CN116341660B

CN116341660B - Information optimization method and server applied to artificial intelligence

Info

Publication number: CN116341660B
Application number: CN202310623878.XA
Authority: CN
Inventors: 蒋天宏; 田凯; 孙凤英
Original assignee: Octopus Artificial Intelligence Technology Changshu Co ltd
Current assignee: Octopus Artificial Intelligence Technology Changshu Co ltd
Priority date: 2023-05-30
Filing date: 2023-05-30
Publication date: 2023-08-01
Anticipated expiration: 2043-05-30
Also published as: CN116341660A

Abstract

According to the information optimization method and the server applied to artificial intelligence, after the user operation behavior information is obtained when the user portrait knowledge set generation is carried out on the cloud service interaction event, the user operation behavior information is generally a behavior preference knowledge relationship network; as the concentration of the active knowledge members is lower for the behavior preference knowledge relationship network, higher precision and timeliness can be ensured when subsequent demand analysis and mining are carried out. Based on the method, the relative relation variable in the behavior preference knowledge relation network is utilized by the active knowledge members in the user operation behavior information, the behavior preference knowledge relation network is optimized, the concentration degree of the active knowledge members in the obtained optimized relation network can be adaptively reduced on the basis of maintaining the original detail characteristics of the user operation behavior information, so that the interference on the optimized relation network during the requirement analysis mining is reduced, and the timeliness and the precision of the feature analysis of the optimized relation network are improved.

Description

Information optimization method and server applied to artificial intelligence

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to an information optimization method and a server applied to artificial intelligence.

Background

From the application situation of the present artificial intelligence products, more intelligent agents will gradually enter the digital industry field in the future, and artificial intelligence will be new kinetic energy developed in the industry field. Under the great tide of the current industry internet development, the landing application of artificial intelligence can be further accelerated. Currently, one of the mainstream applications of artificial intelligence is feature processing, such as knowledge analysis in conjunction with expert systems for user portrayal or behavioral preference processing. However, in practical application, how to reduce the interference in the analysis process and improve the analysis accuracy is a difficult problem to overcome.

Disclosure of Invention

The invention at least provides an information optimization method and a server for artificial intelligence.

The invention provides an information optimization method applied to artificial intelligence, which is applied to an information optimization server, and comprises the following steps: obtaining user operation behavior information corresponding to a cloud service interaction event when a user portrait knowledge set is generated on the cloud service interaction event; the user operation behavior information forms a behavior preference knowledge relation network, and the behavior preference knowledge relation network comprises a plurality of behavior preference knowledge members; combining member contact of each active knowledge member in the behavior preference knowledge relationship network in a plurality of behavior preference knowledge members to determine relative relationship variables of the active knowledge members in the behavior preference knowledge relationship network; and carrying out relational network optimization on the behavior preference knowledge relational network by combining the relative relational variables to obtain an optimized relational network corresponding to the user operation behavior information.

After obtaining the user operation behavior information when the user portrait knowledge set is generated for the cloud service interaction event, the user operation behavior information is generally a behavior preference knowledge relationship network; as the concentration of the active knowledge members is lower for the behavior preference knowledge relationship network, higher precision and timeliness can be ensured when subsequent demand analysis and mining are carried out. Based on the method, the relative relation variable in the behavior preference knowledge relation network is utilized by the active knowledge members in the user operation behavior information, the behavior preference knowledge relation network is optimized, the concentration degree of the active knowledge members in the obtained optimized relation network can be adaptively reduced on the basis of maintaining the original detail characteristics of the user operation behavior information, so that the interference on the optimized relation network during the requirement analysis mining is reduced, and the timeliness and the precision of the feature analysis of the optimized relation network are improved.

In some possible embodiments, further comprising: and determining mapping indicating data between the user operation behavior information and the acquisition offset by combining the optimized relation network and the acquisition offset corresponding to the user operation behavior information.

Therefore, in view of the light weight characteristic of the optimized relation network obtained by the information optimization method applied to the artificial intelligence, the subsequent further feature knowledge analysis is facilitated, and the mapping indication data between the user operation behavior information obtained by using the optimized relation network and the acquisition offset is as accurate as possible based on the light weight characteristic, so that the precision of generating the user image knowledge set for the cloud service interaction event can be improved.

In some possible embodiments, the determining the relative relationship variable of each active knowledge member in the behavior preference knowledge relationship network in combination with the member contact of the active knowledge member in the behavior preference knowledge relationship network includes: disassembling the behavior preference knowledge relation network into a plurality of local knowledge relation chains; each local knowledge relation chain comprises a plurality of behavior preference knowledge members positioned in the local knowledge relation chain; for each local knowledge relation chain in a plurality of local knowledge relation chains, if active knowledge members exist in behavior preference knowledge members in the local knowledge relation chain, determining a relation variable index corresponding to the local knowledge relation chain as a first description value; if the active knowledge members do not exist in the behavior preference knowledge members in the local knowledge relation chain, determining a relation variable index corresponding to the local knowledge relation chain as a second description value; and combining the relation variable indexes respectively corresponding to the local knowledge relation chains to form the relative relation variable.

In some possible embodiments, the disassembling the behavior preference knowledge relationship network into a number of local knowledge relationship chains includes: based on a set dismantling rule, dismantling the behavior preference knowledge relation network into a plurality of local knowledge relation chains; or, determining the dismantling coverage corresponding to the behavior preference knowledge relationship network by combining the window coverage of the behavior preference knowledge relationship network; the window coverage of the behavior preference knowledge relationship network is a set multiple of the dismantling coverage; and disassembling the behavior preference knowledge relationship network into a plurality of local knowledge relationship chains by combining the determined disassembling coverage.

Therefore, the thinking of disassembling the behavior preference knowledge relationship network by adopting the set disassembling rule can adapt to the calculation power of the information optimization server, so that the behavior preference knowledge relationship network can be reasonably disassembled, and the calculation cost of the information optimization server is reduced. The window coverage of the behavior preference knowledge relationship network is used as a guide to determine the dismantling coverage of the local knowledge relationship chain obtained by dismantling, and the concept of dismantling is utilized to ensure the accuracy of the local knowledge relationship chain during dismantling, namely, repeated behavior preference knowledge members of two neighbor local knowledge relationship chains are not existed, and the accuracy and reliability of the optimization of the subsequent relationship network are ensured.

In some possible embodiments, the optimizing the relationship network by combining the relative relationship variables to the behavior preference knowledge relationship network to obtain an optimized relationship network corresponding to the user operation behavior information includes: combining a first region constraint and a second region constraint of each local knowledge relation chain in the behavior preference knowledge relation network to perform region constraint transformation to obtain a plurality of knowledge cores; the x-th knowledge core corresponds to an x-th group of transverse local knowledge relation chains and an x-th group of longitudinal local knowledge relation chains; x is an integer greater than 0 and not greater than P; p reflects the number of the local knowledge relation chains in the first constraint dimension or the second constraint dimension of the behavior preference knowledge relation network; combining the relation variable indexes of the local knowledge relation chains corresponding to the knowledge cores to generate a linkage knowledge relation network reflecting the upstream and downstream relation among a plurality of knowledge cores; the characteristic value of each linkage member in the linkage knowledge relation network reflects whether upstream and downstream connection exists between two knowledge cores corresponding to the linkage member; and carrying out relationship network optimization on the behavior preference knowledge relationship network by combining the linkage knowledge relationship network to obtain the optimized relationship network.

