A vehicle recommendation method and device for an online freight platform based on evaluation tags

Abstract

The present invention relates to a vehicle recommendation method and device for an online freight platform based on evaluation tags. When a cargo owner initiates an order, a tag set is first generated based on the evaluation tags stored in the cargo owner's historical transaction records, and a vehicle set to be recommended is generated based on vehicles that currently meet the rigid requirements of cargo transportation and vehicle type. Then, the total score of each vehicle to be recommended under the cargo owner's tag set is calculated using a scoring discriminant, and all vehicles to be recommended are arranged in descending order according to the total score in the recommendation set. Finally, the recommendation set results are returned to the cargo owner, thereby realizing personalized vehicle recommendations based on the cargo owner's historical transaction evaluation behavior.

Description

Technical Field

[0001] The present invention relates to the field of information system recommendation technology, and in particular to a method and device for recommending vehicles for an online freight platform based on evaluation tags, which is applied to a capacity sharing platform for truckless carriers. Background Art

[0002] Against the backdrop of big data technology development, the emergence of truckless carrier platforms has significantly standardized and integrated the allocation of freight resources across society, effectively avoiding wasteful logistics. As the "Internet Plus" movement enters its second half, platforms are shifting their product focus toward personalized, customized services for customers, prioritizing the user experience. In the actual vehicle-cargo matching process, shippers, while generally meeting their cargo transportation needs, are beginning to develop new demands for service experiences such as truck arrival punctuality, driver responsibility during transportation, and damage control. These preferences often differ between shippers.

[0003] Therefore, the present invention proposes a standardized vehicle recommendation method and device that can fully explore the interests and preferences of different cargo owners and at the same time can highly combine the cargo owner's needs with the actual vehicle information. Summary of the Invention

[0004] The purpose of the present invention is to provide a vehicle recommendation method and device for an online freight platform based on evaluation tags. When a cargo owner initiates an order, a tag set is first generated based on the evaluation tags stored in the cargo owner's historical transaction records, and a vehicle set to be recommended is generated based on vehicles that currently meet the rigid requirements of cargo transportation and vehicle types. Then, the total score of each vehicle to be recommended under the cargo owner's tag set is calculated using a scoring discriminant, and all vehicles to be recommended are arranged in descending order according to the total score in the recommendation set. Finally, the recommendation set results are returned to the cargo owner, thereby realizing personalized vehicle recommendations based on the cargo owner's historical transaction evaluation behavior.

[0005] To achieve the above object, the technical solution of the present invention is as follows:

[0006] As one aspect of the present invention, a vehicle recommendation method for an online freight platform based on evaluation tags is proposed, which is characterized by comprising the following steps:

[0007] Step 1: The cargo owner places an order on the online freight platform. The online freight platform screens vehicles based on the order information and selects qualified vehicles into the initial recommendation set.

[0008] Step 2: Based on the owner's historical transaction evaluation content, segmentation is performed based on grammatical rules to generate positive and negative semantic labels;

[0009] Step 3: The shipper and the recommended vehicle generate a full label set consisting of positive and negative labels based on historical transactions. A reduced label set is generated based on label deduplication in the full label set.

[0010] Step 4: Positive and negative labels are matched forward according to the matching strategy and the total score of the positive label pair and the total score of the negative label pair are calculated respectively;

[0011] The matching strategy process includes finding the intersection of the labels of the consignor and the vehicle to be recommended based on the simplified label set to obtain the maximum label pair, and then counting the number of label pairs in the full label set of the consignor and the recommended vehicle, as well as the number of consignors and vehicles that have the label pair in all the simplified label sets of the online freight platform, and calculating the score of this label pair; after calculating the score of this label pair, the counted number of label pairs is deleted from the full label set of the consignor and the vehicle to be recommended, and based on the updated simplified label set of the consignor and the vehicle to be recommended, the label forward matching is performed again to find the maximum label pair, and the score of each label pair is calculated; the above process is repeated until there are no labels to match;

[0012] Step 5: Store the label score of the vehicle to be recommended in the initial recommendation set and push it to the shipper. The total label score of the vehicle to be recommended is the difference between the total score of the positive label pair and the total score of the negative label pair.

