CN114841637A - HDBSCAN-based logistics distribution center site selection method and system - Google Patents
HDBSCAN-based logistics distribution center site selection method and system Download PDFInfo
- Publication number
- CN114841637A CN114841637A CN202210393042.0A CN202210393042A CN114841637A CN 114841637 A CN114841637 A CN 114841637A CN 202210393042 A CN202210393042 A CN 202210393042A CN 114841637 A CN114841637 A CN 114841637A
- Authority
- CN
- China
- Prior art keywords
- sub
- cluster
- clusters
- site selection
- distribution center
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
- G06Q10/083—Shipping
- G06Q10/0836—Recipient pick-ups
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/231—Hierarchical techniques, i.e. dividing or merging pattern sets so as to obtain a dendrogram
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Entrepreneurship & Innovation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a method and a system for selecting a site of a logistics distribution center based on HDBSCAN, wherein the method comprises the following steps: 1) adjusting the sub-cluster control capability of the HDBSCAN by using a cluster splitting and cluster merging algorithm to balance the distribution amount of distribution centers at the same level, and determining the distribution centers for each distribution range by using a gravity center method; 2) the method is used iteratively to realize balanced addressing layout of the multi-level distribution center. The method utilizes the characteristic that the HDBSCAN algorithm has clustering stability and can process data with different local densities, and is applied to the site selection of the logistics distribution center. And aiming at the problems that the HDBSCAN algorithm lacks the capability of balancing the sizes of sub-clusters and a hierarchical clustering tree generated in the clustering process cannot directly correspond to the center hierarchy of the sub-clusters, a corresponding improvement method is provided, and finally a logistics distribution center site selection scheme considering the hierarchy and the distribution quantity balance of each hierarchical distribution center is established.
Description
Technical Field
The invention belongs to the technical field of geographic information, and particularly relates to a logistics distribution center site selection method and system based on HDBSCAN.
Background
With the rapid development of economic globalization and network technology, the logistics express industry has rapidly developed into an important national economy emerging pillar industry and a new economic growth point. The development of the logistics industry is not only related to the social life quality of residents, but also influences the development of social economy, and is an index for reflecting national comprehensive strength and modernization degree. In recent years, while the logistics industry in China is rapidly developing, the transportation cost and the storage cost are rapidly increased to account for more than one half of the total cost, so that the development of the overall economy and the improvement of the enterprise competitiveness are seriously hindered, and the problem which needs to be solved in the development of enterprises in China is urgent. Therefore, how to effectively reduce the logistics distribution cost and improve the logistics operation efficiency has become a hot issue of research in the logistics distribution field.
The logistics distribution center is an important node in the whole logistics distribution network system, covers most of the basic operation links of logistics, and is the most centralized embodiment of logistics activities. Therefore, the logistics distribution center can effectively reduce the logistics cost and improve the service quality by reasonably selecting the site, and has important significance for improving the utilization efficiency of logistics resources.
The existing logistics distribution center site selection method can be mainly divided into a qualitative type and a quantitative type, wherein the qualitative method generally has the problem of over-strong subjectivity in the site selection process, the quantitative method can be attributed to the solution of an NP-hard mathematical problem, the calculation difficulty is high, a certain solution can not be obtained, and in addition, most of the methods do not consider the multilevel property of the logistics distribution center site selection and the balance characteristic of the logistics distribution amount of the same-level distribution center, so that the logistics distribution center obtained through calculation cannot meet the actual application requirements easily. Aiming at the problems, the invention aims to adopt an improved HDBSCAN algorithm to carry out site selection research of the logistics distribution center so as to solve the problems that the existing site selection process has more subjectivity judgment, high calculation difficulty of the optimal position, incapability of considering multi-level balanced distribution of the logistics distribution center and the like, thereby providing technical support for optimizing the logistics site selection.
Disclosure of Invention
The invention aims to provide a method and a system for selecting a site of a logistics distribution center based on HDBSCAN.
