Disclosure of Invention
An object of the present disclosure is to provide a method and an apparatus for auxiliary analysis of storage, which overcome one or more of the problems due to the limitations and disadvantages of the related art, at least to some extent.
According to one aspect of the present disclosure, there is provided a method for auxiliary analysis of storage sub-warehouse, comprising:
obtaining order information of an area in a historical time period and obtaining single item information in the historical time period from the order information;
establishing a historical three-dimensional incidence matrix, wherein any point H of the historical three-dimensional incidence matrixi(xi,yi,zi) For indicating singlets xiAnd singleton yiAt historical time ziThe correlation of (A) is gi(ii) a Wherein, giIs based on said singlets xiAnd singleton yiAt historical time ziCalculating the single item information;
fitting the time series of each relevance of any two single products, and calculating a future three-dimensional relevance matrix in a future time period according to a fitting result;
adding the future three-dimensional incidence matrix on the historical three-dimensional incidence matrix, and fusing the added three-dimensional incidence matrix on a time dimension to obtain a two-dimensional incidence matrix;
performing spectral clustering on the two-dimensional incidence matrix, and performing graph segmentation on the two-dimensional incidence matrix by taking the minimum relevance between the segmented subgraphs as a target according to a spectral clustering result;
and distributing the single product set to each warehouse according to the graph segmentation result.
In an exemplary embodiment of the present disclosure, the method further comprises:
and performing outlier removal operation on the time series of the relevance through the first moment and the second moment of the time series of the relevance.
In an exemplary embodiment of the present disclosure, the method further comprises:
and further clustering the two-dimensional incidence matrix according to the spectral clustering result, and further evaluating the clustering result.
In an exemplary embodiment of the disclosure, fitting the time series of each correlation of any two of the singlets includes:
wherein x isiIs a value in time series, wiT is the weight of the time sequence, and T is the size of the time window.
In an exemplary embodiment of the present disclosure, fusing the added three-dimensional correlation matrix in the time dimension to obtain a two-dimensional correlation matrix includes:
wherein n is the length of the time series, H
*(x
i,y
i) For any point on the two-dimensional correlation matrix,
is any point on the future three-dimensional correlation matrix.
In an exemplary embodiment of the present disclosure, spectrally clustering the two-dimensional incidence matrix comprises:
establishing a similar graph, wherein a weighted adjacent matrix of the similar graph is W, wherein W is H, and H is the two-dimensional incidence matrix;
calculating a Laplace matrix L, wherein
w
ijJ is any point in the weighted adjacent matrix W, and j is the number of columns in the weighted adjacent matrix W;
decomposing the Laplace matrix L ═ UΛ U-1Wherein U is [ U ═ U1,u2,...,ur],
[u1,u2,...,ur]A characteristic vector value of L, λiIs a characteristic value of L, and in Λ, λ1≤λ2≤...≤λrR is the rank of W;
selecting the first k eigenvector values of the Laplace matrix L to form an r x k matrix, taking each row in the matrix as one vector in a k-dimensional space, and clustering the eigenvectors by using a clustering algorithm, wherein k is the pre-clustering number of spectral clustering.
In an exemplary embodiment of the present disclosure, further clustering the two-dimensional incidence matrix according to the spectral clustering result includes:
wherein S (k, m) is the number of schemes distributed to the single item set, C is a binomial coefficient, m is the number of warehouses needing to participate in calculation, k is the pre-clustering number of spectral clustering, and m is less than k.