It can be seen that the relationships between the local knowledge relationship chains in the behavior preference knowledge relationship network can be expressed by linking the knowledge relationship network. On the basis of ensuring the relationship among the local knowledge relationship chains, the relationship network optimization is carried out on the behavior preference knowledge relationship network through the linkage knowledge relationship network, so that an optimized relationship network with higher feature recognition degree and better light weight degree is obtained.

In some possible embodiments, the generating, by combining the relationship variable indexes of the local knowledge relationship chains corresponding to the knowledge cores, a linkage knowledge relationship network reflecting the upstream and downstream links between the knowledge cores includes: for each linkage member in the linkage knowledge relationship network, determining a target knowledge core corresponding to the linkage member; combining the first region constraint and the second region constraint reflected by the target knowledge core to determine a target knowledge relation chain corresponding to the target knowledge core; combining the relation variable indexes corresponding to the target knowledge relation chain to determine the characteristic value corresponding to the linkage member; combining the characteristic values respectively corresponding to a plurality of linkage members to generate the linkage knowledge relationship network; if the relation variable index corresponding to the target knowledge relation chain comprises a first description value, determining that upstream and downstream relation exists between the two target knowledge cores corresponding to the linkage member; and if the relation variable index corresponding to the target knowledge relation chain comprises a second description value, determining that no upstream and downstream relation exists between the two target knowledge cores corresponding to the linkage member.

Therefore, through the record of the linkage knowledge relation network, other information is not required to be acquired in a targeted way, and the number of other knowledge cores with immediate transmission states for any knowledge core can be directly determined according to the linkage knowledge relation network. Thus, the linkage knowledge relationship network can be flexibly called.

In some possible embodiments, the performing, in combination with the linkage knowledge relationship network, optimization of the relationship network for the behavior preference knowledge relationship network to obtain the optimized relationship network includes: combining the linkage knowledge relation network to determine transmission element characteristics corresponding to a plurality of knowledge cores respectively; the feature of the transmission element corresponding to any knowledge core reflects the number of other knowledge cores with immediate transmission states of the knowledge core; and carrying out relationship network optimization on the behavior preference knowledge relationship network by combining a plurality of transmission element characteristics respectively corresponding to the knowledge cores to obtain the optimized relationship network.

Therefore, the difficulty in adjusting the characteristics of the transmission elements can be reduced, and the timeliness of the whole scheme is ensured.

In some possible embodiments, combining the transmission element features respectively corresponding to the plurality of knowledge cores, performing a relationship network optimization on the behavior preference knowledge relationship network to obtain the optimized relationship network, where the optimizing includes: combining the transmission element characteristics respectively corresponding to the knowledge cores to determine the adjustment rules of the optimized preference knowledge chains respectively corresponding to the knowledge cores; the optimized preference knowledge chain corresponding to any knowledge core comprises the following steps: a sub-relationship network of the behavioral preference knowledge relationship network determined by the first region constraint or the second region constraint of the knowledge core; and carrying out relational network optimization on optimized preference knowledge chains respectively corresponding to the knowledge cores by combining the adjustment rules to obtain the optimized relational network.

In some possible embodiments, the determining, in combination with the transfer element features respectively corresponding to the plurality of knowledge cores, an adjustment rule for the optimized preference knowledge chain respectively corresponding to the plurality of knowledge cores includes: in each of several optimization stages, the following scheme is implemented: from the first knowledge cores without determining the adjustment rule, determining the current first knowledge core by combining the transmission element characteristics corresponding to each first knowledge core; combining the number of the second knowledge cores with the determined adjustment rules, and determining the adjustment rules of the optimized preference knowledge chains corresponding to the current first knowledge core; if the first knowledge core exists currently, modifying the transmission element characteristics of the first knowledge core which exists currently, jumping to the first knowledge core which does not determine the adjustment rule, and combining the transmission element characteristics respectively corresponding to the first knowledge cores to determine the current first knowledge core; and if the first knowledge core does not exist currently, ending the optimization stage to obtain adjustment rules of the knowledge cores corresponding to the optimized preference knowledge chains respectively.

It can be seen that the adjustment rules of the optimized preferred knowledge chain for each knowledge core can be determined in sequence in each optimization stage. Because the updated characteristics of the transfer elements are as concise as possible after an optimization stage, the adjustment rules of the optimized preference knowledge chains corresponding to the knowledge cores are determined sequentially in a plurality of stages based on the updated characteristics, the adjustment time consumption can be reduced, and the timeliness of the whole scheme is ensured.

In some possible embodiments, the determining the current first knowledge core according to the transmission element features corresponding to the first knowledge cores respectively includes: combining the transmission element characteristics respectively corresponding to the first knowledge cores, and performing optimization sequence arrangement on the first knowledge cores to obtain optimization sequence arrangement results respectively corresponding to the first knowledge cores; wherein the transmission element characteristic indicates that the smaller the number of other knowledge cores with immediate transmission states corresponding to the first knowledge core is, the earlier the corresponding optimization sequence is; and combining the optimization sequence arrangement results respectively corresponding to the first knowledge cores, and determining the current first knowledge core from the first knowledge cores.

In some possible embodiments, the optimizing the relationship network for the optimized preference knowledge chain corresponding to each of the plurality of knowledge cores by combining the adjustment rule to obtain the optimized relationship network includes: combining the adjustment rules to determine member connections of optimized preference knowledge chains corresponding to the knowledge cores in the optimized relationship network; and combining the member connection and optimized preference knowledge chains respectively corresponding to the knowledge cores to form the optimized relation network.

The invention also provides an information optimization server, which comprises a processor and a memory; the processor is in communication with the memory, and the processor is configured to read and execute a computer program from the memory to implement the method described above.

The present invention also provides a computer readable storage medium having stored thereon a computer program which, when run, implements the method described above.

For a description of the effects of the above information optimization server, computer-readable storage medium, see the description of the above method.

In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are necessary for the embodiments to be used are briefly described below, the drawings being incorporated in and forming a part of the description, these drawings showing embodiments according to the present invention and together with the description serve to illustrate the technical solutions of the present invention. It is to be understood that the following drawings illustrate only certain embodiments of the invention and are therefore not to be considered limiting of its scope, for the person of ordinary skill in the art may admit to other equally relevant drawings without inventive effort.

Fig. 1 is a block diagram of an information optimization server according to an embodiment of the present invention.

FIG. 2 is a flow chart of an information optimization method applied to artificial intelligence according to an embodiment of the present invention.