[0013] As another aspect disclosed in the present invention, a vehicle recommendation device for an online freight platform based on evaluation tags is proposed, comprising:

[0014] The initial recommendation set module is applicable to cargo owners who place orders on online freight platforms. The online freight platform will initially screen vehicles based on the order information and select qualified vehicles into the initial recommendation set.

[0015] The semantic label generation module is used to generate positive and negative semantic labels based on the owner's historical transaction evaluation content and grammatical rules;

[0016] The label set module is used by shippers and recommended vehicles to generate a full label set consisting of positive and negative labels based on historical transactions, and to generate a reduced label set based on label deduplication in the full label set.

[0017] A matching strategy module, adapted to calculate the total score of positive label pairs, the total score of negative label pairs, and the total score of inverse label pairs according to the matching strategy for positive labels, negative labels, and inverse labels; the matching strategy is a matching strategy in a vehicle recommendation method for an online freight platform based on evaluation labels;

[0018] The vehicle push module is suitable for storing the label scores of the vehicles to be recommended into the initial recommendation set and pushing them to the shippers. The total label score of the vehicles to be recommended is the difference between the sum of the total score of the positive label pair and the total score of the inverse label pair and the total score of the negative label pair.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] This invention mines and statistically analyzes transportation service data in shippers' behavior tags, performs forward and reverse matching based on matching strategies, and recommends a set of freight vehicles that best suits shippers' preferences. This improves the shippers' experience when using freight services and enables personalized vehicle recommendations based on their user profiles.

[0021] Based on historical transaction data and combined with the idea of ​​TF-IDF in the field of system recommendation, this paper establishes a complete method system for extracting, storing and utilizing user behavior tags, and proposes an effective recommendation scoring discriminant. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Among them:

[0023] Figure 1 Flowchart of a method for recommending vehicles on a network freight platform based on evaluation tags according to an embodiment of the present invention;

[0024] Figure 2 It is a structural diagram of a vehicle recommendation device for an online freight platform based on evaluation tags according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The technical solution of the present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0026] Figure 1 A flowchart of a vehicle recommendation method for an online freight platform based on evaluation tags according to an embodiment of the present invention is shown, including the following steps:

[0027] Step 1: The cargo owner places an order on the online freight platform. The online freight platform conducts an initial screening of vehicles based on the order information, and vehicles that meet the recommendation qualifications are selected into the initial recommendation set.

[0028] Specifically, recommendation eligibility primarily examines a vehicle's cargo carrying capacity and order fulfillment capabilities. Cargo carrying capacity includes whether the vehicle's model, load capacity, volume, length, width, and height meet cargo transportation requirements. Order fulfillment capabilities include whether the vehicle is idle within a specified timeframe and whether the distance and travel time to the freight destination are within reasonable ranges as determined by the system.

[0029] Step 2: Based on the owner's historical transaction evaluation content, the entire text is divided into multiple corpora based on punctuation and grammatical rules. Valuable corpora are screened out based on modifiers with practical descriptive meaning and obvious semantic positive and negative. They are then classified according to semantic positive and negative to generate positive and negative semantic labels.

[0030] For example, after a transaction is completed, the cargo owner comments that "the driver is very careful and responsible. Although the cargo could not be delivered on time due to heavy rain, the cargo was not damaged at all." The preprocessed corpus is "driver", "very", "careful", "responsible", "heavy rain", "although", "cargo", "failed to deliver on time", "but", "cargo", "not damaged at all".

[0031] Among them, only four corpora, namely "careful", "responsible", "failed to deliver on time", and "not damaged at all", have practical significance. At the same time, "careful", "responsible", and "not damaged at all" are positive semantic corpora and are classified into the positive semantic label type, while "failed to deliver on time" is negative semantic corpora and is classified into the negative semantic label type.

[0032] Furthermore, the generated positive semantic labels and negative semantic labels are replaced by synonyms and / or near-synonyms in the standard vocabulary and saved as the initial labels of positive semantic labels and negative semantic labels; for labels that belong to the standard vocabulary or have no calibrated words, they are directly saved as the initial labels of positive semantic labels or negative semantic labels.

