An order batching method and device based on an improved K-means algorithm

Abstract

The application relates to the technical field of clustering analysis, in particular to an order batching method and equipment based on an improved K-means algorithm, wherein the method firstly adopts one-hot coding to map orders into high-dimensional feature vectors based on an EIQ analysis framework, and constructs a low-dimensional feature subset through two-stage feature selection to relieve the dimension disaster and improve feature representation; secondly, aiming at the sensitivity of the traditional K-means to the initial center and the limitation of the Euclidean distance in high-dimensional sparse data, on one hand, a roulette strategy is used to optimize the selection of the clustering center to reduce the risk of local optimization, and on the other hand, the cosine distance is used to replace the Euclidean distance to measure the SKU overlap degree between orders according to the direction similarity, so that orders similar in demand can be more accurately aggregated into the same batch.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of clustering analysis, in particular to an order batching method and device based on an improved K-means algorithm. BACKGROUND

[0002] In China, with the deepening of medical and pharmaceutical policy reforms such as hierarchical diagnosis and treatment, drug zero addition, and national centralized drug procurement, medical and pharmaceutical distribution enterprises are facing a business structure change of increasing disassembled orders and decreasing whole-piece orders. The traditional person-to-goods picking mode has been difficult to adapt to the new demand due to high error rate and high labor cost, prompting the rapid development of goods-to-person picking systems. As a representative of the goods-to-person mode, the multi-layer shuttle vehicle system has the characteristics of high-density storage and high efficiency of warehouse in and out. The multi-layer shuttle vehicle system is composed of multi-layer storage shelves, shuttle vehicles, elevators, conveying lines, and control systems. The out-of-warehouse material box is completed by the cooperation of the elevator and the shuttle vehicle, and the subsequent order picking operation is carried out on the picking table. However, the physical space of the picking table is limited, and it is not possible to allocate independent box positions for each order, so the order batching method is needed to realize the recycling of box positions. If orders are processed centrally, it may lead to congestion of the conveying line due to the centralized out-of-warehouse of material boxes, while decentralized processing may reduce equipment utilization due to the distribution of SKUs across lanes. Therefore, developing an efficient order batching algorithm has become a key link in optimizing the operation of the multi-layer shuttle vehicle system. A reasonable order batching method can optimize the overlap of SKU demand and reduce the number of repeated out-of-warehouse of material boxes, thereby significantly improving the overall efficiency of the system.

[0003] The traditional K-means algorithm is highly sensitive to the initial clustering center, and is easily trapped in a local optimal solution due to random selection of the initial center, leading to instability of the clustering result. These challenges are particularly prominent in the medical and pharmaceutical logistics field. With the increase in the number of order types and the expansion of the number of SKUs, traditional batching methods are increasingly difficult to cope with. Therefore, there is an urgent need for a new algorithm that combines feature selection and clustering optimization to solve this problem and meet the specific needs of the medical and pharmaceutical logistics industry. SUMMARY

[0004] Therefore, the purpose of the present application is to provide an order batching method and device based on an improved K-means algorithm to overcome the problems existing in the prior art.

[0005] To achieve the above purpose, the present application adopts the following technical solutions:

[0006] On the one hand, the present application provides an order batching method based on an improved K-means algorithm, comprising:

[0007] obtaining order information;

[0008] extracting features from the order information through one-hot encoding;

[0009] The extracted features are selected by IK frequency and Pearson correlation coefficient to generate a low-dimensional feature set;

[0010] The cluster centers are selected from the low-dimensional feature set by a roulette strategy;

[0011] A low-dimensional feature vector of the order is constructed according to the low-dimensional feature set, and a cosine distance between the low-dimensional feature vectors is calculated;

[0012] Orders are clustered and batched by an improved K-means order batching algorithm according to the cluster centers and the cosine distance.

[0013] Further, the above-mentioned method, the feature extraction of the order information by one-hot encoding, comprises:

[0014] The SKU feature set is extracted from the order information, and all different SKUs are extracted;

[0015] Each order is one-hot encoded according to all different SKUs to generate a feature vector corresponding to the order.