Detailed Description

Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the invention. Rather, they are merely examples of apparatus and methods consistent with aspects of the invention as detailed in the accompanying claims.

Fig. 1 is a schematic diagram of an information optimization server 10 according to an embodiment of the present invention, including a processor 102, a memory 104, and a bus 106. The memory 104 is used for storing execution instructions, including a memory and an external memory, where the memory may also be understood as an internal memory, and is used for temporarily storing operation data in the processor 102 and data exchanged with the external memory such as a hard disk, where the processor 102 exchanges data with the external memory through the memory, and when the information optimization server 10 operates, the processor 102 and the memory 104 communicate with each other through the bus 106, so that the processor 102 executes the information optimization method applied to artificial intelligence according to the embodiment of the present invention.

Referring to fig. 2, fig. 2 is a schematic flow chart of an information optimization method applied to artificial intelligence, which is applied to an information optimization server and may include the descriptions of steps 101-103 according to an embodiment of the present invention.

Step 101: and obtaining user operation behavior information corresponding to the cloud service interaction event when the user portrait knowledge set is generated on the cloud service interaction event.

The user operation behavior information forms a behavior preference knowledge relation network, and the behavior preference knowledge relation network comprises a plurality of behavior preference knowledge members.

Step 102: and combining member contact of each active knowledge member in the behavior preference knowledge relationship network in the plurality of behavior preference knowledge members to determine relative relationship variables of the active knowledge members in the behavior preference knowledge relationship network.

Step 103: and carrying out relational network optimization on the behavior preference knowledge relational network by combining the relative relational variables to obtain an optimized relational network corresponding to the user operation behavior information.

The information optimization method applied to artificial intelligence provided by the invention is applied to steps 101-103, after user operation behavior information when user portrait knowledge set generation is performed on cloud service interaction events is obtained, a behavior preference knowledge relationship network to be processed can be determined by utilizing the user operation behavior information, behavior preference knowledge members in the behavior preference knowledge relationship network can reflect space positioning variables (space coordinate values) of user behavior habit items corresponding to the behavior preference knowledge members in an AI feature space, and based on the space positioning variables (space coordinate values), the corresponding relationship between the user operation behavior information and operation behavior items can be reflected. When the feature analysis is carried out on the behavior preference knowledge relationship network, the relation of each active knowledge member in the behavior preference knowledge relationship network is considered to determine the relative relationship variable of the active knowledge member in the behavior preference knowledge relationship network. For the behavior preference knowledge relation network, the lower the concentration of the active knowledge members is, the higher the precision and timeliness can be ensured when the subsequent demand analysis and mining are carried out. Based on the method, the relational network optimization is carried out on the behavior preference knowledge relational network by utilizing the relative relational variables, so that the feature recognition degree and the light weight degree of the optimized relational network can be improved on the basis of maintaining the corresponding relation of the original detail information, and the timeliness of the whole scheme is ensured.

For the step 101, the information optimization method applied to artificial intelligence provided in the embodiment of the present invention may be applied to an application scenario of a digital service, for example. For example, when determining a cloud service interaction event for generating a user portrait knowledge set, such as an interaction event of electronic commerce, an intelligent internet of things, a remote government enterprise service, big data security, etc., the big data interaction can be obtained by detecting the cloud service interaction event. The interaction behavior vector can be obtained in the interaction big data, and the interaction behavior vector can comprise a space positioning variable in the AI feature space. When determining the AI feature space, feature space construction can be performed according to actual situations, such as feature projection processing by artificial intelligence technology, so as to obtain a relatively perfect feature space (i.e. a feature coordinate system) capable of performing feature processing analysis.

In some examples, through transformation of the relational network of the interaction behavior vectors, user operation behavior information corresponding to the interaction behavior vectors, that is, user operation behavior information corresponding to the cloud service interaction event described in the embodiment of the present invention, can be obtained. The user operation behavior information can form a behavior preference knowledge relation network, the behavior preference knowledge relation network comprises a plurality of behavior preference knowledge members, and each behavior preference knowledge member reflects a space positioning variable of an interaction behavior vector corresponding to the behavior preference knowledge member in an AI feature space. Further, the behavior preference knowledge relation network can be understood as a feature map distribution or a feature map matrix formed by the behavior features of the user operation, and can be realized by utilizing an expert system branch in the artificial intelligence technology, so that the behavior preference knowledge relation network can be ensured to be matched with the behavior preference vector under the actual experience level as much as possible.

In the behavioral preference knowledge relationship network, a plurality of behavioral preference knowledge members can be included, and after the behavioral preference knowledge members are transformed (feature graph processing) on the relationship network of the interaction behavior vector, the behavioral preference knowledge members can include inactive knowledge members (knowledge points with low feature heat) and active knowledge members (knowledge points with high feature heat). The window coverage of the behavior-preference knowledge relationship network (which may be understood as the size or scale of the behavior-preference knowledge relationship network) is determined by the number of interaction behavior vectors, i.e. the number of behavior-preference knowledge members in the behavior-preference knowledge relationship network is determined by the number of interaction behavior vectors. In some examples, if the number of interaction behavior vectors in a process is greater, the window coverage of the behavior-preference knowledge relationship network is greater; if the number of interactive behavior vectors in one process is small, the window coverage of the behavior preference knowledge relationship network is small.

For the step 102, after the behavior preference knowledge relationship network is formed by using the user operation behavior information, in order to improve timeliness, the behavior preference knowledge relationship network can be optimized (such as relationship network reconstruction) by determining the relative relationship variables (such as the position features in the behavior preference knowledge relationship network) of the active knowledge members in the behavior preference knowledge relationship network, so as to obtain an optimized relationship network that is convenient for feature mining analysis.

When the relative relation variable of the active knowledge members in the behavior preference knowledge relation network is obtained, the relative relation variable can be determined by combining member relation of each active knowledge member in the behavior preference knowledge relation network in a plurality of behavior preference knowledge members, and the member relation can be understood as the relative position relation of each active knowledge member in the behavior preference knowledge relation network.

In some examples, the following section describes schemes for determining relative relationship variables.

Step 201: disassembling the behavior preference knowledge relation network into a plurality of local knowledge relation chains; each local knowledge relationship chain contains a plurality of behavior preference knowledge members within the local knowledge relationship chain.

For this step, the behavioral preference knowledge relationship network can be disassembled in, but is not limited to, the following two ways.

In the first way, the behavior preference knowledge relationship network is disassembled into a plurality of local knowledge relationship chains based on a set disassembly rule.

In some examples, if the computational power of the information optimization server is not high, the information optimization server can only allow the behavior preference knowledge relationship network to be disassembled into a limited number of local knowledge relationship chains for processing, or call performance of behavior preference knowledge members in the behavior preference knowledge relationship network to be disassembled is poor, then the set disassembly rule can be determined according to the actual requirement of the information optimization server.