According to another aspect of the present disclosure, there is provided a device for auxiliary analysis of warehouses, including:
an information acquisition module: the system comprises a data processing system and a data processing system, wherein the data processing system is used for acquiring order information of an area in a historical time period and acquiring single item information in the historical time period from the order information;
a first incidence matrix module: is used for establishing a historical three-dimensional incidence matrix, and any point H of the historical three-dimensional incidence matrixi(xi,yi,zi) For indicating singlets xiAnd singleton yiAt historical time ziThe correlation of (A) is gi(ii) a Wherein, giIs based on said singlets xiAnd singleton yiAt historical time ziCalculating the single item information;
a fitting module: fitting a time series of each said association for any two of said singlets;
a second incidence matrix module: the three-dimensional incidence matrix calculation module is used for calculating a future three-dimensional incidence matrix in a future time period according to the fitting result;
a third correlation matrix module: for adding the future three-dimensional correlation matrix to the historical three-dimensional correlation matrix;
a fusion module: the system is used for fusing the added three-dimensional incidence matrix on a time dimension to obtain a two-dimensional incidence matrix;
a spectral clustering module: the system is used for performing spectral clustering on the two-dimensional incidence matrix;
a graph partitioning module: the two-dimensional incidence matrix is used for carrying out graph segmentation on the two-dimensional incidence matrix according to a spectral clustering result by taking the minimum relevance between the segmented subgraphs as a target;
a unit dispensing module: the system is used for distributing the single product set to each warehouse according to the result of the graph segmentation.
In an exemplary embodiment of the present disclosure, the apparatus further includes:
outlier removal module: for performing a de-outlier operation on the time-series of correlations by the first and second moments of the time-series of correlations.
In an exemplary embodiment of the present disclosure, the apparatus further includes:
the first clustering module: the two-dimensional incidence matrix is used for further clustering according to the spectral clustering result;
an evaluation module: for further evaluation of the clustering results.
In an exemplary embodiment of the present disclosure, fitting the time series of each correlation of any two of the singlets includes:
wherein x isiIs a value in time series, wiT is the weight of the time sequence, and T is the size of the time window.
In an exemplary embodiment of the present disclosure, the added three-dimensional correlation matrix is fused in the time dimension to obtain a two-dimensional correlation matrix, where the fusing includes:
wherein n is the length of the time series, H
*(x
i,y
i) For any point on the two-dimensional correlation matrix,
is any point on the future three-dimensional correlation matrix.
In an exemplary embodiment of the present disclosure, the spectral clustering module includes:
the similar graph establishing module: the method comprises the steps of establishing a similarity graph, wherein a weighted adjacency matrix of the similarity graph is W, wherein W is H, and H is the two-dimensional incidence matrix;
a Laplace matrix calculation module: for calculating the laplacian matrix L, wherein,
w
ijj is any point in the weighted adjacent matrix W, and j is the number of columns in the weighted adjacent matrix W;
a Laplace matrix decomposition module: for decomposing the Laplace matrix L ═ UΛ U-1Wherein U is [ U ═ U1,u2,...,ur],
[u1,u2,...,ur]A characteristic vector value of L, λiIs a characteristic value of L, and in Λ, λ1≤λ2≤...≤λrR is the rank of W;
a second type of module: and the method is used for selecting the first k characteristic vector values of the Laplace matrix L to form an r x k matrix, taking each row in the matrix as a vector in a k-dimensional space, and clustering the characteristic vectors by using a clustering algorithm, wherein k is the pre-clustering number of the spectral clustering.
In an exemplary embodiment of the present disclosure, the two-dimensional incidence matrix is further clustered according to the spectral clustering result, and the clustering includes:
wherein S (k, m) is the number of schemes distributed to the single item set, C is a binomial coefficient, m is the number of warehouses needing to participate in calculation, k is the pre-clustering number of spectral clustering, and m is less than k.
