CN118134358A - Wisdom logistics distribution data management platform - Google Patents

Abstract

The invention discloses an intelligent logistics distribution data management platform, in particular to the field of logistics distribution management, which is used for solving the problem of order distribution data processing. And dynamically constructing a logistics distribution demand sequence by using a hidden Markov chain model, and predicting the order importance by using a state transition probability and observation probability matrix, thereby improving distribution efficiency and accuracy. Screening distribution resources meeting requirements, checking operation cost according to order importance, selecting resources with lowest cost to process orders, reducing cost and improving efficiency. The data analysis, the prediction model and the resource optimization are integrated, the comprehensive support is provided for the logistics distribution system, the customer requirements can be predicted and met more accurately, and the business development and the service quality improvement are promoted.

Description

Wisdom logistics distribution data management platform

Technical Field

The invention relates to the field of logistics distribution management, in particular to an intelligent logistics distribution data management platform.

Background

Existing order distribution processes rely primarily on whether distribution resources within the enterprise are adequate, and lack a way to systematically integrate and analyze completed distribution resource data to categorize the order for risk and importance. This approach presents a series of potential risks, most notably insufficient prediction accuracy. Due to the lack of careful sorting and analysis of orders, the urgency and importance of the orders cannot be accurately assessed, resulting in increased uncertainty in delivery predictions. This may lead to enterprises having difficulty in meeting customer demands in time, affecting timeliness and efficiency of logistics distribution. In addition, due to the failure to fully consider the risk and importance of the order, the cost estimate may also deviate, resulting in additional expense to the enterprise during distribution, affecting financial performance and profitability. Moreover, this approach based on simple self-resource decisions may also result in decisions that are not scientific enough and data-based, and may deviate from customer needs or market trends, thereby affecting the long-term evolution of the enterprise. Most seriously, the lack of the full utilization of the existing distribution resource data leads to low resource utilization efficiency, and resource waste or unreasonable distribution arrangement can occur, thereby affecting the operation efficiency and competitiveness of enterprises.

In order to solve the above problems, a technical solution is now provided.

Disclosure of Invention

In order to overcome the defects in the prior art, the embodiment of the invention provides an intelligent logistics distribution data management platform, which firstly classifies and extracts characteristics of historical order data through a clustering algorithm of machine learning, combines a correlation rule mining technology, determines analysis priority of the historical data, constructs a training data set and provides effective support for subsequent decisions. Secondly, a logistics distribution demand sequence of a customer is dynamically constructed by using a hidden Markov chain model, the importance of the order is predicted by using a state transition probability and an observation probability matrix, the priority classification of the order is realized, and the logistics distribution efficiency and accuracy are improved. Finally, the distribution resources meeting the requirements are screened out by comparing the key characteristics with the logistics distribution standard, and the operation cost is checked according to the importance degree of the order, so that the distribution resources with the lowest cost after adjustment are selected for order processing, thereby being beneficial to reducing the cost and improving the efficiency. And further, data analysis, prediction model and resource optimization are effectively integrated, and comprehensive support is provided for management and decision-making of the logistics distribution system, so that cost control, efficiency improvement and customer service level improvement are realized, and the problems in the background technology are solved.

In order to achieve the above purpose, the present invention provides the following technical solutions: the system comprises a data management module, a severe test and prediction module and a demand ontology construction module;

The data management module automatically classifies historical order data through a clustering algorithm of machine learning, extracts effective features to construct feature vectors, and simultaneously utilizes an association rule mining technology to mine association relations among the data, so that the analyzable priority of the historical data is determined, a training data set is formed, and the training data set is sent to the severe test and prediction module;

The severe testing and predicting module dynamically constructs a logistics distribution demand sequence of a customer according to the characteristics and the historical data of the order by using a hidden Markov chain model, predicts the importance of the order by using a state transition probability and an observation probability matrix, realizes order priority classification, and sends a processing result to the demand ontology constructing module;

The demand body construction module screens the existing delivery resources meeting the requirements by comparing the key characteristics with the degree of the logistics delivery standard, checks the operation cost according to the importance degree of the order, and finally selects the delivery resource with the lowest cost after adjustment to process the order.