For example, when the window coverage of the behavior preference knowledge relationship network is a1×a1, if the information optimization server processes the a2×a2 local knowledge relationship chains only in the following steps, the set disassembly rule may be determined as a3×a3 in advance, so that the behavior preference knowledge relationship network is disassembled into the a2×a2 local knowledge relationship chains by using the set disassembly rule.

Therefore, the computing power of the information optimization server is considered, so that the behavior preference knowledge relationship network can be reasonably disassembled, and the computing cost of the information optimization server is reduced.

In a second mode, determining a dismantling coverage corresponding to the behavior preference knowledge relationship network by combining the window coverage of the behavior preference knowledge relationship network; the window coverage of the behavior preference knowledge relationship network is a set multiple of the dismantling coverage; and disassembling the behavior preference knowledge relationship network into a plurality of local knowledge relationship chains by combining the determined disassembling coverage.

For example, the behavior preference knowledge relationship network is a feature graph with consistent length and width, when the behavior preference knowledge relationship network is disassembled into a plurality of local knowledge relationship chains, the local knowledge relationship chains are not overlapped, and the disassembled plurality of local knowledge relationship chains can be distributed on the behavior preference knowledge relationship network in order. Based on the above, the dismantling coverage can be determined according to the window coverage of the behavior preference knowledge relation network, so that the dismantled local knowledge relation chain meets the requirements.

For example, when the window coverage of the behavior preference knowledge relationship network is a1×a1, it can be determined that the resolution coverage is a2×a2, a3×a3, a4×a4, or the like. Because the window coverage of the behavior preference knowledge relation network is a set multiple of the dismantling coverage, a plurality of local knowledge relation chains obtained after the behavior preference knowledge relation network is dismantled based on the window coverage of the behavior preference knowledge relation network can be orderly distributed on the behavior preference knowledge relation network. For example, on the basis of determining that the disassembling coverage is a2×a2, a3×a3 local knowledge relation chains can be obtained from the behavior preference knowledge relation network, and the local knowledge relation chains are not intersected with each other but have upstream and downstream links.

Based on the method, the window coverage of the behavior preference knowledge relation network is used as a guide to determine the dismantling coverage of the local knowledge relation chain obtained by dismantling, and the idea of dismantling is utilized to ensure the accuracy of the local knowledge relation chain during dismantling, namely, repeated behavior preference knowledge members of two neighbor local knowledge relation chains are not existed, so that the accuracy and reliability of the optimization of the subsequent relation network are ensured.

In addition, in the subsequent process of optimizing the relationship network of the behavior preference knowledge relationship network, the characteristics of whether active knowledge members exist in the local knowledge relationship chains or not are specifically related to the local knowledge relationship chains, the number of the local knowledge relationship chains can influence the number of optimization stages in the relationship network optimization process, and based on the characteristics, consideration can be performed in advance when the behavior preference knowledge relationship network is disassembled to obtain a plurality of local knowledge relationship chains.

For example, on the basis of the optimized relationship network to be processed more conveniently, the local knowledge relationship chain (local behavior preference knowledge relationship network) obtained by disassembling can be set to have smaller corresponding disassembling coverage, so that the difference between different local knowledge relationship chains can be conveniently highlighted on the characteristics of whether active knowledge members exist and related to the local knowledge relationship chain. Or when the number of optimization stages in the optimization process of the relational network is to be reduced, a larger dismantling coverage can be set for the local knowledge relation chain, so that the corresponding optimization stages are correspondingly reduced when the relational network is optimized.

In combination with the step 201, for each of the plurality of local knowledge relationship chains, the method further includes the following steps 202 and 203.

Step 202: and if active knowledge members exist in the behavior preference knowledge members in the local knowledge relationship chain, determining a relationship variable index corresponding to the local knowledge relationship chain as a first description value.

Step 203: and if the active knowledge members do not exist in the behavior preference knowledge members in the local knowledge relation chain, determining the relation variable index corresponding to the local knowledge relation chain as a second description value.

Wherein the first description value and the second description value may be 1 and 0, respectively.

Step 204: and combining the relation variable indexes respectively corresponding to the local knowledge relation chains to form the relative relation variable.

For example, after determining the relationship variable indexes corresponding to the local knowledge relationship chains in the behavior preference knowledge relationship network in the step 202 and the step 203, the corresponding relative relationship variables may be obtained. Wherein, in order to highlight the relation variable index in which the first description value is displayed, the first description value is annotated only in the corresponding region of the local knowledge relation chain to which the first description value corresponds, and the second description value is not annotated.

In other words, the relative relationship variables may reflect the location of the local knowledge relationship chain comprising inactive knowledge members in the behavior preference knowledge relationship network after the disassembly of the local knowledge relationship chain for the behavior preference knowledge relationship network.

For the step 103, on the basis of determining the relative relationship variables, the relationship network optimization can be performed on the behavior preference knowledge relationship network by using the relative relationship variables, so as to obtain an optimized relationship network corresponding to the user operation behavior information.

In some independent embodiments, the following sections illustrate solutions for determining an optimized relationship network.

Step 501: and carrying out region constraint transformation by combining a first region constraint and a second region constraint of each local knowledge relation chain in the behavior preference knowledge relation network to obtain a plurality of knowledge cores.

The x-th knowledge core corresponds to an x-th group of transverse local knowledge relation chains and an x-th group of longitudinal local knowledge relation chains; x is an integer greater than 0 and not greater than P; p reflects the number of the local knowledge relation chains in the first constraint dimension or the second constraint dimension of the behavior preference knowledge relation network.

For example, corresponding first and second region constraints may be determined separately for each local knowledge relationship chain. The first region constraint includes, for example, a transverse constraint value HA-HF, and the second region constraint correspondingly includes a longitudinal constraint value VA-VF, and after the region constraint transformation is performed, the obtained knowledge cores include a knowledge core CA-knowledge core CF. The first knowledge core (which may be understood as a node or a knowledge cluster unit) of the plurality of knowledge cores, i.e., the knowledge core CA, corresponds to a first set of local knowledge relation chains in a horizontal direction and also corresponds to a first set of local knowledge relation chains in a vertical direction. In determining the number of knowledge cores, the number of the first constraint dimensions or the second constraint dimensions of the behavior preference knowledge relation network can be determined according to the local knowledge relation chain, in this example, 6 groups of transverse local knowledge relation chains exist in the first constraint dimensions, 6 groups of longitudinal local knowledge relation chains exist in the second constraint dimensions, the number of the knowledge cores determined based on the determination is 6, and the knowledge cores are correspondingly represented by the knowledge cores CA-knowledge cores CF.

Step 502: and generating a linkage knowledge relation network reflecting the upstream and downstream relation among a plurality of knowledge cores by combining the relation variable indexes of the local knowledge relation chains corresponding to the knowledge cores.

The feature value of each behavior preference knowledge member in the linkage knowledge relationship network reflects whether an upstream-downstream relationship (such as a neighbor relationship or an adjacent relationship) exists between two knowledge cores corresponding to the behavior preference knowledge member.

