Integrated Management Platform System for Warehouse Applications Based on AI Cloud Computing

Abstract

The integrated management platform system for warehousing applications based on AI cloud computing provided by the present invention relates to the field of artificial intelligence technology. In the present invention, first, a target warehousing solution relationship network is determined; secondly, at least one first member link with a specified warehousing item member as the starting relationship network member is extracted from the target warehousing solution relationship network, and at least one second member link with the warehousing solution member reflected by a pending warehousing solution as the starting relationship network member is extracted; then, the relationship network member semantic vectors corresponding to the specified warehousing item member and the relationship network member semantic vectors corresponding to each pending warehousing solution are mined; finally, a target warehousing solution is determined based on the relationship network member semantic vectors corresponding to the specified warehousing item member and each pending warehousing solution. Based on the above, the problem of relatively low reliability of warehousing management existing in the prior art can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and more particularly, to a comprehensive management platform system for warehousing applications based on AI cloud computing. Background Art

[0002] Cloud computing is a technology that provides computing resources over the Internet, allowing users to access and use shared computing resources on demand, including servers, storage, databases, networks, software, and analysis tools, etc. It has changed the traditional IT infrastructure model, enabling enterprises and individuals to use technology resources in a more cost-effective, flexible, and efficient manner. Among them, through the application of cloud computing in warehousing, the convenience of warehousing monitoring can be improved. However, for warehousing management, such as how to store items (i.e., determine the warehousing plan), there are still problems with relatively low reliability. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a comprehensive management platform system for warehousing applications based on AI cloud computing, so as to improve the relatively low reliability problem existing in the prior art in warehousing management.

[0004] To achieve the above purpose, the embodiments of the present invention adopt the following technical solutions:

[0005] A comprehensive management platform system for warehousing applications based on AI cloud computing, including a plurality of warehousing front-end devices and a back-end cloud computing device communicatively connected to the plurality of warehousing front-end devices. The back-end cloud computing device is used to execute a comprehensive management method for warehousing applications based on AI cloud computing. Among them, the comprehensive management method for warehousing applications based on AI cloud computing includes:

[0006] Determine a target warehousing plan relationship network, and determine the member description data corresponding to each relationship network member in the target warehousing plan relationship network. The target warehousing plan relationship network includes a plurality of relationship network members. For two relationship network members determined to have a relevant relationship, relationship line segments are configured in the target warehousing plan relationship network. The plurality of relationship network members include a plurality of warehousing item members reflecting warehousing items and a plurality of warehousing plan members reflecting pending warehousing plans. And the plurality of warehousing item members include designated warehousing item members reflecting designated warehousing items. The member description data corresponding to the warehousing item members is formed based on information collection of warehousing items by the warehousing front-end devices, and the member description data corresponding to the warehousing plan members is used to describe the warehousing plan;

[0007] Extract at least one first member link that takes the specified warehousing item member as the starting member of the relationship network from the target warehousing solution relationship network, and extract at least one second member link that takes the warehousing solution member reflected by the to-be-determined warehousing solution as the starting member of the relationship network;

[0008] Load the member description data corresponding to each first member link of the specified warehousing item member and the relationship network members on each first member link, so that the data mining network mines the relationship network member semantic vectors corresponding to the specified warehousing item member, and load the member description data corresponding to each second member link of each to-be-determined warehousing solution and the relationship network members on each second member link, so that the data mining network mines the relationship network member semantic vectors corresponding to each to-be-determined warehousing solution;

[0009] Based on the relationship network member semantic vectors corresponding to the specified warehousing item member and the relationship network member semantic vectors corresponding to each to-be-determined warehousing solution, determine the target warehousing solution corresponding to the specified warehousing item among multiple to-be-determined warehousing solutions, where the target warehousing solution is used as the basis for storing the specified warehousing item.

[0010] In some preferred embodiments, in the above-mentioned integrated management platform system for warehousing applications based on AI cloud computing, the step of determining the target warehousing solution corresponding to the specified warehousing item among multiple to-be-determined warehousing solutions based on the relationship network member semantic vectors corresponding to the specified warehousing item member and the relationship network member semantic vectors corresponding to each to-be-determined warehousing solution includes:

[0011] When the relationship network members connected by the relationship line segments of the specified warehousing item member all belong to warehousing item members, mark all the first member links of the specified warehousing item as first target member links, and use the semantic conversion network to perform a conversion operation on the relationship network member semantic vectors corresponding to the specified warehousing item member, output the corresponding relationship network member conversion vectors, and for each to-be-determined warehousing solution, determine the warehousing matching parameter between the specified warehousing item and the to-be-determined warehousing solution based on the relationship network member conversion vector and the relationship network member semantic vector corresponding to the to-be-determined warehousing solution;

[0012] When the relationship network members connected by the relationship line segments of the specified warehousing item member include warehousing solution members and warehousing item members, mark all the first member links of the specified warehousing item as including first target member links and second target member links, and for each to-be-determined warehousing solution, determine the warehousing matching parameter between the specified warehousing item and the to-be-determined warehousing solution based on the relationship network member semantic vectors corresponding to the specified warehousing item and the to-be-determined warehousing solution;

[0013] Based on the warehousing matching parameters corresponding to each of the to-be-determined warehousing solutions, among the various to-be-determined warehousing solutions, a target warehousing solution corresponding to the specified warehousing item is analyzed.

[0014] In some preferred embodiments, in the above-mentioned integrated management platform system for warehousing applications based on AI cloud computing, the warehousing application comprehensive management method based on AI cloud computing further includes:

[0015] Determine a training warehousing solution relationship network, and determine member description data corresponding to each relationship network member in the training warehousing solution relationship network, where the training warehousing solution relationship network includes multiple relationship network members. For two relationship network members determined to have a relevant relationship, a relationship line segment is configured in the warehousing solution relationship network. The multiple relationship network members include multiple warehousing item members reflecting warehousing items and multiple warehousing solution members reflecting to-be-determined warehousing solutions. Among the multiple warehousing item members, there is at least one target warehousing item member, and the target warehousing item member is respectively connected to at least one warehousing solution member and at least one other warehousing item member through relationship line segments;

[0016] In the training warehousing solution relationship network, determine at least one training relationship network member, and for each training relationship network member, in the training warehousing solution relationship network, extract at least one member link with the training relationship network member as the starting relationship network member. Among them, a member link whose connected relationship network members all belong to warehousing item members is marked as a first target member link, and a member link whose connected relationship network members all belong to warehousing solution members is marked as a second target member link. The connected relationship network members refer to the relationship network members having a relationship line segment with the starting relationship network member. The at least one training relationship network member includes at least one target warehousing item member, and each member link of the target warehousing item member includes at least one first target member link and at least one second target member link;

[0017] Based on the member description data corresponding to the relational network members on each member link of each of the training relational network members, perform at least one update operation on the candidate data mining network and the candidate semantic conversion network, and output the corresponding data mining network and semantic conversion network. Wherein, the target update error of the update operation includes the first local update error corresponding to each target storage item member, and the first local update error is used to reflect the difference information between the semantic vector output by the corresponding target storage item member in each update operation and the relational network member conversion vector output by using the candidate semantic conversion network based on the first semantic vector corresponding to the corresponding target storage item member in each update operation. The output semantic vector belongs to the relational network member semantic vector or the second semantic vector. The candidate data mining network is used to load the member description data corresponding to the relational network members on each member link of the training relational network members, and mine the relational network member semantic vectors corresponding to the training relational network members. And when each member link of the training relational network members includes at least one first target member link, the relational network member semantic vector includes the first semantic vector mined based on the member description data corresponding to the relational network members on each first target member link. When each member link of the training relational network members includes at least one second target member link, the relational network member semantic vector includes the second semantic vector mined based on the member description data corresponding to the relational network members on each second target member link.