The technical scheme for solving the technical problems is as follows:
a logistics distribution center site selection method based on HDBSCAN comprises the following steps:
step 1, receiving address distribution information of an area is obtained, wherein the receiving address distribution information comprises a plurality of data points, and each data point is receiving address information;
step 2, dividing all data points into a plurality of disjoint sub-clusters by using an HDBSCAN algorithm;
step 3, carrying out balance adjustment on the sub-clusters by using a cluster splitting and cluster merging algorithm to obtain final sub-clusters;
step 4, assuming that the transportation cost is in direct proportion to the transportation distance, replacing the transportation cost by the transportation distance, obtaining coordinates of all data points in each sub-cluster by using a gravity center method, averaging to obtain a sub-cluster center, and obtaining a site selection result of the distribution center of the level by taking all the sub-cluster centers as the site selection of the distribution center;
and 5, judging whether higher-level addressing needs exist, if so, turning to the step 1, replacing the distribution information of the addressees of the areas in the step 1 with the addressing result of the distribution center obtained in the step 4, calculating the higher-level addressing result, and if not, ending the addressing.
Further, the step 2 specifically includes the following steps:
step 2.1, calculating the reachable distance between every two data points, taking the data points as vertexes, and constructing a distance weighted graph, wherein the weight of each side is equal to the mutual reachable distance of the vertexes;
step 2.2, calculating a minimum spanning tree of the distance weighted graph by using a prim algorithm, sequencing edges of the tree in an increasing order according to the distance, creating a new sub-cluster for each edge through a parallel set, and constructing a sub-cluster hierarchical structure;
step 2.3, compressing the cluster hierarchical structure according to the minimum sub-cluster size, traversing the hierarchical structure, deleting all edges from the minimum spanning tree according to the descending order of the weight, each time deleting the edges to split the sub-clusters, judging whether data points in the new connected components created by deleting the edges are less than those in the minimum sub-clusters, if so, declaring the new connected components as false connected components, marking the false components as noise, and adjusting the sub-clusters; if all components generated after the sub-cluster edge deletion are false, deleting the sub-cluster, if both false components and real components exist in the components generated after the sub-cluster edge deletion, retaining the original sub-cluster label of the component, namely retaining the sub-cluster before the edge removal, and if the components generated after the sub-cluster edge deletion are not false, allocating a new sub-cluster label to each component, namely successfully splitting the original sub-cluster into new sub-clusters;
and 2.4, traversing the tree in a reverse topological sorting order, judging whether the stability of the father and son clusters of each son cluster is greater than the sum of the son nodes of the father and son clusters, if so, declaring the son clusters as selected son clusters and merging all descendants of the selected son clusters, and obtaining all selected son clusters after reaching the root node.
Further, the step 3 specifically includes the following steps:
and traversing all the selected sub-clusters, combining the sub-clusters with the sub-cluster size smaller than v/2 with the nearest sub-cluster according to a preset capacity value v, splitting the sub-clusters with the sub-cluster size larger than v, and traversing for multiple times until the sub-clusters are not changed, wherein the rest sub-clusters are the final sub-clusters.
The invention has the beneficial effects that: the invention provides corresponding improvement aiming at the problems that an HDBSCAN algorithm lacks the capability of balancing the sizes of sub-clusters and a hierarchical clustering tree generated in the clustering process cannot directly correspond to the center hierarchy of the sub-clusters so as to realize a logistics distribution center site selection scheme considering the hierarchy and the distribution quantity balance of each hierarchical distribution center. The main advantages of the invention are as follows: 1) adjusting the sub-cluster control capability of the HDBSCAN by using a cluster splitting and cluster merging algorithm to balance the distribution amount of distribution centers at the same level, and determining the distribution centers for each distribution range by using a gravity center method; 2) the method is used iteratively to realize balanced addressing layout of the multi-level distribution center.
Drawings
FIG. 1 is a schematic overall flow diagram of the present invention;
FIG. 2 is a schematic diagram of the clustering effect obtained by the method of the present invention;
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
Through the development of many years, most logistics enterprises adopt a multi-stage sorting and distribution system: the express is delivered to a demand point from the highest-level regional transfer warehouse sequentially through the first-level regional sorting center, the second-level regional sorting center and the terminal express station. According to the system structure of the logistics network and the modern logistics distribution process, goods are transported to a sorting center from a highest-level area-level transfer warehouse in the logistics distribution process, transported to a terminal express station after sorting is completed, and finally distributed to a customer group. The goods are configured with agent factories in each area as much as possible, the goods required by customers are produced by factories and then are delivered to transfer warehouses of corresponding large areas, after the goods are distributed according to the requirements of the customers, the express finally reaches express stations or express cabinets of customer residential areas through sorting centers of provincial and urban areas, and cross-level transportation of cross-logistics distribution centers can not occur in the process of operation and distribution of logistics express.