[0033] The standard vocabulary is pre-designed by the online freight platform. Each positive word has a unique corresponding negative word, and is regularly maintained by platform staff. For example, when entering "cheap freight" into the standard vocabulary, "expensive freight" needs to be entered as the opposite word.

[0034] For example, in the positive semantic label type, "responsible" belongs to the standard vocabulary and can directly generate a positive label; "careful" has a synonym "careful" in the standard vocabulary, and "no damage at all" has a synonym "undamaged" in the standard vocabulary, which can be replaced with the positive label of the standard vocabulary; in the negative semantic corpus, "failed to deliver on time" has a synonym "not on time" in the standard vocabulary, which can be replaced with the negative label of the standard vocabulary.

[0035] Step 3: After the above steps, the shipper and the recommended vehicle generate a full label set consisting of positive and negative labels based on historical transactions. The format is as follows:

[0036] Shipowner:{(positive label),(negative label)}

[0037] Vehicle 1 to be recommended: {(positive label), (negative label)}

[0038] Vehicle 2 to be recommended: {(positive label), (negative label)}

[0039] …

[0040] In order to further improve the matching accuracy between car owners and recommended vehicles, labels will be divided into mainstream labels and personalized labels according to their commonality before matching. When mainstream labels exist in the label set, they will be included in the mainstream label set; otherwise, they will be included in the personalized label set.

[0041] For example, mainstream tags are the most frequently used words in the standard vocabulary of online freight platforms over the past six months. Initially, these tags are manually entered by sales staff, including five positive semantic terms ("careful," "punctual," "responsible," "no damage," and "cheap freight") and five negative semantic terms ("careless," "unpunctual," "irresponsible," "damaged," and "expensive freight"). These tags are then continuously updated as the business develops and user usage evolves.

[0042] All words other than mainstream tags are personalized tags. After classification by tag type, we get the full tag set for both the vehicle and the cargo. The format is as follows:

[0043] Owner: {[Mainstream positive label], [Personalized positive label], [Mainstream negative label], [Personalized negative label]}

[0044] Vehicle 1 to be recommended: {[mainstream positive label], [personal positive label], [mainstream negative label], [personal negative label]}

[0045] Vehicle 2 to be recommended: {[mainstream positive label], [personal positive label], [mainstream negative label], [personal negative label]}

[0046] …

[0047] It should be noted that due to the different contents of cargo owners' evaluations, the same word is allowed to appear multiple times in the full tag set, and certain types of tags are also allowed to be missing, which does not affect the subsequent score calculation.

[0048] Generate a simplified label set based on the label deduplication of the full label set, for example, the full label set of the shipper's historical transaction summary:

[0049] {[punctual, responsible, cheap shipping, punctual], [enthusiastic], [expensive shipping, damaged], [breach of contract]}

[0050] Among them, “on time” appears twice in the mainstream positive labels. After deduplication, the simplified label set of the consignor is obtained:

[0051] {[punctual, responsible, cheap shipping], [enthusiastic], [expensive shipping, damaged], [breach of contract]}

[0052] The label set of the vehicle to be recommended is processed in the same way as the consignor label set. For example, the full label set of vehicle 1 to be recommended is:

[0053] {[responsible, careful, attentive, no damage, cheap shipping],[skilled, trustworthy],[expensive shipping],[]}

[0054] Among them, “careful” appears twice in the mainstream positive labels. After deduplication, the simplified label set of vehicle 1 to be recommended is obtained:

[0055] {[responsible, careful, no damage, cheap shipping],[skilled, trustworthy],[expensive shipping],[]}.

[0056] In step 4, the positive and negative labels are forward matched according to the matching strategy and the total score of the positive label pair and the total score of the negative label pair are calculated respectively.

[0057] Among them, the matching strategy process includes finding the intersection of the labels of the consignor and the vehicle to be recommended based on the simplified label set to obtain the maximum label pair, and counting the number of label pairs in the full label set of the consignor and the recommended vehicle, as well as the number of consignors and vehicles that have the label pair in all simplified label sets of the online freight platform, and calculating the score of this label pair; after calculating the score of this label pair, the counted number of label pairs will be deleted from the full label set of the consignor and the vehicle to be recommended, and based on the updated simplified label set of the consignor and the vehicle to be recommended, the label forward matching will be performed again to find the maximum label pair, and the score of each label pair will be calculated; the above process is repeated until there are no labels to match.