[0018] In some preferred embodiments, in the above-mentioned integrated management platform system for warehouse applications based on AI cloud computing, the at least one training relational network member further includes a warehouse plan member, and the target update error further includes the second local update error corresponding to each target storage item member. Wherein, in each update operation, the second local update error corresponding to each target storage item member is determined based on the following steps:

[0019] In the training warehouse plan relational network, determine the warehouse plan members connected to the target storage item member through relationship line segments, and mark them as relevant warehouse plan members. And in the training warehouse plan relational network, determine the warehouse plan members not connected to the target storage item member through relationship line segments, and mark them as non-relevant warehouse plan members;

[0020] Poll each relevant warehouse plan member and each non-relevant warehouse plan member to form corresponding multiple combinations of warehouse plan members, where each combination of warehouse plan members includes a relevant warehouse plan member and a non-relevant warehouse plan member;

[0021] For each combination of warehousing solution members, based on the semantic vectors of the relationship network members of the target warehousing item members and the relevant warehousing solution members in the current update operation in the combination of the warehousing solution members, determine the first warehousing matching parameter between the target warehousing item members and the relevant warehousing solution members, and based on the semantic vectors of the relationship network members of the target warehousing item members and the non-relevant warehousing solution members in the combination of the warehousing solution members in the current update operation, determine the second warehousing matching parameter between the target warehousing item members and the non-relevant warehousing solution members, and further, based on the difference information between the first warehousing matching parameter and the second warehousing matching parameter, determine the warehousing solution matching sub-error of the target warehousing item members based on the combination of the warehousing solution members;

[0022] Based on the warehousing solution matching sub-errors of the target warehousing item members for all combinations of the warehousing solution members, determine the second local update error of the target warehousing item members.

[0023] In some preferred embodiments, in the above-mentioned integrated management platform system for warehousing applications based on AI cloud computing, the target update error further includes a third local update error corresponding to each training relationship network member belonging to the warehousing item members, wherein the third local update error is determined based on the following steps:

[0024] For each update operation, determine the set center semantic vector of the set center of each vector clustering set and the semantic vectors of the relationship network members of each warehousing item member in the current update operation, wherein the set center semantic vector of the set center of each vector clustering set at the first clustering belongs to a pre-configured reference vector;

[0025] For each warehousing item member, based on the semantic vectors of the relationship network members of the warehousing item member in the current update operation and the set center semantic vector of each set center, determine the warehousing matching parameters between the warehousing item member and each set center respectively;

[0026] Based on the warehousing matching parameters between each warehousing item member and each set center respectively, determine the assignment probability of each warehousing item member to each vector clustering set;

[0027] For each warehousing item member, based on the difference information between the warehousing matching parameters between the warehousing item member and each set center respectively and the assignment probability of the warehousing item member assigned to the corresponding vector clustering set, determine the third local update error of the warehousing item member.

[0028] In some preferred embodiments, in the above-mentioned integrated management platform system for warehousing applications based on AI cloud computing, the step of determining the assignment probability of each warehousing item member to each vector clustering set based on the warehousing matching parameters respectively possessed by each warehousing item member and each aggregation center includes:

[0029] For each aggregation center, calculate the sum value of the warehousing matching parameters corresponding to each training warehousing item member for this aggregation center, and obtain the clustering representation parameter of the vector clustering set corresponding to this aggregation center;

[0030] For each warehousing item member, based on the warehousing matching parameter between this warehousing item member and an aggregation center, and in combination with the clustering representation parameter of the vector clustering set corresponding to this aggregation center, determine the set contribution degree of this warehousing item member in this vector clustering set;

[0031] For each warehousing item member, calculate the ratio between the set contribution degree of this warehousing item member in a vector clustering set and the set contribution degrees of this warehousing item member in all vector clustering sets, and obtain the assignment probability of this warehousing item member to this vector clustering set.

[0032] In some preferred embodiments, in the above-mentioned integrated management platform system for warehousing applications based on AI cloud computing, each update operation further includes:

[0033] For each warehousing item member, determine the target aggregation center as the aggregation center corresponding to the warehousing matching parameter with the maximum value of this warehousing item member, and mark the vector clustering set corresponding to this target aggregation center, so that this vector clustering set is marked as the vector clustering set corresponding to this warehousing item member in the next update operation;

[0034] For each vector clustering set corresponding to the next update operation, based on the target update error, adjust the set center semantic vector of the aggregation center corresponding to this vector clustering set.

[0035] In some preferred embodiments, in the above-mentioned integrated management platform system for warehousing applications based on AI cloud computing, the first local update error corresponding to the target warehousing item member is determined based on the following steps:

[0036] Based on the vector difference between each set center and the output semantic vector of the target warehousing item member in the current update operation, aggregate and calculate the set center semantic vector of each set center, output the corresponding first aggregated semantic vector, and perform a semantic enhancement operation on the output semantic vector based on the first aggregated semantic vector to output the output semantic vector after the semantic enhancement operation. Moreover, mark the output semantic vector after the semantic enhancement operation to make it the warehousing item enhancement vector of the target warehousing item member in the current update operation;

[0037] Based on the vector difference between each set center and the first semantic vector of the target warehousing item member in the current update operation, aggregate and calculate the set center semantic vector of each set center, output the corresponding second aggregated semantic vector, and perform a semantic enhancement operation on the first semantic vector based on the second aggregated semantic vector to output the first semantic vector after the semantic enhancement operation. Moreover, use the candidate semantic conversion network to perform a conversion operation on the first semantic vector after the semantic enhancement operation to output the relational network member conversion vector of the target warehousing item member in the current update operation;

[0038] Based on the vector difference between the warehousing item enhancement vector and the relational network member conversion vector of the target warehousing item member in the current update operation, determine the first local update error corresponding to the target warehousing item member.

[0039] In some preferred embodiments, in the above-mentioned integrated management platform system for warehousing applications based on AI cloud computing, the multiple warehousing item members include at least one non-related warehousing item member, and the non-related warehousing item member refers to a warehousing item member that does not have a relationship line segment with the warehousing plan member;

[0040] At least one training relational network member also includes non-related warehousing item members. The member links of the non-related warehousing item members all belong to the first target member link, and the relational network member semantic vectors of the non-related warehousing item members belong to the first semantic vector.

[0041] In some preferred embodiments, in the above-mentioned integrated management platform system for warehousing applications based on AI cloud computing, the first semantic vector or the second semantic vector corresponding to each training relational network member is determined according to the following steps:

[0042] In each member link where the training relational network member belongs to the starting relational network member, determine each comparison member link, where the comparison member link belongs to the first target member link or the second target member link;

[0043] For each comparison member link, starting from the network member with the second-lowest relevance in the comparison member link, based on the member transfer vectors of each network member with lower relevance connected by the network member and the member description data corresponding to the network member, sequentially transfer to the starting network member to determine the member transfer vector that the starting network member has on the comparison member link. Among them, the member transfer vector of the network member with the lowest relevance is mined and formed based on the member description data corresponding to the network member with the lowest relevance;

[0044] Perform an association focus fusion operation on the member transfer vectors of the training network member on each comparison member link, and output the semantic vector corresponding to the comparison member link of the training network member. Among them, when the comparison member link belongs to the first target member link, the semantic vector corresponding to the comparison member link belongs to the first semantic vector, and when the comparison member link belongs to the second target member link, the semantic vector corresponding to the comparison member link belongs to the second semantic vector.