The invention mainly aims at the requirement of the site selection problem of a logistics distribution center, and provides a HDBSCAN-based logistics distribution center site selection method, as shown in figure 1, which comprises the following steps:
first, address selection is carried out to the distribution center of the bottommost layer (k layer)
1. Data set acquisition
And the verification data set adopts an official data set provided by the 2018 global operational research optimization challenge match, and comprises a serial number, longitude and latitude information, package information and receiving time window information of a merchant.
2. Partitioning distribution coverage using modified HDBSCAN algorithm
Constructing a data weighted graph, wherein data points are vertexes, the weight of an edge between any two points is equal to the mutual reachable distance between the points, then constructing a minimum spanning tree of the distance weighted graph, sequencing the edges of the minimum spanning tree in an increasing order according to the distance, then iterating, and creating a new sub-cluster for each edge through parallel searching; compressing the hierarchical structure of the cluster according to the minimum sub-cluster size, traversing the hierarchical structure, judging when splitting the sub-cluster each time, and only reserving the split meeting the minimum sub-cluster size; and finally, extracting stable clusters from the compressed tree, traversing the tree in a reverse topological sorting order, declaring the child clusters as selected child clusters and merging all descendants thereof if the stability of the parent child clusters is greater than the sum of the child nodes of the parent child clusters, and taking the currently selected clusters as clustering results after the root nodes are reached.
Adjusting the sub-clusters according to the input parameter capacity v, traversing all generated sub-clusters after the HDBSCAN algorithm extracts stable clusters from the compressed tree, merging the sub-clusters with the sub-cluster size smaller than v/2 with the nearest sub-cluster according to a given capacity value v, and splitting the sub-clusters with the sub-cluster size larger than v. Because both the cluster merging and the cluster splitting algorithm are likely to generate new sub-clusters with sub-cluster sizes not within the [ v/2, v ] interval, the balanced clustering result needs to be obtained through traversing for many times.
3. Gravity center method calculation logistics distribution center
And determining the logistics distribution center corresponding to each sub-cluster, namely the distribution range, by using a gravity center method. Using the transport distance instead of the transport cost, the transport cost can be expressed as:
and (3) solving a partial derivative of x and y, solving a point with the partial derivative being 0, knowing that the transportation cost is minimum at the moment according to a maximum theorem, and solving the coordinate average value of the data point in each sub-cluster, namely the center of the solved sub-cluster, namely the address of the logistics distribution center.
Secondly, address selection is carried out on the distribution center of higher level
And taking the site selection result of the k layer as input data of site selection of a higher layer (the k-1 layer), and repeatedly performing two steps of dividing a radiation range and selecting a distribution center site in the radiation range by using an iteration method, so that a site selection scheme of the multi-layer logistics distribution center can be obtained from bottom to top.
Third, experimental results
Experiments were conducted using the official dataset provided by the global operational optimization challenge. Only the latitude information is taken, and the 1091 data is used as the address of the bottom logistics distribution center (express cabinet or express post). The HDBSCAN-based multi-level logistics distribution center site selection method is used for site selection, a two-stage distribution center needs to be decided in the case, according to the comparison of the bearing capacity of a first-stage logistics distribution center and a second-stage logistics distribution center, namely a tail-end sorting center and a tail-end express station or an express cabinet, when the site selection of the second-stage logistics center (a second-stage regional sorting center) is carried out, the capacity value is set to be 120, and when the site selection of the first-stage logistics distribution center (a first-stage regional sorting center) is carried out, the capacity value is set to be 3. Finally, the decision schemes of 9 secondary region sorting centers and 4 primary region sorting centers are obtained. Because the second layer selects four primary logistics distribution centers according to nine secondary logistics distribution centers simply, whether the site selection of the secondary logistics distribution centers is reasonable or not is mainly analyzed, and the distribution range division of the secondary logistics distribution centers, namely the first clustering result of algorithm iteration, is as follows:
TABLE 1 results of the Algorithm
It can be known from table 1 that the results obtained by the method of the present invention indicate that except for the excessive number of points in one sub-cluster, other sub-clusters are relatively balanced, and the size of the sub-cluster represents the number of the required points in the distribution range of the distribution center and is in direct proportion to the distribution amount of the distribution center, so that the distribution amount of the logistics distribution center generated by the method of the present invention is relatively balanced, the distribution efficiency is high after the logistics distribution network is put into use, and the distribution resources of each logistics node are not wasted. The address and the cluster to which each data point belongs are visualized, and the result is shown in fig. 2 below.
