Platform-based methods, systems, devices, and storage media for clustering similar sub-markets.

Abstract

This invention provides a platform-based method, system, device, and storage medium for clustering similar sub-markets. The method includes the following steps: establishing a three-dimensional route vector for each sub-market, where the maximum values ​​of the three dimensions are the total number of departure cities, the total number of destination cities, and the total number of vehicle classes representing vehicle carrying capacity, respectively; defining a training dataset based on the characteristics of the sub-markets; slicing the training data into individual sub-markets based on each record in the training dataset; grouping all sub-markets into m similar sub-market clusters and training an algorithm model in each cluster; and obtaining at least one route task belonging to the same sub-market based on the user's historical route data and pushing it to the user. This invention can provide users with transportation routes corresponding to similar sub-markets through deep learning feature embedding methods, greatly improving the accuracy of vehicle-cargo matching.

Description

Technical Field

[0001] This invention relates to the field of vehicle-cargo matching technology, and in particular to a platform-based method, system, device, and storage medium for clustering similar sub-markets. Background Technology

[0002] With the development of mobile internet technology, the integration of location-based services (LBS) with real-world business scenarios is becoming increasingly feasible and important. The digital freight market is a typical LBS service and a typical two-sided market, serving both shippers and drivers. Shippers publish their freight needs on the platform, with key information including the origin, destination, type of goods, weight, volume, packaging method, and loading time. Drivers, when searching for suitable freight, typically prioritize loading and unloading locations, weight, volume, type of goods, packaging method, and loading time. They also generally consider the "empty driving distance" from their current location to the next shipment. These factors determine the driver's per-unit-time economic efficiency in freight operations and the risks they bear.

[0003] Moreover, logistics is a complex system. To ensure stable income, drivers are reluctant to accept new transport tasks. Current recommendation methods mainly push similar tasks based on drivers' historical data, which easily creates information silos. This recommendation method also leads to an overabundance of drivers and a shortage of cargo on some popular routes, while some areas (possibly newly developed regions) have an overabundance of cargo and a relative shortage of drivers. If platforms need to provide additional rewards to encourage drivers to accept orders, all of these factors expose the shortcomings and vulnerabilities of existing logistics matching algorithms.

[0004] In view of this, the present invention provides a platform-based method, system, device and storage medium for clustering similar sub-markets.

[0005] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of the present invention, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] To address the problems in the prior art, the present invention aims to provide a platform-based method, system, device, and storage medium for clustering similar sub-markets. This overcomes the difficulties of the prior art and can provide users with transportation routes corresponding to similar sub-markets through deep learning feature embedding methods, greatly improving the accuracy of vehicle-cargo matching.

[0007] Embodiments of the present invention provide a platform-based method for clustering similar sub-markets, comprising the following steps:

[0008] A three-dimensional route vector is established for each sub-market. The maximum values ​​of the three dimensions of the route vector are the total number of departure cities, the total number of destination cities, and the total number of vehicle classes representing vehicle carrying capacity, respectively. The one-hot encoding of the k-th sub-market is defined as SM(k).

[0009] Define the training dataset as

[0010] T: = {(y(i)|x1(i), ...,x N (i), SM(i))}

[0011] Where y(i) represents the label value of the i-th training data, x1(i), ..., x N (i) represents the value of each feature, N represents the total number of features in the training dataset, SM(i) represents the sub-market, i = 1, ..., |T|, and |T| is defined as the size of the training dataset;

[0012] Based on the sub-market of each record in the training dataset, the training data can be sliced ​​into various sub-markets:

[0013] T(k)∶={(y(i)|x t (i), ..., x N (i), SM(i) = SM(k))}

[0014] |T(k)| is defined as the size of the training data subset within each sub-market. The training data for each sub-market is sampled proportionally to the actual transaction data volume in each sub-market.

[0015] All submarkets {SM(1), ..., SM(k), ..., SM(K)} are grouped into M similar submarket clusters {G(1), ..., G(m), ..., G(M)}, and an algorithm model is trained in each cluster.

[0016] Y = FG m (X1, ..., X) N )

[0017] Where M is a preset parameter, m is a natural number between 1 and M, G(m) is the m-th sub-market cluster, K is a preset parameter, k is a natural number between 1 and K, and FG m It is an algorithmic model function in clustering, X N These are the values ​​of each feature in the function;

[0018] Based on the user's historical route data, at least one route task belonging to the same submarket is obtained and pushed to the user.

[0019] Preferably, the step of establishing a three-dimensional route vector for each sub-market, wherein the maximum values ​​of the three dimensions of the route vector are respectively the total number of departure cities, the total number of destination cities, and the total number of vehicle classes representing vehicle carrying capacity, and defining the one-hot encoding of the k-th sub-market as SM(k), further includes:

[0020] The total number of departure cities and the total number of destination cities refer to the total number of cities in the target area, and the total number of vehicle classes are divided into 5 intervals based on vehicle length.

[0021] Preferably, the step of establishing a three-dimensional route vector for each sub-market, wherein the maximum values ​​of the three dimensions of the route vector are respectively the total number of departure cities, the total number of destination cities, and the total number of vehicle classes representing vehicle carrying capacity, and defining the one-hot encoding of the k-th sub-market as SM(k), further includes:

[0022] The one-hot encoding is SM(k), which represents (0, 0, 0, ..., 1, 0, 0, ... 0), where only the k-th bit is 1 and the rest are 0.

[0023] Preferably, the step of grouping all sub-markets {SM(1), ..., SM(k), ..., SM(K)} into M similar sub-market clusters {G(1), ..., G(m), ..., G(M)}, and training an algorithm model in each cluster, includes:

[0024] Define the full training data T: = {(y(i)|x1(i), ...,x} N SM(i) is the default clustering, and a default model is trained:

[0025] Y = FG0(X1, ..., X) N )

[0026] FG0 is a default model function in clustering. Based on the data volume distribution |T(k)| in each sub-market of the training data, it selects the M sub-markets with the largest data volume and trains a model G(m) = T(k), m = 0, ..., M, for each sub-market. The training result is Y = FG0. m (X1, ..., X) N );

[0027] Based on the average error of the training data across M+1 algorithm models, the sub-market training data is assigned to the sub-market cluster with the smallest average error.