[0058] The specific process of positive label pair forward matching and calculating the positive label pair score includes the following steps:

[0059] S41, positive tag forward matching. Based on the reduced tag set, the intersection of the positive tags of the vehicle owner and the vehicle to be recommended is calculated to obtain the maximum possible positive tag pair tag_1_i_p_q that can successfully match. 1 indicates the matching type is positive tag forward matching, i is the positive tag pair number, p is the number of mainstream tags in the tag pair, and q is the number of personalized tags in the tag pair.

[0060] Taking the above-mentioned cargo owner and vehicle 1 to be recommended as an example, the cargo owner and the vehicle to be recommended have only two matching mainstream positive labels and no matching individual positive labels. Therefore, the maximum positive label pair tag_1_1_2_0 is obtained:

[0061] {[Responsible, cheap shipping],[]}

[0062] Count the number of {[responsible, cheap freight], []} label pairs owned by the shipper in the full label set:

[0063] user_tags[u,t]=∑tag_1_1_2_0=1

[0064] Count the number of {[responsible, cheap freight], []} label pairs in the full set of labels for recommended vehicles:

[0065] vehicles_tags[v,t]=∑tag_1_1_2_0=1

[0066] Among them, u represents the owner's serial number, v represents the vehicle serial number, and t represents the tag serial number.

[0067] S42, count the number of shippers who have this tag pair in all simplified tags on the online freight platform:

[0068] tag_users[t]=∑{1 if ∑tag_1_1_2_0>0 else 0}

[0069] Count the number of vehicles with this label pair in all simplified label sets of the online freight platform:

[0070] tag_vehicles[t]=∑{1 if ∑tag_1_1_2_0>0 else 0}

[0071] Where u represents the owner's serial number, v represents the vehicle serial number, and t represents the tag serial number.

[0072] Assume that the platform has 10,000 vehicles and 2,000 cargo owners with a simplified tag set containing the tag [responsible, cheap freight], then tag_users[1] = 2,000, tag_vehicles[1] = 10,000.

[0073] S43, positive label pair score calculation:

[0074]

[0075] Among them, u represents the serial number of the cargo owner, v represents the serial number of the vehicle, 1 represents the matching type is positive tag pair forward matching, i represents the serial number of the positive tag pair, p represents the number of mainstream tags in the tag pair, and q represents the number of personalized tags in the tag pair. and It is the absolute weight adjustment coefficient introduced in the score calculation process. The more positive label pairs there are, the higher the similarity between the owner and the vehicle, and the higher the weight coefficient of the label pair score should be.

[0076] In order to ensure the effectiveness and accuracy of the matching strategy, it is stipulated that δ2>δ1>1, that is, a higher weight is given to the matching results of personality tags. and This is a relative weight adjustment factor introduced during the scoring process. Its purpose is to avoid homogenizing recommendation results due to highly prevalent tags while simultaneously amplifying the role of less prevalent tags in recommending matches between shippers and vehicles, thereby ensuring personalized matching strategies. This discriminant still holds mathematical significance when a tag pair appears only once.

[0077] For example, when δ1=1.1 and δ2=1.5, the score of tag_1_2_0 in step S42 is:

[0078]

[0079] S44: Tag set update. After calculating the score of the current tag pair, to avoid the influence of counted tags on uncounted tags, the tags contained in the tag pair are deleted from the shipper's full tag set by user_tags[u,t] times, and from the full tag set of the recommended vehicle by vehicles_tags[v,t] times. Then, the full tag sets and the reduced tag sets of both parties are updated.

[0080] For example, after completing step S43, the shipper's full label set is:

[0081] {[On time],[On time],[Enthusiastic],[Expensive shipping, Damaged],[Breach of contract]}

[0082] The simplified label set of the consignor is:

[0083] {[punctual],[enthusiastic],[expensive shipping, damaged],[breach of contract]}

[0084] The full label set of vehicle 1 to be recommended is:

[0085] {[Careful, careful, no damage],[Expert, trustworthy],[Expensive shipping],[]}

[0086] The simplified label set of vehicle 1 to be recommended is:

[0087] {[Careful, no damage],[Expert, trustworthy],[Expensive shipping],[]}

[0088] At this point, since there is no positive label to match between the simplified label set of the cargo owner and the simplified label set of the vehicle 1 to be recommended, the positive label matching ends.