Experimental results show that the multi-level logistics distribution center site selection scheme generated based on the HDBSCAN clustering algorithm has good capability of controlling the distribution quantity balance of the same-level logistics distribution center, and can solve the problem of multi-level site selection of the logistics distribution center.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (4)
1. A logistics distribution center site selection method based on HDBSCAN is characterized by comprising the following steps:
step 1, acquiring recipient address distribution information of an area, wherein the recipient address distribution information comprises a plurality of data points, and each data point is recipient address information;
step 2, dividing all data points into a plurality of disjoint sub-clusters by using an HDBSCAN algorithm;
step 3, carrying out balance adjustment on the sub-clusters by using a cluster splitting and cluster merging algorithm to obtain final sub-clusters;
step 4, assuming that the transportation cost is in direct proportion to the transportation distance, replacing the transportation cost by the transportation distance, obtaining coordinates of all data points in each sub-cluster by using a gravity center method, averaging to obtain a sub-cluster center, and obtaining a site selection result of the distribution center of the level by taking all the sub-cluster centers as the site selection of the distribution center;
and 5, judging whether higher-level addressing needs exist, if so, turning to the step 1, replacing the distribution information of the addressees of the areas in the step 1 with the addressing result of the distribution center obtained in the step 4, calculating the higher-level addressing result, and if not, ending the addressing.
2. The HDBSCAN-based logistics distribution center site selection method of claim 1 wherein said step 2 specifically comprises the steps of:
step 2.1, calculating the reachable distance between every two data points, taking the data points as vertexes, and constructing a distance weighted graph, wherein the weight of each side is equal to the mutual reachable distance of the vertexes;
step 2.2, calculating a minimum spanning tree of the distance weighted graph by using a prim algorithm, sequencing edges of the tree in an increasing order according to the distance, creating a new sub-cluster for each edge through a parallel set, and constructing a sub-cluster hierarchical structure;
step 2.3, compressing the cluster hierarchical structure according to the minimum sub-cluster size, traversing the hierarchical structure, deleting all edges from the minimum spanning tree according to the descending order of the weight, each time deleting the edges to split the sub-clusters, judging whether data points in the new connected components created by deleting the edges are less than those in the minimum sub-clusters, if so, declaring the new connected components as false connected components, marking the false components as noise, and adjusting the sub-clusters; if all components generated after the edge of the sub-cluster is deleted are false, deleting the sub-cluster, if the components generated after the edge of the sub-cluster is deleted have false components and real components at the same time, keeping the original sub-cluster label of the sub-cluster, namely keeping the sub-cluster before the removed edge, and if the components generated after the edge of the sub-cluster is deleted are not false, allocating a new sub-cluster label to each component, namely successfully splitting the original sub-cluster into new sub-clusters;
and 2.4, traversing the tree in a reverse topological sorting order, judging whether the stability of the father and son clusters of each son cluster is greater than the sum of the son nodes of the father and son clusters, if so, declaring the son clusters as selected son clusters and merging all descendants of the selected son clusters, and obtaining all selected son clusters after reaching the root node.
3. The HDBSCAN-based logistics distribution center site selection method of claim 1 wherein said step 3 specifically comprises the steps of:
and traversing all the selected sub-clusters, combining the sub-clusters with the sub-cluster size smaller than v/2 with the nearest sub-cluster according to a preset capacity value v, splitting the sub-clusters with the sub-cluster size larger than v, and traversing for multiple times until the sub-clusters are not changed, wherein the rest sub-clusters are the final sub-clusters.