[0028] Preferably, the step of selecting the M sub-markets with the largest data volume based on the data volume distribution |T(k)| in each sub-market, and training a model G(m) = T(k) for each sub-market, further includes:

[0029] Calculate the average error when the training data in each sub-market k=1,...K are predicted using these M+1 models respectively, where m=0 is the default model.

[0030] Preferably, the step of assigning the sub-market training data to the sub-market cluster with the smallest average error based on the average error of the training data across M+1 algorithm models includes:

[0031] Iterate through each algorithm model, and initially set the sub-market clustering to an empty set;

[0032] Iterate through each submarket and match an algorithm model with the minimum error for each submarket;

[0033] The sub-market data is incorporated into the sub-market cluster corresponding to the algorithm model;

[0034] The training error of this sub-market is included in the total error.

[0035] Repeat the training process until the total error is minimized.

[0036] Obtain M similar sub-market clusters {G(1), ..., G(m), ..., G(M)}.

[0037] Preferably, obtaining at least one route task belonging to the same submarket based on the user's historical route data and pushing it to the user includes:

[0038] Obtain statistics on sub-market clusters corresponding to each user's historical route data;

[0039] The set of sub-markets clustered with the highest frequency is used as the set of candidate routes;

[0040] Based at least on the user's current location, at least one transportation route task corresponding to a submarket is matched from the set of alternative routes and fed back to the user.

[0041] Preferably, the step of obtaining at least one route task belonging to the same submarket based on the user's historical route data and pushing it to the user further includes:

[0042] The system monitors the popularity of transportation routes and tasks in each sub-market within the alternative route set, prioritizing the matching of transportation routes and tasks with low success rates. This ensures a more even distribution of transportation capacity across different regions and prevents a few areas from remaining unattended for extended periods. As drivers receive new transportation tasks that are highly similar to their familiar historical transportation tasks, this system also protects their interests, but it is not the only measure of its effectiveness.

[0043] Embodiments of the present invention also provide a platform-based similar submarket clustering system for implementing the above-described platform-based similar submarket clustering method. The platform-based similar submarket clustering system includes:

[0044] The route vectorization module establishes a three-dimensional route vector for each sub-market. The maximum values ​​of the three dimensions of the route vector are the total number of departure cities, the total number of destination cities, and the total number of vehicle classes representing vehicle carrying capacity, respectively. The unique hot code of the k-th sub-market is defined as SM(k).

[0045] The training dataset module defines the training dataset as follows:

[0046] T: = {(y(i)|x1(i), ...,x N (i), SM(i))}

[0047] Where y(i) represents the label value of the i-th training data, x1(i), ..., x N (i) represents the value of each feature, N represents the total number of features in the training dataset, SM(i) represents the sub-market, i = 1, ..., |T|, and |T| is defined as the size of the training dataset;

[0048] The training data slicing module can slice the training data into various sub-markets based on the sub-market of each record in the training dataset:

[0049] T(k):={(y(i)|x1(i),…,x N (i), SM(i) = SM(k))}

[0050] |T(k)| is defined as the size of the training data subset within each sub-market. The training data for each sub-market is sampled proportionally to the actual transaction data volume in each sub-market.

[0051] The algorithm model training module will train all sub-markets.

[0052] {SM(1), …, SM(k), …, SM(K)} are grouped into M similar sub-market clusters {G(1), …, G(m), …, G(M)}, and an algorithm model is trained in each cluster.

[0053] Y = FG m (X1, ..., X) N )

[0054] Where M is a preset parameter, m is a natural number between 1 and M, G(m) is the m-th sub-market cluster, K is a preset parameter, k is a natural number between 1 and K, and FG m It is an algorithmic model function in clustering, X N These are the values ​​of each feature in the function;

[0055] The route task push module obtains at least one route task belonging to the same submarket based on the user's historical route data and pushes it to the user.

[0056] Embodiments of the present invention also provide a platform-based similar submarket clustering device, comprising:

[0057] processor;

[0058] Memory, which stores the processor's executable instructions;

[0059] The processor is configured to execute the steps of the platform-based similar submarket clustering method described above by executing executable instructions.

[0060] Embodiments of the present invention also provide a computer-readable storage medium for storing a program that, when executed, implements the steps of the platform-based similar submarket clustering method described above.

[0061] The platform-based similar sub-market clustering method, system, device, and storage medium of the present invention can provide users with transportation routes corresponding to similar sub-markets through deep learning feature embedding methods, greatly improving the accuracy of vehicle-cargo matching. Attached Figure Description

[0062] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0063] Figure 1 This is a flowchart of the platform-based similar submarket clustering method of the present invention.

[0064] Figure 2 This is a schematic diagram illustrating an implementation scenario of the platform-based similar sub-market clustering method of the present invention.

[0065] Figure 3 This is a schematic diagram of the structure of the platform-based similar sub-market clustering system of the present invention.

[0066] Figure 4 This is a schematic diagram of the platform-based similar sub-market clustering device of the present invention.

[0067] Figure 5 This is a schematic diagram of the structure of a computer-readable storage medium according to an embodiment of the present invention. Detailed Implementation

[0068] The following specific examples illustrate the implementation methods of this application. Those skilled in the art can easily understand the other advantages and effects of this application from the content disclosed herein. This application can also be implemented or applied through other different specific embodiments, and various details in this application can be modified or changed according to different viewpoints and application systems without departing from the spirit of this application. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0069] The embodiments of this application will now be described in detail with reference to the accompanying drawings, so that those skilled in the art can easily implement the application. This application may be embodied in many different forms and is not limited to the embodiments described herein.