[0045] The integrated management platform system for warehousing applications based on AI cloud computing provided by the embodiments of the present invention, first, determines the target warehousing scheme network; secondly, extracts at least one first member link with the specified warehousing item member as the starting network member from the target warehousing scheme network, and extracts at least one second member link with the warehousing scheme member reflected by the to-be-determined warehousing scheme as the starting network member; then, mines the semantic vector of the network member corresponding to the specified warehousing item member, and the semantic vector of the network member corresponding to each to-be-determined warehousing scheme; finally, determines the target warehousing scheme based on the semantic vectors of the network members corresponding to the specified warehousing item member and each to-be-determined warehousing scheme. Based on the above content, since the semantic vector of the network member corresponding to the specified warehousing item member not only includes the semantic information of the member description data corresponding to the specified warehousing item member itself, but also includes the semantic information of the member description data corresponding to the network members on the first member link, and, the semantic vector of the network member corresponding to the to-be-determined warehousing scheme also not only includes the semantic information of the member description data corresponding to the to-be-determined warehousing scheme itself, but also includes the semantic information of the member description data corresponding to the network members on the second member link, making the semantic representation ability of the mined semantic vector of the network member better and the semantic information more abundant, so that the scheme determination based on the semantic vector of the network member can be more reliable, that is, ensuring that the obtained target warehousing scheme has a higher reliability, thereby improving the problem of relatively low reliability of warehousing management existing in the prior art.

[0046] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following specifically enumerates preferred embodiments and, in conjunction with the accompanying drawings, makes a detailed description as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a structural block diagram of the integrated management platform system for warehousing applications based on AI cloud computing provided by an embodiment of the present invention.

[0048] Figure 2 It is a schematic flow diagram of each step included in the integrated management method for warehousing applications based on AI cloud computing provided by an embodiment of the present invention.

[0049] Figure 3 It is a schematic diagram of the target warehousing solution relationship network provided by an embodiment of the present invention.

[0050] Figure 4 It is a schematic diagram of the relevant process of the first local update error provided by an embodiment of the present invention.

[0051] Figure 5 It is a schematic diagram of the formation process of the member transfer vector provided by an embodiment of the present invention. Detailed implementation manners

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Usually, the components of the embodiments of the present invention described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0053] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents the selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0054] As Figure 1 shown, an embodiment of the present invention provides an integrated management platform system for warehousing applications based on AI cloud computing. Among them, the integrated management platform system for warehousing applications based on AI cloud computing may include a plurality of warehousing front-end devices (such as terminal devices) and a back-end cloud computing device (such as a server) communicatively connected to the plurality of warehousing front-end devices.

[0055] Specifically, the backend cloud computing device may include a memory and a processor. The memory and the processor are electrically connected directly or indirectly to achieve data transmission or interaction. For example, they may be electrically connected through one or more communication buses or signal lines. The memory may store at least one software functional module (computer program) that can exist in the form of software or firmware. The processor may be used to execute the executable computer program stored in the memory, thereby implementing the comprehensive management method for warehousing applications based on AI cloud computing provided by the embodiments of the present invention (as described later).

[0056] Optionally, the memory may be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc. The processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on chip (SoC), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0057] Combined Figure 2 , the embodiments of the present invention further provide a comprehensive management method for warehousing applications based on AI cloud computing, which can be applied to the above backend cloud computing device. Among them, the method steps defined by the processes related to the comprehensive management method for warehousing applications based on AI cloud computing can be implemented by the backend cloud computing device. The following will Figure 2 elaborate in detail on the specific process shown.

[0058] Step S110, determine the target warehousing scheme relationship network, and determine the member description data corresponding to each relationship network member in the target warehousing scheme relationship network.

[0059] In an embodiment of the present invention, the backend cloud computing device may determine a target warehousing solution relationship network, and determine member description data corresponding to each relationship network member in the target warehousing solution relationship network. Among them, in combination with Figure 3 , the target warehousing solution relationship network includes a plurality of relationship network members. For two relationship network members determined to have a relevant relationship, relationship line segments are configured in the target warehousing solution relationship network. The plurality of relationship network members include a plurality of warehousing item members reflecting warehousing items and a plurality of warehousing solution members reflecting pending warehousing solutions. And the plurality of warehousing item members include designated warehousing item members reflecting designated warehousing items. The member description data corresponding to the warehousing item members is formed based on information collection of the corresponding warehousing items by the warehousing front-end device (for example, the warehousing front-end device includes image collection devices such as cameras, which are used to collect images of warehousing items to obtain corresponding images as the corresponding member description data. And, object recognition can also be performed on the images to obtain information such as the object type or product description of the warehousing items). The member description data corresponding to the warehousing solution members is used to describe the corresponding warehousing solutions (for example, including how to store, for example, how to transport the warehousing items to what positions for storage, and what storage conditions are required, such as the limitation of light, temperature, humidity and other conditions). In addition, that there is a relationship line segment between warehousing item members may mean that the corresponding warehousing items have a relevant relationship, such as belonging to the same category of items (such as being fragile items). And, that there is a relationship line segment between a warehousing item member and a warehousing solution member may mean that the corresponding warehousing item has a relevant relationship with the warehousing solution, such as the warehousing item is stored based on the warehousing solution. The designated warehousing item member may refer to the warehousing item member corresponding to the warehousing item for which the warehousing solution has not been determined, such as Figure 3 the warehousing item member i in, which may belong to the same category of items as the warehousing item member h and is characterized by a relationship line segment.

[0060] Step S120: Extract at least one first member link with the designated warehousing item member as the starting relationship network member and at least one second member link with the warehousing solution member reflecting the pending warehousing solution as the starting relationship network member from the target warehousing solution relationship network.

[0061] In an embodiment of the present invention, the backend cloud computing device may extract at least one first member link with the designated warehousing item member as the starting relationship network member and at least one second member link with the warehousing solution member reflecting the pending warehousing solution as the starting relationship network member from the target warehousing solution relationship network. For example, Figure 3The storage item member i in it, as the designated storage item member, can determine the first member link "storage item member i, storage item member h, storage plan member 2, storage item member g", "storage item member i, storage item member f, storage item member c, storage plan member 1, storage item member a", etc. Figure 3 The storage plan member e in it, as the starting relationship network member, can determine the second member link "storage item member e, storage plan member 2, storage item member g", etc.

[0062] Step S130, load the member description data corresponding to each first member link of the designated storage item member and the relationship network members on each of the first member links, so that the data mining network mines out the relationship network member semantic vectors corresponding to the designated storage item member, and load the member description data corresponding to each second member link of each of the to-be-determined storage plans and the relationship network members on each of the second member links, so that the data mining network mines out the relationship network member semantic vectors corresponding to each of the to-be-determined storage plans.

[0063] In the embodiment of the present invention, the backend cloud computing device can load the member description data corresponding to each first member link of the designated storage item member and the relationship network members on each of the first member links, so that the data mining network mines out the relationship network member semantic vectors corresponding to the designated storage item member (which can be used to represent the semantic information of the corresponding member description data), and load the member description data corresponding to each second member link of each of the to-be-determined storage plans and the relationship network members on each of the second member links, so that the data mining network mines out the relationship network member semantic vectors corresponding to each of the to-be-determined storage plans (which can be used to represent the semantic information of the corresponding member description data). In addition, the data mining network can be a neural network formed by training and updating, and the update process is as described later.

[0064] Step S140, based on the relationship network member semantic vectors corresponding to the designated storage item member and the relationship network member semantic vectors corresponding to each of the to-be-determined storage plans, determine the target storage plan corresponding to the designated storage item among the multiple to-be-determined storage plans.

[0065] In an embodiment of the present invention, the backend cloud computing device may determine, among multiple to-be-determined storage plans, a target storage plan corresponding to the specified storage item based on the relationship network member semantic vectors corresponding to the specified storage item members and the relationship network member semantic vectors corresponding to each of the to-be-determined storage plans. For example, the to-be-determined storage plan with the smallest distance (or the largest similarity) between the semantic vectors is determined as the target storage plan. Wherein, the target storage plan serves as the basis for storing the specified storage item.

[0066] Based on the above, since the relationship network member semantic vectors corresponding to the specified storage item members not only include the semantic information of the member description data corresponding to the specified storage item members themselves, but also include the semantic information of the member description data corresponding to the relationship network members on the first member link, and the relationship network member semantic vectors corresponding to the to-be-determined storage plans not only include the semantic information of the member description data corresponding to the to-be-determined storage plans themselves, but also include the semantic information of the member description data corresponding to the relationship network members on the second member link, the semantic representation ability of the mined relationship network member semantic vectors is better and the semantic information is more abundant. Therefore, the determination of the plan based on the relationship network member semantic vectors can be more reliable, that is, it ensures that the obtained target storage plan has a high reliability, thereby improving the problem of relatively low reliability in warehouse management existing in the prior art.