In the warehousing warehouse-by-warehouse auxiliary analysis method and device of the exemplary embodiment, on one hand, the order information of the user is abstracted into the commodity incidence matrix, so that the calculation of commodity distribution is completely based on the incidence matrix, the user order is not required to be traversed, and the calculation efficiency is improved; on the other hand, the incidence matrix in a short period can be predicted through analysis of the time sequence, so that the bill splitting condition after commodity warehouse splitting can be predicted, an enterprise is assisted to realize better warehouse splitting, and the cost brought to enterprise operation by high bill splitting rate can be reduced to a great extent.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Detailed Description
Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many different forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the disclosure. One skilled in the relevant art will recognize, however, that the subject matter of the present disclosure can be practiced without one or more of the specific details, or with other methods, components, devices, steps, and the like. In other instances, well-known technical solutions have not been shown or described in detail to avoid obscuring aspects of the present disclosure.
Furthermore, the drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus their repetitive description will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor devices and/or microcontroller devices.
In the present exemplary embodiment, a warehouse and warehouse separation auxiliary analysis method is provided, where the warehouse and warehouse separation auxiliary analysis method may start from data analysis of a customer order, start from a purchasing behavior of a user for different regions and different time periods, analyze a purchasing relevance of a commodity, and split n warehouses into m warehouses by planning a commodity layout. Referring to fig. 1, the warehouse sub-warehouse auxiliary analysis method may include the following steps:
in step S110, order information of a region in a historical time period is obtained and item information in the historical time period is obtained from the order information. As shown in fig. 2, step S110 in the present exemplary embodiment may include, for example, steps S112 to S116 described below. Wherein:
in step S112, order information within a period of time is acquired from the hadoop cluster.
The Hadoop cluster is a distributed system infrastructure developed by the Apache foundation, and a user can develop a distributed program without knowing details of a distributed bottom layer, and fully utilizes the advantages of the cluster to carry out high-speed operation and storage. In this example embodiment, the order information obtained from the hadoop cluster may include: parent order number, child order number, purchase order (sku or class) information, warehouse information (warehouse information includes warehouse name, the area and the time of delivery), order placing time, order valid identifier, etc., but the disclosure is not limited thereto. In addition, in other exemplary embodiments of the present disclosure, the order information may also be obtained in other manners according to circumstances, and this is not particularly limited in this exemplary embodiment.
In step S114, the order information is sorted.
In this example embodiment, the sorting of the order information may include, for example, an invalid order exclusion, an order type identifier, a primary attribute information identifier, and the like, and the disclosure is not limited thereto. In this exemplary embodiment, step S114 can implement: filtering out invalid orders (order valid identification is invalid); filtering out orders without warehouse information; identifying the order according to the order service type: such as may be identified as: self-operated orders, third-party orders, self-operated and third-party mixed orders and the like, wherein the order service type can be deduced from the order number, or order service type identification is added into the order information; and adding attribute information to the order: for example, attribute information can be added to distinguish cross-regional orders from cross-product orders.
In step S116, the individual information is collated.
In this exemplary embodiment, the singleton itemiMay be the goods in the order, which may be sku (store ordering unit) or categories, etc.; sku is the minimum storage unit in warehouse management, and the categories can be divided into multiple stages according to different services: for example Yanjing beer and Qingdao beer, the two single products are calculated according to sku, but the two single products are both beer according to three grades and are single products; for another example, Yanjing beer, Qingdao beer and Feitian Maotai, three single products are provided according to sku, but the three-grade product is two single products of beer and white spirit, and one single product is provided according to the second-grade product. And selecting the processing granularity (sku or class) according to the actual application requirement when the single-product information is arranged.
After obtaining the order information of each region, in this exemplary embodiment, the order information of the region may be combined into one item set I for each region, where I includes all items appearing in the order in the region for a period of time, such as: i ═ item
i}. In addition, in the present exemplary embodiment, order information may also be integrated, and the sub-orders of the same parent order need to be merged to generate an order set; such as order o
1Including sub-orders
And sub-order
Wherein
The order includes an item
1,item
2,item
3),
The order comprises an item
4,item
5) I.e. by
Post-consolidation order o
1Is o
1{item
1,item
2,item
3,item
4,item
5}; wherein item
1,item
2,item
3,item
4,item
5Is a single product.