[0089] It should be noted that if there are still matching positive labels, S41-S44 are repeated until there are no other positive labels that can be matched.

[0090] In practical applications, for example, when the full label set of the consignor is {[responsible, responsible, responsible], [enthusiastic]}, and the full label set of the vehicle 2 to be recommended is {[responsible, responsible, responsible], [enthusiastic, enthusiastic]}, the maximum positive label pair tag_1_1_1_1 of the first intersection matching of the reduced label sets of both parties is {[responsible], [enthusiastic]}. The full label set of the consignor contains 1 such label pair, and the full label set of the vehicle 2 to be recommended contains 2 such label pairs. After the first score calculation, the remaining labels of the full label set of the consignor are {[responsible, responsible]}, and the remaining labels of the full label set of the vehicle are {[responsible]}.

[0091] The maximum positive label pair tag_1_2_1_0 found by the second intersection of the simplified label sets of the consignor and the vehicle 2 to be recommended is {[responsible]}. Both the vehicle and the cargo have only one label pair. The consignor's full label set contains two of this label pair, and the vehicle 2 to be recommended has one of this label pair. Neither the consignor nor the vehicle 2 to be recommended has any remaining available labels, and the loop ends.

[0092] After completing the positive tag forward matching, the matching strategy is used to perform negative tag forward matching. Based on the reduced tag set, the negative tags of both the shipper and the recommended vehicle are forward matched to find the largest negative tag pair tag_2_j_p_q that can successfully match. Where 2 represents the negative tag forward match, j represents the negative tag sequence, p represents the number of mainstream tags in the tag pair, and q represents the number of unique tags in the tag pair. The score calculation and tag set update process for score(u,v)_2_j_p_q are similar to those for positive tag forward matching and are not further described here.

[0093] To improve matching accuracy, as described in step 2, all words in the standard vocabulary have a unique corresponding opposite semantic word. Using this opposite semantic word to perform reverse tag matching can complement the forward matching effect. Find the largest inverse tag pair tag_k_p_q that can successfully match in the reduced tag set, where k is the inverse tag pair index, p is the number of mainstream tags in the tag pair, and q is the number of personalized tags in the tag pair.

[0094] For example, after the positive matching of the vehicle and cargo label sets is completed, the shipper's full label set and simplified label set are both:

[0095] {[punctuality],[enthusiasm],[damage],[breach of contract]}

[0096] The full label set of vehicle 1 to be recommended is:

[0097] {[Careful, careful, no damage],[Skilled, trustworthy],[],[]}

[0098] The simplified label set of vehicle 1 to be recommended is:

[0099] {[Careful, No Damage],[Expert, Trustworthy],[],[]}

[0100] Since "breach of contract" and "keeping promises" are a pair of personalized tags with opposite semantics in the standard vocabulary, the maximum inverse tag pair tag_1_0_1 is represented by {[keeping promises]} using positive semantic vocabulary. The remaining calculation process is the same as that for forward matching and is not repeated here. The inverse tag pair score score(u,v)_1_p_q is obtained.

[0101] Finally, the total label score of the vehicle to be recommended is calculated.

[0102]

[0103] Among them, u represents the serial number of the cargo owner, v represents the serial number of the vehicle, i represents the serial number of the positive label pair, j represents the serial number of the negative label pair, k represents the serial number of the reverse label pair, p represents the number of mainstream labels in the label pair, and q represents the number of personalized labels in the label pair.

[0104] Repeat the process in step 4 until all recommended vehicles have completed the scoring. For vehicles with no historical transaction history on the online freight platform and all tags are empty, the system defaults to a recommendation score of 0.

[0105] Step 5: Store the label scores of the vehicles to be recommended into the initial recommendation set and push them to the cargo owner.