[0067] Regarding step S140, it should be noted that the specific manner of determining the target storage plan corresponding to the specified storage item is not limited. For example, in an alternative embodiment, in order to improve the reliability of the determined target storage plan, step S140 described above may include:

[0068] First, when all the relationship network members connected by the relationship line segments of the specified storage item members belong to storage item members (that is, all the relationship network members directly connected by the relationship line segments of the specified storage item members belong to storage item members, and a specific application scenario may be that the storage item corresponding to the specified storage item member has similar storage items but no possible matching storage plans and needs to be further determined), all the first member links of the specified storage item are marked as the first target member links, and using a semantic conversion network, a conversion operation is performed on the relationship network member semantic vectors corresponding to the specified storage item members, and the corresponding relationship network member conversion vectors are output. For each of the to-be-determined storage plans, based on the relationship network member conversion vectors and the relationship network member semantic vectors corresponding to the to-be-determined storage plans, the storage matching parameters between the specified storage item and the to-be-determined storage plan are determined, such as calculating the cosine similarity between the semantic vectors.

[0069] Secondly, when the members of the relationship network connected by the relationship line segments of the specified warehousing item members include warehousing plan members and warehousing item members (specific application scenarios can be that the warehousing items corresponding to the specified warehousing item members have similar warehousing items, and some possible matching warehousing plans are configured manually or the like. Or, the warehousing plans corresponding to similar warehousing items can also be used as possible matching warehousing plans), mark each first member link of the specified warehousing item as including a first target member link and a second target member link, and for each pending warehousing plan, based on the semantic vectors of the relationship network members corresponding to the specified warehousing item and this pending warehousing plan, determine the warehousing matching parameters between the specified warehousing item and this pending warehousing plan;

[0070] Finally, based on the warehousing matching parameters corresponding to each of the pending warehousing plans, among the various pending warehousing plans, analyze the target warehousing plan corresponding to the specified warehousing item. For example, determine the pending warehousing plan with the maximum warehousing matching parameter as the target warehousing plan corresponding to the specified warehousing item.

[0071] Based on the above embodiments, in order to enable the data mining network to reliably mine the corresponding semantic vectors and enable the semantic conversion network to have reliable semantic conversion capabilities, the comprehensive warehousing application management method based on AI cloud computing may further include steps of updating and training to form the data mining network and the semantic conversion network, such as step S150, step S160, and step S170. The specific content of each step is described as follows.

[0072] Step S150, determine the training warehousing plan relationship network, and determine the member description data corresponding to each relationship network member in the training warehousing plan relationship network.

[0073] In the embodiments of the present invention, the backend cloud computing device can determine the training warehousing plan relationship network (as described in the previous related description), and determine the member description data corresponding to each relationship network member in the training warehousing plan relationship network. Among them, the training warehousing plan relationship network includes multiple relationship network members. For two relationship network members determined to have a relevant relationship, relationship line segments are configured in the warehousing plan relationship network. The multiple relationship network members include multiple warehousing item members reflecting warehousing items and multiple warehousing plan members reflecting pending warehousing plans. The multiple warehousing item members include at least one target warehousing item member, and the target warehousing item member is connected to at least one warehousing plan member and at least one other warehousing item member respectively through relationship line segments.

[0074] Step S160: In the training storage solution relationship network, determine at least one training relationship network member, and for each training relationship network member, extract at least one member link in the training storage solution relationship network that takes the training relationship network member as the starting relationship network member.

[0075] In an embodiment of the present invention, the backend cloud computing device may determine at least one training relationship network member in the training storage solution relationship network, and for each training relationship network member, extract at least one member link in the training storage solution relationship network that takes the training relationship network member as the starting relationship network member. Among them, a member link whose connected relationship network members all belong to storage item members is marked as a first target member link, a member link whose connected relationship network members all belong to storage solution members is marked as a second target member link. The connected relationship network members refer to relationship network members having relationship line segments with the starting relationship network member. The at least one training relationship network member includes at least one target storage item member, and each member link of the target storage item member includes at least one first target member link and at least one second target member link.

[0076] Step S170: Based on the member description data corresponding to the relationship network members on each member link of each training relationship network member, perform at least one update operation on the candidate data mining network and the candidate semantic conversion network, and output the corresponding data mining network and semantic conversion network.

[0077] In an embodiment of the present invention, the backend cloud computing device may perform at least one update operation on the candidate data mining network and the candidate semantic conversion network based on the member description data corresponding to the relationship network members on each member link of each of the training relationship network members, and output the corresponding data mining network and semantic conversion network. Wherein, the target update error of the update operation includes a first local update error corresponding to each target storage item member, and the first local update error is used to reflect the difference information between the semantic vector output by the corresponding target storage item member in each update operation and the relationship network member conversion vector converted and output by using the candidate semantic conversion network based on the first semantic vector corresponding to the corresponding target storage item member in each update operation. The output semantic vector belongs to the relationship network member semantic vector or the second semantic vector. The candidate data mining network is used to load the member description data corresponding to the relationship network members on each member link of the training relationship network members, and mine the relationship network member semantic vector corresponding to the training relationship network members. And when each member link of the training relationship network members includes at least one first target member link, the relationship network member semantic vector includes a first semantic vector mined based on the member description data corresponding to the relationship network members on each first target member link. When each member link of the training relationship network members includes at least one second target member link, the relationship network member semantic vector includes a second semantic vector mined based on the member description data corresponding to the relationship network members on each second target member link. It should be noted that the end condition of the at least one update operation may be that the number of update operations reaches a preset number, or the current target update error converges, and specifically can be configured according to actual requirements, and no specific limitation is made here.

[0078] To facilitate the understanding of the determination process of the above first local update error, as shown in Figure 4 , the data corresponding to the first target member link and the second target member link can be respectively loaded into the candidate data mining network, and then mined respectively in the candidate data mining network to output the first semantic vector and the second semantic vector. Among them, the first semantic vector and the second semantic vector can be fused by operations such as splicing and mean calculation to form the corresponding relationship network member semantic vector. Then, on the one hand, the first semantic vector can be loaded into the candidate semantic conversion network to output the corresponding relationship network member conversion vector. On the other hand, the second semantic vector can be used as the output semantic vector, or the relationship network member semantic vector can be used as the output semantic vector. Finally, the difference between the relationship network member conversion vector and the output semantic vector can be calculated to obtain the first local error. Then, the network parameters of the candidate data mining network and the candidate semantic conversion network can be updated and adjusted based on the first local error.

[0079] It is understandable that in an alternative implementation, in order to further improve the reliability of the network formed by training, the at least one training relationship network member further includes a warehousing plan member, and the target update error further includes a second local update error corresponding to each target warehousing item member. Wherein, in each update operation, the second local update error corresponding to each target warehousing item member can be determined based on the following steps:

[0080] First, in the training warehousing plan relationship network, determine the warehousing plan members connected to the target warehousing item member through relationship line segments, and mark them as relevant warehousing plan members. And in the training warehousing plan relationship network, determine the warehousing plan members not connected to the target warehousing item member through relationship line segments, and mark them as non-relevant warehousing plan members;

[0081] Second, poll each relevant warehousing plan member and each non-relevant warehousing plan member to form corresponding multiple warehousing plan member combinations, where each warehousing plan member combination includes a relevant warehousing plan member and a non-relevant warehousing plan member;

[0082] Then, for each warehousing plan member combination, based on the relationship network member semantic vectors of the target warehousing item member and the relevant warehousing plan member in the current update operation, determine the first warehousing matching parameter (such as the cosine similarity between the corresponding relationship network member semantic vectors) between the target warehousing item member and the relevant warehousing plan member, and based on the relationship network member semantic vectors of the target warehousing item member and the non-relevant warehousing plan member in the current update operation, determine the second warehousing matching parameter (such as the cosine similarity between the corresponding relationship network member semantic vectors) between the target warehousing item member and the non-relevant warehousing plan member. And, based on the difference information between the first warehousing matching parameter and the second warehousing matching parameter, determine the warehousing plan matching sub-error of the target warehousing item member based on the warehousing plan member combination (for example, the difference between the first warehousing matching parameter and the second warehousing matching parameter can be calculated, then, the difference is activated through an activation function to obtain the corresponding activation value, and, the activation value is logarithmically processed to obtain the corresponding logarithmic value, and finally, the logarithmic value can be negated to obtain the corresponding warehousing plan matching sub-error);

[0083] Finally, based on the storage solution matching sub-errors of all storage solution member combinations with respect to the target storage item member, the second local update error of the target storage item member is determined. For example, the sum of the storage solution matching sub-errors can be calculated and used as the second local update error of the target storage item member.