In step S120, a historical three-dimensional correlation matrix is established, and any point H of the historical three-dimensional correlation matrixi(xi,yi,zi) For indicating singlets xiAnd singleton yiAt historical time ziThe correlation of (A) is gi. Wherein, giIs based on said singlets xiAnd singleton yiAt historical time ziAnd calculating the single item information.
Referring to fig. 3(a), in the present exemplary embodiment, for each region, correlation analysis may be performed starting from order information, and statistics may be performed on a daily basis; obtaining a three-dimensional incidence matrix, namely a data cube H (X, Y, Z), wherein the data cube records the incidence of the commodities by taking days as granularity; referring to fig. 3(b), each XY layer H (, z) of each correlation matrix represents a point at the same time.
In step S130, a time series of each of the correlations of any two of the singlets is fitted, and a future three-dimensional correlation matrix in a future time period is calculated according to the fitting result.
Referring to FIG. 4(a), each point H (x) on the matrix1,y1,z1) (i.e., A in the figure) represents the correlation of two singles at a certain time point, and the data of the point along the z-axis is the time sequence H (x) of the correlation of the two singles1,y1,*). In this exemplary embodiment, fitting the time series of correlations may be calculated by the following formula:
wherein x isiIs a value in time series, wiT is the weight of the time sequence, and T is the size of the time window.
Referring to fig. 4(b), fig. 4(b) shows a time series of any two-commodity correlation, in which N1 represents true data, N2 represents fitted data, and N3 region represents prediction expectation. In step S140, the future three-dimensional correlation matrix is added to the historical three-dimensional correlation matrix, and the added three-dimensional correlation matrices are fused in the time dimension to obtain a two-dimensional correlation matrix. However, it is easily understood by those skilled in the art that in other exemplary embodiments of the present disclosure, the future three-dimensional correlation matrix in the future time period may be calculated in other manners, which is not particularly limited in the present exemplary embodiment.
Referring to fig. 5, fig. 5 is different from fig. 3(a) in that a correlation matrix of prediction time periods is added; wherein C1 is a future three-dimensional correlation matrix, a historical three-dimensional correlation matrix at C2; the three-dimensional correlation matrix in fig. 5 is fused in the time dimension to form a new two-dimensional correlation matrix H*. In this exemplary embodiment, the fusion mode may be as follows:
where n is the length of the time series.
The above type will be squareThe arrays are summed in the time dimension to form a new matrix H*Unlike matrix H, matrix H*Is a two-dimensional matrix, H*Each element in the matrix represents the overall relevance of some two singlets over a period of time.
In step S150, performing spectral clustering on the two-dimensional incidence matrix, and performing graph segmentation on the two-dimensional incidence matrix according to a spectral clustering result with the minimum relevance between the segmented subgraphs as a target. Referring to fig. 6, in the present exemplary embodiment, spectral clustering may include steps S602 to S608. Wherein:
in step S602, a similarity graph (similarity graph) is created. In this exemplary embodiment, the weighted adjacency matrix of the similarity graph is W, where W is H, and H is the above two-dimensional correlation matrix.
In step S604, a non-normalized graph (unnormalized graph) Laplace matrix L is calculated, wherein
J is an arbitrary point in the weighted adjacent matrix W, and j is the number of columns in the weighted adjacent matrix W.
In step S606, the laplacian matrix L is subjected to SVD decomposition. For example:
L=UΛU-1;
wherein the content of the first and second substances,
[u
1,u
2,...,u
r]a characteristic vector value of L, λ
iIs a characteristic value of L, and in Λ, λ
1≤λ
2≤...≤λ
rAnd r is the rank of W.
In step S608, the first k eigenvector values of the laplacian matrix L are selected to form an r × k matrix, each row in the matrix is used as one vector in a k-dimensional space, the eigenvectors are clustered by using a clustering algorithm, and k is the pre-clustering number of spectral clustering.
Selecting the first k minimum feature vectors (the same as the pre-clustering number k of the spectral clustering), wherein the k value selection needs to fully consider the bin dividing condition; for example, if 2 bins are divided into 3 bins (i.e., a new bin is created, and the number of bins to be involved in calculation is 3), then the k value is selected to be 15; for another example, if 4 bins are to be changed into 6 bins (i.e. 2 bins are newly built, and 6 bins are needed to participate in the calculation), the k value needs to be set to about 30; typically, the k value is selected to be at least 5 times the number of bins and not greater than H, but the disclosure is not so limited.
Arranging the k eigenvectors together to form an r x k matrix, taking each row as one vector in a k-dimensional space, and clustering by using a k-means algorithm; in other exemplary embodiments of the present disclosure, clustering may also be performed according to other manners, which is not particularly limited in this exemplary embodiment.
In step S160, the singleton sets are assigned to the respective warehouses according to the result of the graph division.
According to the storage sub-warehouse auxiliary analysis method and device, on one hand, the order information of the user is abstracted into the commodity incidence matrix, so that the calculation of commodity distribution is completely based on the incidence matrix, the user order is not required to be traversed, and the calculation efficiency is improved; on the other hand, the incidence matrix in a short term can be predicted through the analysis of the time sequence, so that the bill splitting condition after the commodities are separated in bins is predicted.
According to the other storage warehouse-dividing auxiliary analysis method and device, the influence of sales promotion on the relevance of the single product is filtered through the outlier of the time sequence, the accuracy of the relevance is improved, the rate of splitting the single product is reduced, and the cost is greatly reduced.
In other embodiments of the present disclosure, the warehouse sub-warehouse auxiliary analysis method further includes: and performing outlier removal operation on the time series of the relevance through the first moment and the second moment of the time series of the relevance.
Referring to fig. 7, the step of removing outliers may include:
in step S702, the correlation T distribution of any two items is used to calculate the time series of any two items in the unit timeFirst order matrix M1And a second order matrix M2。
In step S704, a confidence level of 95%, that is, α is 0.05, and the degree of freedom is the number of samples minus 1, that is, p-1(p is the range of the time series, in days).
In the above step S704, α is selected to be 0.05, but the present disclosure is not limited thereto, and may be selected according to the actual situation.
In step S706, the critical value c ═ t is calculated(1-α/2)(p-1) wherein t(1-α/2)(p-1) can be obtained by looking up the t distribution table.
In step S708, selection is performed
For the confidence interval, and for each value x in the time series,
if x' is in the interval
If not, the data is retained, otherwise, the data is removed.
By removing outliers, order information about promotions is removed, increasing the accuracy of the correlation. Furthermore, the above-mentioned outlier removing method is merely exemplary, and outlier removing schemes implemented by other means are also within the scope of the present disclosure.
In other embodiments of the present disclosure, the warehouse sub-warehouse auxiliary analysis method further includes: and cleaning the two-dimensional incidence matrix. For example:
referring to fig. 8, the first action is the correlation between the item a and other items (a, b, c, d, e, f), and the darker the color represents the strength of the correlation, the stronger the correlation; at this time, the scheme only reserves the first m associated data (m) of each row>k, k is the number of spectral clusters to be clustered), and the rest is set to zero. Further, H after washing*Still a symmetric matrix.
And the clustering effect is ensured by cleaning the data with weak relevance.
In other embodiments of the present disclosure, the warehouse sub-warehouse auxiliary analysis method further includes: and further clustering the two-dimensional incidence matrix according to the spectral clustering result, and further evaluating the clustering result.
In the scheme, the related single item sets are divided into k types, and the number of warehouses needing to participate in calculation is m; since k is larger than m, further clustering is needed, and if k is clustered into m classes, there are S (k, m) kinds of possibilities, as shown in the following formula, each class in the last m classes represents the class stored in the warehouse after warehouse separation.
Where C is a binomial coefficient.
In the above step, through clustering again, S (k, m) possible single-item layout schemes are generated, and in this step, S (k, m) schemes need to be evaluated to select an optimal scheme; the excellent single-product layout scheme can effectively reduce the occurrence of the single-product splitting situation, so the contribution degree of the commodity layout scheme to the single-product splitting is used as an evaluation criterion.
The evaluation step comprises:
S(k,m)={s1,s2,s3...,sll represents the length of the set S (k, m) with l combination schemes;
si={m1,m2,m3,...,mm};
wherein s is
iFor a singleton layout scheme of S (k, m) schemes, each S
iIn which there are m classes, each of which is m
iConsisting of single articles, and one s
iThe individual items in the m classifications of (1) are not duplicative, i.e., are not duplicated
Representative item
i,item
jThe relevance of (a) to (b),
represents H
*One line above, representing item
iAssociation with all merchandise;
as mentioned above, item
iAt H
*In which there is a corresponding row
Value representation item on corresponding line
iAssociation with all merchandise; if item
iAnd item
jThe bill dismantling can be reduced when the warehouse is placed in the same warehouse; on the contrary, if item
iAnd item
jPlacing in different warehouses necessarily increases the quantity of orders broken.
Thus, evaluation scheme s
iWhen the quality of s is good, s needs to be calculated
iEach class m of
iContribution score _ m to the sheet splitting
iTo m, to m
iChinese single item
iCorresponding to
The rows are summed. item
iAnd item
jAre all in m
iIn (1), then: score _ m
iPlus with
The two single products are put together to reduce the number of the split single products;
on the contrary, if item
iAnd item
jIs not in m
iIn (1), then: score _ m
iPlus with
Indicating that the two singlets are not put together, the rate of singling is increased.
Will calculate the score _ miSumming to obtain siIs evaluated byi_score;si_scoreRepresenting the influence of the distribution of the single products on the order, wherein the larger the value is, the more the commodity layout scheme contributes to the order splitting; otherwise, the better the scheme is, the lower the contribution to the bill splitting is; if s isi_scoreThe negative value of (A) indicates that the scheme can effectively reduce the bill breaking. At all siTo select the minimum si_scoreNamely the final single-product layout scheme. Through the re-clustering and the combination of the evaluation parameters, the binning scheme can be further optimized, and the optimal binning scheme is further realized.
The following are embodiments of the disclosed apparatus that may be used to perform embodiments of the disclosed methods. For details not disclosed in the embodiments of the apparatus of the present disclosure, refer to the embodiments of the method of the present disclosure.
The present exemplary embodiment further provides a warehouse and warehouse separation auxiliary analysis device, which starts from data analysis of a customer order, starts from a purchasing behavior of a user for different regions and different time periods, analyzes a purchasing relevance of a commodity, and splits n warehouses into m warehouses by planning a commodity layout. Referring to fig. 9, the warehousing sub-warehouse auxiliary analysis device includes: the system comprises an information acquisition module 900, a first incidence matrix module 910, a second incidence matrix module 920, a third incidence matrix module 930, a fusion module 940, a spectral clustering module 950, a graph segmentation module 960 and a single item distribution module 970; wherein the content of the first and second substances,
the information acquisition module 900: the system comprises a data processing system and a data processing system, wherein the data processing system is used for acquiring order information of an area in a historical time period and acquiring single item information in the historical time period from the order information;
the first incidence matrix module 910: is used for establishing a historical three-dimensional incidence matrix, and any point H of the historical three-dimensional incidence matrixi(xi,yi,zi) For indicating singlets xiAnd singleton yiAt historical time ziThe correlation of (A) is gi(ii) a Wherein, giIs based on said singlets xiAnd singleton yiAt historical time ziCalculating the single item information;
a fitting module: fitting a time series of each said association for any two of said singlets;
the second incidence matrix module 920: the three-dimensional incidence matrix calculation module is used for calculating a future three-dimensional incidence matrix in a future time period according to the fitting result;
the third correlation matrix module 930: for adding the future three-dimensional correlation matrix to the historical three-dimensional correlation matrix;
a fusion module 940: the system is used for fusing the added three-dimensional incidence matrix on a time dimension to obtain a two-dimensional incidence matrix;
the spectral clustering module 950: the system is used for performing spectral clustering on the two-dimensional incidence matrix;
the graph partitioning module 960: the two-dimensional incidence matrix is used for carrying out graph segmentation on the two-dimensional incidence matrix according to a spectral clustering result by taking the minimum relevance between the segmented subgraphs as a target;
the individual product dispensing module 970: the system is used for distributing the single product set to each warehouse according to the result of the graph segmentation.
In other embodiments of the present disclosure, the apparatus further comprises:
outlier removal module: for performing a de-outlier operation on the time-series of correlations by the first and second moments of the time-series of correlations.
In other embodiments of the present disclosure, the apparatus further comprises:
the first clustering module: the two-dimensional incidence matrix is used for further clustering according to the spectral clustering result;
an evaluation module: for further evaluation of the clustering results.
In other embodiments of the present disclosure, fitting the time series of each correlation for any two of the singlets comprises:
wherein x isiIs a value in time series, wiT is the weight of the time sequence, and T is the size of the time window.
In other embodiments of the present disclosure, the added three-dimensional correlation matrix is fused in the time dimension to obtain a two-dimensional correlation matrix, where the fusing includes:
wherein n is the length of the time series, H
*(x
i,y
i) For any point on the two-dimensional correlation matrix,
is any point on the future three-dimensional correlation matrix.
In other embodiments of the present disclosure, the spectral clustering module includes:
the similar graph establishing module: the method comprises the steps of establishing a similarity graph, wherein a weighted adjacency matrix of the similarity graph is W, wherein W is H, and H is the two-dimensional incidence matrix;
a Laplace matrix calculation module: for calculating the laplacian matrix L, wherein,
w
ijj is any point in the weighted adjacent matrix W, and j is the number of columns in the weighted adjacent matrix W;
a Laplace matrix decomposition module: for decomposing the Laplace matrix L ═ UΛ U-1Wherein U is [ U ═ U1,u2,...,ur],
[u1,u2,...,ur]A characteristic vector value of L, λiIs a characteristic value of L, and in Λ, λ1≤λ2≤...≤λrR is the rank of W;
a second type of module: and the method is used for selecting the first k characteristic vector values of the Laplace matrix L to form an r x k matrix, taking each row in the matrix as a vector in a k-dimensional space, and clustering the characteristic vectors by using a clustering algorithm, wherein k is the pre-clustering number of the spectral clustering.
In other embodiments of the present disclosure, the two-dimensional incidence matrix is further clustered according to the spectral clustering result, where the clustering includes:
wherein S (k, m) is the number of schemes distributed to the single item set, C is a binomial coefficient, m is the number of warehouses needing to participate in calculation, k is the pre-clustering number of spectral clustering, and m is less than k.
Since each functional module of the warehousing sub-warehouse auxiliary analysis device in the embodiment of the invention is the same as that in the embodiment of the method, the description is omitted.
It should be noted that although in the above detailed description several modules or units of the device for action execution are mentioned, such a division is not mandatory. Indeed, the features and functionality of two or more modules or units described above may be embodied in one module or unit, according to embodiments of the present disclosure. Conversely, the features and functions of one module or unit described above may be further divided into embodiments by a plurality of modules or units.
Moreover, although the steps of the methods of the present disclosure are depicted in the drawings in a particular order, this does not require or imply that the steps must be performed in this particular order, or that all of the depicted steps must be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step execution, and/or one step broken down into multiple step executions, etc.
Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein may be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (which may be a CD-ROM, a usb disk, a removable hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which may be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.