In a preferred embodiment, the data management module comprises the following processing:

Step 1-1, extracting features according to features and attributes of order data;

Step 1-2, fusing the extracted features, normalizing the different features according to equal proportions, and then splicing the features into a comprehensive feature vector to construct the comprehensive feature vector;

normalizing the comprehensive feature vector into a unit vector, calculating an included angle cosine value between the vectors, and obtaining a similarity matrix between orders;

step 1-3, mapping the high-dimensional feature space to a low-dimensional space by using a t-SNE algorithm by taking the standardized comprehensive feature vector as input, and finally obtaining the coordinate of each order in the low-dimensional space;

And step 1-4, taking the feature vector after dimension reduction as input, and applying a selected clustering algorithm to divide the order data into a plurality of clusters, wherein each cluster contains orders with similar features.

In a preferred embodiment, step 2-1, extracting features from the order data of each cluster using Apriori algorithm, looking for frequent item sets;

step 2-2, generating an order importance index based on the frequent item set;

The calculation formula of the order importance index is as follows: ；

Wherein, Representing an order importance index;

is the support of the order, and represents the frequency of the order in all orders;

information entropy of an order represents information quantity contained in the order;

The distance of the order represents the minimum value in the distances between the order and all orders;

the relevance of orders represents the value with the largest relevance degree between the orders and other orders;

Representing a positive constant to prevent the denominator from being zero;

and extracting orders with the order importance indexes exceeding an importance threshold value in each cluster to form a training data set.

In a preferred embodiment, the weight testing and predicting module comprises the following:

feature extraction is carried out on the training data set, and a logistics distribution demand sequence of a customer is constructed:

Dividing the logistics distribution demands of customers into different states, taking the characteristics of orders as observation variables, defining a state transition probability matrix, and representing transition probabilities among the different states, wherein a calculation formula is as follows: ；

Wherein, Representing slave states/>Transition to State/>Probability of (2);

representing slave states/> Transition to State/>Is a number of times (1);

Defining an observation probability matrix: representing the probability that each order feature is observed in each state, the specific steps are as follows:

carrying out normalization processing on each order feature, dividing the order amount into a plurality of intervals, and converting the place information into distance grade;

Let the hidden state set be Wherein/>Represents the/>A hidden state;

Let the set of observation variables be Wherein/>Represents the/>The number of observation variables, namely order characteristics;

for each hidden state, calculating a probability that each order feature is observed in that state;

All the observation probabilities are combined into one observation probability matrix.

In a preferred embodiment, a state transition probability matrix and an observation probability matrix are randomly initialized, wherein the transition probability matrix is A, and the observation probability matrix is B;

Step 3-1, for each order sequence, calculating joint probabilities of the observation sequences, namely the probabilities of observing the order sequence given the current parameters A and B, by using a forward algorithm;

according to the current parameters A and B, forward and backward probabilities of each order sequence in each hidden state are calculated by using a backward algorithm, specifically:

Forward probability Expressed at time/>In state/>And observe observation sequence/>Probability of (2);

Backward probability Expressed at time/>In state/>And observe observation sequence/>Probability of (2);

Calculating probability distribution of each order sequence in each hidden state by utilizing forward and backward probabilities, namely, giving the probability of the order sequence in each hidden state under the given current parameters;

step 3-2, updating the state transition probability matrix A and the observation probability matrix B;

the update formula of the state transition probability matrix A is as follows: ;

Wherein, Time/>In state/>And at time/>Transition to State/>Probability of (2);

the update formula of the observation probability matrix B is as follows: ；

Wherein, Time/>Observed value of/>Represents the/>Observed value/>Time/>In a state ofProbability of (2);

repeating the step 3-1 and the step 3-2 until the maximum iteration times are reached;

Step 3-3, calculating the probability that the observed sequence appears in each state;

Calculating the importance of each state according to the prior probability, wherein the prior probability can be estimated according to historical data or domain knowledge;

the posterior probability, i.e., the importance of the order, at each state is calculated.

In a preferred embodiment, the demand assignment module includes the following:

step 4-1, counting the available logistics resources at present, and counting the corresponding distribution standard of each available logistics resource;

Step 4-2, extracting key features from the newly generated orders of the customers;

Is provided with Is the/>, of the orderKey features,/>For logistics deliveryItem criteria;

converting key features and logistics distribution standards into vector form, and setting key features The value range of (2) is/>，/>The value range of (2) is/>；

The extent to which each key feature belongs to each standard in the logistics distribution is:

；

Wherein, Indicating the degree to which each key feature belongs to each standard in the logistics distribution, i.e. the degree to which the key feature belongs to the standard.

In a preferred embodiment, the key feature belonging to the standard degree is obtained by calculation, the key feature belonging to the standard degree and the corresponding belonging degree threshold value of each key feature are compared, if the key feature belonging to the standard degree is greater than the corresponding belonging degree threshold value, the corresponding key feature in the order of the current customer is indicated to conform to the corresponding delivery standard of the currently available delivery resource, and so on, whether each key feature conforms to the delivery standard corresponding to the currently available delivery resource is checked, then the existing delivery resource conforming to the condition is counted, the operation cost is checked by using the importance degree of the order, the adjusted operation cost is obtained, and the calculation is performed by the following formula:；

Wherein, Representing the adjusted operating cost of the collated delivery resource,/>Representing the correction value;

sorting all the distribution resources meeting the conditions according to the adjustment of the operation cost from small to large, selecting the distribution resource with the first sorting, and processing the newly generated orders of the customers.

The intelligent logistics distribution data management platform has the technical effects and advantages that:

1. According to the invention, the historical order data is automatically classified by using a clustering algorithm of machine learning, effective features are extracted from the order data in the classifying process, and are fused into comprehensive feature vectors, and the classifying result of the order is obtained by a dimension reduction and clustering algorithm. Further, potential association relations of order data in each cluster are mined through association rule mining technology, and the internal rules of the order data are understood. And determining the priority of the order by calculating the order importance index so as to screen out data with analysis value, thereby optimizing business decision and improving the overall efficiency. The comprehensive processing process enables the historical order data to have higher analysis value, and provides powerful support for optimization and decision of the business.

2. The invention utilizes the hidden Markov chain model to construct the customer logistics distribution demand sequence based on the order feature, thereby accurately predicting the importance of the customer demand. The customer behavior mode can be understood through the state transition probability matrix and the observation probability matrix, and future demands are predicted according to historical data. After the matrix is randomly initialized, the order probability distribution is estimated by utilizing forward and backward algorithms through iterative optimization parameters, so that the prediction of the order importance is realized. The logistics company can more accurately predict and meet the demands of clients, and the business development and the service quality improvement are promoted.

3. The invention can quantify the matching degree of the order and the distribution standard by counting the available logistics resources and the distribution standard thereof and converting the key characteristics of the new order into a vector form. Whether the order accords with the distribution standard of the available resources can be automatically judged, so that an intelligent distribution decision is realized. And according to the importance of the order, the operation cost of the distribution resources is checked, and the effective utilization and cost efficiency of the resources are further ensured. And further, the accuracy and the efficiency of logistics distribution are improved, the operation cost can be reduced, the enterprise competitiveness is enhanced, and the customer satisfaction is improved.

Drawings

FIG. 1 is a schematic diagram of a smart logistics distribution data management platform according to the present invention;

FIG. 2 is a schematic flow chart of classifying historical order data of an intelligent logistics distribution data management platform according to the present invention;

FIG. 3 is a flow chart of a demand distribution module of the intelligent logistics distribution data management platform according to the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

Example 1

FIG. 1 shows an intelligent logistics distribution data management platform of the present invention, comprising: the system comprises a data management module, a severe test and prediction module and a demand ontology construction module;

The data management module automatically classifies historical order data through a clustering algorithm of machine learning, extracts effective features to construct feature vectors, and simultaneously utilizes an association rule mining technology to mine association relations among the data, so that the analyzable priority of the historical data is determined, a training data set is formed, and the training data set is sent to the severe test and prediction module;

The severe testing and predicting module dynamically constructs a logistics distribution demand sequence of a customer according to the characteristics and the historical data of the order by using a hidden Markov chain model, predicts the importance of the order by using a state transition probability and an observation probability matrix, realizes order priority classification, and sends a processing result to the demand ontology constructing module;

The demand body construction module screens the existing delivery resources meeting the requirements by comparing the key characteristics with the degree of the logistics delivery standard, checks the operation cost according to the importance degree of the order, and finally selects the delivery resource with the lowest cost after adjustment to process the order.

In the field of data analysis, data quality and feature selection are critical to building accurate analytical models. Historical data often contains a large amount of redundant or analytically meaningless information that can interfere with the model training and prediction process. In order to solve the problem, a clustering layering strategy is needed to divide the data into different layers according to the similarity. In this way, the inherent structure of the data can be better understood, thereby enabling targeted feature extraction and selection. In each level, the most representative and salient features are selected to ensure the quality and information content of the data used by the model. The optimization method not only improves the accuracy and generalization capability of the model, but also can better adapt to analysis requirements under different data situations, and provides more reliable support and guidance for business decisions. By carefully processing the historical data, a more reliable and accurate analytical model can be established, thereby providing a more competitive data-driven decision support for the enterprise.

The data management module comprises the following processing contents:

Preprocessing historical order data, including data cleaning, outlier removal, feature selection, etc., to prepare the data for input of a classification algorithm;

Step 1, based on a clustering algorithm of machine learning, an automatic data classification algorithm is introduced, and historical order data is dynamically classified according to the characteristics and the attributes of the order data, so that the data management efficiency and the accuracy are improved.

As shown in fig. 2, the following are included:

And step 1-1, extracting effective features according to the features and the attributes of the order data. This may include characteristics of the order such as time, place, type of goods, amount of order, etc.;

step 1-2, fusing the characteristics of time, place, goods type, order amount and the like of the order to construct a comprehensive characteristic;

For the time feature, converting the time into a period to reflect the time period feature of the order;

for the location features, converting the location information into a distance of a geographic location to reflect similarity or association between locations;

for the characteristics of the cargo types, calculating the similarity between different cargo types to reflect the association degree between the cargo types;

for the order amount feature, the order amount is subjected to standardized processing so as to reflect the relative size of the order amount;

Fusing the characteristics of time, place, goods type, order amount and the like of the order, normalizing different characteristics according to equal proportion, then splicing the characteristics into a comprehensive characteristic vector, and constructing the comprehensive characteristic vector;

and normalizing the comprehensive feature vector into a unit vector, calculating an included angle cosine value between the vectors, and obtaining a similarity matrix between orders.

Step 1-3, taking the standardized comprehensive feature vector as input, mapping the high-dimensional feature space into a low-dimensional space by using a t-SNE algorithm, and finally obtaining the coordinates of each order in the low-dimensional space so as to facilitate subsequent cluster analysis;

step 1-4, determining the optimal clustering number by using methods such as an elbow method, a contour coefficient method, interval statistics and the like;

taking the feature vector after dimension reduction as input, and applying a selected clustering algorithm (such as a k-means algorithm) to perform clustering analysis;

The clustering algorithm divides the order data into a plurality of clusters, each cluster containing orders with similar characteristics;

And step 1-5, evaluating the clustering result by using an internal evaluation index (such as a contour coefficient) or an external evaluation index (such as comparison with the existing label).

And selecting the optimal clustering number and the optimal clustering result as a final model according to the model evaluation result and the business requirement, thereby completing automatic classification of the order data.

And 2, discovering potential association relations in the historical order data of each cluster by using an association rule mining technology, so that the value of the order data is better understood. An association rule mining algorithm such as an Apriori algorithm or an FP-Growth algorithm can be used for mining out hidden association relations in order data;

the method comprises the following steps:

Step 2-1, extracting features such as product type, customer location, order quantity, etc. from the order data of each cluster;

an association rule mining algorithm (Apriori algorithm or FP-Growth algorithm) is used to find the frequent item set.

The core idea of the Apriori algorithm is to iteratively generate a candidate set and reduce the search space with frequent properties.

Step 2-2, generating an order importance index based on the frequent item set;

The calculation formula of the order importance index is as follows: ；

Wherein, Representing an order importance index;

is the support of the order, and represents the frequency of the order in all orders;

information entropy of an order represents information quantity contained in the order;

The distance of the order, which represents the minimum value in the distance between the order and all orders, can be calculated by using a distance measurement method, such as Euclidean distance, cosine similarity and the like;

the relevance of an order, which represents the value of the greatest degree of relevance between the order and other orders, may be calculated using a relevance coefficient, such as pearson's relevance coefficient, or the like.

A positive constant is represented to prevent the denominator from being zero.

The order importance index is used to measure the importance or salience of an order throughout the historical dataset. A larger index value indicates a more important or significant order, which affects more in the entire dataset; conversely, a smaller index value indicates a lower importance of the order, with relatively less impact. The calculation of the order importance index can help determine the priority or importance of the order, so that support and guidance can be helped when screening data with analytical value, and overall efficiency and decision accuracy are improved.

And extracting orders with the order importance indexes exceeding an importance threshold value in each cluster to form a training data set.

According to the invention, the historical order data is automatically classified by using a clustering algorithm of machine learning, effective features are extracted from the order data in the classifying process, and are fused into comprehensive feature vectors, and the classifying result of the order is obtained by a dimension reduction and clustering algorithm. Further, potential association relations of order data in each cluster are mined through association rule mining technology, and the internal rules of the order data are understood. And determining the priority of the order by calculating the order importance index so as to screen out data with analysis value, thereby optimizing business decision and improving the overall efficiency. The comprehensive processing process enables the historical order data to have higher analysis value, and provides powerful support for optimization and decision of the business.

Analyzing the importance of customer orders based on historical order data is of great importance, and is of great importance to formulating distribution resource schemes. First, the historical order data reflects the purchasing behavior and preference of the customer, and by analyzing the importance of the order, the change trend and priority of the customer demand can be better understood, so that the distribution resources can be adjusted to meet different demands. Secondly, distribution resources can be effectively distributed by determining the importance level of the order, more resources are used for processing the important order, and distribution efficiency and customer satisfaction are improved. In addition, analyzing the order importance based on the historical data is also helpful to identify and optimize bottlenecks and shortages in the distribution process, purposefully improve distribution strategies and processes, reduce cost and improve distribution efficiency. In summary, analyzing the importance of the customer order based on the historical order data has important significance in optimizing the allocation of the distribution resources, improving the distribution efficiency and the customer satisfaction, and can help the enterprise to better cope with market changes and improve the competitiveness.

The severe test and prediction module comprises the following contents:

extracting characteristics of the training data set, including order time, place, goods type, order amount and the like;

Constructing a logistics distribution demand sequence of a customer according to the historical order data:

The logistics distribution demands of customers are divided into different states, such as very urgent, normal and low priority;

Taking the characteristics of the order as an observation variable, defining a state transition probability matrix, representing transition probabilities among different states, and calculating the formula as follows: ；

Wherein, Representing slave states/>Transition to State/>Probability of (2);

representing slave states/> Transition to State/>Is a number of times (1).

Defining an observation probability matrix: representing the probability that each order feature is observed in each state, the specific steps are as follows:

Each order feature is normalized to facilitate modeling. For example, dividing the order amount into a plurality of sections, and converting the location information into a distance level;

Let the hidden state set be Wherein/>Represents the/>A hidden state;

Let the set of observation variables be Wherein/>Represents the/>The number of observation variables, namely order characteristics;

for each hidden state, a probability is calculated that each order feature is observed in that state. May be estimated by statistical history data, or may be calculated by other methods, such as inferring based on correlation between features;

All the observation probabilities are combined into one observation probability matrix.

The observation probability matrix and the state transition probability matrix play a key role in the hidden Markov chain model. The observation probability matrix is used to describe the probability that each order feature is observed in each hidden state, thereby helping the model calculate the likelihood that a certain order feature is observed in a given state. The method is favorable for predicting and judging the state of the model according to actual observation data, and further predicting the importance of the logistics distribution demands of customers. The state transition probability matrix is used for representing the transition probability between different states, namely the probability that the logistics distribution demand of the customer is transited from one state to another state in a given state. Through the two matrixes, the hidden Markov chain model can learn the behavior mode and the demand change rule of the customer from the historical order data, so that the importance of the future customer demand can be predicted more accurately.

Randomly initializing a state transition probability matrix and an observation probability matrix, and setting the transition probability matrix as A and the observation probability matrix as B;

Step 3-1, for each order sequence, calculating joint probabilities of the observation sequences, namely the probabilities of observing the order sequence given the current parameters A and B, by using a forward algorithm;

And according to the current parameters A and B, calculating the forward and backward probabilities of each order sequence in each hidden state by using a backward algorithm. Specifically:

Forward probability Expressed at time/>In state/>And observe observation sequence/>Probability of (2);

Backward probability Expressed at time/>In state/>And observe observation sequence/>Probability of (2);

the probability distribution of each order sequence in each hidden state is calculated by using the forward and backward probabilities, namely the probability that the order sequence is in each hidden state given the current parameters.

Through such processing, probability estimation can be performed for each order sequence according to the current parameters A and B, thereby providing important information for parameter estimation of the hidden Markov chain model.

The state transition probability matrix is used to describe the transition probabilities between different hidden states. The method is mainly used for determining the evolution rule between hidden states, so that the change trend of the logistics distribution demands of customers is better understood. Through the state transition probability matrix, the transition probability among different states can be analyzed, so that the evolution path and the change trend of the customer demand can be inferred. In this way, the observation probability matrix and the state transition probability matrix together form the basic components of the hidden Markov chain model, which helps to better understand and predict the logistics distribution needs of customers.

Step 3-2, updating the state transition probability matrix A and the observation probability matrix B;

the update formula of the state transition probability matrix A is as follows: ;

Wherein, Time/>In state/>And at time/>Transition to State/>Probability of (2);

the update formula of the observation probability matrix B is as follows: ；

Wherein, Time/>Observed value of/>Represents the/>Observed value/>Time/>In state/>Probability of (2);

and repeatedly executing the step 3-1 and the step 3-2 until the maximum iteration number is reached.

The Baum-Welch algorithm or the EM algorithm can effectively estimate parameters of the hidden Markov chain model, so that importance of logistics distribution demands of customers can be predicted more accurately.

Predicting importance of logistics distribution demands of customers according to the trained hidden Markov chain model:

step 3-3, firstly, calculating the probability of the appearance of the sequencing in each state;

Next, the importance of each state is calculated based on the prior probability. The prior probability may be estimated based on historical data or domain knowledge.

Finally, the posterior probability, i.e., the importance of the order, for each state is calculated.

The importance of an order represents the importance or priority of the order throughout the logistics distribution system, with more important orders having a greater impact on the system requiring more preferential processing and scheduling. Categorizing importance may help logistics companies to more effectively manage orders and resource allocation to meet the needs of different orders.

One common method of classifying order importance is to divide orders into several classes or categories according to the magnitude of their importance values. For example, orders can be categorized into the following categories:

Very urgent: including those orders that have a significant impact on customer production or business, require immediate processing and distribution, and typically have the highest importance values.

Emergency: the method has a certain influence on customer business, and orders which are required to be processed and distributed in a short time have high importance.

Common: the method has small influence on customer business, can process and distribute orders in a long time, and has moderate importance value.

Low priority: the method has little influence on customer business, can arrange orders for processing and distribution in a longer time, and has low importance value.

According to the importance level of the orders, different processing strategies and priority arrangement can be formulated by a logistics company, so that the orders with high importance are ensured to be processed in time and distributed preferentially, and the customer satisfaction degree and the overall efficiency are improved.

The invention utilizes the hidden Markov chain model to construct the customer logistics distribution demand sequence based on the order feature, thereby accurately predicting the importance of the customer demand. The customer behavior mode can be understood through the state transition probability matrix and the observation probability matrix, and future demands are predicted according to historical data. After the matrix is randomly initialized, the order probability distribution is estimated by utilizing forward and backward algorithms through iterative optimization parameters, so that the prediction of the order importance is realized. The logistics company can more accurately predict and meet the demands of clients, and the business development and the service quality improvement are promoted.

The demand assignment module as shown in fig. 3 includes the following:

step 4-1, counting the available logistics resources at present, and counting the corresponding distribution standard of each available logistics resource;

Step 4-2, extracting key features from the newly generated orders of customers, including time, place, goods type, etc.;

Is provided with Is the/>, of the orderKey features,/>For logistics deliveryItem criteria;

converting key features and logistics distribution standards into vector form, and setting key features The value range of (2) is/>，The value range of (2) is/>；

The extent to which each key feature belongs to each standard in the logistics distribution is:

；

Wherein, Indicating the degree to which each key feature belongs to each standard in the logistics distribution, i.e. the degree to which the key feature belongs to the standard.

The greater the degree of the key feature belonging to the standard means that the corresponding key feature of the order meets the corresponding delivery standard.

Step 4-3, calculating to obtain the standard degree of the key feature, comparing the standard degree of the key feature of each key feature with the corresponding degree threshold, if the standard degree of the key feature of each key feature is larger than the corresponding degree threshold, indicating that the corresponding key feature in the order of the current customer accords with the corresponding distribution standard of the existing available distribution resources, and so on, checking whether each key feature accords with the distribution standard corresponding to the existing available distribution resources, then counting the existing distribution resources which accord with the conditions, checking the operation cost by using the importance degree of the order, and obtaining the adjusted operation cost, for example, calculating by the following formula:；

Wherein, Representing the adjusted operating cost of the collated delivery resource,/>Representing the correction value;

And sequencing all the distribution resources meeting the conditions according to the adjustment operation cost from small to large. And selecting the distribution resource with the first ranking to process the newly generated order of the customer.

The invention can quantify the matching degree of the order and the distribution standard by counting the available logistics resources and the distribution standard thereof and converting the key characteristics of the new order into a vector form. Whether the order accords with the distribution standard of the available resources can be automatically judged, so that an intelligent distribution decision is realized. And according to the importance of the order, the operation cost of the distribution resources is checked, and the effective utilization and cost efficiency of the resources are further ensured. And further, the accuracy and the efficiency of logistics distribution are improved, the operation cost can be reduced, the enterprise competitiveness is enhanced, and the customer satisfaction is improved.

The above formulas are all formulas with dimensionality removed and numerical calculation, the formulas are formulas with the latest real situation obtained by software simulation through collecting a large amount of data, and preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

The above embodiments may be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above-described embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the processes or functions described in accordance with embodiments of the present application are produced in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains one or more sets of available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium may be a solid state disk.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

In the several embodiments provided in the present application, it should be understood that the disclosed system and apparatus may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other forms.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, or the like) to perform all or part of the steps described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, or other various media capable of storing program codes.

The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Finally: the foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and principles of the invention are intended to be included within the scope of the invention.

Claims (7)

1. The intelligent logistics distribution data management platform is characterized by comprising a data management module, a severe testing and predicting module and a demand body construction module;

The data management module automatically classifies historical order data through a clustering algorithm of machine learning, extracts effective features to construct feature vectors, and simultaneously utilizes an association rule mining technology to mine association relations among the data, so that the analyzable priority of the historical data is determined, a training data set is formed, and the training data set is sent to the severe test and prediction module;

The severe testing and predicting module dynamically constructs a logistics distribution demand sequence of a customer according to the characteristics and the historical data of the order by using a hidden Markov chain model, predicts the importance of the order by using a state transition probability and an observation probability matrix, realizes order priority classification, and sends a processing result to the demand ontology constructing module;

The demand body construction module screens the existing delivery resources meeting the requirements by comparing the key characteristics with the degree of the logistics delivery standard, checks the operation cost according to the importance degree of the order, and finally selects the delivery resource with the lowest cost after adjustment to process the order.

2. The intelligent logistics distribution data management platform of claim 1, wherein:

the data management module comprises the following processing contents:

Step 1-1, extracting features according to features and attributes of order data;

Step 1-2, fusing the extracted features, normalizing the different features according to equal proportions, and then splicing the features into a comprehensive feature vector to construct the comprehensive feature vector;

normalizing the comprehensive feature vector into a unit vector, calculating an included angle cosine value between the vectors, and obtaining a similarity matrix between orders;

step 1-3, mapping the high-dimensional feature space to a low-dimensional space by using a t-SNE algorithm by taking the standardized comprehensive feature vector as input, and finally obtaining the coordinate of each order in the low-dimensional space;

And step 1-4, taking the feature vector after dimension reduction as input, and applying a selected clustering algorithm to divide the order data into a plurality of clusters, wherein each cluster contains orders with similar features.

3. An intelligent logistics distribution data management platform in accordance with claim 2 wherein:

Step 2-1, extracting features from order data of each cluster by using an Apriori algorithm, and searching frequent item sets;

step 2-2, generating an order importance index based on the frequent item set;

The calculation formula of the order importance index is as follows: ；

Wherein, Representing an order importance index;

is the support of the order, and represents the frequency of the order in all orders;

information entropy of an order represents information quantity contained in the order;

The distance of the order represents the minimum value in the distances between the order and all orders;

the relevance of orders represents the value with the largest relevance degree between the orders and other orders;

Representing a positive constant to prevent the denominator from being zero;

and extracting orders with the order importance indexes exceeding an importance threshold value in each cluster to form a training data set.

4. A smart logistics distribution data management platform of claim 3 wherein:

the severe test and prediction module comprises the following contents:

feature extraction is carried out on the training data set, and a logistics distribution demand sequence of a customer is constructed:

Dividing the logistics distribution demands of customers into different states, taking the characteristics of orders as observation variables, defining a state transition probability matrix, and representing transition probabilities among the different states, wherein a calculation formula is as follows: ；

Wherein, Representing slave states/>Transition to State/>Probability of (2);

representing slave states/> Transition to State/>Is a number of times (1);

Defining an observation probability matrix: representing the probability that each order feature is observed in each state, the specific steps are as follows:

carrying out normalization processing on each order feature, dividing the order amount into a plurality of intervals, and converting the place information into distance grade;

Let the hidden state set be Wherein/>Represents the/>A hidden state;

Let the set of observation variables be Wherein/>Represents the/>The number of observation variables, namely order characteristics;

for each hidden state, calculating a probability that each order feature is observed in that state;

All the observation probabilities are combined into one observation probability matrix.

5. The intelligent logistics distribution data management platform of claim 4, wherein:

randomly initializing a state transition probability matrix and an observation probability matrix, and setting the transition probability matrix as A and the observation probability matrix as B;

Step 3-1, for each order sequence, calculating joint probabilities of the observation sequences, namely the probabilities of observing the order sequence given the current parameters A and B, by using a forward algorithm;

according to the current parameters A and B, forward and backward probabilities of each order sequence in each hidden state are calculated by using a backward algorithm, specifically:

Forward probability Expressed at time/>In state/>And observe observation sequence/>Probability of (2);

Backward probability Expressed at time/>In state/>And observe observation sequence/>Probability of (2);

Calculating probability distribution of each order sequence in each hidden state by utilizing forward and backward probabilities, namely, giving the probability of the order sequence in each hidden state under the given current parameters;

step 3-2, updating the state transition probability matrix A and the observation probability matrix B;

the update formula of the state transition probability matrix A is as follows: ;

Wherein, Time/>In state/>And at time/>Transition to State/>Probability of (2);

the update formula of the observation probability matrix B is as follows: ；

Wherein, Time/>Observed value of/>Represents the/>Observed value/>Time/>In state/>Probability of (2);

repeating the step 3-1 and the step 3-2 until the maximum iteration times are reached;

Step 3-3, calculating the probability that the observed sequence appears in each state;

Calculating the importance of each state according to the prior probability;

the posterior probability, i.e., the importance of the order, at each state is calculated.

6. The intelligent logistics distribution data management platform of claim 5, wherein:

the demand distribution module comprises the following contents:

step 4-1, counting the available logistics resources at present, and counting the corresponding distribution standard of each available logistics resource;

Step 4-2, extracting key features from the newly generated orders of the customers;

Is provided with Is the/>, of the orderKey features,/>For logistics deliveryItem criteria;

converting key features and logistics distribution standards into vector form, and setting key features The value range of (2) is/>，/>The value range of (2) is/>；

The extent to which each key feature belongs to each standard in the logistics distribution is:

；

Wherein, Indicating the degree to which each key feature belongs to each standard in the logistics distribution, i.e. the degree to which the key feature belongs to the standard.

7. The intelligent logistics distribution data management platform of claim 6, wherein:

Calculating to obtain the standard degree of the key features, comparing the standard degree of the key features of each key feature with the corresponding degree threshold, if the standard degree of the key features of each key feature is larger than the corresponding degree threshold, indicating that the corresponding key features in the order of the current customer accord with the corresponding delivery standard of the available delivery resources, and so on, checking whether each key feature accords with the delivery standard corresponding to the available delivery resources, counting the available delivery resources which accord with the conditions, checking the operation cost by using the importance degree of the order, and obtaining the adjustment operation cost, wherein the calculation is performed by the following formula: ；

Wherein, Representing the adjusted operating cost of the collated delivery resource,/>Representing the correction value;

sorting all the distribution resources meeting the conditions according to the adjustment of the operation cost from small to large, selecting the distribution resource with the first sorting, and processing the newly generated orders of the customers.