[0084] It can be understood that in an alternative implementation, to further improve the reliability of the trained network, the target update error further includes the third local update error corresponding to each training relationship network member belonging to the storage item member. The third local update error can be determined based on the following steps:

[0085] First, for each update operation, determine the set center semantic vector of the set center of each vector clustering set and the relationship network member semantic vector of each storage item member in the current update operation. The set center semantic vector of the set center of each vector clustering set at the first clustering belongs to a pre-configured reference vector (such as a randomly generated vector, which can be updated accordingly in subsequent processing).

[0086] Second, for each storage item member, based on the relationship network member semantic vector of the storage item member in the current update operation and the set center semantic vector of each set center, determine the storage matching parameters of the storage item member with each set center. For example, the cosine similarity between the relationship network member semantic vector and the set center semantic vector can be calculated.

[0087] Then, based on the storage matching parameters of each storage item member with each set center respectively, determine the assignment probability of each storage item member to each vector clustering set, that is, determine the probability of belonging the storage item member to the corresponding vector clustering set.

[0088] Finally, for each storage item member, based on the difference information between the storage matching parameters of the storage item member with each set center respectively and the assignment probability of the storage item member to the corresponding vector clustering set, determine the third local update error of the storage item member. In this way, the third local update error can be used to reflect the accuracy of the clustering result. Specifically, the ratio of the assignment probability to the storage matching parameter can be calculated, then, the logarithm operation is performed on the ratio, and, the product of the assignment probability and the result of the logarithm operation is calculated. Finally, the sum of the corresponding products can be calculated to obtain the corresponding third local update error.

[0089] It can be understood that in the above steps, the specific manner of determining the allocation probability is not limited. For example, in an alternative implementation manner, it can be determined based on the following steps:

[0090] First, for each set center, calculate the sum value of the warehousing matching parameters corresponding to each training warehousing item member for this set center, and obtain the clustering characterization parameter of the vector clustering set corresponding to this set center;

[0091] Second, for each warehousing item member, based on the warehousing matching parameter between this warehousing item member and a set center, and in combination with the clustering characterization parameter of the vector clustering set corresponding to this set center, determine the set contribution degree of this warehousing item member in this vector clustering set; Exemplarily, the ratio between the square of the warehousing matching parameter and the clustering characterization parameter can be calculated to obtain the set contribution degree. In addition, the set center refers to the center formed by clustering, and can also be called the clustering center;

[0092] Then, for each warehousing item member, calculate the ratio between the set contribution degree of this warehousing item member in a vector clustering set and the sum value of the set contribution degrees of this warehousing item member in all vector clustering sets, and obtain the allocation probability of this warehousing item member being assigned to this vector clustering set.

[0093] It can be understood that in the above steps, the specific manner of performing the update operation is not limited. For example, in an alternative implementation manner, in order to ensure the reliability of the update operation, each update operation may further include the following steps:

[0094] First, for each warehousing item member, determine the target set center as the set center with the maximum warehousing matching parameter corresponding to this warehousing item member, and mark the vector clustering set corresponding to this target set center, so that this vector clustering set is marked as the vector clustering set corresponding to this warehousing item member in the next update operation; that is to say, for each warehousing item member, by counting the warehousing matching parameters of this warehousing item member corresponding to each set center, the set center with the maximum warehousing matching parameter is used as the target set center, and further the vector clustering set corresponding to the target set center is used as the vector clustering set to which the warehousing item member belongs in the next update operation. For example, if there are 3 vector clustering sets, and the warehousing matching parameters of the warehousing item member with the respective set centers of the 3 vector clustering sets are 98%, 70%, and 60% respectively, then the vector clustering set corresponding to the warehousing matching parameter 98% with the maximum value is used as the vector clustering set of the warehousing item member in the next update operation;

[0095] Secondly, for each vector clustering set corresponding to the next update operation, based on the target update error, adjust the set center semantic vector of the set center corresponding to this vector clustering set. That is to say, since the initial set center semantic vector is preset, in each subsequent update operation, the set center semantic vector needs to be adjusted accordingly to achieve the update. That is to say, the set center semantic vector can be used as a network parameter of the candidate data mining network and is updated as the candidate data mining network is updated. The specific update process will not be elaborated here one by one, that is, update and adjust the network parameters along the direction of reducing the target update error.

[0096] It can be understood that in the above steps, the specific method for determining the first local update error is not limited. For example, in an alternative implementation, in order to improve the reliability of the determined first local update error, the first local update error corresponding to the target storage item member can be determined based on the following steps:

[0097] First, based on the vector difference between each set center and the output semantic vector of the target storage item member in the current update operation, perform an aggregation calculation on the set center semantic vector of each set center to output the corresponding first aggregated semantic vector, and perform a semantic enhancement operation on the output semantic vector based on the first aggregated semantic vector to output the output semantic vector after the semantic enhancement operation. In addition, mark the output semantic vector after the semantic enhancement operation so that it is marked as the storage item enhancement vector of the target storage item member in the current update operation; Exemplarily, the output semantic vector can be multiplied by the transposed vector of the set center semantic vector, and then the result of the multiplication is divided by the 0.5th power of the dimension number of the output semantic vector to obtain a weighting parameter. Then, multiply the weighting parameter by the set center semantic vector and perform a normalization process on the result of the multiplication to obtain the first aggregated semantic vector. It should be noted that the dimension number of the output semantic vector is equal to the dimension number of the set center semantic vector; In addition, the first aggregated semantic vector and the output semantic vector can be multiplied to implement the semantic enhancement operation on the output semantic vector, so that the output semantic vector after the semantic enhancement operation also carries the semantic information carried by the first aggregated semantic vector;

[0098] Secondly, based on the vector difference between each set center and the first semantic vector of the target warehousing item member in the current update operation, the set center semantic vectors of each set center are aggregated and calculated to output corresponding second aggregated semantic vectors. Then, a semantic enhancement operation is performed on the first semantic vector based on the second aggregated semantic vector to output the first semantic vector after the semantic enhancement operation. Moreover, the candidate semantic conversion network is used to perform a conversion operation on the first semantic vector after the semantic enhancement operation to output the relationship network member conversion vector of the target warehousing item member in the current update operation. It should be noted that the candidate semantic conversion network may include a parameter matrix, and the first semantic vector after the semantic enhancement operation can be multiplied by the parameter matrix to achieve the conversion operation and obtain the corresponding relationship network member conversion vector;

[0099] Then, based on the vector difference between the warehousing item enhancement vector and the relationship network member conversion vector of the target warehousing item member in the current update operation, the first local update error corresponding to the target warehousing item member is determined. For example, the vector distance between the warehousing item enhancement vector and the relationship network member conversion vector can be used as the first local update error. Based on this, the first local update error can not only consider the error in the conversion output of the first semantic vector of the target warehousing item member, but also take into account the clustering loss in the clustering of the target warehousing item member. Thus, the relationship network member semantic vector obtained can also express more information.

[0100] It can be understood that in an alternative implementation, the multiple warehousing item members include at least one non-related warehousing item member. The non-related warehousing item member refers to a warehousing item member that does not have a relationship line segment with the warehousing plan member. At least one of the training relationship network members also includes non-related warehousing item members. The member links of the non-related warehousing item members all belong to the first target member link, and the relationship network member semantic vectors of the non-related warehousing item members belong to the first semantic vector. Based on this, the first semantic vector or the second semantic vector corresponding to each training relationship network member can be determined according to the following steps:

[0101] First, in each member link where the training relationship network member belongs to the starting relationship network member, each comparison member link is determined, where the comparison member link belongs to the first target member link or the second target member link, that is, either all are the first target member links or all are the second target member links;

[0102] Secondly, for each comparison member link, starting from the relationship network member with the second-lowest relevance in the comparison member link (relevance can refer to the number of relationship network members separated from the starting relationship network member on the corresponding comparison member link. The larger this number, the lower the relevance. For example, the relationship network member with the second-lowest relevance refers to the relationship network member with the second-largest number of relationship network members separated from the starting relationship network member), based on the member transfer vectors of each relationship network member with lower relevance connected to the relationship network member (i.e., the relationship network member with the largest number of relationship network members separated from the starting relationship network member) and the member description data corresponding to the relationship network member, sequentially transfer to the starting relationship network member to determine the member transfer vector that the starting relationship network member has on the comparison member link. Among them, the member transfer vector of the relationship network member with the lowest relevance is mined and formed based on the member description data corresponding to the relationship network member with the lowest relevance. Specifically, if the member description belongs to a warehousing item image, the convolutional model included in the candidate data mining network can be used to perform convolutional processing on the warehousing item image to obtain the corresponding convolutional vector as the corresponding member transfer vector; if the member description data belongs to a warehousing plan text, the word embedding model included in the candidate data mining network can be used to perform word embedding processing on the warehousing plan text to obtain the corresponding word embedding vector as the corresponding member transfer vector. For example, if the warehousing plan text is "The storage conditions include a storage temperature below 0 degrees...", in this way, "storage conditions" can be embedded as:

[0103] “0.12, −0.45, 0.23, 0.56, −0.11, 0.34, −0.25, 0.67, −0.18, 0.39, 0.44, −0.30, 0.29, 0.10, −0.55, 0.26, 0.12, 0.04, 0.49, −0.14, 0.58, −0.39, 0.33, 0.50, −0.21, 0.19, 0.11, −0.60, 0.47, 0.23, −0.01, 0.35, 0.18, −0.22, 0.15, 0.09, −0.28, 0.45, 0.07, −0.36, 0.53, −0.08, 0.40, 0.22, 0.01, −0.27...”

[0104] "include" can be embedded as:

[0105] “0.05, 0.11, -0.10, 0.30, 0.21, -0.15, 0.09, 0.22, 0.18, -0.12, 0.25, -0.19, 0.30, -0.23, 0.16, 0.21, 0.05, 0.03, -0.10, 0.12, 0.40, -0.07, 0.25, 0.04, 0.14, 0.05, -0.18, 0.20, 0.11, -0.12, 0.08, 0.03, 0.01, 0.14, 0.03, 0.09, -0.04, 0.22, 0.06, 0.15, 0.02, -0.10, 0.25, 0.09, -0.12...”;

[0106] “Below 0 degrees” can be embedded as:

[0107] “0.40, 0.12, -0.11, 0.20, 0.03, -0.30, 0.50, -0.09, 0.31, 0.25, -0.05, 0.09, 0.22, 0.14, -0.01, 0.18, 0.29, -0.15, 0.24, 0.35, 0.02, -0.17, 0.05, 0.03, 0.11, 0.10, 0.15, -0.08, 0.22, 0.12, -0.13, 0.07, 0.10, 0.14, 0.08, 0.20, 0.25, 0.17, -0.19, 0.21, 0.05, -0.03, 0.16, 0.09, -0.25...”;

[0108] “Storage temperature” can be embedded as:

[0109] “0.60, 0.20, -0.15, 0.25, 0.30, -0.20, 0.15, 0.40, -0.12, 0.35, 0.02, 0.10, 0.21, 0.15, -0.05, 0.14, 0.18, 0.09, -0.11, 0.24, 0.30, -0.08, 0.20, 0.12, 0.03, -0.10, 0.14, 0.19, 0.25, -0.01, 0.18, 0.11, 0.05, 0.30, 0.02, -0.16, 0.13, 0.01, 0.10, 0.17, -0.12, 0.05, 0.21, 0.07, -0.09...”;

[0110] Then, perform an associated focus fusion operation on the member transfer vectors of the training relationship network members on each comparison member link, and output the semantic vectors corresponding to the comparison member links of the training relationship network members. Among them, when the comparison member link belongs to the first target member link, the semantic vector corresponding to the comparison member link belongs to the first semantic vector; when the comparison member link belongs to the second target member link, the semantic vector corresponding to the comparison member link belongs to the second semantic vector. Among them, the specific processing process of the associated focus fusion operation may include: First, pairwise combine the member transfer vectors of the training relationship network members on each comparison member link. Each combination includes a first vector and a second vector. Then, multiply the first vector and the second vector to obtain the corresponding similarity parameter, and then weight the second vector based on the similarity parameter to obtain the corresponding associated vector. Finally, superimpose, splice, or calculate the mean value of the associated vectors corresponding to each combination to obtain the corresponding semantic vector to achieve the associated focus fusion.

[0111] Among them, it can be understood that in the above steps, the specific manner of sequentially transmitting to the starting relationship network member is not limited. For example, in an alternative implementation manner, it can be sequentially transmitted to the starting relationship network member based on the following steps:

[0112] First, for any relationship network member in the current transmission stage, based on the member description data corresponding to the relationship network member, mine out the member representation vector corresponding to the relationship network member, such as the convolution vector and the embedding vector as described above.

[0113] Second, for any relationship network member in the current transmission stage, mark each relationship network member with a smaller correlation that has a relationship segment with the relationship network member, so that it is marked as a comparison relationship network member (such as a relationship network member farther from the starting relationship network member on the comparison member link), and determine the member transfer vector corresponding to the comparison relationship network member.

[0114] Then, for any relationship network member in the current transmission stage, based on the member representation vector corresponding to the relationship network member and the member transfer vector corresponding to each comparison relationship network member, determine the member association parameter between the relationship network member and each comparison relationship network member, such as the cosine similarity between the member representation vector and the member transfer vector.

[0115] Finally, for any member of the current transmission stage in the relationship network, based on the member association parameters corresponding to each of the comparison relationship network members, calculate the weighted sum of the member transmission vectors corresponding to each of the comparison relationship network members, and, fuse the vector obtained from the weighted sum calculation and the member representation vector corresponding to the relationship network member, and output the member transmission vector corresponding to the relationship network member; exemplarily, the member association parameters can be used as weight coefficients to calculate the weighted sum of each member transmission vector, and, superimpose the result of the weighted sum calculation on the member representation vector, and then normalization processing can be performed to obtain the corresponding member transmission vector.

[0116] To facilitate the understanding of the above transmission process, in combination with Figure 5 , relationship network member 1 is used as the starting relationship network member, and the corresponding relationship network member link also includes relationship network member 2, relationship network member 3, relationship network member 4, and relationship network member 5. First, for the branch of relationship network member 2 and relationship network member 3, the member representation vector of relationship network member 3 can be used as the corresponding member transmission vector, and then, based on the member transmission vector corresponding to relationship network member 3 and the member representation vector of relationship network member 2, determine the member transmission vector corresponding to relationship network member 2, and, based on the member transmission vector corresponding to relationship network member 2 and the member representation vector of relationship network member 1, determine the member transmission vector that relationship network member 1 has on this branch. For the branch of relationship network member 4 and relationship network member 5, the member representation vector of relationship network member 5 can be used as the corresponding member transmission vector, and then, based on the member transmission vector corresponding to relationship network member 5 and the member representation vector of relationship network member 4, determine the member transmission vector corresponding to relationship network member 4, and, based on the member transmission vector corresponding to relationship network member 4 and the member representation vector of relationship network member 1, determine the member transmission vector that relationship network member 1 has on this branch. Then, the two member transmission vectors that relationship network member 1 has on these two branches can be subjected to an associated focus fusion operation to output the semantic vector corresponding to the comparison member link of the training relationship network member, that is, obtain the first semantic vector or the second semantic vector.