[0016] Further, the above-mentioned method, the selection of the extracted features by IK frequency and Pearson correlation coefficient to generate a low-dimensional feature subset, comprises:

[0017] The subscription frequency of each SKU in the order information is calculated, and the SKUs are sorted from high to low according to the subscription frequency;

[0018] According to the sorting result, a preset number of SKUs ranked at the top are selected as a feature subset to construct a feature sub-vector of the order and an order sample vector;

[0019] The Pearson correlation coefficient between the features in the order sample vector is calculated by a Pearson correlation coefficient calculation formula, and if the correlation coefficient is less than a preset threshold, the feature with a small subscription frequency is recorded as a to-be-removed feature in the two features participating in the calculation;

[0020] The to-be-removed feature is removed from the feature sub-vector of the order to construct a low-dimensional feature set of the order.

[0021] Further, the above-mentioned method, the selection of the cluster centers from the low-dimensional feature set by a roulette strategy, comprises:

[0022] A sample point is randomly selected in the low-dimensional feature set as a first cluster center;

[0023] A sample that is not selected as a center is randomly selected in the low-dimensional feature set, and a minimum value of the distance between the sample and all selected centers is calculated;

[0024] The probability of selecting the next cluster center is calculated using a probability formula based on the minimum value of the center distance.

[0025] Based on the probability that the next cluster center is selected, the next cluster center is selected from the samples in the low-dimensional feature set that have not been selected as centers.

[0026] The selection probability of the cluster center is recalculated, and new cluster centers are selected again until a preset number of cluster centers are selected.

[0027] Furthermore, in the above-described method, the step of constructing a low-dimensional feature vector of the order based on the low-dimensional feature set and calculating the cosine distance between the low-dimensional feature vectors includes:

[0028] Construct a low-dimensional feature vector for the order based on the low-dimensional feature set;

[0029] The cosine similarity between the low-dimensional feature vectors is calculated using the cosine similarity calculation formula.

[0030] The cosine distance between the low-dimensional feature vectors is calculated according to the cosine similarity and cosine distance calculation formulas.

[0031] Furthermore, in the method described above, the formula for calculating the Pearson correlation coefficient is as follows:

[0032]

[0033] in, The Pearson correlation coefficient between the features. Features The average value across N orders. Features The average value across N orders. For order sample vectors The i-th feature, For order sample vectors The i-th feature.

[0034] Furthermore, in the method described above, the probability formula is:

[0035]

[0036] Wherein, P(x=C) next Let D(X) represent the probability that the next cluster center is selected. 2 The minimum distance between the sample and all currently selected centers. It is the sum of the minimum squared distances of all sample points z in the dataset.

[0037] Further, the above-mentioned method, the cosine similarity calculation formula is:

[0038]

[0039] Wherein, the cosine similarity between the low-dimensional feature vector X and the low-dimensional feature vector Y is cos (X, Y), The Euclidean norm of the low-dimensional feature vector X is ||X||, The Euclidean norm of the low-dimensional feature vector Y is ||Y||.

[0040] Further, the above-mentioned method, the cosine distance calculation formula is:

[0041]

[0042] Wherein, The cosine distance between the low-dimensional feature vector X and the low-dimensional feature vector Y is d (X, Y).

[0043] On the other hand, the application provides an order batching device based on improved K-means algorithm, comprising a processor and a memory, the processor is connected with the memory:

[0044] The processor is used for calling and executing the program stored in the memory;

[0045] The memory is used for storing the program, and the program is used for executing at least the order batching method based on the improved K-means algorithm.

[0046] The beneficial effects of the application are:

[0047] The application first acquires order information, extracts features of the order information through one-hot encoding, selects the extracted features through IK frequency and Pearson correlation coefficient, generates a low-dimensional feature set, then selects a cluster center in the low-dimensional feature set through roulette strategy, constructs a low-dimensional feature vector of the order according to the low-dimensional feature set, calculates the cosine distance between the low-dimensional feature vectors, and finally clusters and batches the orders through the improved K-means order batching algorithm according to the cluster center and the cosine distance. In the application, first, based on the EIQ analysis framework, the order is mapped to a high-dimensional feature vector by using one-hot encoding, and a low-dimensional feature subset is constructed through two-stage feature selection to alleviate the "dimension disaster" and improve the feature representativeness. Secondly, in view of the sensitivity of traditional K-means to initial center and the limitation of Euclidean distance in high-dimensional sparse data, on the one hand, the roulette strategy is used to optimize the selection of cluster center, and on the other hand, the cosine distance is used to replace the Euclidean distance to measure the SKU overlap degree between orders in terms of direction similarity, so that the orders with similar demand are more accurately aggregated into the same batch. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a flowchart provided by an embodiment of an order batching method based on an improved K-means algorithm according to the present invention;