For this step, the linkage knowledge relation network reflecting the upstream and downstream links between the knowledge cores can be generated by using the relation variable indexes of the local knowledge relation chains corresponding to the knowledge cores.

For some examples, a linked knowledge relationship network (neighbor relationship network) can be generated according to the following: for each linkage member in the linkage knowledge relationship network, determining a target knowledge core corresponding to the linkage member; combining the first region constraint and the second region constraint reflected by the target knowledge core to determine a target knowledge relation chain corresponding to the target knowledge core; combining the relation variable indexes corresponding to the target knowledge relation chain to determine the characteristic value corresponding to the linkage member; combining the characteristic values respectively corresponding to a plurality of linkage members to generate the linkage knowledge relationship network; if the relation variable index corresponding to the target knowledge relation chain comprises a first description value, determining that upstream and downstream relation exists between the two target knowledge cores corresponding to the linkage member; and if the relation variable index corresponding to the target knowledge relation chain comprises a second description value, determining that no upstream and downstream relation exists between the two target knowledge cores corresponding to the linkage member.

Wherein the window coverage of the linkage knowledge relationship network is related to, for example, the size of the local knowledge relationship chain after the local knowledge relationship chain is disassembled by the behavior preference knowledge relationship network. After the partial knowledge relation chains are disassembled from the behavior preference knowledge relation network, A7-A7 partial knowledge relation chains can be obtained. Then A7 linkage members may be included for the linkage knowledge relationship network.

For each linkage member, a target knowledge core corresponding to the linkage member can be determined. In some examples, for linkage members on a first set grid path of the linkage knowledge relationship network, the corresponding first region constraint and second region constraint indicate the same knowledge core, based on which there is one corresponding target knowledge core; on another possible basis, for linkage members on a second set grid path in the linkage knowledge relationship network, the corresponding first region constraint and second region constraint indicate two different knowledge cores, and two corresponding target knowledge cores are based on the two different knowledge cores.

On the basis of determining the target knowledge core corresponding to the linkage member, the corresponding target knowledge relation chain can be determined for the target knowledge core by utilizing the first region constraint and the second region constraint reflected by the target knowledge core. For example, for a linkage member of horizontal 1 and vertical 1 in the linkage knowledge relationship network, the corresponding first region constraint and second region constraint indicate the same knowledge core CA, and based on this, it is determined that the target knowledge core corresponding to the linkage member is the knowledge core CA. For linkage members of horizontal 1 and vertical 2 in the linkage knowledge relation network, the corresponding first region constraint indicates a knowledge core CA, the corresponding second region constraint indicates a knowledge core CB, and the corresponding target knowledge core comprises the knowledge core CA and the knowledge core CB.

For linkage members of horizontal 1 and vertical 1 in the linkage knowledge relation network, the corresponding target knowledge core is a knowledge core CA. For linkage members with the horizontal direction 1 and the vertical direction 2 in the linkage knowledge relationship network, the corresponding target knowledge cores are a knowledge core CA and a knowledge core CB, and other target knowledge relationship chains can be determined correspondingly according to the corresponding relations of the corresponding rows of the knowledge core CA and the corresponding columns of the knowledge core CB.

And determining the characteristic value corresponding to the linkage member by utilizing the relation variable index corresponding to the target knowledge relation chain. For example, a relationship variable index corresponding to the target knowledge relationship chain can be obtained on the basis of determining the target knowledge relationship chain for the linkage members. In some examples, if it is determined that the relationship variable index corresponding to the target knowledge relationship chain includes the first description value, it is determined that there is an upstream-downstream relationship between two target knowledge cores corresponding to the target knowledge relationship chain, for example, it may be determined that the feature value corresponding to the linkage member is "1". On the other possible basis, if it is determined that the relationship variable index corresponding to the target knowledge relationship chain includes the second description value, it is determined that there is an upstream-downstream relationship between the two target knowledge cores corresponding to the target knowledge relationship chain, for example, it may be determined that the feature value corresponding to the linkage member is "0".

Thus, the linkage knowledge relationship network can be generated and obtained by determining the characteristic values respectively corresponding to a plurality of linkage members in the linkage knowledge relationship network.

Step 503: and carrying out relationship network optimization on the behavior preference knowledge relationship network by combining the linkage knowledge relationship network to obtain the optimized relationship network.

When the linkage knowledge relationship network is utilized to optimize the relationship network of the behavior preference knowledge relationship network, the following thinking can be adopted: combining the linkage knowledge relation network to determine transmission element characteristics corresponding to a plurality of knowledge cores respectively; the feature of the transmission element corresponding to any knowledge core reflects the number of other knowledge cores with immediate transmission states of the knowledge core; and carrying out relationship network optimization on the behavior preference knowledge relationship network by combining a plurality of transmission element characteristics respectively corresponding to the knowledge cores to obtain the optimized relationship network.

For example, for a knowledge core CA, a knowledge core CB directly connected to the knowledge core CA has a corresponding determination transfer element characteristic of 1. Other knowledge cores can also use the same ideas to determine the corresponding transfer element features.

For the information optimization server, other information can be obtained without pertinence, and the number of other knowledge cores with immediate transmission states for any knowledge core can be directly determined according to the linkage knowledge relation network.

When determining the transmission element characteristics corresponding to the knowledge cores respectively, performing relationship network optimization on the behavior preference knowledge relationship network to obtain an optimized relationship network, determining the adjustment rule of the optimized preference knowledge chain corresponding to the knowledge cores respectively by combining the transmission element characteristics corresponding to the knowledge cores respectively; the optimized preference knowledge chain corresponding to any knowledge core comprises the following steps: a sub-relationship network of the behavioral preference knowledge relationship network determined by the first region constraint or the second region constraint of the knowledge core; and carrying out relational network optimization on optimized preference knowledge chains respectively corresponding to the knowledge cores by combining the adjustment rules to obtain the optimized relational network.

In other words, when the relationship network optimization is performed on the behavior preference knowledge relationship network, the adjustment rule of the optimized preference knowledge chain can be determined, so that the relationship network optimization is performed on the optimized preference knowledge chain by using the adjustment rule. The optimized preferred knowledge chain corresponding to the knowledge core described herein can be a sub-relationship network of the behavioral preference knowledge relationship network determined from the first region constraint of the knowledge core.

For example, for the knowledge core CA, a first set of local knowledge relationship chains in the lateral direction, that is, sub-relationship networks of the behavior preference knowledge relationship network defined by the first region constraint of the knowledge core CA as described herein, when the optimized preference knowledge chain may include 3×a5 behavior preference knowledge members of the behavior preference knowledge relationship network; alternatively, the sub-relationship network of the behavior preference knowledge relationship network may be determined according to the second region constraint of the knowledge core, and for the knowledge core CA, the first set of longitudinal local knowledge relationship chains, that is, the sub-relationship network of the behavior preference knowledge relationship network determined by the second region constraint of the knowledge core CA described herein, may be used as the optimized preference knowledge chain, where the optimized preference knowledge chain may include a5×3 behavior preference knowledge members of the behavior preference knowledge relationship network.