[0070] In this application, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics represented in connection with that embodiment or example, which are included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics represented may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate different embodiments or examples represented in this application, as well as features of different embodiments or examples.

[0071] Furthermore, the terms "first" and "second" are used for illustrative purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the representation of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0072] For the purpose of clearly describing this application, devices that are not relevant to the description are omitted, and the same or similar components throughout the specification are given the same reference numerals.

[0073] Throughout this specification, when it is said that a device is "connected" to another device, this includes not only "direct connection" but also "indirect connection" by placing other components in between. Furthermore, when it is said that a device "comprises" a certain constituent element, unless otherwise stated otherwise, this does not exclude other constituent elements, but rather implies that other constituent elements may be included.

[0074] When we say that a device is "above" another device, this can mean that it is directly above the other device, or it can mean that other devices are present in between. Conversely, when we say that a device is "directly" "above" another device, there are no other devices present in between.

[0075] While the terms first, second, etc., are used in some instances to denote various elements in this invention, these elements should not be limited by these terms. These terms are used only to distinguish one element from another. For example, first interface and second interface, etc., are used. Furthermore, as used herein, the singular forms “a,” “an,” and “the” are intended to also include the plural forms, unless the context indicates otherwise. It should be further understood that the terms “comprising,” “including,” indicate the presence of features, steps, operations, elements, components, items, kinds, and / or groups, but do not exclude the presence, occurrence, or addition of one or more other features, steps, operations, elements, components, items, kinds, and / or groups. The terms “or” and “and / or” as used herein are to be interpreted inclusively, or mean any one or any combination thereof. Therefore, “A, B, or C” or “A, B, and / or C” means “any one of: A; B; C; A and B; A and C; B and C; A, B, and C.” An exception to this definition will only occur if the combination of elements, functions, steps, or operations is inherently mutually exclusive in some way.

[0076] The technical terms used herein are for reference only to specific embodiments and are not intended to limit the scope of this application. The singular form used herein includes the plural form unless the statement explicitly indicates otherwise. The word "comprising" as used in the specification means to specify a particular characteristic, region, integer, step, operation, element, and / or component, and does not exclude the presence or addition of other characteristics, regions, integers, steps, operations, elements, and / or components.

[0077] Although not explicitly defined, all terms, including technical and scientific terms used herein, shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. Terms defined in commonly used dictionaries shall be further interpreted as having a meaning consistent with the relevant technical literature and the content of this present application, and shall not be over-interpreted as having an ideal or overly formulaic meaning unless otherwise defined.

[0078] Figure 1 This is a flowchart of the platform-based similar sub-market clustering method of the present invention. Figure 1 As shown, an embodiment of the present invention provides a platform-based method for clustering similar sub-markets, including the following steps:

[0079] S110. Establish a three-dimensional route vector for each sub-market. The maximum values ​​of the three dimensions of the route vector are the total number of departure cities, the total number of destination cities, and the total number of vehicle classes representing vehicle carrying capacity, respectively. Define the one-hot encoding of the k-th sub-market as SM(k).

[0080] S120, Define the training data set as...

[0081] T: = {(y(i)|x1(i), ...,x N (i), SM(i))}

[0082] Where y(i) represents the label value of the i-th training data, x1(i), ..., x N (i) represents the value of each feature, N represents the total number of features in the training dataset, SM(i) represents the sub-market, i = 1, ..., |T|, and |T| is defined as the size of the training dataset.

[0083] S130. Based on the sub-market of each record in the training dataset, the training data can be sliced ​​into various sub-markets:

[0084] T(k):={(y(i)|x1(i),…,x N (i), sm(i) = SM(k))}

[0085] |T(k)| is defined as the size of the training data subset within each sub-market. The training data for each sub-market is sampled proportionally to the actual transaction data volume in each sub-market.

[0086] S140. All submarkets {SM(1), ..., SM(k), ..., SM(K)} are grouped into M similar submarket clusters {G(1), ..., G(m), ..., G(M)}, and an algorithm model is trained in each cluster.

[0087] Y = FG m (X1, ..., X) N )

[0088] Where M is a preset parameter, m is a natural number between 1 and M, G(m) is the m-th sub-market cluster, K is a preset parameter, k is a natural number between 1 and K, and FG m It is an algorithmic model function in clustering, X N These are the values ​​of the various features in the function.

[0089] S150: Based on the user's historical route data, obtain at least one route task belonging to the same sub-market and push it to the user.

[0090] In a preferred embodiment, step S110 further includes:

[0091] The total number of departure cities and the total number of destination cities are both the total number of cities in the target area, and the total number of vehicle classes are divided into 5 intervals based on vehicle length.

[0092] In a preferred embodiment, step S110 further includes:

[0093] One-hot encoding is SM(k) representing (0, 0, 0, ..., 1, 0, 0, ... 0), where only the k-th bit is 1 and the rest are 0.

[0094] In a preferred embodiment, step S130 includes:

[0095] S131. Based on the sub-market of each record in the training dataset, the training data can be sliced ​​into various sub-markets according to the corresponding departure city code, destination city code, and vehicle level code corresponding to the vehicles in the training data:

[0096] T(k):={(y(i)|x1(i),…,x N (i), sm(i) = SM(k))}

[0097] |T(k)| is defined as the size of the training data subset within each sub-market. The training data for each sub-market is sampled proportionally to the actual transaction data volume in each sub-market.

[0098] In a preferred embodiment, step S140 includes:

[0099] S141. Define the full training data T: = {(y(i)|x1(i), ...,x...} N SM(i) is the default clustering, and a default model is trained:

[0100] Y = FG0(X1, ..., X) N )

[0101] FG0 is a default model function in clustering;

[0102] S142. Based on the data volume distribution |T(k)| in each sub-market, select the M sub-markets with the largest data volume, and train a model G(m)=T(k) for each sub-market, m=0,...M, and train Y=FG. m (X1, ..., X) N ).