[0050] Figure 2 This is a schematic diagram of the structure of an embodiment of an order batching device based on an improved K-means algorithm according to the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0052] Figure 1 This is a flowchart illustrating an embodiment of an order batching method based on an improved K-means algorithm according to the present invention. Please refer to [link / reference]. Figure 1 This embodiment may include the following steps:

[0053] S1. Obtain order information;

[0054] S2. Extract features from order information using one-hot encoding;

[0055] S3. Select the extracted features using IK frequency and Pearson correlation coefficient to generate a low-dimensional feature set;

[0056] S4. Select cluster centers in a low-dimensional feature set using a roulette wheel strategy;

[0057] S5. Construct low-dimensional feature vectors for orders based on the low-dimensional feature set, and calculate the cosine distance between the low-dimensional feature vectors.

[0058] S6. Cluster and batch orders based on cluster centers and cosine distance using an improved K-means order batching algorithm.

[0059] It can be understood that the embodiment first acquires order information, extracts features of the order information through one-hot encoding, selects the extracted features through IK frequency and Pearson correlation coefficient, generates a low-dimensional feature set, selects a cluster center in the low-dimensional feature set through a roulette strategy, constructs a low-dimensional feature vector of the order according to the low-dimensional feature set, calculates the cosine distance between the low-dimensional feature vectors, and finally clusters and batches the orders according to the cluster center and the cosine distance through an improved K-means order batching algorithm. In the embodiment, first, based on the EIQ analysis framework, the order is mapped to a high-dimensional feature vector by using one-hot encoding, and a low-dimensional feature subset is constructed through two-stage feature selection to alleviate the "dimension disaster" and improve the feature representativeness; second, in view of the sensitivity of the traditional K-means to the initial center and the limitation of the Euclidean distance in high-dimensional sparse data, on the one hand, the roulette strategy is used to optimize the selection of the cluster center to reduce the risk of local optimization, and on the other hand, the cosine distance is used to replace the Euclidean distance to measure the SKU overlap degree between the orders according to the direction similarity, so that the orders with similar demands are more accurately aggregated to the same batch.

[0060] Preferably, step S2 comprises:

[0061] Extracting a SKU feature set from the order information, extracting all different SKUs;

[0062] One-hot encoding each order according to all different SKUs to generate a feature vector corresponding to the order.

[0063] It can be understood that, based on the high-dimensional, sparse and discrete characteristics of the medical order, one-hot encoding is very suitable for mapping each order to a 0 / 1 vector in several possible SKU dimensions.

[0064] For an order set , all different SKUs involved are denoted as , and theoretically there may be thousands or even tens of thousands of SKUs without any screening.

[0065] For any order , if it contains , the vector is 1 in the th dimension, otherwise it is 0. After encoding, the feature vector of each order can be represented as

[0066]

[0067] At this time may be very large, so the feature selection process is needed to reduce the dimension to avoid "dimension disaster".

[0068] Preferably, step S3 includes:

[0069] Calculate the order frequency of each SKU in the order information, and sort the SKUs from high to low according to the order frequency;

[0070] Based on the ranking results, a predetermined number of top-ranked SKUs are selected as feature subsets to construct the feature sub-vectors of orders and the order sample vectors.

[0071] The Pearson correlation coefficient is calculated using the Pearson correlation coefficient formula. If the correlation coefficient is less than a preset threshold, the feature with the lower order frequency among the two features involved in the calculation is recorded as the feature to be removed.

[0072] Remove the features to be removed from the feature subvectors of the order to construct a low-dimensional feature set for the order.

[0073] Understandably, to improve clustering performance, it is necessary to filter existing one-hot features, removing unimportant or redundant features to form a smaller and better low-dimensional feature set. This specifically includes two parts: important feature analysis and feature correlation analysis.

[0074] Statistical analysis of order frequency for each SKU This refers to how many orders this SKU appears in. All SKUs are categorized based on... Sort in descending order of size:

[0075]

[0076] Set feature extraction threshold That is, select the top n% of SKUs as the feature subset and construct the feature sub-vector of the order.

[0077] when When n is small, the feature subset is too small to accurately characterize the differences between orders; when n is large, the "curse of dimensionality" and noise interference increase; the selection should allow the total number of times the material bin is dispatched to vary with... The changes tend to be stable and the clustering effect is ideal. Value or range.

[0078] Even if the former Even among individual SKUs, some features may still provide almost identical "interpretations" of the same batch of orders, exhibiting high redundancy. To further eliminate these redundant features, the Pearson correlation coefficient is used to measure the linear correlation between features.