[0106] Figure 2 Another embodiment of the present invention provides a vehicle recommendation device for an online freight platform based on evaluation tags, including:

[0107] The initial recommendation set module is applicable to cargo owners who place orders on online freight platforms. The online freight platform will initially screen vehicles based on the order information and select qualified vehicles into the initial recommendation set.

[0108] The semantic label generation module is used to generate positive and negative semantic labels based on the owner's historical transaction evaluation content and grammatical rules;

[0109] The label set module is used by shippers and recommended vehicles to generate a full label set consisting of positive and negative labels based on historical transactions, and to generate a reduced label set based on label deduplication in the full label set.

[0110] Matching strategy module, which is applicable to positive labels, negative labels and inverse labels to calculate the total score of positive label pairs, the total score of negative label pairs and the total score of inverse label pairs according to the matching strategy;

[0111] The specific matching strategy process includes finding the intersection of the labels of the consignor and the vehicle to be recommended based on the simplified label set to obtain the maximum label pair, and counting the number of label pairs in the full label set of the consignor and the recommended vehicle, as well as the number of consignors and vehicles that have the label pair in all the simplified label sets of the online freight platform, and calculating the score of this label pair; after calculating the score of this label pair, the counted number of labels will be deleted from the full label set of the consignor and the vehicle to be recommended, and based on the updated simplified label set of the consignor and the vehicle to be recommended, label forward matching will be performed again to obtain label pairs, and the score of each label pair will be calculated; the above process is repeated until there are no labels to match.

[0112] The vehicle push module is suitable for storing the label scores of the vehicles to be recommended into the initial recommendation set and pushing them to the shippers. The total label score of the vehicles to be recommended is the difference between the sum of the total score of the positive label pair and the total score of the inverse label pair and the total score of the negative label pair.

[0113] The device of this embodiment can be used to perform Figure 1 The implementation principles and technical effects of the steps in the method embodiment shown are similar and will not be repeated here.

[0114] Another embodiment of the present invention provides a hardware schematic diagram of a vehicle recommendation device for an online freight platform based on evaluation tags. The device includes a processor, a memory, and a computer program. The memory is used to store the computer program, which can also be flash memory. The computer program is, for example, an application program or functional module that implements the above-mentioned method.

[0115] The processor is configured to execute the computer program stored in the memory to implement the various steps performed by the device in the above method. For details, please refer to the relevant description in the above method embodiment.

[0116] Optionally, the memory can be independent or integrated with the processor.

[0117] When the memory is a device independent of the processor, the apparatus may further include a bus for connecting the memory and the processor.

[0118] The present invention also provides a readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, it is used to implement the methods provided in the various embodiments described above.

[0119] Among them, the readable storage medium can be a computer storage medium or a communication medium. Communication media include any medium that facilitates the transmission of computer programs from one place to another. Computer storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer. For example, a readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be an integral part of the processor. The processor and the readable storage medium can be located in an application-specific integrated circuit. In addition, the ASIC can be located in a user device. Of course, the processor and the readable storage medium can also exist as discrete components in a communication device. The readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0120] The present invention also provides a program product, which includes execution instructions stored in a readable storage medium. At least one processor of a device can read the execution instructions from the readable storage medium, and at least one processor executes the execution instructions so that the device implements the methods provided in the various embodiments described above.

[0121] In the above-described device embodiments, it should be understood that the processor may be a central processing unit, or may be other general-purpose processors, digital signal processors, application-specific integrated circuits, etc. The general-purpose processor may be a microprocessor, or the processor may be any conventional processor, etc. The steps of the method disclosed in the present invention may be directly implemented and executed by a hardware processor, or by a combination of hardware and software modules in the processor.