For some examples, when determining the adjustment rule for the optimized preference knowledge chain corresponding to the plurality of knowledge cores respectively in combination with the transfer element features corresponding to the plurality of knowledge cores respectively, the following scheme can be implemented in each of the plurality of optimization stages: from the first knowledge cores without determining the adjustment rule, determining the current first knowledge core by combining the transmission element characteristics corresponding to each first knowledge core; combining the number of the second knowledge cores with the determined adjustment rules, and determining the adjustment rules of the optimized preference knowledge chains corresponding to the current first knowledge core; if the first knowledge core exists currently, modifying the transmission element characteristics of the first knowledge core which exists currently, jumping to the first knowledge core which does not determine the adjustment rule, and combining the transmission element characteristics respectively corresponding to the first knowledge cores to determine the current first knowledge core; and if the first knowledge core does not exist currently, ending the optimization stage to obtain adjustment rules of the knowledge cores corresponding to the optimized preference knowledge chains respectively.

In the first optimization phase (processing cycle), no adjustment rule is determined for all the determined first knowledge cores, on the basis of which the current first knowledge core can be determined from all the first knowledge cores.

When the current first knowledge core is determined by combining the transmission element characteristics corresponding to each first knowledge core, the following thought can be adopted: combining the transmission element characteristics (knowledge connection attributes, which can be understood as connection characteristics) corresponding to the first knowledge cores respectively, and performing optimization sequence arrangement on the first knowledge cores to obtain optimization sequence arrangement results corresponding to the first knowledge cores respectively; wherein the transmission element characteristic indicates that the smaller the number of other knowledge cores with immediate transmission states corresponding to the first knowledge core is, the earlier the corresponding optimization sequence is; and combining the optimization sequence arrangement results respectively corresponding to the first knowledge cores, and determining the current first knowledge core from the first knowledge cores.

For example, for the knowledge core CA-knowledge core CF described in the above example, it can be determined that the transmission element features corresponding to the knowledge core CC are 1 for all of the knowledge cores CA, CD, CE, and CF except for the transmission element feature corresponding to the knowledge core CC being 4 and the transmission element feature corresponding to the knowledge core CB being 2. When the first knowledge cores are sorted according to the optimization sequence of the transfer element features, a sorting mode that the smaller the value of the transfer element features is, the earlier the corresponding optimization sequence is can be followed.

Based on the above, in the embodiment of the invention, the thought that the corresponding optimization sequence is more advanced is determined when the number of other knowledge cores with immediate transmission states of the transmission element characteristic indication and the corresponding first knowledge core is smaller is selected, so that the difficulty in updating the transmission element characteristic is reduced, and the timeliness of the whole scheme is ensured.

In the first optimization stage, for the knowledge core with the highest optimization order, the knowledge core CA, the knowledge core CD, the knowledge core CE, and the knowledge core CF may be included. When determining the current first knowledge core selected from the current optimization stage, one first knowledge core can be selected as the current first knowledge core based on the fact that the optimization sequences of the four knowledge cores are consistent. For example, the knowledge core CD is taken as the current first knowledge core. Or, the first knowledge core with the front queue priority can be selected as the current first knowledge core according to the queue priorities of the four knowledge cores. For example, the knowledge core CA is taken as the current first knowledge core. For example, in the first optimization stage, the current first knowledge core is determined to be the knowledge core CA.

In the first optimization phase, since there is no second knowledge core already determined for the adjustment rule, the number of second knowledge cores based on which the adjustment rule is determined to be 0. Based on this, it can be determined that the current first knowledge core, i.e., the knowledge core CA, has an adjustment rule of 1 for the optimized preference knowledge chain corresponding to the knowledge core CA, which indicates that the optimized preference knowledge chain corresponding to the knowledge core CA is the optimized preference knowledge chain that is optimized for the first time.

After determining the current first knowledge core, the first knowledge core also exists at the moment, wherein the first knowledge core comprises a knowledge core CB-knowledge core CF, and the modification processing is carried out on the transmission element characteristics corresponding to the knowledge core CB-knowledge core CE based on the first knowledge core. Before modification, the knowledge core CA is directly connected with the knowledge core CB, and the condition that the knowledge core CA is used as a transition knowledge core to indirectly connect the two knowledge cores does not exist, so that the connection relation between the knowledge core CA and the knowledge core CB only needs to be cleaned in the transmission element feature.

And then, performing a processing step of a second optimization stage, namely jumping to a step of determining the current first knowledge core by combining the transmission element characteristics corresponding to the first knowledge cores respectively from the first knowledge cores without determining the adjustment rule. For example, the current first knowledge core determined in the second optimization stage is knowledge core CB. Since there is a second knowledge core with determined adjustment rules, namely, knowledge core CA, only one, based on which it can be determined that the adjustment rule of the optimized preference knowledge chain corresponding to knowledge core CB is 2, the optimized preference knowledge chain corresponding to knowledge core CB is the optimized preference knowledge chain optimized for the first time. And then, modifying the transmission element characteristics corresponding to the rest of the first knowledge cores. Since there is still a first knowledge core for which no adjustment rules are determined, the processing steps of the third optimization phase continue to be executed on the basis of this.

Further, according to the description of the first optimizing stage and the second optimizing stage, if a plurality of optimizing stages are considered, if two first knowledge cores exist, after one optimizing stage, the current first knowledge core selected in the optimizing stage can be determined, and the next knowledge core, namely the last knowledge core, which is not selected in the optimizing stage can be determined. Based on this, on the basis of the existence of P knowledge cores, at least (P-1) optimization phases will be performed.

For example, for the knowledge core CA-knowledge core CF, in five optimization stages, for example, it may be determined that an adjustment rule of an optimized preferred knowledge chain corresponding to the knowledge core CA is 1, an adjustment rule of an optimized preferred knowledge chain corresponding to the knowledge core CB is 2, an adjustment rule of an optimized preferred knowledge chain corresponding to the knowledge core CD is 3, an adjustment rule of an optimized preferred knowledge chain corresponding to the knowledge core CE is 4, an adjustment rule of an optimized preferred knowledge chain corresponding to the knowledge core CC is 5, and an adjustment rule of an optimized preferred knowledge chain corresponding to the knowledge core CF is 6.

After the adjustment rule corresponding to the optimized preference knowledge chain of each knowledge core is determined, the relationship network optimization can be carried out on the optimized preference knowledge chains respectively corresponding to the plurality of knowledge cores by utilizing the adjustment rule, so as to obtain an optimized relationship network.

For some examples, in determining an optimized relationship network, the following ideas can be employed: combining the adjustment rules to determine member connections of optimized preference knowledge chains corresponding to the knowledge cores in the optimized relationship network; and combining the member connection and optimized preference knowledge chains respectively corresponding to the knowledge cores to form the optimized relation network.