[0079] For any two features and The corresponding order sample vector can be denoted as:

[0080]

[0081] where if the order contains the feature then 0 otherwise.

[0082] Compute the Pearson correlation coefficient between features:

[0083]

[0084] where is the average value of feature over N orders.

[0085] Set a feature elimination threshold m, if , it means that the two features are almost highly positively correlated on the order data, and the feature with higher IK frequency can be selected to be retained, and the other feature with lower IK frequency can be eliminated. Repeat the execution until the correlation coefficient between all retained features is not more than .

[0086] Through the important feature analysis and the correlation analysis between features described above, a more representative and more mutually non-redundant feature subset can be selected from the massive SKU, laying an efficient foundation for subsequent clustering processing.

[0087] Preferably, step S4 comprises: randomly selecting a sample point in the low-dimensional feature set as a first clustering center;

[0088] Randomly selecting a sample in the low-dimensional feature set which is not selected as a center, and calculating the minimum value of the distance between the sample and all selected centers;

[0089] According to the minimum value of the center distance, the probability of selecting the next clustering center is calculated by a probability formula;

[0090] According to the probability of selecting the next clustering center, the next clustering center is selected from the samples in the low-dimensional feature set which are not selected as a center;

[0091] The selection probability of the clustering center is recalculated, and a new clustering center is selected again until a preset number of clustering centers are selected.

[0092] It can be understood that after the selection of the feature subset is completed, each order (sample) will be represented as a relatively low-dimensional 0 / 1 feature vector:

[0093]

[0094] where is much smaller than the original dimension At this time, improved K-means can be used for clustering batch.

[0095] To reduce the sensitivity of the algorithm to the selection of initial centers and avoid falling into local optimum, a roulette wheel idea is used to select the clustering center.

[0096] First, a sample point is randomly selected as the first center; then, for any sample not selected as the center, the minimum value of the distance between the sample and all selected centers is calculated, and is recorded as , and the probability of the next center being selected is:

[0097]

[0098] and the roulette wheel (Roulette Wheel) method is used to extract ;

[0099] Finally, iterate until initial clustering centers are selected.

[0100] Preferably, step S5 comprises:

[0101] constructing a low-dimensional feature vector of the order according to the low-dimensional feature set;

[0102] calculating the cosine similarity between the low-dimensional feature vectors by a cosine similarity calculation formula;

[0103] calculating the cosine distance between the low-dimensional feature vectors according to a cosine similarity and cosine distance calculation formula.

[0104] It can be understood that the Euclidean distance measures the absolute distance of each point in space, while the cosine similarity measures the angle between the space vectors. The numerical result of the former is affected by the specific values on each coordinate axis, while the latter focuses on the difference in direction. Combined with the characteristics of the feature subset in this paper, which is a 0 / 1 vector, the difference between different features is emphasized, so the calculation of the similarity distance from the perspective of cosine similarity is more suitable for the needs of this paper. The cosine distance converts the similarity into a distance measure by focusing on the directional difference between vectors, so this paper selects the cosine distance as the distance measure.

[0105] For the feature vectors of two orders: and , the Euclidean distance can be represented as:

[0106]

[0107] The cosine similarity can be represented as:

[0108]

[0109] wherein and is the Euclidean norm of a vector, respectively:

[0110]

[0111] The value range of cosine similarity is [-1, 1], the closer to 1 indicates that the two vectors are more similar, and the closer to -1 indicates that the two vectors are less similar.

[0112] The cosine distance can be expressed as

[0113]

[0114] The value range of cosine distance is [0, 2], when the angle between two vectors is 0 (i.e. they are exactly the same), the cosine distance is 0; when the angle between them is 180° (i.e. they are completely opposite), the cosine distance is 2.

[0115] In some optional embodiments, the improved K-means clustering algorithm pseudo code is shown in Table 1.

[0116] Table 1 Improved K-means clustering algorithm pseudo code

[0117]

[0118] The application also provides an order batching device based on an improved K-means algorithm, which is used to implement the above method embodiments. Figure 2 is a structural schematic diagram provided by an embodiment of the order batching device based on the improved K-means algorithm. As shown in Figure 2 the order batching device based on the improved K-means algorithm includes a processor 21 and a memory 22, and the processor 21 is connected with the memory 22. The processor 21 is used to call and execute a program stored in the memory 22; the memory 22 is used to store the program, and the program is used to at least execute the order batching method based on the improved K-means algorithm in the above embodiments.