[0117] In addition, the target warehousing scheme relationship network can be formed on the basis of the training warehousing scheme relationship network. For example, by adding a specified warehousing item member to the training warehousing scheme relationship network, the corresponding target warehousing scheme relationship network can be obtained, and then, based on the finally determined target warehousing scheme, update the target warehousing scheme relationship network, so as to form a new target warehousing scheme relationship network. In this way, through the continuous application and update of the target warehousing scheme relationship network, the information included in the target warehousing scheme relationship network can be gradually enriched to ensure its reliability.

[0118] In summary, for the integrated management platform system of warehousing applications based on AI cloud computing provided by the present invention, first, a target warehousing solution relationship network is determined; second, at least one first member link with a specified warehousing item member as the starting relationship network member is extracted from the target warehousing solution relationship network, and at least one second member link with a warehousing solution member reflected by a pending warehousing solution as the starting relationship network member is extracted; then, the relationship network member semantic vector corresponding to the specified warehousing item member is mined, and the relationship network member semantic vector corresponding to each pending warehousing solution is mined through network mining; finally, a target warehousing solution is determined based on the relationship network member semantic vectors corresponding to the specified warehousing item member and each pending warehousing solution. Based on the above, since the relationship network member semantic vector corresponding to the specified warehousing item member not only includes the semantic information of the member description data corresponding to the specified warehousing item member itself, but also includes the semantic information of the member description data corresponding to the relationship network members on the first member link, and the relationship network member semantic vector corresponding to the pending warehousing solution not only includes the semantic information of the member description data corresponding to the pending warehousing solution itself, but also includes the semantic information of the member description data corresponding to the relationship network members on the second member link, the semantic representation ability of the mined relationship network member semantic vector is better and the semantic information is more abundant. Therefore, the solution determination based on the relationship network member semantic vector can be more reliable, that is, it ensures that the obtained target warehousing solution has a high reliability, thereby improving the problem of relatively low reliability of warehousing management existing in the prior art.

[0119] In several embodiments provided by the embodiments of the present invention, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of devices, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0120] In addition, in each embodiment of the present invention, each functional module may be integrated together to form an independent part, or each module may exist alone, or two or more modules may be integrated to form an independent part.

[0121] If the above-mentioned function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, an electronic device, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes. It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.

[0122] The foregoing are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A warehousing application integrated management platform system based on AI cloud computing, characterized in that: The system comprises a plurality of warehouse front-end devices and a back-end cloud computing device in communication with the plurality of warehouse front-end devices, wherein the back-end cloud computing device is used to execute a warehouse application comprehensive management method based on AI cloud computing, wherein the warehouse application comprehensive management method based on AI cloud computing comprises: Determine a target storage solution relationship network, and determine member description data corresponding to each relationship network member in the target storage solution relationship network, wherein the target storage solution relationship network includes multiple relationship network members, and relationship line segments are configured in the target storage solution relationship network for two relationship network members that are determined to have a relevant relationship, the multiple relationship network members include multiple storage item members reflecting storage items and multiple storage solution members reflecting pending storage solutions, and the multiple storage item members include designated storage item members reflecting designated storage items, the member description data corresponding to the storage item members are formed based on the storage front-end device collecting information on the storage items, and the member description data corresponding to the storage solution members are used to describe the storage solution; Extracting at least one first member link having the designated storage item member as a starting network member from the target storage solution relationship network, and extracting at least one second member link having the storage solution member reflected by the pending storage solution as a starting network member; Loading member description data corresponding to each first member link corresponding to the designated storage item member and each network member on the first member link, so that the data mining network mines out the network member semantic vector corresponding to the designated storage item member, and loading member description data corresponding to each second member link corresponding to each pending storage plan and each network member on the second member link, so that the data mining network mines out the network member semantic vector corresponding to each pending storage plan; Based on the semantic vector of the relationship network member corresponding to the designated storage item member and the semantic vector of the relationship network member corresponding to each of the pending storage solutions, a target storage solution corresponding to the designated storage item is determined from multiple pending storage solutions, wherein the target storage solution serves as a basis for storing the designated storage item, the target storage solution is determined by a semantic conversion network, the data mining network and the semantic conversion network are formed by updating the candidate data mining network and the candidate semantic conversion network, the updated error includes a first local update error, the first local update error is used to reflect the difference information between the output semantic vector of the target storage item member and the relationship network member conversion vector output by converting the first semantic vector corresponding to the target storage item member using the candidate semantic conversion network, the output semantic vector belongs to For the semantic vector of the relationship network member or the second semantic vector, the candidate data mining network is used to mine the semantic vector of the relationship network member corresponding to the training relationship network member, and when each member link of the training relationship network member includes at least one first target member link, the relationship network member semantic vector includes the first semantic vector mined based on the member description data corresponding to the relationship network member on each first target member link, and when each member link of the training relationship network member includes at least one second target member link, the relationship network member semantic vector includes the second semantic vector mined based on the member description data corresponding to the relationship network member on each second target member link, and the member links connecting the relationship network members that all belong to the storage item members are marked as the first target member links, and the member links that all belong to the storage solution members are marked as the second target member links.

2. The AI ​​cloud computing-based warehousing application integrated management platform system according to claim 1, characterized in that: The step of determining a target storage solution corresponding to the designated storage item from among a plurality of the pending storage solutions based on the relationship network member semantic vector corresponding to the designated storage item member and the relationship network member semantic vector corresponding to each of the pending storage solutions comprises: When all the relationship network members connected by the relationship line segments of the designated storage item member belong to the storage item members, each first member link of the designated storage item is marked as a first target member link, and the relationship network member semantic vector corresponding to the designated storage item member is converted by using a semantic conversion network, and the corresponding relationship network member conversion vector is output, and for each of the pending storage solutions, based on the relationship network member conversion vector and the relationship network member semantic vector corresponding to the pending storage solution, the storage matching parameters of the designated storage item and the pending storage solution are determined; When the relationship network members connected by the designated storage item member through the relationship line segment include storage solution members and storage item members, each first member link of the designated storage item is marked as including a first target member link and a second target member link, and for each pending storage solution, based on the relationship network member semantic vectors corresponding to the designated storage item and the pending storage solution, the storage matching parameters between the designated storage item and the pending storage solution are determined; Based on the storage matching parameters corresponding to each of the pending storage solutions, a target storage solution corresponding to the designated storage items is analyzed in each of the pending storage solutions.