[0122] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A vehicle recommendation method for an online freight platform based on evaluation tags, characterized in that: The steps include: Step 1: The cargo owner places an order on the online freight platform. The online freight platform screens vehicles based on the order information and selects qualified vehicles into the initial recommendation set. Step 2: Based on the owner's historical transaction evaluation content, segmentation is performed based on grammatical rules to generate positive and negative semantic labels; Step 3: The shipper and the recommended vehicle generate a full label set consisting of positive and negative labels based on historical transactions. A reduced label set is generated based on label deduplication in the full label set. According to the universality of the label, it is divided into mainstream labels and personalized labels. Words outside the mainstream labels are all personalized labels. Step 4: Positive and negative labels are matched forward according to the matching strategy and the total score of the positive label pair and the total score of the negative label pair are calculated respectively; Among them, the matching strategy process includes finding the intersection of the labels of the consignor and the vehicle to be recommended based on the simplified label set to obtain the maximum label pair, and then counting the number of label pairs in the full label set of the consignor and the recommended vehicle, as well as the number of consignors and vehicles that have the label pair in all the simplified label sets of the online freight platform, and calculating the score of this label pair; after calculating the score of this label pair, the counted number of label pairs will be deleted from the full label set of the consignor and the vehicle to be recommended, and based on the updated simplified label set of the consignor and the vehicle to be recommended, the label forward matching will be performed again to find the maximum label pair, and the score of each label pair will be calculated; the above process is repeated until there are no labels to match; The formula for calculating the label pair score is: Among them, u represents the serial number of the cargo owner, v represents the serial number of the vehicle, Indicates the absolute weight adjustment coefficient of mainstream tags, represents the absolute weight adjustment coefficient of the personalized tag, user_tags[u, t] is the number of matching tag pairs in the full tag set of the shipper, vehicles_tags[v, t] is the number of matching tag pairs in the full tag set of the recommended vehicle, tag_users[t] is the number of shippers with matching tag pairs in all simplified tag sets of the online freight platform, and tag_vehicles[t] is the number of vehicles with matching tag pairs in all simplified tag sets of the online freight platform. and is the relative weight adjustment coefficient; Step 5: Store the label score of the vehicle to be recommended in the initial recommendation set and push it to the shipper. The total label score of the vehicle to be recommended is the difference between the total score of the positive label pair and the total score of the negative label pair.

2. The vehicle recommendation method according to claim 1, characterized in that: The positive and negative semantic labels generated in step 2 are replaced by synonyms and / or near-synonyms in the standard vocabulary and saved as the initial positive and negative semantic labels. Labels that belong to the standard vocabulary or have no calibration words are directly saved as the initial positive or negative semantic labels.

3. The vehicle recommendation method according to claim 1, characterized in that: The positive tag pair is marked as tag_1_i_p_q, and the negative tag pair is marked as tag_2_j_p_q, where 1 indicates a positive tag positive match, 2 indicates a negative tag positive match, i indicates the positive tag pair number, j indicates the negative tag pair number, p indicates the number of mainstream tags in the tag pair, and q indicates the number of personalized tags in the tag pair.

4. The vehicle recommendation method according to claim 1, characterized in that: The matching weight δ2 of the individual tag is greater than the matching weight δ1 of the mainstream tag.

5. The vehicle recommendation method according to claim 1, characterized in that: Step 4 also includes supplementing the positive label with an inverse label by a word with opposite semantics, performing reverse matching on the inverse label according to the matching strategy and calculating the total score of the inverse label pair.

6. The vehicle recommendation method according to claim 5, characterized in that: The total label score of the vehicle to be recommended is the difference between the sum of the total score of the positive label pair and the total score of the inverse label pair and the total score of the negative label pair.

7. A device for the vehicle recommendation method according to any one of claims 1 to 6, characterized in that: include: The initial recommendation set module is applicable to cargo owners who place orders on online freight platforms. The online freight platform will initially screen vehicles based on the order information and select qualified vehicles into the initial recommendation set. The semantic label generation module is used to generate positive and negative semantic labels based on the owner's historical transaction evaluation content and grammatical rules; The label set module is used by shippers and recommended vehicles to generate a full label set consisting of positive and negative labels based on historical transactions, and to generate a reduced label set based on label deduplication in the full label set. Matching strategy module, which is applicable to positive labels, negative labels and inverse labels to calculate the total score of positive label pairs, the total score of negative label pairs and the total score of inverse label pairs according to the matching strategy; The vehicle push module is suitable for storing the label scores of the vehicles to be recommended into the initial recommendation set and pushing them to the shippers. The total label score of the vehicles to be recommended is the difference between the sum of the total score of the positive label pair and the total score of the inverse label pair and the total score of the negative label pair.

Images