[0103] S143. Based on the average error of the training data across M+1 algorithm models, assign the sub-market training data to the sub-market cluster with the smallest average error.

[0104] In a preferred embodiment, step S142 further includes:

[0105] Calculate the average error when the training data in each sub-market k=1,...K are predicted using these M+1 models respectively, where m=0 is the default model.

[0106] In a preferred embodiment, step S143 includes:

[0107] S1431. Traverse each algorithm model, and initially set the sub-market clustering to an empty set.

[0108] S1432. Traverse each submarket and match an algorithm model with the minimum error for each submarket.

[0109] S1433, Incorporate the sub-market data into the sub-market cluster corresponding to the algorithm model.

[0110] S1434, Include the training error of this sub-market in the total error.

[0111] S1435. Repeat the training until the total error is minimized.

[0112] S1436. Obtain M similar sub-market clusters {G(1), ..., G(m), ..., G(M)}.

[0113] In a preferred embodiment, step S150 includes:

[0114] S151. Obtain statistics on sub-market clusters corresponding to each user's historical route data.

[0115] S152. The set of the top few sub-markets with the highest frequency is used as the set of candidate routes.

[0116] S153. Based at least on the user's current location, match at least one transportation route task corresponding to a sub-market from the set of available routes and provide feedback to the user.

[0117] In a preferred embodiment, the popularity of transportation route tasks in each sub-market of the alternative route set is monitored, and transportation route tasks with low matching success rates are matched first, thereby making the capacity matching in different regions more even and avoiding a situation where no one accepts orders in a few areas for a long time. As drivers receive new transportation tasks that are still highly similar to their familiar historical transportation tasks, the interests of drivers are also protected, but this is not the only limitation.

[0118] The platform-based similar submarket clustering method of the present invention can provide users with transportation routes corresponding to similar submarkets through deep learning feature embedding, which greatly improves the accuracy of vehicle-cargo matching.

[0119] From a macro perspective, the national highway freight market is a vast, interconnected network. However, in actual operation, the substitutability between various transport products and routes (origin + destination) and typical vehicles (primarily characterized by length and type) operating on a given route is limited. For example, a driver operating a 9.6-meter van between point A and point B typically chooses this route and purchases this type of vehicle primarily because of relatively stable freight resources along the route, familiarity with the roads and parking locations, and historical experience indicating a profitable business cycle. Changing routes easily is difficult because it involves increased costs and other risks.

[0120] Within a specific timeframe, the national road freight market clearly exhibits characteristics determined by transportation routes, whether in terms of the composition of cargo demand, the quantity and distribution of transport capacity, or even transportation prices and their fluctuation trends. Extensive data demonstrates that while drivers' chosen routes may involve some wandering, most drivers prefer to haul goods on routes they are familiar with. The primary reason behind this phenomenon lies in the fact that road freight is essentially a B2B (business-to-business) business, and the fundamental profit model of B2B services is "repeated successful actions." Repetition leads to increased efficiency, savings in time and other costs, and reduced risk, thereby ensuring profitability.

[0121] Based on the above observations, the entire road freight market can be defined as a collection of numerous "partially substitutable submarkets," with the granularity of a submarket defined as "route (origin city + destination city) x vehicle length." Currently, there are 333 prefecture-level cities and 4 municipalities directly under the central government in China. If vehicle length is divided into 5 segments based on typical data, then there are 337 * 337 * 5 = 567,845 submarkets. Throughout the entire lifecycle of a transportation demand (freight information), from its posting on the platform to its acceptance by a driver, the distribution of supply and demand, as well as prices, varies significantly depending on the submarket it belongs to. This has a profound impact on driver order-accepting decisions and platform scheduling.

[0122] From a data perspective, the national highway freight market exhibits significant differentiation across its various sub-markets. Some sub-markets experience robust and relatively balanced supply and demand, resulting in abundant data. Others suffer from severe supply-demand imbalances, while still others see only sporadic freight transactions over the years, leading to highly sparse data. In sub-markets with sparse data, it is difficult to generate effective statistical values ​​and reach a consensus (standardization) based on these statistics. Therefore, it is essential to merge sub-markets with similar behaviors to ensure the rationality of statistical values ​​in terms of both quantity and quality.

[0123] Because the road freight market is a typical "route-level market," the composition of freight demand, the quantity and distribution of transport capacity, and even transport prices and their fluctuation trends clearly demonstrate characteristics determined by the routes. As mentioned above, if we define the granularity of a sub-market as "route (origin city + destination city) x vehicle length," then there are 337 * 337 * 5 = 567,845 sub-markets nationwide. Except for a few sub-markets with relatively rich business data, the data for most sub-markets is highly sparse.

[0124] On the other hand, current machine learning and deep learning algorithms used in industry are generally based on massive features and massive amounts of data for modeling. When the model encounters certain features with a large number of unordered discrete values, also known as "high cardinality" features, such as the origin city ID and destination city ID in a route, the algorithm heavily relies on increasing model complexity and training dataset size. This results in a large-scale model with poor interpretability and huge server computing power consumption. The discovery and merging of similar sub-markets can avoid introducing such high cardinality features into the model, ensuring that the model size is controlled, the physical meaning is clearer, and the interpretability is improved. This is the main technical deficiency faced by current algorithms that this invention aims to address.

[0125] The implementation process of this invention is as follows:

[0126] (1) One-hot encoding in the road freight submarket

[0127] A sub-market's granularity is defined as "route (origin city + destination city) x vehicle length". The primary key of such a sub-market is uniquely determined by the value of (from_city_id, to_city_id, truck_length_id). Currently, there are 333 prefecture-level cities and 4 municipalities directly under the central government in China. If the vehicle length is divided into 5 segments based on typical data, then there are 337 * 337 * 5 = 567,845 sub-markets.

[0128] One-hot encoding means that each submarket is represented by a 567,845-dimensional vector. For a specific submarket, 567,844 elements are 0, and only one dimension has a value of 1, representing the current submarket. For the k-th submarket, its one-hot encoding is as follows: (0,0,0,...,1,0,0,...0), where only the k-th bit is 1, and the rest are 0. The one-hot encoding for each submarket is unique.

[0129] The one-hot encoding of the k-th sub-market is defined as SM(k), and the global market is defined as the set of all sub-markets.

[0130] GM:={SM(1),…,SM(k),…,SM(K)}

[0131] Therefore, K = |GM| = 567845 is the total number of sub-markets that need to be considered.

[0132] 2) Supervised training dataset

[0133] The goal of machine learning is to predict an output value using a set of input features. In the context of road freight, the output result may be the probability of a transaction [0,1], whether the transaction is successful (0 / 1), or the price of the transaction [0,] etc.

[0134] Since the road freight market is a typical "route-level market" with weak substitutability among different vehicle lengths, the submarket SM(k) defined above is an important input feature.

[0135] The goal of supervised learning is to build an algorithmic model.

[0136] Y = F(X1,…,X) N ,SM)

[0137] Where X1,…,X N SM is the set of input features for model training, and Y is the output. SM represents the sub-market in which the data exists, and is a high-cardinality feature. If SM is used as an input feature for modeling, the resulting algorithm model will have high complexity and poor interpretability.

[0138] Define the training dataset as

[0139] T: = {(y(i)|x1(i),…,x} N (i),sm(i))}

[0140] Where y(i) represents the label value of the i-th training data, x1(i),…,x N (i) represents the values ​​of each feature, and sm(i) represents the sub-market. |T| is the size of the training dataset, i = 1, ..., |T|. Based on the sub-market of each record in the training dataset, the training data can be sliced ​​into various sub-markets:

[0141] T(k): = {(y(i)|x1(i),…,x N (i),sm(i)=SM(k))}

[0142] |T(k)| is defined as the size of the subset of training data within each sub-market. The training data for each sub-market is sampled proportionally to the actual transaction data volume in each sub-market.

[0143] (3) Special considerations for training features

[0144] Since the national highway freight market is a natural LBS (Location Based Service), the distribution of actual geographical locations, such as the region, province, and city of origin and the region, province, and city of destination, has a significant impact on the similarity between various sub-markets. Based on the above assumptions, in the feature set {X1,…,X...} of the training data... N In this context, continuous numerical features such as the latitude and longitude of the origin, the latitude and longitude of the destination, and the polar coordinates from the origin to the destination are consciously introduced.

[0145] To further characterize the sub-markets, it is necessary to incorporate time-specific sub-market demand and supply statistics, departure city demand and supply statistics, and departure province demand and supply statistics into the training data. In practice, these statistics are highly correlated with the predicted output Y.

[0146] (4) Model training process

[0147] As mentioned above, the goal of supervised learning is to build an algorithmic model.

[0148] Y = F(X1,…,X) N ,SM)

[0149] The goal of this invention is to group all submarkets {SM(1),…,SM(k),…,SM(K)} into M similar submarket clusters {G(1),…,G(m),…,G(M)}, and to train an algorithm model in each cluster.

[0150] Y = FG m (X1,…,X N )

[0151] Where m = 1, ..., M.

[0152] If we simply use an unsupervised clustering algorithm to obtain these M sub-market clusters beforehand, it's difficult to clearly define the so-called "similarity" according to the final objective function. Here, "similarity" is defined based on the accuracy of predicting Y. The following is the algorithm of this invention:

[0153] Step 0, Train the default model

[0154] Initially, the full training data T is defined as: = {(y(i)|x1(i),…,x...} N (i),SM(i))} represents the default clustering, and a default model is trained:

[0155] Y = FG0(X1,…,X) N )

[0156] Depending on the desired output, such as the probability of a transaction [0,1], whether a transaction was completed (0 / 1), or the price of the transaction [0,], the training method can employ linear or nonlinear classification or clustering algorithms used in traditional machine learning and deep learning.

[0157] Step 1, train M models (M is a selected constant, such as 300).

[0158] Based on the data volume distribution |T(k)| in each sub-market of the training data, select the M sub-markets with the largest data volume, where M is the expected number of clusters to be generated. Train a model for each sub-market.

[0159] For m=1,...M

[0160] G(m)=T(k)##Using the M sub-market data with the largest training data volume

[0161] Training Y=FG m (X1,…,X N Training the model

[0162] This results in M+1 trained algorithm models, m = 0, ..., M, where m = 0 is the default model. The average error is calculated when these M+1 models are used to predict the training data in each sub-market k = 1, ..., K.

[0163]

[0164] In this invention, an existing set of training models can be used to train the data of each sub-market. By statistically analyzing the errors of different training models, the training model with the smallest error is selected as the training model for that sub-market. This will not be elaborated further here.

[0165] Step 2: Classify data and calculate overall error.

[0166] Based on the average error of the training data across the aforementioned M+1 algorithm models, the sub-market training data is assigned to the sub-market cluster with the smallest error.

[0167]

[0168]

[0169] Step 3: Repeat the training until the error stops decreasing.

[0170]

[0171] (5) Model Output

[0172] As can be seen from the training process described above, this method will ultimately yield M similar sub-market clusters {G(1),…,G(m),…,G(M)}, and one default cluster. Each sub-market cluster corresponds to an algorithm model.

[0173] Y = FG m (X1,…,X N ), m=0,…M

[0174] The term "similarity" here refers to minimizing the overall error on the defined error function.

[0175] Figure 2 This is a schematic diagram illustrating an implementation scenario of the platform-based similar sub-market clustering method of the present invention. Figure 2 As shown, through the above steps, several clusters were obtained in the custom training dataset 2. Then, based on the order information 11, 12, 13, 14, 15, and 16 of vehicle 1's historical data, the number of times each sub-market cluster was completed was statistically analyzed. Among them, vehicle 1 completed clusters 21 and 22 the most times, so the set of sub-market clusters corresponding to clusters 21 and 22 was used as the candidate route set 3. Finally, a corresponding transportation route task 31 was selected from the candidate route set 3. At least based on the user's current location, a transportation route task 31 was matched from the candidate route set 3 and fed back to the user of vehicle 1. This allows the user to be matched with a new transportation task that is highly similar to their familiar historical transportation tasks, avoiding the risk of the user randomly selecting a new transportation task and improving the accuracy of customer matching, but this is not the only limitation.

[0176] In another variation, by monitoring the popularity of transportation route tasks in alternative route set 3, some less popular tasks can be sent to suitable drivers, thereby making the capacity matching in different regions more even and avoiding a situation where no one accepts orders in a few areas for a long time. As drivers receive new transportation tasks that are still highly similar to the historical transportation tasks they are familiar with, the interests of drivers are also protected, but this is not the only way.

[0177] This solution involves simultaneously maintaining M+1 sub-market clusters and M+1 algorithm models. The algorithm heavily relies on the ability to process massive amounts of data, placing extremely high demands on computing power. The machine learning method used for each sub-market cluster can be either complex or relatively simple. By eliminating high cardinality features in the overall model, the complexity of each model decreases, and interpretability is improved.

[0178] From a computational complexity perspective, since the total number of clusters M is fixed, the computational cost for estimating the error in each iteration increases linearly from |P| calculations for a single model to M·|P| calculations. Because each model is trained only on sub-market clusters, the overall computational cost increase is minimal. By appropriately choosing M, similar sub-markets can be effectively identified, and efficient training can be ensured.

[0179] The basis of this invention is that the platform has sufficient cargo volume and a large amount of driver behavior data, which is a natural barrier. Without sufficient network density, or for freight lines with only a few routes in operation, splitting sub-markets and clustering will become almost meaningless.

[0180] The problem addressed by this invention can be partially solved using deep learning feature embedding methods. However, ordinary feature embeddings often struggle to establish a direct correlation with the final predicted output value. Deep learning algorithms incorporating feature embeddings inherently suffer from poor interpretability. This invention offers an end-to-end training method, resulting in a simpler model with significantly improved interpretability.

[0181] As an algorithm-based patent, the actual effectiveness of this invention is highly dependent on the correlation between the problem to be solved and the lines and line clustering, as well as the selected set of input features and the output value to be predicted. This requires continuous analysis and exploration of massive amounts of real data. Therefore, this invention has a certain natural barrier to simple infringement.

[0182] Because the road freight market is a typical "route-level market," the actual operation of offline business is usually organized according to routes. The composition of freight demand, the quantity and distribution of transport capacity, and even the transport price and price fluctuation trend clearly show characteristics determined by the routes, and the substitutability of transport capacity across routes and vehicle lengths is relatively low. On the other hand, although the total volume of national road transport is huge, it is dispersed across a large number of routes, and there are many routes with extremely sparse data, making it difficult to build mathematical models for vehicle-freight matching and price prediction on these routes based on this data.

[0183] This invention proposes an effective end-to-end training method that, through a controllable increase in computational load, can obtain reasonable estimates of similar sub-markets in the road freight market. In the similar sub-market clustering generated by this algorithm, the problem of data sparsity is effectively solved, the performance of vehicle-cargo matching and freight pricing algorithms is significantly improved, and the interpretability of the algorithm model is also enhanced, which is of great value to the platform's core business.

[0184] Figure 3 This is a schematic diagram of the platform-based similar sub-market clustering system of the present invention. Figure 3As shown, the platform-based similar sub-market clustering system 5 of the present invention includes:

[0185] The route vectorization module 51 establishes a three-dimensional route vector for each sub-market. The maximum values ​​of the three dimensions of the route vector are the total number of departure cities, the total number of destination cities, and the total number of vehicle classes representing vehicle carrying capacity, respectively. The unique hot code of the k-th sub-market is defined as SM(k).

[0186] Training data set module 52, defines the training data set as follows

[0187] P: = {(y(i)|x1(i), ...,x} N (i), SM(i))}

[0188] Where y(i) represents the label value of the i-th training data, x1(i), ..., x N (i) represents the value of each feature, N represents the total number of features in the training dataset, SM(i) represents the sub-market, i = 1, ..., |T|, and |T| is defined as the size of the training dataset;

[0189] The training data slicing module 53 can slice the training data into various sub-markets based on the sub-market of each record in the training dataset:

[0190] T(k):={(y(i)|x1(i),…,x N (i), sm(i) = SM(k))}

[0191] |T(k)| is defined as the size of the training data subset within each sub-market. The training data for each sub-market is sampled proportionally to the actual transaction data volume in each sub-market.

[0192] The algorithm model training module 54 groups all sub-markets {SM(1), ..., SM(k), ..., SM(K)} into M similar sub-market clusters {G(1), ..., G(m), ..., G(M)}, and trains an algorithm model in each cluster.

[0193] Y = FG m (X1, ..., X) N )

[0194] Where M is a preset parameter, m is a natural number between 1 and M, G(m) is the m-th sub-market cluster, K is a preset parameter, k is a natural number between 1 and K, and FG m It is an algorithmic model function in clustering, X N These are the values ​​of each feature in the function;

[0195] The route task push module 55 obtains at least one route task belonging to the same sub-market based on the user's historical route data and pushes it to the user.

[0196] In a preferred embodiment, the total number of departure cities and the total number of target cities in the route vectorization module 51 are both the total number of cities in the target area, and the total number of vehicle levels are 5 intervals divided according to the vehicle length.

[0197] In a preferred embodiment, the one-hot encoding in the route vectorization module 51 is SM(k) representing (0, 0, 0, ..., 1, 0, 0, ... 0), where only the k-th bit is 1 and the rest are 0.

[0198] In a preferred embodiment, the algorithm model training module 54 is configured to define the full training data T: = {(y(i)|x1(i), ...,x...} N SM(i) is the default clustering, and a default model is trained:

[0199] Y = FG0(X1, ..., X) N )

[0200] FG0 is a default model function in clustering;

[0201] Based on the data volume distribution |T(k)| in each sub-market, select the M sub-markets with the largest data volume, and train a model G(m)=T(k) for each sub-market, m=0,...M, and train Y=FG. m (X1, ..., X) N );

[0202] Based on the average error of the training data across M+1 algorithm models, the sub-market training data is assigned to the sub-market cluster with the smallest average error.

[0203] In a preferred embodiment, the algorithm model training module 54 calculates the average error when the training data in each sub-market k=1,...K is predicted using these M+1 models, where m=0 is the default model.

[0204] In a preferred embodiment, the algorithm model training module 54 is further configured to traverse each algorithm model, with the sub-market cluster initially set to an empty set; traverse each sub-market and match an algorithm model with the smallest error for each sub-market; incorporate the sub-market data into the sub-market cluster corresponding to the algorithm model; include the training error of the sub-market in the total error; perform iterative training until the total error is minimized; and obtain M similar sub-market clusters {G(1), ..., G(m), ..., G(M)}.

[0205] In a preferred embodiment, the route task push module 55 is configured to obtain statistics on sub-market clusters corresponding to each of the user's historical route data; take the set of the top several sub-market clusters with the highest frequency as the candidate route set; and, based at least on the user's current location, match at least one transportation route task corresponding to a sub-market from the candidate route set and feed it back to the user.

[0206] The platform-based similar sub-market clustering system of the present invention can provide users with transportation routes corresponding to similar sub-markets through deep learning feature embedding methods, which greatly improves the accuracy of vehicle-cargo matching.

[0207] This invention also provides a platform-based similar submarket clustering device, including a processor and a memory storing executable instructions for the processor. The processor is configured to execute steps of a platform-based similar submarket clustering method via the executable instructions.

[0208] As described above, the platform-based similar sub-market clustering device of the present invention can provide users with transportation routes corresponding to similar sub-markets through deep learning feature embedding methods, which greatly improves the accuracy of vehicle-cargo matching.

[0209] Those skilled in the art will understand that various aspects of the present invention can be implemented as systems, methods, or program products. Therefore, various aspects of the present invention can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software aspects, collectively referred to herein as a "circuit," "module," or "platform."

[0210] Figure 4 This is a schematic diagram of the platform-based similar sub-market clustering device of the present invention. See below for reference. Figure 4 To describe an electronic device 600 according to this embodiment of the present invention. Figure 4 The electronic device 600 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0211] like Figure 4 As shown, the electronic device 600 is presented in the form of a general-purpose computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including storage unit 620 and processing unit 610), a display unit 640, etc.

[0212] The storage unit stores program code, which can be executed by the processing unit 610 to perform the steps described in the above-described section on the electronic prescription transfer processing method according to various exemplary embodiments of the present invention. For example, the processing unit 610 can perform actions such as... Figure 1 The steps are shown in the figure.

[0213] Storage unit 620 may include a readable medium in the form of a volatile storage unit, such as random access memory (RAM) 6201 and / or cache memory 6202, and may further include a read-only memory (ROM) 6203.

[0214] Storage unit 620 may also include a program / utility 6204 having a set (at least one) program module 6205, such program module 6205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0215] Bus 630 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0216] Electronic device 600 can also communicate with one or more external devices 700 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 600, and / or with any device that enables electronic device 600 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 650. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 660. Network adapter 660 can communicate with other modules of electronic device 600 via bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0217] This invention also provides a computer-readable storage medium for storing a program that, when executed, implements the steps of a platform-based similar submarket clustering method. In some possible implementations, various aspects of the invention can also be implemented as a program product comprising program code that, when run on a terminal device, causes the terminal device to perform the steps described in the above-described electronic prescription processing method section of this specification according to various exemplary embodiments of the invention.

[0218] As shown above, when the program of the computer-readable storage medium of this embodiment is executed, it can provide users with transportation routes corresponding to similar sub-markets through a deep learning feature embedding method, which greatly improves the accuracy of vehicle-cargo matching.

[0219] Figure 5 This is a schematic diagram of the structure of the computer-readable storage medium of the present invention. (Reference) Figure 5 As shown, a program product 800 for implementing the above-described method according to an embodiment of the present invention is described. This product may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0220] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0221] Computer-readable storage media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transfer a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0222] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0223] In summary, the platform-based similar sub-market clustering method, system, device, and storage medium of the present invention can provide users with transportation routes corresponding to similar sub-markets through deep learning feature embedding methods, greatly improving the accuracy of vehicle-cargo matching.