4. A logistics distribution center site selection system based on HDBSCAN is characterized by comprising an information acquisition module, a sub-cluster division module, a sub-cluster adjustment module, a distribution center site selection module and a judgment module, wherein the information acquisition module is used for acquiring recipient address distribution information of an area, the recipient address distribution information comprises a plurality of data points, and each data point is recipient address information; the sub-cluster dividing module is used for dividing all data points into a plurality of disjoint sub-clusters by using an HDBSCAN algorithm; the sub-cluster adjusting module is used for carrying out balance adjustment on the sub-clusters by using a cluster splitting and cluster merging algorithm to obtain final sub-clusters; the distribution center site selection module is used for replacing the transportation cost with the transportation distance, obtaining the coordinates of all data points in each sub-cluster by using a gravity center method, obtaining the center of the sub-cluster by taking the average value, and obtaining the site selection result of the distribution center of the level by taking all the centers of the sub-clusters as the distribution center site selection; the judging module is used for judging whether higher-level site selection is required after site selection of the local-level distribution center is finished, if yes, the obtained site selection result of the local-level distribution center is used for replacing address distribution information of the area in the information obtaining module, and the higher-level site selection result is calculated, otherwise, site selection is finished.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210393042.0A CN114841637A (en) | 2022-04-14 | 2022-04-14 | HDBSCAN-based logistics distribution center site selection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210393042.0A CN114841637A (en) | 2022-04-14 | 2022-04-14 | HDBSCAN-based logistics distribution center site selection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114841637A true CN114841637A (en) | 2022-08-02 |
Family
ID=82566496
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210393042.0A Pending CN114841637A (en) | 2022-04-14 | 2022-04-14 | HDBSCAN-based logistics distribution center site selection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114841637A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116484260A (en) * | 2023-04-28 | 2023-07-25 | 南京信息工程大学 | Semi-supervised log anomaly detection method based on bidirectional time convolution network |
| CN118069778A (en) * | 2024-04-24 | 2024-05-24 | 成都锋卫科技有限公司 | LLM model-based similar asset fingerprint extraction method |
-
2022
- 2022-04-14 CN CN202210393042.0A patent/CN114841637A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116484260A (en) * | 2023-04-28 | 2023-07-25 | 南京信息工程大学 | Semi-supervised log anomaly detection method based on bidirectional time convolution network |
| CN116484260B (en) * | 2023-04-28 | 2024-03-19 | 南京信息工程大学 | Semi-supervised log anomaly detection method based on bidirectional time convolution network |
| CN118069778A (en) * | 2024-04-24 | 2024-05-24 | 成都锋卫科技有限公司 | LLM model-based similar asset fingerprint extraction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114841637A (en) | HDBSCAN-based logistics distribution center site selection method and system | |
| CN111639237B (en) | Electric power communication network risk assessment system based on clustering and association rule mining | |
| CN108053037B (en) | Power distribution network emergency repair strategy making method and device based on two-network fusion | |
| CN105005570A (en) | Method and apparatus for mining massive intelligent power consumption data based on cloud computing | |
| CN102393937A (en) | Address matching method and system of address tree based on backward production | |
| CN109858858A (en) | A kind of classification of underground logistic network nodal points and site selection system and method | |
| CN108256969A (en) | A kind of public bicycles lease point dispatcher-controlled territory division methods | |
| Fang et al. | Research on Improved NSGA‐II Algorithm and Its Application in Emergency Management | |
| CN112015733A (en) | Method for storing and rapidly inquiring mass data of electric power customer service operation and distribution service | |
| CN111882123A (en) | Intelligent community logistics terminal site selection method for big data intelligent decision | |
| Cheng et al. | Design of a support system for complicated logistics location integrating big data | |
| CN111309810B (en) | Library and table dividing method and system based on small sample KEY value | |
| Ge et al. | Proactive Two‐Level Dynamic Distribution Routing Optimization Based on Historical Data | |
| CN115630868A (en) | Supplier selection method under external environment interference | |
| Jianzhu et al. | Decision-making for location of manufacturing bases in an uncertain demand situation | |
| CN106953323A (en) | A kind of transformer station's supply district optimization method for adapting to load seasonal variety | |
| Wang | Planning and layout of intelligent logistics park based on improved genetic algorithm | |
| McKinnon | Logistics and land: the changing land use requirements of logistical activity | |
| Dai et al. | Establishing a network planning model of urban-rural logistics based on hub-and-spoke network | |
| CN110504703A (en) | The preferred method and apparatus of energy source station in a kind of distributed energy energy supply network | |
| CN104123605B (en) | The method for digging of a kind of electrical network production data and factor of influence correlationship | |
| Wei et al. | Multi-objective location model of urban terminal distribution network | |
| Sutapa et al. | Determining the Number and Location of Warehouses to Minimize Logistics Costs of Business to Consumer (B2C) Distribution | |
| CN117195010A (en) | Method for carrying out logistics center site selection based on hybrid clustering algorithm | |
| CN106651176A (en) | Electricity consumption analysis method for service industry |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |