision information.
Wherein the second decision algorithm is a second tree structure;
the second tree structure includes: each first tree structure and each leaf node obtained by each data source end; each leaf node corresponds to a prediction result;
each decision branch is an edge in the second tree structure;
the association relation among the decision branches is the connection relation of each edge in the second tree structure;
a data source provides local decision information comprising: the data source end selects the information of the decision edge from each edge of the first tree structure.
Wherein the obtaining of the joint prediction result comprises:
determining the position and the connection relation of each decision edge in the second tree structure according to the information of each decision edge provided by each data source end;
obtaining a path from a root node to a leaf node in the second tree structure according to the determined position and the connection relation;
and determining the obtained leaf node on the path as the joint prediction result.
According to a third aspect, there is provided a joint prediction apparatus, provided in a data source, including:
a first decision algorithm saving module configured to obtain a first decision algorithm; the first decision algorithm comprises an attribute judgment method for the private data of the data source end;
the decision acquisition module is configured to perform attribute judgment on local private data according to a first decision algorithm to obtain local decision information;
a decision providing module configured to provide the local decision information to the decision-making party.
Wherein the content of the first and second substances,
the first decision algorithm is a first tree structure; the first tree structure is generated by utilizing at least one node corresponding to the data source end and at least two edges connected to each node; wherein each node characterizes: judging the attribute of a private data of the data source end; different edges represent different attribute judgment results.
Wherein the decision obtaining module is configured to execute, for each node corresponding to the data source end in the first tree structure:
obtaining an attribute judgment result according to the attribute judgment method represented by the node and local private data; and
selecting a decision edge from at least two edges of the node for connecting the next-level node according to the currently obtained attribute judgment result; the attribute judgment result of the decision edge representation is the same as the currently obtained attribute judgment result;
and obtaining local decision information according to the obtained decision edges.
The decision obtaining module is configured to add the serial numbers of the decision edges to a branch set in sequence according to the sequence of the decision edges in the first tree structure from the root node to the leaf node, and determine the branch set as the local decision information.
The decision obtaining module is configured to determine whether at least two adjacent nodes corresponding to the data source end exist in the first tree structure, and if so, add, to the branch set, only the number of the decision edge connected to the last node in the at least two adjacent nodes with respect to the number of each decision edge connected to the at least two adjacent nodes.
Wherein the first tree structure comprises:
the nodes corresponding to all the data source ends participating in the joint prediction and at least two edges connected to each node;
alternatively, the first and second electrodes may be,
and the nodes corresponding to the data source end and at least two edges connected on each node.
According to a fourth aspect, there is provided a joint prediction apparatus, provided on a decision-making side, comprising:
a second decision algorithm saving module configured to obtain a second decision algorithm; the second decision algorithm comprises decision branches corresponding to the data source ends participating in the joint prediction and incidence relations among the decision branches; different decision branches represent different attribute judgment results;
the decision summarizing module is configured to acquire local decision information provided by the data source ends;
and the prediction module is configured to obtain a joint prediction result according to the second decision algorithm and the acquired local decision information.
Wherein the second decision algorithm is a second tree structure; the second tree structure includes: each first tree structure and each leaf node obtained by each data source end; each leaf node corresponds to a prediction result;
each decision branch is an edge in the second tree structure;
the association relation among the decision branches is the connection relation of each edge in the second tree structure;
a data source provides local decision information comprising: the data source end selects the information of the decision edge from each edge of the first tree structure.
Wherein the prediction module is configured to perform: determining the position and the connection relation of each decision edge in the second tree structure according to the information of each decision edge provided by each data source end; obtaining a path from a root node to a leaf node in the second tree structure according to the determined position and the connection relation; and determining the obtained leaf node on the path as the joint prediction result.
According to a fifth aspect, there is provided a computing device comprising a memory having stored therein executable code and a processor that, when executing the executable code, implements a method as described in any of the embodiments of the present specification.
According to the joint prediction method and device provided by the embodiment of the specification, the data source end obtains the decision algorithm needed to be used when the attribute judgment is carried out on the private data of the data source end, and carries out the attribute judgment on the private data owned by the data source end locally to obtain the local decision information related to the data source end. Therefore, the plurality of data source ends participating in the joint prediction all send the local decision information to the decision-making party, and the decision-making party can obtain a final joint prediction result according to the received local decision information. Each data source end only needs to send the local decision information obtained by calculation, and does not send private data, so that the leakage of the private data can be avoided, and the safety of joint prediction is improved.
Drawings
In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the description below are some embodiments of the present specification, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a schematic diagram of a system architecture applied in the embodiment of the present specification.
Fig. 2 is a flow diagram of a method for implementing joint prediction in a data source in one embodiment of the present description.
FIG. 3 is a structural diagram of a complete decision tree in one embodiment of the present disclosure.
Fig. 4 is a schematic diagram of a tree structure 1 cut from a decision tree in one embodiment of the present description.
FIG. 5 is a flow diagram of a method for implementing joint prediction in a decision-maker in one embodiment of the present description.
FIG. 6 is a schematic diagram of a tree structure 2 cut from a decision tree in one embodiment of the present description.
Fig. 7 is a schematic structural diagram of a joint prediction device provided in a data source end in one embodiment of the present specification.
Fig. 8 is a schematic structural diagram of a joint prediction apparatus provided in a decision-making party in one embodiment of the present specification.
Detailed Description
The scheme provided by the specification is described below with reference to the accompanying drawings.
First, in this specification, a data source end includes a computing device of a party that owns the data source.
At least two data source ends are involved in the joint prediction service, each data source end has respective private data, for example, the data source end 1 has data of age, weight, sex and the like of a user, the data source end 2 has data of academic calendar, occupation, salary and the like of the user, and the data source end 3 has data of deposit, historical financing record and the like of the user. If the user applies for the loan, the loan risk of the user can be jointly predicted by jointly calculating according to the private data of the 3 data sources, so as to determine whether the loan can be provided for the user.
In the prior art, the private data of 3 data sources are collected to one decision maker, so that the decision maker has the private data of 3 data sources at the same time, and can perform decision calculation according to the private data of 3 data sources and a decision algorithm to obtain a joint prediction result. However, this method may cause leakage of private data, which reduces security.
Analyzing the characteristics of the joint prediction service can know that if the data source end can obtain a relevant decision algorithm for predicting local private data and the decision-making party can obtain a relevant decision algorithm for joint prediction according to the decision information of each data source end, that is, both the data source end and the decision-making party can obtain the relevant decision algorithm required by each joint prediction processing, the data source end does not need to send the private data to the decision-making party, and only needs to send the local decision information to the decision-making party, so that the decision-making party can obtain a final joint prediction result according to each received local decision information. Therefore, the leakage of private data can be avoided, and the safety of joint prediction is improved.
Specific implementations of the above concepts are described below.
To facilitate understanding of the present specification, a system architecture to which the present specification applies will be described first. As shown in fig. 1, the system architecture mainly includes at least two data sources participating in joint prediction and a decision maker.
Each data source end has private data which needs to be used for carrying out joint prediction once. The decision-making party may be one of the data source ends participating in the joint prediction, for example, the decision-making party may be the data source end 1 in fig. 1 or a newly added data source end 4, and of course, the decision-making party may also be an independent device instead of the data source end and only performs the calculation of the final joint prediction.
The parties involved in the joint prediction interact over the network. The network may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.
The number of data sources shown in fig. 1 is merely an example, and there may be other numbers of data sources according to actual service needs.
In this specification, the method of joint prediction relates to the processing of each data source end and the processing of the decision maker, and the following describes the processing of the data source end and the processing of the decision maker respectively by using different embodiments.
First, the processing of the data source end in the joint prediction process will be described.
FIG. 2 shows a flow diagram of a joint prediction method in one embodiment of the present description. The execution subject of the method is a data source end, and the data source end can be any device, equipment, platform or equipment cluster with computing and processing capabilities. For example, taking the data source end 1 as an example, referring to fig. 2, the method includes:
step 201: the data source end 1 obtains a first decision algorithm; the first decision algorithm includes an attribute determination method for private data of the data source end 1.
Step 203: the data source end 1 performs attribute judgment on the local private data according to a first decision algorithm to obtain local decision information.
Step 205: the data source end 1 provides local decision information to the decision maker.
As can be seen from the flow shown in fig. 2, the data source obtains the decision algorithm that needs to be used when performing attribute judgment on the private data of the data source, and performs attribute judgment on the private data owned by the data source locally to obtain local decision information related to the data source. After each data source end participating in joint prediction executes the processing shown in fig. 2, the respective local decision information can be sent to the decision-making party, and the decision-making party can obtain a final joint prediction result according to the received respective local decision information. Each data source end only needs to send the local decision information obtained by calculation, and does not send private data, so that the leakage of the private data can be avoided, and the safety of joint prediction is improved.
Each step in the flow shown in fig. 2 will be described separately below.
Firstly, in step 201, a data source end 1 obtains a first decision algorithm; the first decision algorithm includes an attribute judgment method for the private data of the data source end 1.
The first Decision algorithm may be in various forms, for example, it may be a tree structure, such as a tree structure in the form of gbdt (gradient Boosting Decision tree).
The first decision algorithm may also be an algorithm obtained from a Garbled Circuit (GC).
Because the number of data sources involved in joint prediction is multiple, it is necessary to combine various private data of multiple data sources and various attribute judgment methods to perform comprehensive judgment to obtain a joint prediction result. Therefore, in actual services, a decision tree is usually generated for all data sources involved in joint prediction, as shown in fig. 3, the purpose of the decision tree is to perform attribute judgment according to attribute values of nodes { a, B, C, D, E, F, G } and find a path, an end point of the path is a certain leaf node of the decision tree, and a value of the leaf node is an output of the decision tree, that is, a joint prediction result. In fig. 3, it is assumed that circular nodes, i.e., node C, node D, and node E, are nodes corresponding to the data source end 1, square nodes, i.e., node a, node B, and node F, are nodes corresponding to the data source end 2, and a diamond node, i.e., node G, is node data corresponding to the data source end 3. Each node represents a method for judging the attribute of a piece of private data of the data source end corresponding to the node, for example, the node a represents: it is determined whether the value of a piece of private data a in the data source terminal 2 is greater than 10. Different edges in the decision tree represent different attribute judgment results, for example, the node a is respectively connected to the next two nodes, namely the node B and the node C, through the two edges, the edge with the number of 1 represents that the value of the private data a is greater than 10, and the edge with the number of 2 represents that the value of the private data a is not greater than 10.
The decision tree shown in fig. 3 includes nodes and edges corresponding to all data sources participating in joint prediction. According to the decision tree shown in fig. 3, when the first decision algorithm obtained by the data source end 1 in step 201 is a tree structure, the tree structure may have multiple forms:
the first form: the complete decision tree, i.e. the tree structure 1 obtained by the data source end 1, includes: and the nodes corresponding to all the data sources participating in the joint prediction and at least two edges connected to each node.
When this form one is used, the data source end 1 can be made to obtain a decision tree such as that shown in fig. 3. In this way, the data source end 1 can also obtain information of nodes and edges corresponding to other data source ends, and can determine attribute judgment methods of other data source ends, but because only the attribute judgment methods of other data source ends are obtained, and original private data in other data source ends are not obtained, a security problem is not caused.
In the second form, the part corresponding to the data source end 1, which is intercepted from the decision tree, i.e. the tree structure 1 obtained by the data source end 1 includes only: the nodes corresponding to the data source end 1 and at least two edges connected to each node.
When this form two is adopted, the data source terminal 1 can be made to obtain a tree structure as shown in fig. 4, for example. The data source end 1 does not obtain nodes and edges corresponding to other data source ends, and cannot determine attribute judgment methods of other data source ends, so that the safety is further improved.
Of course, there may be other forms of tree structures 1, for example, the tree structure 1 obtained by the data source end 1 includes only: the part corresponding to the data source end 1 and the part corresponding to the data source end 2 are intercepted from the decision tree.
As can be seen, no matter what form of the tree structure 1 is obtained by the data source end 1, the tree structure 1 is generated by using at least one node corresponding to the data source end 1 and at least two edges connected to each node; wherein each node characterizes: a method for judging the attribute of a piece of private data of the data source end 1; different edges represent different attribute judgment results.
Next, in step 203, according to the first decision algorithm, the attribute of the local private data is determined, so as to obtain local decision information.
When the data source end 1 acquires the tree structure 1 in step 201, the local decision information acquired in this step 203 includes: information of the decision edge selected from the tree structure 1.
In an embodiment of the present specification, a specific implementation process of step 203 includes:
for each node corresponding to the data source end 1, executing:
obtaining an attribute judgment result according to the attribute judgment method represented by the node and local private data; and
selecting a decision edge from at least two edges of the node for connecting the next-level node according to the currently obtained attribute judgment result; the attribute judgment result of the decision edge representation is the same as the currently obtained attribute judgment result;
and obtaining local decision information, namely information of each decision side according to each obtained decision side.
The implementation of step 203 is illustrated by taking the decision tree shown in fig. 3 as an example. Assuming that the values of the private data C, the private data d, and the private data e in the data source end 1 are 10, 3, and 8, respectively, first, for a node C corresponding to the data source end 1, an attribute determination method characterized by the node C is as follows: judging whether the value of the private data C is greater than 0, obtaining that the attribute judgment result is that the value of the private data C is greater than 0 by the data source end 1 according to the attribute judgment method and the value 10 of the private data C, wherein the node C is connected with two edges of a next-level node, the edge with the number of 5 on the left represents that the value of the private data C is greater than 0, and the edge with the number of 6 on the right represents that the value of the private data C is not greater than 0, so that the edge with the number of 5 is selected as a decision edge. Similarly, for node D, the edge numbered 8 on the right, i.e., the edge representing the private data D with a value not greater than 5, is selected as another decision edge, and for node E, the edge numbered 9 on the left, i.e., the edge representing the private data E with a value greater than 5, is selected as yet another decision edge. The data source end 1 obtains local decision information according to the obtained 3 decision edges (edges numbered 5,8 and 9).
In an embodiment of the present specification, when local decision information is obtained according to each obtained decision edge, a branch set is generated by specifically using the number of each decision edge, and the branch set is used as the local decision information. For example, the data source end 1 generates the numbers of decision edges into a branch set {5,8,9 }.
When generating the branch set, the numbers 5,8, and 9 of the decision edges may be sequentially added to the branch set according to the sequence of the decision edges in the tree structure 1 from the root node to the leaf node. In the subsequent process, the decision party obtaining the branch set searches the path according to the sequence from the root node to the leaf node, so that the search efficiency of the decision party can be improved by sequentially adding the numbers 5,8 and 9 of the decision edges into the branch set.
In order to further save network overhead, in an embodiment of the present specification, a specific implementation process of sequentially adding the numbers of the decision edges to the branch set includes: judging whether at least two adjacent nodes corresponding to the data source end exist in the tree structure 1, if so, adding the serial number of the decision edge connected with the last level node in the at least two adjacent nodes into the branch set according to the serial number of each decision edge connected with the at least two adjacent nodes.
Wherein, any two adjacent nodes refer to: the two nodes are connected through one edge in the tree structure. The last level node of the at least two neighboring nodes refers to: and according to the sequence of the tree structure in the direction from the root node to the leaf node, the lowest node in at least two adjacent nodes.
For example, in the decision tree shown in fig. 3, node a and node B are two neighboring nodes corresponding to the data source end 2, but all three nodes C, D, E corresponding to the data source end 1 are not neighboring nodes. Therefore, when the data source end 1 performs step 203 shown in fig. 2, the numbers 5,8, and 9 of the decision edges are all added to the branch set, and when the data source end 2 performs step 203 shown in fig. 2, the number of only one decision edge of the lowest node B, such as the number 3, of two adjacent nodes a and B, may be added to the branch set, and the number of the decision edge of node a, such as the number 1, is no longer added to the branch set.
Next, the data source end 1 provides local decision information to the decision maker in step 205.
As mentioned above, because any one of the multiple data sources participating in the joint prediction may serve as a decision-making party, in step 205, if the decision-making party is a data source, for example, the decision-making party is exactly the data source 1 itself, the data source 1 may provide the local decision-making information to a corresponding processing module inside itself.
In this step 205, the data source end 1 only needs to provide the local decision information, such as the numbers of the decision edges, to the decision party, and does not need to provide the attribute determination result in a complex form, so that the network overhead is reduced, and the implementation process is simplified.
Further, in each joint prediction method provided in the embodiments of the present specification, since the data source end only needs to provide the local decision information to the decision maker, and does not need to perform complex processing such as encryption and decryption on private data, interaction of decision algorithms between different data source ends, and the like, the processing performance is greatly improved.
The following is a description of the processing of the decision-maker in the joint prediction process.
FIG. 5 is a flow diagram of a method for implementing joint prediction in a decision-maker in one embodiment of the present description. The execution subject of the method is a decision-making party, and the decision-making party can be any device, equipment, platform and equipment cluster with computing and processing capabilities. Referring to fig. 5, the method includes:
step 501: the decision party obtains a second decision algorithm; the second decision algorithm comprises decision branches corresponding to all data source ends participating in the joint prediction and incidence relations among the decision branches; different decision branches represent different attribute judgment results.
Step 503: the decision-making party obtains each local decision-making information provided by each data source terminal.
In this step 503, each piece of local decision information is obtained by the data source end by using the method according to any embodiment of this specification and provided to the decision maker.
Step 505: and obtaining a joint prediction result according to the second decision algorithm and the obtained local decision information.
As can be seen from the flow shown in fig. 5, the decision-making party can obtain the final joint prediction result only by obtaining the local decision information from each data source terminal, but not by obtaining the private data in each data source terminal. Because each data source end only needs to send the local decision information obtained by calculation instead of private data, the leakage of the private data can be avoided, and the safety of joint prediction is improved.
Each step in the flow shown in fig. 5 will be described separately below.
Firstly, in step 501, a decision-making party obtains a second decision-making algorithm; the second decision algorithm comprises decision branches corresponding to all data source ends participating in the joint prediction and incidence relations among the decision branches; different decision branches represent different attribute judgment results.
The second decision algorithm includes all the decision branches and their incidence relations, which represent all the possible decision information corresponding to each data source end and the incidence relations among the various decision information.
The second Decision algorithm may also be in various forms, for example, it may be a tree structure, such as gbdt (gradient Boosting Decision tree).
The second decision algorithm may also be an algorithm obtained from a Garbled Circuit (GC).
Because the decision maker needs to comprehensively judge the local decision information of each data source end, when the second decision algorithm is a tree structure and is marked as a tree structure 2, the tree structure 2 needs to include: each tree structure 1 and each leaf node obtained from each data source end participating in joint prediction; each leaf node corresponds to a prediction result;
each decision branch in step 501 is an edge in the tree structure 2;
in step 501, the association relationship among the decision branches is the connection relationship of each edge in the tree structure 2;
a data source provides local decision information comprising: the data source end selects the information of the decision edge from each edge of the tree structure 1.
When the second decision algorithm obtained by the decision-making party in step 501 is the tree structure 2, the tree structure 2 may have multiple forms:
form 1: a complete decision tree comprising: and the nodes corresponding to all the data source ends participating in the joint prediction and at least two edges connected to each node.
When this form 1 is adopted, a decision-making party can be made to obtain a decision tree such as that shown in fig. 3.
When this form 2 is adopted, a decision maker can be made to obtain a tree structure as shown in fig. 6, for example. Because the attribute judgment of the private data is finished in each data source end, the decision-making party can judge the attribute of a certain private data without acquiring node information in a tree structure, and only needs to know the position and the connection relation of each edge and each leaf node, thereby finally finding a path and determining a leaf node.
Of course, there may be other forms of the tree structure 2, for example, a part cut from the decision tree includes only the node and edge corresponding to a part of the data source end and the leaf node, but does not include the node corresponding to another part of the data source end.
Next, in step 503, the decision-making party obtains each local decision information provided by each data source terminal.
When the second decision algorithm obtained by the decision maker is the tree structure 2 in step 501, the local decision information obtained from each data source end is: the data source side selects the information of the decision edge from its local tree structure 1.
For example, the characteristic values of the data of the user 1 are { a:20, b:5, C:10, D:3, E:8, f:12, g:9}, referring to fig. 3, the data source end 1 corresponds to a node C, a node D and a node E, and the owned private data comprises { C, D, E }; the data source end 2 corresponds to a node A, a node B and a node F, and the owned private data comprises { a, B, F }; the data source end 3 corresponds to a node G, and the owned private data comprises { G }. After the 3 data source ends respectively adopt the joint prediction method of the data source end in the present specification, the data source end 1 provides information of the decision edge thereof, for example, a branch set {5,8,9} to the decision party, the data source end 2 provides information of the decision edge thereof, for example, a branch set {1,3,11} to the decision party, and the data source end 3 provides information of the decision edge thereof, for example, a branch set {14}, i.e., the decision party obtains 3 branch sets.
Next, in step 505, the decision-making party obtains a joint prediction result according to a second decision-making algorithm and the obtained local decision-making information.
In an embodiment of this specification, a specific implementation process of this step 505 includes: determining the position and the connection relation of each decision edge in the second tree structure according to the information of each decision edge provided by each data source end; obtaining a path from a root node to a leaf node in the second tree structure according to the determined position and the connection relation; and determining the prediction result corresponding to the obtained leaf node as a joint prediction result.
In step 503, the decision-making party obtains 3 branch sets {5,8,9}, {1,3,11}, and {14 }. Taking the decision tree shown in fig. 3 as an example, the decision party first traverses from the position of the root node, and for the two edges numbered 1 and 2 of the first level, since the decision side includes the number 1 of the decision side in the 3 branch sets, the direction of the side with the number 1 is selected to continue downwards, for the two edges numbered 3 and 4 at the next level, because the number 3 of the decision edge is included in the 3 branch sets obtained by the decision party, the direction of the edge with the number 3 is selected to continue downwards, for the two edges numbered 7 and 8 at the next level, since the number 8 of the decision edge is included in the 3 branch sets obtained by the decision party, the direction of the edge with the number 8 is selected to continue downwards, the leaf node 4 on the path is found, and therefore the prediction result (for example, the audit is passed and the amount is 5 thousand) corresponding to the leaf node 4 is determined as the joint prediction result.
An embodiment of the present specification further provides a joint prediction apparatus, disposed in a data source end, referring to fig. 7, where the apparatus 700 includes:
a first decision algorithm saving module 701 configured to obtain a first decision algorithm; the first decision algorithm comprises an attribute judgment method for the private data of the data source end;
a decision obtaining module 702, configured to perform attribute judgment on local private data according to a first decision algorithm to obtain local decision information;
a decision providing module 703 configured to provide local decision information to the decision party.
In one embodiment of the apparatus of the present specification, the first decision algorithm is a first tree structure; the first tree structure is generated by utilizing at least one node corresponding to the data source end and at least two edges connected to each node; wherein each node characterizes: judging the attribute of a private data of the data source end; different edges represent different attribute judgment results;
the local decision information includes: information of decision edges selected from respective edges of the first tree structure.
In an embodiment of the apparatus of the present specification, the decision obtaining module 702 is configured to, for each node corresponding to the data source end in the first tree structure, perform:
obtaining an attribute judgment result according to the attribute judgment method represented by the node and local private data; and
selecting a decision edge from at least two edges of the node for connecting the next-level node according to the currently obtained attribute judgment result; the attribute judgment result of the decision edge representation is the same as the currently obtained attribute judgment result;
and obtaining the information of each decision side.
In an embodiment of the apparatus in this specification, the decision obtaining module 702 is configured to add the serial numbers of the decision edges to a branch set in sequence according to the sequence of the decision edges in the first tree structure from the root node to the leaf node, and determine the branch set as the obtained information of each decision edge.
In an embodiment of the present specification apparatus, the decision obtaining module 702 is configured to determine whether there are at least two adjacent nodes corresponding to the data source end in the first tree structure, and if so, add, to the branch set, only the number of the decision edge connected to the last-level node in the at least two adjacent nodes, with respect to the number of each decision edge connected to the at least two adjacent nodes.
In one embodiment of the apparatus of the present specification, the first tree structure comprises:
the nodes corresponding to all the data source ends participating in the joint prediction and at least two edges connected to each node;
alternatively, the first and second electrodes may be,
and the nodes corresponding to the data source end and at least two edges connected on each node.
An embodiment of the present disclosure further provides a joint prediction apparatus, disposed on a decision-making party, referring to fig. 8, where the apparatus 800 includes:
a second decision algorithm saving module 801 configured to obtain a second decision algorithm; the second decision algorithm comprises decision branches corresponding to the data source ends participating in the joint prediction and incidence relations among the decision branches; different decision branches represent different attribute judgment results;
a decision summarizing module 802 configured to obtain each local decision information provided by each data source end;
and the prediction module 803 is configured to obtain a joint prediction result according to the second decision algorithm and the obtained local decision information.
In one embodiment of the apparatus of the present specification, the second decision algorithm is a second tree structure; the second tree structure includes: each first tree structure and each leaf node obtained by each data source end; each leaf node corresponds to a prediction result;
each decision branch is an edge in the second tree structure;
the association relation among the decision branches is the connection relation of each edge in the second tree structure;
a data source provides local decision information comprising: the data source end selects the information of the decision edge from each edge of the first tree structure.
In one embodiment of the apparatus of the present specification, the prediction module 803 is configured to perform: determining the position and the connection relation of each decision edge in the second tree structure according to the information of each decision edge provided by each data source end; obtaining a path from a root node to a leaf node in the second tree structure according to the determined position and connection relation; and determining the obtained prediction result corresponding to the leaf node on the path as a joint prediction result.
An embodiment of the present specification provides a computer-readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method of any of the embodiments of the specification.
One embodiment of the present specification provides a computing device comprising a memory and a processor, the memory having stored therein executable code, the processor implementing a method in accordance with any one of the embodiments of the specification when executing the executable code.
It is understood that the illustrated structure of the embodiments of the present disclosure does not constitute a specific limitation on the warehouse cargo measuring device. In other embodiments of the specification, the bin load measuring device may include more or fewer components than shown, or some components may be combined, some components may be separated, or a different arrangement of components. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.
For the information interaction, execution process and other contents between the modules in the above-mentioned apparatus and system, because the same concept is based on the embodiment of the method in this specification, specific contents may refer to the description in the embodiment of the method in this specification, and are not described herein again.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, as for the apparatus embodiment, since it is substantially similar to the method embodiment, the description is relatively simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
Those skilled in the art will recognize that, in one or more of the examples described above, the functions described in this disclosure may be implemented in hardware, software, hardware, or any combination thereof. When implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium.
The above-mentioned embodiments, objects, technical solutions and advantages of the present invention are further described in detail, it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made on the basis of the technical solutions of the present invention should be included in the scope of the present invention.
Claims (15)
1. The joint prediction method is applied to a data source end and comprises the following steps:
obtaining a first decision algorithm; the first decision algorithm comprises an attribute judgment method for the private data of the data source end;
according to a first decision algorithm, carrying out attribute judgment on local private data to obtain local decision information;
providing the local decision information to the decision-making party;
wherein the first decision algorithm is a first tree structure;
the first tree structure is generated by utilizing at least one node corresponding to the data source end and at least two edges connected to each node; wherein each node characterizes: judging the attribute of a private data of the data source end; different edges represent different attribute judgment results;
the local decision information includes: information of decision edges selected from each edge of the first tree structure;
and the nodes and edges in the first tree structure owned by the data source end are determined according to the private data owned by the data source end.
2. The method of claim 1, wherein the deriving local decision information comprises:
for each node corresponding to the data source end in the first tree structure, executing:
obtaining an attribute judgment result according to the attribute judgment method represented by the node and local private data; and
selecting a decision edge from at least two edges of the node for connecting the next-level node according to the currently obtained attribute judgment result; the attribute judgment result of the decision edge representation is the same as the currently obtained attribute judgment result;
and obtaining the information of each decision side.
3. The method of claim 2, wherein the obtaining information of each decision edge comprises:
sequentially adding the serial numbers of the decision edges into a branch set according to the sequence of the decision edges in the first tree structure from the root node to the leaf node;
the set of branches is determined as information for each resulting decision edge.
4. The method of claim 3, wherein said adding the number of each decision edge to the set of branches in turn comprises:
and judging whether at least two adjacent nodes corresponding to the data source end exist in the first tree structure, if so, adding the serial number of the decision edge connected with the last level node in the at least two adjacent nodes into the branch set according to the serial number of each decision edge connected with the at least two adjacent nodes.
5. The method of any of claims 1-4, wherein the first tree structure comprises:
the nodes corresponding to all the data source ends participating in the joint prediction and at least two edges connected to each node;
alternatively, the first and second electrodes may be,
and the nodes corresponding to the data source end and at least two edges connected on each node.
6. The joint prediction method is applied to a decision-making party and comprises the following steps:
obtaining a second decision algorithm; the second decision algorithm comprises decision branches corresponding to the data source ends participating in the joint prediction and incidence relations among the decision branches; different decision branches represent different attribute judgment results;
acquiring local decision information provided by each data source end; local decision information provided by each data source is obtained by the method of any one of claims 1 to 5;
obtaining a joint prediction result according to a second decision algorithm and the obtained local decision information;
wherein the second decision algorithm is a second tree structure;
the second tree structure includes: each first tree structure and each leaf node obtained by each data source end; each leaf node corresponds to a prediction result; the node and the edge in the first tree structure owned by each data source end are determined according to the private data owned by the data source end;
each decision branch is an edge in the second tree structure;
the association relationship between the decision branches is the connection relationship of the edges in the second tree structure.
7. The method of claim 6, wherein said deriving a joint prediction result comprises:
determining the position and the connection relation of each decision edge in the second tree structure according to the information of each decision edge provided by each data source end;
obtaining a path from a root node to a leaf node in the second tree structure according to the determined position and connection relation;
and determining the prediction result corresponding to the obtained leaf node as the joint prediction result.
8. The joint prediction device is arranged in the data source end and comprises:
a first decision algorithm saving module configured to obtain a first decision algorithm; the first decision algorithm comprises an attribute judgment method for private data of the data source end;
the decision acquisition module is configured to perform attribute judgment on local private data according to a first decision algorithm to obtain local decision information;
a decision providing module configured to provide local decision information to a decision-making party;
wherein, the first and the second end of the pipe are connected with each other,
the first decision algorithm is a first tree structure; the first tree structure is generated by utilizing at least one node corresponding to the data source end and at least two edges connected to each node; wherein each node characterizes: judging the attribute of a private data of the data source end; different edges represent different attribute judgment results; the nodes and edges in the first tree structure owned by the data source end are determined according to the private data owned by the data source end;
the local decision information includes: information of decision edges selected from among the respective edges of the first tree structure.
9. The apparatus of claim 8, wherein the decision obtaining module is configured to perform, for each node corresponding to the data source end in the first tree structure:
obtaining an attribute judgment result according to the attribute judgment method represented by the node and local private data; and
selecting a decision edge from at least two edges of the node for connecting the next-level node according to the currently obtained attribute judgment result; the attribute judgment result of the decision edge representation is the same as the currently obtained attribute judgment result;
and obtaining the information of each decision side.
10. The apparatus according to claim 9, wherein the decision obtaining module is configured to add the number of each decision edge to a branch set in sequence according to the precedence order of each decision edge in the first tree structure from the root node to the leaf node, and determine the branch set as the obtained information of each decision edge.
11. The apparatus according to claim 10, wherein the decision obtaining module is configured to determine whether there are at least two neighboring nodes corresponding to the data source end in the first tree structure, and if so, add, to the branch set, only the number of the decision edge connected to the last node of the at least two neighboring nodes for the number of each decision edge connected to the at least two neighboring nodes.
12. The apparatus of any of claims 8-11, wherein the first tree structure comprises:
the nodes corresponding to all the data source ends participating in the joint prediction and at least two edges connected to each node;
alternatively, the first and second electrodes may be,
and the nodes corresponding to the data source end and at least two edges connected on each node.
13. The joint prediction device is arranged on a decision-making party and comprises:
a second decision algorithm saving module configured to obtain a second decision algorithm; the second decision algorithm comprises decision branches corresponding to all data source ends participating in the joint prediction and incidence relations among the decision branches; different decision branches represent different attribute judgment results;
the decision summarizing module is configured to acquire each piece of local decision information provided by each data source end; wherein each local decision information is obtained and transmitted by the joint prediction device provided at the data source end in any one of claims 8 to 12;
the prediction module is configured to obtain a joint prediction result according to the second decision algorithm and the acquired local decision information;
wherein the second decision algorithm is a second tree structure; the second tree structure includes: each first tree structure and each leaf node obtained by each data source end; each leaf node corresponds to a prediction result; the node and the edge in the first tree structure owned by each data source end are determined according to the private data owned by the data source end;
each decision branch is an edge in the second tree structure;
the association relationship between the decision branches is the connection relationship of the edges in the second tree structure.
14. The apparatus of claim 13, wherein the prediction module is configured to perform: determining the position and the connection relation of each decision edge in the second tree structure according to the information of each decision edge provided by each data source end; obtaining a path from a root node to a leaf node in the second tree structure according to the determined position and connection relation; and determining the obtained prediction result corresponding to the leaf node as the combined prediction result.