[0224] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A platform-based clustering method for similar sub-markets, characterized in that, Includes the following steps: A three-dimensional route vector is established for each sub-market. The maximum values ​​of the three dimensions of the route vector are the total number of departure cities, the total number of destination cities, and the total number of vehicle classes representing vehicle carrying capacity, respectively. The unique hot code of the k-th sub-market is defined as SM(k). The total number of departure cities and the total number of destination cities are the total number of cities in the target area. The total number of vehicle classes is divided into 5 intervals based on vehicle length. The unique hot code SM(k) represents (0, 0, 0, ..., 1, 0, 0, ... 0), with only the k-th bit being 1 and the rest being 0. Define the training data set as T: = {(y(i)|x1(i), ...,x...} N (i), SM(i))} Where y(i) represents the label value of the i-th training data, x1(i), ..., x N (i) represents the value of each feature, N represents the total number of features in the training dataset, SM(i) represents the sub-market, i = 1, ..., |T|, and |T| is defined as the size of the training dataset; Based on the sub-market of each record in the training dataset, the training data can be sliced ​​into various sub-markets: T(k)∶={(y(i)|x1(i),…,x N (i),SM(i)=SM(k))} |T(k)| is defined as the size of the subset of training data within each sub-market; All submarkets {SM(1), ..., SM(k), ..., SM(K)} are grouped into M similar submarket clusters {G(1), ..., G(m), ..., G(M)}, and an algorithm model is trained in each cluster. Y=FG m (X1,…,X N ) Where M is a preset parameter, m is a natural number between 1 and M, G(m) is the m-th sub-market cluster, K is a preset parameter, k is a natural number between 1 and K, and FG m It is an algorithmic model function in clustering, X N These are the values ​​of each feature in the function; Obtain statistics on sub-market clusters corresponding to each of the user's historical route data; select the set of the top few sub-market clusters with the highest frequency as the candidate route set; monitor the popularity of transportation route tasks in each sub-market of the candidate route set; based at least on the user's current location, match at least one transportation route task corresponding to a sub-market from the candidate route set and feed it back to the user, prioritizing the matching of transportation route tasks with low matching success rates.

2. The platform-based similar sub-market clustering method as described in claim 1, characterized in that, The method involves grouping all submarkets {SM(1), ..., SM(k), ..., SM(K)} into M similar submarket clusters {G(1), ..., G(m), ..., G(M)}, and training an algorithm model in each cluster, including: Define the full training data T: = {(y(i)|x1(i), ...,x} N SM(i) is the default clustering, and a default model is trained: Y=FG " (X1,…,X N ) FG0 is a default model function in clustering. Based on the data volume distribution |T(k)| in each sub-market, it selects the M sub-markets with the largest data volume and trains a model G(m) = T(k) for each sub-market, and trains Y = FG0. m (X1, ..., X) N ); Based on the average error of the training data across M+1 algorithm models, the sub-market training data is assigned to the sub-market cluster with the smallest average error.

3. The platform-based similar sub-market clustering method as described in claim 2, characterized in that, The step of selecting the M sub-markets with the largest data volume based on the data volume distribution |T(k)| in each sub-market according to the training data, and training a model G(m) = T(k) for each sub-market, also includes: Calculate the average error when using these M+1 models to predict the training data in each sub-market, where m=0 is the default model.

4. The platform-based similar sub-market clustering method as described in claim 3, characterized in that, The step of assigning the sub-market training data to the sub-market cluster with the smallest average error based on the average error of the training data across M+1 algorithm models includes: Iterate through each algorithm model, and initially set the sub-market clustering to an empty set; Iterate through each submarket and match an algorithm model with the minimum error for each submarket; The sub-market data is incorporated into the sub-market cluster corresponding to the algorithm model; The training error of this sub-market is included in the total error. Repeat the training process until the total error is minimized. Obtain M similar sub-market clusters {G(1), ..., G(m), ..., G(M)}.

5. A platform-based clustering system for similar sub-markets, characterized in that, The system includes: The route vectorization module establishes a three-dimensional route vector for each sub-market. The maximum values ​​of the three dimensions of the route vector are the total number of departure cities, the total number of destination cities, and the total number of vehicle classes representing vehicle carrying capacity, respectively. The unique hot code of the k-th sub-market is defined as SM(k). The total number of departure cities and the total number of destination cities are the total number of cities in the target area. The total number of vehicle classes is divided into 5 intervals based on vehicle length. The unique hot code SM(k) represents (0, 0, 0, ..., 1, 0, 0, ... 0), with only the k-th bit being 1 and the rest being 0. The training dataset module defines the training dataset as T: = {(y(i)|x1(i), ..., x...} N (i), SM(i))} Where y(i) represents the label value of the i-th training data, x1(i), ..., x N (i) represents the value of each feature, N represents the total number of features in the training dataset, SM(i) represents the sub-market, i = 1, ..., |T|, and |T| is defined as the size of the training dataset; The training data slicing module can slice the training data into various sub-markets based on the sub-market of each record in the training dataset: T(k)∶={(y(i)|x1(i),…,x N (i),SM(i)=SM(k))} |T(k)| is defined as the size of the training data subset within each sub-market. The training data for each sub-market is sampled proportionally to the actual transaction data volume in each sub-market. The algorithm model training module categorizes all sub-markets {SM(1), ..., SM(k), ..., SM(K)} into M similar sub-market clusters {G(1), ..., G(m), ..., G(M)}, and trains an algorithm model in each cluster. Y=FG m (X1,…,X N ) Where M is a preset parameter, m is a natural number between 1 and M, G(m) is the m-th sub-market cluster, K is a preset parameter, k is a natural number between 1 and K, and FG m It is an algorithmic model function in clustering, X N These are the values ​​of each feature in the function; The route task push module obtains statistics on sub-market clusters corresponding to each of the user's historical route data; it takes the set of the top few sub-market clusters with the highest frequency as the candidate route set; it monitors the popularity of transportation route tasks in each sub-market of the candidate route set, and at least based on the user's current location, it matches at least one transportation route task corresponding to a sub-market from the candidate route set and feeds it back to the user, prioritizing the matching of transportation route tasks with low matching success rates.

6. A platform-based similar sub-market clustering device, characterized in that, include: processor; A memory in which executable instructions of the processor are stored; The processor is configured to perform the steps of the platform-based similar submarket clustering method according to any one of claims 1 to 4 by executing the executable instructions.

7. A computer-readable storage medium for storing a program, characterized in that, When the program is executed, it implements the steps of the platform-based similar submarket clustering method according to any one of claims 1 to 4.

Images