3. The AI ​​cloud computing-based warehousing application integrated management platform system according to claim 2, characterized in that: The AI ​​cloud computing-based warehousing application comprehensive management method also includes: Determine a training storage solution relationship network, and determine member description data corresponding to each relationship network member in the training storage solution relationship network, wherein the training storage solution relationship network includes a plurality of relationship network members, and relationship line segments are configured in the storage solution relationship network for two relationship network members that are determined to have a relevant relationship, the plurality of relationship network members include a plurality of storage item members reflecting storage items and a plurality of storage solution members reflecting pending storage solutions, the plurality of storage item members include at least one target storage item member, and the target storage item member is connected to at least one storage solution member and at least one other storage item member through relationship line segments respectively; In the training storage solution relationship network, at least one training relationship network member is determined, and for each training relationship network member, at least one member link that uses the training relationship network member as a starting relationship network member is extracted, wherein a member link that connects relationship network members that are all storage item members is marked as a first target member link, and a member link that connects relationship network members that are all storage solution members is marked as a second target member link, the connecting relationship network member refers to a relationship network member that has a relationship line segment with the starting relationship network member, the at least one training relationship network member includes at least one target storage item member, and each member link of the target storage item member includes at least one first target member link and at least one second target member link; Based on the member description data corresponding to the relationship network member on each member link of each training relationship network member, the candidate data mining network and the candidate semantic conversion network are updated at least once, and the corresponding data mining network and semantic conversion network are output, wherein the target update error of the update operation includes a first local update error corresponding to each target storage item member, and the first local update error is used to reflect the difference between the semantic vector output by the corresponding target storage item member in each update operation and the relationship network member conversion vector output by using the candidate semantic conversion network based on the first semantic vector corresponding to the corresponding target storage item member in each update operation. The output semantic vector belongs to the relationship network member semantic The candidate data mining network is used to load the member description data corresponding to the network member on each member link of the training network member, and mine the network member semantic vector corresponding to the training network member, and when each member link of the training network member includes at least one first target member link, the network member semantic vector includes the first semantic vector mined based on the member description data corresponding to the network member on each first target member link, and when each member link of the training network member includes at least one second target member link, the network member semantic vector includes the second semantic vector mined based on the member description data corresponding to the network member on each second target member link.

4. The AI ​​cloud computing-based warehousing application integrated management platform system according to claim 3 is characterized in that: The at least one training relationship network member further includes a storage solution member, and the target update error further includes a second local update error corresponding to each target storage item member, wherein in each update operation, the second local update error corresponding to each target storage item member is determined based on the following steps: In the training storage solution relationship network, storage solution members connected to the target storage item member through relationship line segments are determined and marked as related storage solution members, and storage solution members not connected to the target storage item member through relationship line segments are determined in the training storage solution relationship network and marked as non-related storage solution members; Polling each relevant storage solution member and each non-relevant storage solution member to form a corresponding plurality of storage solution member combinations, wherein each of the storage solution member combinations includes a relevant storage solution member and a non-relevant storage solution member; For each storage solution member combination, based on the relationship network member semantic vectors of the target storage item member and the related storage solution members in the storage solution member combination in the current update operation, determine the first storage matching parameter of the target storage item member and the related storage solution members, and based on the relationship network member semantic vectors of the target storage item member and the non-related storage solution members in the storage solution member combination in the current update operation, determine the second storage matching parameter of the target storage item member and the non-related storage solution members, and, based on the difference information between the first storage matching parameter and the second storage matching parameter, determine the storage solution matching sub-error of the target storage item member based on the storage solution member combination; Based on the storage solution matching sub-error of the target storage item member to all storage solution member combinations, the second local update error of the target storage item member is determined.

5. The AI ​​cloud computing-based warehousing application integrated management platform system according to claim 3, characterized in that: The target update error also includes a third local update error corresponding to each training relationship network member belonging to the storage item member, wherein the third local update error is determined based on the following steps: For each update operation, determine the set center semantic vector of the set center of each vector cluster set in the current update operation and the relationship network member semantic vector of each storage item member in the current update operation, wherein the set center semantic vector of the set center of each vector cluster set when clustering for the first time belongs to the pre-configured reference vector; For each storage item member, based on the relationship network member semantic vector of the storage item member in the current update operation and the collection center semantic vector of each collection center, determine the storage matching parameters that the storage item member has with each collection center respectively; Based on the storage matching parameters of each storage item member and each collection center, the probability of each storage item member being assigned to each vector clustering set is determined; For each storage item member, the third local update error of the storage item member is determined based on the difference information between the storage matching parameters respectively possessed by the storage item member and each collection center and the allocation probability of the storage item member being allocated to the corresponding vector clustering set.

6. The AI ​​cloud computing-based warehousing application integrated management platform system according to claim 5, characterized in that: The step of determining the probability of each storage item member being assigned to each vector clustering set based on the storage matching parameters respectively possessed by each storage item member and each collection center comprises: For each collection center, calculate the sum of the storage matching parameters of each training storage item member corresponding to the collection center, and obtain the clustering representation parameters of the vector clustering set corresponding to the collection center; For each storage item member, based on the storage matching parameters between the storage item member and a collection center, and in combination with the cluster representation parameters of the vector cluster set corresponding to the collection center, determine the collection contribution of the storage item member in the vector cluster set; For each storage item member, the ratio between the set contribution of the storage item member in a vector cluster set and the set contribution of the storage item member in all vector cluster sets is calculated to obtain the allocation probability of the storage item member to the vector cluster set.

7. The AI ​​cloud computing-based warehousing application integrated management platform system according to claim 5, characterized in that: Each update operation also includes: For each storage item member, the set center of the storage matching parameter with the maximum value corresponding to the storage item member is determined as the target set center, and the vector cluster set corresponding to the target set center is marked, so that the vector cluster set is marked as the vector cluster set corresponding to the storage item member in the next update operation; For each vector cluster set corresponding to the next update operation, based on the target update error, the set center semantic vector of the set center corresponding to the vector cluster set is adjusted.

8. The AI ​​cloud computing-based warehousing application integrated management platform system according to claim 5, characterized in that: The first local update error corresponding to the target storage item member is determined based on the following steps: Based on the vector difference between the output semantic vectors of each collection center and the target stored item member in the current update operation, the collection center semantic vectors of each collection center are aggregated and calculated, and the corresponding first aggregated semantic vector is output, and based on the first aggregated semantic vector, a semantic enhancement operation is performed on the output semantic vector, and the output semantic vector after the semantic enhancement operation is output, and the output semantic vector after the semantic enhancement operation is marked so that it is marked as the storage item enhancement vector of the target stored item member in the current update operation; Based on the vector difference between the first semantic vectors of each collection center and the target storage item member in the current update operation, the collection center semantic vector of each collection center is aggregated and calculated, and the corresponding second aggregate semantic vector is output, and the first semantic vector is semantically enhanced based on the second aggregate semantic vector, and the first semantic vector after the semantic enhancement operation is output, and the first semantic vector after the semantic enhancement operation is converted using the candidate semantic conversion network, and the relationship network member conversion vector of the target storage item member in the current update operation is output; Based on the vector difference between the storage item reinforcement vector and the relationship network member conversion vector of the target storage item member in the current update operation, a first local update error corresponding to the target storage item member is determined.

9. The AI ​​cloud computing-based warehousing application integrated management platform system according to claim 3, characterized in that: The plurality of storage item members include at least one non-related storage item member, wherein the non-related storage item member refers to a storage item member having no relationship line segment with a storage solution member; At least one training relationship network member also includes a non-related storage item member, the member links of the non-related storage item member all belong to the first target member link, and the relationship network member semantic vector of the non-related storage item member belongs to the first semantic vector.

10. The AI ​​cloud computing-based warehousing application integrated management platform system according to claim 9, characterized in that: The first semantic vector or the second semantic vector corresponding to each of the training relationship network members is determined based on the following steps: In each member link of the training relationship network member belonging to the initial relationship network member, each comparison member link is determined, wherein the comparison member link belongs to the first target member link or the second target member link; For each comparison member link, starting from the penultimate network member with the least correlation in the comparison member link, based on the member transfer vectors of each network member with less correlation connected to the network member and the member description data corresponding to the network member, the member transfer vectors are sequentially transferred to the starting network member to determine the member transfer vector of the starting network member on the comparison member link, wherein the member transfer vector of the network member with the least correlation is formed based on the member description data corresponding to the network member with the least correlation; The member transfer vectors of the training relationship network members on each comparison member link are subjected to an associative focusing fusion operation, and a semantic vector corresponding to the comparison member link of the training relationship network member is output, wherein when the comparison member link belongs to the first target member link, the semantic vector corresponding to the comparison member link belongs to the first semantic vector, and when the comparison member link belongs to the second target member link, the semantic vector corresponding to the comparison member link belongs to the second semantic vector.

Images