15. A computing device comprising a memory having executable code stored therein and a processor that, when executing the executable code, implements the method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110447583.2A CN113177212B (en) | 2021-04-25 | 2021-04-25 | Joint prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110447583.2A CN113177212B (en) | 2021-04-25 | 2021-04-25 | Joint prediction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113177212A CN113177212A (en) | 2021-07-27 |
| CN113177212B true CN113177212B (en) | 2022-07-19 |
Family
ID=76925510
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110447583.2A Active CN113177212B (en) | 2021-04-25 | 2021-04-25 | Joint prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113177212B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104111920A (en) * | 2013-04-16 | 2014-10-22 | 华为技术有限公司 | Decision-making tree based prediction method and device |
| CN110222527A (en) * | 2019-05-22 | 2019-09-10 | 暨南大学 | A kind of method for secret protection |
| CN110427969A (en) * | 2019-07-01 | 2019-11-08 | 阿里巴巴集团控股有限公司 | Data processing method, device and electronic equipment |
| EP3570575A1 (en) * | 2018-05-16 | 2019-11-20 | INESC TEC - Instituto de Engenharia de Sistemas e Computadores, Tecnologia e Ciência | Internet of things security with multi-party computation (mpc) |
| CN110874481A (en) * | 2018-08-31 | 2020-03-10 | 阿里巴巴集团控股有限公司 | GBDT model-based prediction method and device |
| CN110944011A (en) * | 2019-12-16 | 2020-03-31 | 支付宝(杭州)信息技术有限公司 | Joint prediction method and system based on tree model |
| CN111242164A (en) * | 2019-12-27 | 2020-06-05 | 天津幸福生命科技有限公司 | Decision result determination method, device and equipment |
| CN111639367A (en) * | 2020-07-31 | 2020-09-08 | 支付宝(杭州)信息技术有限公司 | Tree model-based two-party combined classification method, device, equipment and medium |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110457912B (en) * | 2019-07-01 | 2020-08-14 | 阿里巴巴集团控股有限公司 | Data processing method and device and electronic equipment |
| CN111783810B (en) * | 2019-09-24 | 2023-12-08 | 北京沃东天骏信息技术有限公司 | Method and device for determining attribute information of user |
-
2021
- 2021-04-25 CN CN202110447583.2A patent/CN113177212B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104111920A (en) * | 2013-04-16 | 2014-10-22 | 华为技术有限公司 | Decision-making tree based prediction method and device |
| EP3570575A1 (en) * | 2018-05-16 | 2019-11-20 | INESC TEC - Instituto de Engenharia de Sistemas e Computadores, Tecnologia e Ciência | Internet of things security with multi-party computation (mpc) |
| CN110874481A (en) * | 2018-08-31 | 2020-03-10 | 阿里巴巴集团控股有限公司 | GBDT model-based prediction method and device |
| CN110222527A (en) * | 2019-05-22 | 2019-09-10 | 暨南大学 | A kind of method for secret protection |
| CN110427969A (en) * | 2019-07-01 | 2019-11-08 | 阿里巴巴集团控股有限公司 | Data processing method, device and electronic equipment |
| CN110944011A (en) * | 2019-12-16 | 2020-03-31 | 支付宝(杭州)信息技术有限公司 | Joint prediction method and system based on tree model |
| CN111242164A (en) * | 2019-12-27 | 2020-06-05 | 天津幸福生命科技有限公司 | Decision result determination method, device and equipment |
| CN111639367A (en) * | 2020-07-31 | 2020-09-08 | 支付宝(杭州)信息技术有限公司 | Tree model-based two-party combined classification method, device, equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113177212A (en) | 2021-07-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8694979B2 (en) | Efficient egonet computation in a weighted directed graph | |
| US8024281B2 (en) | Alpha node hashing in a rule engine | |
| CN109951547B (en) | Transaction request parallel processing method, device, equipment and medium | |
| CN109508326B (en) | Method, device and system for processing data | |
| CN113037877B (en) | Optimization method for time-space data and resource scheduling under cloud edge architecture | |
| US8959229B1 (en) | Intelligently provisioning cloud information services | |
| CN113157981B (en) | Graph network relation diffusion method and device | |
| CN111083013A (en) | Test method and device based on flow playback, electronic equipment and storage medium | |
| Baldacci et al. | Pricing strategies for capacitated ring-star problems based on dynamic programming algorithms | |
| Wang et al. | Optimal tree topology for a submarine cable network with constrained internodal latency | |
| CN108470251B (en) | Community division quality evaluation method and system based on average mutual information | |
| CN110719200A (en) | Information identification method and device | |
| CN113255950B (en) | Method and device for optimizing logistics network | |
| CN113177212B (en) | Joint prediction method and device | |
| CN112966054A (en) | Enterprise graph node relation-based ethnic group division method and computer equipment | |
| CN111951112A (en) | Intelligent contract execution method based on block chain, terminal equipment and storage medium | |
| CN113850675A (en) | Information processing method and device for enterprise transaction relation data | |
| CN117076563A (en) | Pruning method and device applied to blockchain | |
| CN104598661A (en) | Method and device for change request analysis of project related to IT system | |
| CN114492844A (en) | Method and device for constructing machine learning workflow, electronic equipment and storage medium | |
| CN113010745A (en) | Cash management network map construction method and device, electronic equipment and medium | |
| US20230017693A1 (en) | Topology design apparatus, topology design method and program | |
| CN112560213A (en) | System modeling method and system based on model system engineering and hyper-network theory | |
| CN116402432B (en) | Route planning method, device, equipment and storage medium | |
| CN113204535B (en) | Routing method and device, electronic equipment 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 |