[0119] The specific implementation of the order batching device based on the improved K-means algorithm provided by the embodiments of the application can refer to the implementation of the order batching method based on the improved K-means algorithm of any of the above embodiments, which will not be repeated here.

[0120] It can be understood that the same or similar parts in the above embodiments can be mutually referred to, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0121] It should be noted that in the description of this invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this invention, unless otherwise stated, "a plurality of" means at least two.

[0122] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0123] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0124] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0125] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0126] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0127] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the specification, the illustrative description of the above terms does not necessarily mean the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0128] Although the embodiments of the present application have been shown and described above, it is understood that the above-described embodiments are exemplary, and cannot be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above-described embodiments within the scope of the present application.

Claims

1. An order batching method based on an improved K-means algorithm, characterized in that, include: Get order information; Feature extraction is performed on the order information using one-hot encoding; Calculate the order frequency of each SKU in the order information, and sort the SKUs from high to low according to the order frequency; Based on the ranking results, a preset number of top-ranked SKUs are selected as a feature subset to construct the feature sub-vector and order sample vector of the order. The Pearson correlation coefficient is calculated using the Pearson correlation coefficient formula. If the correlation coefficient is less than a preset threshold, the feature with the lower order frequency is recorded as the feature to be removed from the two features involved in the calculation. Remove the features to be removed from the feature subvectors of the order to construct a low-dimensional feature set for the order; Cluster centers are selected from the low-dimensional feature set using a roulette wheel strategy; Construct low-dimensional feature vectors for the orders based on the low-dimensional feature set, and calculate the cosine distance between the low-dimensional feature vectors; Orders are clustered and batched using an improved K-means order batching algorithm based on the cluster centers and the cosine distance.

2. The method according to claim 1, characterized in that, The feature extraction of the order information using one-hot encoding includes: Extract the complete set of SKU features from the order information, and extract all different SKUs; Each order is one-hot encoded based on all different SKUs to generate a feature vector for that order.

3. The method according to claim 2, characterized in that, The selection of cluster centers in the low-dimensional feature set using a roulette wheel strategy includes: A sample point is randomly selected from the low-dimensional feature set as the first cluster center. Randomly select a sample that has not been selected as a center from the low-dimensional feature set, and calculate the minimum distance between the sample and all currently selected centers; The probability of selecting the next cluster center is calculated using a probability formula based on the minimum value of the center distance. Based on the probability that the next cluster center is selected, the next cluster center is selected from the samples in the low-dimensional feature set that have not been selected as centers. The selection probability of the cluster center is recalculated, and new cluster centers are selected again until a preset number of cluster centers are selected.

4. The method according to claim 3, characterized in that, The step of constructing low-dimensional feature vectors for the orders based on the low-dimensional feature set and calculating the cosine distance between the low-dimensional feature vectors includes: Construct a low-dimensional feature vector for the order based on the low-dimensional feature set; The cosine similarity between the low-dimensional feature vectors is calculated using the cosine similarity calculation formula. The cosine distance between the low-dimensional feature vectors is calculated according to the cosine similarity and cosine distance calculation formulas.

5. The method according to claim 4, characterized in that, The formula for calculating the Pearson correlation coefficient is as follows: in, The Pearson correlation coefficient between the features. Features The average value across N orders. Features The average value across N orders. For order sample vectors The i-th feature, For order sample vectors The i-th feature.

6. The method according to claim 5, characterized in that, The probability formula is: Wherein, P(x=C) next Let D(X) represent the probability that the next cluster center is selected. 2 The minimum distance between the sample and all currently selected centers. It is the sum of the minimum squared distances of all sample points z in the dataset.

7. The method according to claim 6, characterized in that, The formula for calculating cosine similarity is: Where cos(X,Y) is the cosine similarity between low-dimensional feature vectors X and Y. Let X be the Euclidean norm of the low-dimensional eigenvector X. Let be the Euclidean norm of the low-dimensional eigenvector Y.

8. The method according to claim 7, characterized in that, The formula for calculating the cosine distance is: in, Let X be the cosine distance between the low-dimensional eigenvectors X and Y.

9. An order batching device based on an improved K-means algorithm, characterized in that, It includes a processor and a memory, wherein the processor is connected to the memory: The processor is used to call and execute the program stored in the memory; The memory is used to store the program, which is at least used to execute the order batching method based on the improved K-means algorithm as described in any one of claims 1-8.

Images