Taking the sub-relationship network of the behavioral preference knowledge relationship network determined by taking the optimized preference knowledge chain as the first region constraint of the knowledge core as an example for explanation. The optimized relation network can be obtained by optimizing the relation network of the optimized preference knowledge chain in the behavior preference knowledge relation network, and the description value of the behavior preference knowledge members in the optimized preference knowledge chain is not changed, and the number or the position of the behavior preference knowledge members is not changed in the process of optimizing the relation network. Based on this, the window coverage of the optimized relationship network and the behavior preference knowledge relationship network are the same.

For example, in determining an optimized relationship network, it can be done by, but not limited to, two different ideas as described below.

Under a possible thinking, after determining the current first knowledge core and the adjustment rule corresponding to the optimized preference knowledge chain in each optimization stage, the optimized preference knowledge chain can be determined to be in contact with a member in the optimized relationship network based on the adjustment rule, and in the last optimization stage of a plurality of optimization stages, the optimized preference knowledge chain corresponding to the last knowledge core is optimized.

Under another possible thinking, after the adjustment rules of the optimized preference knowledge chains corresponding to the knowledge cores are respectively determined in a plurality of optimization stages, the member connection of each optimized preference knowledge chain in the optimized relation network is determined through the adjustment rules corresponding to the optimized preference knowledge chains, and then the optimized relation network is directly determined. Under two different ideas, the optimized preference knowledge chains are optimized in the optimized relationship network in different orders, but the obtained optimized relationship network is the same.

In another embodiment of the present invention, after obtaining the optimized relationship network, mapping indication data between the user operation behavior information and the acquisition offset may be determined by combining the optimized relationship network and the acquisition offset corresponding to the user operation behavior information.

For example, after performing a relational network optimization on a behavioral preference knowledge relational network formed by user operation behavior information, the obtained optimized relational network is represented by a list L, an acquisition offset may be obtained, and a corresponding list is represented by a vector vec, for example, a mapping indication between the optimized relational network and the acquisition offset may be determined as follows: lt=vec.

The vector T is the mapping instruction data described above. And (3) utilizing the mapping indication data obtained by determination to perform mining analysis on the operation behavior matters of the user operation behavior information, thereby performing deeper requirement mining and processing.

Further, according to the introduction of the information optimization method applied to artificial intelligence, after the relative relation variable of the active knowledge member in the behavior preference knowledge relation network is determined, a linkage knowledge relation network reflecting upstream and downstream relations among a plurality of knowledge cores can be correspondingly obtained, the upstream and downstream relations among the knowledge cores in the linkage knowledge relation network are used as the basis for optimization, and in a plurality of optimization stages for optimizing the relation network, the optimized relation network with fewer active knowledge members can be obtained by reserving the relations among the knowledge cores, namely, the behavior preference knowledge relation network with higher simplification degree is obtained, so that the subsequent requirement mining analysis is facilitated.

Based on the foregoing, in some independent embodiments, after performing the optimization of the relational network of behavior preference knowledge in conjunction with the relative relational variables to obtain an optimized relational network corresponding to the user operation behavior information in step 103, the method may further include the following: and carrying out user demand mining on the cloud service interaction event based on the optimized relation network to obtain a user demand mining result.

In some independent embodiments, the user demand mining of the cloud service interaction event based on the optimized relationship network, to obtain a user demand mining result may include the following contents: determining an interest knowledge trend field sequence from the optimized relation network, wherein the interest knowledge trend field sequence comprises N groups of continuous interest knowledge trend fields, and N is an integer greater than or equal to 1; acquiring a demand trend field sequence according to the interest knowledge trend field sequence, wherein the demand trend field sequence comprises N groups of continuous demand trend fields; acquiring an interest knowledge vector distribution sequence through a first depth residual layer included in a full convolution neural network based on the interest knowledge trend field sequence, wherein the interest knowledge vector distribution sequence comprises N interest knowledge vector distributions; acquiring a required knowledge vector distribution sequence through a second depth residual layer included in the full convolution neural network based on the required trend field sequence, wherein the required knowledge vector distribution sequence comprises N required knowledge vector distributions; acquiring multiple regression scores (classification probability values) corresponding to the interest knowledge trend fields through a demand item classification layer included in the full convolution neural network based on the interest knowledge vector distribution sequence and the demand knowledge vector distribution sequence; and determining a user demand mining result of the interest knowledge trend field sequence according to the multiple regression score.

Thus, the accuracy and the credibility of the mining result required by the user can be ensured.

In some independent embodiments, the obtaining, based on the interest knowledge vector distribution sequence and the demand knowledge vector distribution sequence, a multiple regression score corresponding to the interest knowledge trend field sequence through a demand item classification layer included in the full convolutional neural network includes: based on the interest knowledge vector distribution sequence, acquiring N first array expressions through a first feature mining unit included in the full convolution neural network, wherein each first array expression corresponds to one interest knowledge vector distribution; based on the required knowledge vector distribution sequence, acquiring N second group expressions through a second feature mining unit included in the full convolution neural network, wherein each second group expression corresponds to one required knowledge vector distribution; splicing the N first array expressions and the N second array expressions to obtain N target array expressions, wherein each target array expression comprises a first array expression and a second array expression; and acquiring multiple regression scores corresponding to the interest knowledge trend field sequences through the demand item classification layer included in the full convolution neural network based on the N target array expressions.

In this way, noise interference in the determination of multiple regression scores may be minimized.

Further, there is also provided a readable storage medium having stored thereon a program which when executed by a processor implements the above-described method.

It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described system and apparatus may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein. In the several embodiments provided in the present invention, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be other manners of division in actual implementation, and for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, device or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form.

Claims

1. An information optimization method applied to artificial intelligence, characterized in that the method is applied to an information optimization server, and the method comprises the following steps:

obtaining user operation behavior information corresponding to a cloud service interaction event when a user portrait knowledge set is generated on the cloud service interaction event; the user operation behavior information forms a behavior preference knowledge relation network, and the behavior preference knowledge relation network comprises a plurality of behavior preference knowledge members;

combining member contact of each active knowledge member in the behavior preference knowledge relationship network in a plurality of behavior preference knowledge members to determine relative relationship variables of the active knowledge members in the behavior preference knowledge relationship network;

carrying out relational network optimization on the behavior preference knowledge relational network by combining the relative relational variables to obtain an optimized relational network corresponding to the user operation behavior information;

combining member contact of each active knowledge member in the behavior preference knowledge relationship network in the plurality of behavior preference knowledge members, and determining a relative relationship variable of the active knowledge member in the behavior preference knowledge relationship network, wherein the method comprises the following steps:

disassembling the behavior preference knowledge relation network into a plurality of local knowledge relation chains; each local knowledge relation chain comprises a plurality of behavior preference knowledge members positioned in the local knowledge relation chain;

For each local knowledge relation chain in a plurality of local knowledge relation chains, if active knowledge members exist in behavior preference knowledge members in the local knowledge relation chain, determining a relation variable index corresponding to the local knowledge relation chain as a first description value;

if the active knowledge members do not exist in the behavior preference knowledge members in the local knowledge relation chain, determining a relation variable index corresponding to the local knowledge relation chain as a second description value;

and combining the relation variable indexes respectively corresponding to the local knowledge relation chains to form the relative relation variable.

2. The method as recited in claim 1, further comprising: and determining mapping indicating data between the user operation behavior information and the acquisition offset by combining the optimized relation network and the acquisition offset corresponding to the user operation behavior information.

3. The method according to claim 1, wherein the disassembling the behavior preference knowledge relationship network into a number of local knowledge relationship chains comprises any one of the following:

based on a set dismantling rule, dismantling the behavior preference knowledge relation network into a plurality of local knowledge relation chains;

Combining the window coverage of the behavior preference knowledge relationship network, and determining the dismantling coverage corresponding to the behavior preference knowledge relationship network; the window coverage of the behavior preference knowledge relationship network is a set multiple of the dismantling coverage; and disassembling the behavior preference knowledge relationship network into a plurality of local knowledge relationship chains by combining the determined disassembling coverage.

4. The method according to claim 1, wherein the optimizing the relationship network of the behavior preference knowledge relationship network in combination with the relative relationship variables to obtain an optimized relationship network corresponding to the user operation behavior information comprises:

combining a first region constraint and a second region constraint of each local knowledge relation chain in the behavior preference knowledge relation network to perform region constraint transformation to obtain a plurality of knowledge cores; the x-th knowledge core corresponds to an x-th group of transverse local knowledge relation chains and an x-th group of longitudinal local knowledge relation chains; x is an integer greater than 0 and not greater than P; p reflects the number of the local knowledge relation chains in the first constraint dimension or the second constraint dimension of the behavior preference knowledge relation network;

Combining the relation variable indexes of the local knowledge relation chains corresponding to the knowledge cores to generate a linkage knowledge relation network reflecting the upstream and downstream relation among a plurality of knowledge cores; the characteristic value of each linkage member in the linkage knowledge relation network reflects whether upstream and downstream connection exists between two knowledge cores corresponding to the linkage member;

and carrying out relationship network optimization on the behavior preference knowledge relationship network by combining the linkage knowledge relationship network to obtain the optimized relationship network.

5. The method of claim 4, wherein the generating a linkage knowledge relationship network reflecting the upstream and downstream links between the knowledge cores by combining the relationship variable indexes of the local knowledge relationship chains corresponding to the knowledge cores comprises:

for each linkage member in the linkage knowledge relationship network, determining a target knowledge core corresponding to the linkage member;

combining the first region constraint and the second region constraint reflected by the target knowledge core to determine a target knowledge relation chain corresponding to the target knowledge core;

combining the relation variable indexes corresponding to the target knowledge relation chain to determine the characteristic value corresponding to the linkage member;

Combining the characteristic values respectively corresponding to a plurality of linkage members to generate the linkage knowledge relationship network;

if the relation variable index corresponding to the target knowledge relation chain comprises a first description value, determining that upstream and downstream relation exists between the two target knowledge cores corresponding to the linkage member; and if the relation variable index corresponding to the target knowledge relation chain comprises a second description value, determining that no upstream and downstream relation exists between the two target knowledge cores corresponding to the linkage member.

6. The method of claim 4, wherein the performing, in conjunction with the linkage knowledge relationship network, a relationship network optimization on the behavior preference knowledge relationship network to obtain the optimized relationship network comprises: combining the linkage knowledge relation network to determine transmission element characteristics corresponding to a plurality of knowledge cores respectively; the feature of the transmission element corresponding to any knowledge core reflects the number of other knowledge cores with immediate transmission states of the knowledge core; combining a plurality of transmission element characteristics respectively corresponding to the knowledge cores to optimize the relationship network of the behavior preference knowledge relationship network, so as to obtain the optimized relationship network;

And carrying out relationship network optimization on the behavior preference knowledge relationship network by combining a plurality of transmission element characteristics respectively corresponding to the knowledge cores to obtain the optimized relationship network, wherein the method comprises the following steps: combining the transmission element characteristics respectively corresponding to the knowledge cores to determine the adjustment rules of the optimized preference knowledge chains respectively corresponding to the knowledge cores; the optimized preference knowledge chain corresponding to any knowledge core comprises the following steps: a sub-relationship network of the behavioral preference knowledge relationship network determined by the first region constraint or the second region constraint of the knowledge core; and carrying out relational network optimization on optimized preference knowledge chains respectively corresponding to the knowledge cores by combining the adjustment rules to obtain the optimized relational network.

7. The method of claim 6, wherein the determining, in combination with the transfer element features respectively corresponding to the plurality of knowledge cores, an adjustment rule for the optimized preferred knowledge chain respectively corresponding to the plurality of knowledge cores comprises: in each of several optimization stages, the following scheme is implemented: from the first knowledge cores without determining the adjustment rule, determining the current first knowledge core by combining the transmission element characteristics corresponding to each first knowledge core; combining the number of the second knowledge cores with the determined adjustment rules, and determining the adjustment rules of the optimized preference knowledge chains corresponding to the current first knowledge core; if the first knowledge core exists currently, modifying the transmission element characteristics of the first knowledge core which exists currently, jumping to the first knowledge core which does not determine the adjustment rule, and combining the transmission element characteristics respectively corresponding to the first knowledge cores to determine the current first knowledge core; if the first knowledge core does not exist currently, ending the optimization stage, and obtaining adjustment rules of a plurality of knowledge cores corresponding to optimized preference knowledge chains respectively;

The determining the current first knowledge core by combining the transmission element characteristics corresponding to the first knowledge cores respectively includes: combining the transmission element characteristics respectively corresponding to the first knowledge cores, and performing optimization sequence arrangement on the first knowledge cores to obtain optimization sequence arrangement results respectively corresponding to the first knowledge cores; wherein the transmission element characteristic indicates that the smaller the number of other knowledge cores with immediate transmission states corresponding to the first knowledge core is, the earlier the corresponding optimization sequence is; and combining the optimization sequence arrangement results respectively corresponding to the first knowledge cores, and determining the current first knowledge core from the first knowledge cores.

8. The method of claim 6, wherein the optimizing the relationship network for the optimized preference knowledge chain corresponding to each of the plurality of knowledge cores in combination with the adjustment rule to obtain the optimized relationship network comprises:

combining the adjustment rules to determine member connections of optimized preference knowledge chains corresponding to the knowledge cores in the optimized relationship network;

and combining the member connection and optimized preference knowledge chains respectively corresponding to the knowledge cores to form the optimized relation network.

9. An information optimization server, comprising a processor and a memory; the processor being communicatively connected to the memory, the processor being adapted to read a computer program from the memory and execute it to carry out the method of any of the preceding claims 1-8.