Driver clue scoring method and device based on data fusion, equipment and medium

Abstract

The invention provides a driver clue scoring method and device based on data fusion, equipment and a medium. The method comprises the steps of obtaining multi-channel data of a driver to be scored; obtaining clue features and periodic features of the driver according to the multi-channel data; crosswise combining the clue features of the drivers according to a preset rule to obtain combined features of the drivers; and inputting the combined features and the periodic features of the driver into a trained clue scoring model to obtain a clue score of the driver. According to the method, multi-channel data from different sources are integrated, driver clue features are generated, driver feature data are enriched through feature combination, periodic feature design and other methods, finally, the driver feature data are processed into driver clues based on a model, the driver clues are converted into platform transport capacity possibility scores, the scores can visually and accurately show the quality of the driver clues, and the driver clue quality is improved. When the inviter makes an invitation for the driver clues with high scores, more purposiveness is achieved, and the conversion rate can be increased on the basis that the workload is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of online car-hailing big data processing, and in particular to a driver clue scoring method and device, equipment, and medium based on data fusion. Background Art

[0002] With the development of Internet technology, more services have been expanded on the basis of Internet technology. Among them, online booking of taxis or freight trucks provides people with more abundant and convenient options for travel. On the platform that provides services, users can register as drivers by themselves. After becoming a driver, users can become drivers who can accept orders by uploading certificate information (such as driver's license, ID card), paying a deposit, etc. For the platform, there are a large number of drivers who have registered but have not been certified as platform capacity. These are collectively referred to as driver leads.

[0003] At present, in addition to waiting for users to actively authenticate themselves as drivers for platform capacity, you can also invite these registered but unauthenticated drivers to authenticate through inviters. For inviters, every driver lead has the potential to be converted into platform capacity, but most of them are invalid leads with a very low conversion rate. It is difficult to cover all leads with manpower alone and the efficiency is low.

[0004] In order to improve the efficiency of the inviter, the current general practice is to trigger the invitation based on certain behavioral events of the driver. For example, when a driver recently logs into the platform, the inviter will send an invitation to the driver. This approach can improve the conversion rate to a certain extent, but the improvement effect is not significant, and a large number of drivers who have not logged in will be missed. Summary of the invention

[0005] In view of the above defects in the prior art, the present invention provides a driver lead scoring method and device, equipment, and medium based on data fusion to solve the technical problems of low driver conversion rate, low efficiency, and omissions caused by unclear invitation targets when improving platform capacity.

[0006] To achieve the above-mentioned purpose and other related purposes, the present invention provides a driver clue scoring method based on data fusion, a driver clue scoring method based on data fusion, comprising: obtaining multi-channel data of a driver to be scored; obtaining clue characteristics and periodic characteristics of the driver based on the multi-channel data; cross-combining the clue characteristics of the driver according to preset rules to obtain the combined characteristics of the driver; inputting the combined characteristics and periodic characteristics of the driver into a trained clue scoring model to obtain the clue score of the driver.

[0007] In one embodiment of the present invention, in the step of cross-combining the clue features according to preset rules to obtain the combined features of the driver, the preset rules are obtained according to the following steps: cross-combining the clue features of all drivers to obtain multiple first features; using the multiple first features of the driver as input features and whether the driver is converted into platform capacity as a label, using a logistic regression model to calculate the importance of each of the first features; and obtaining the preset rules based on the combination of a preset number of first features ranked top in importance.

[0008] In one embodiment of the present invention, the multi-channel data includes manual outbound call data, artificial intelligence outbound call data, and historical behavioral event data of the driver on the freight platform.

[0009] In one embodiment of the present invention, obtaining the periodic characteristics of the driver based on the multi-channel data includes: extracting dimensional characteristics of a preset period from the multi-channel data to obtain the periodic characteristics.

[0010] In one embodiment of the present invention, after the step of cross-combining the clue features of the driver according to preset rules to obtain the combined features of the driver, the step also includes: obtaining the city features of the driver based on the order volume, order acceptance rate and driver activity characteristics of the city where the driver is located on the freight platform; inputting the combined features and periodic features of the driver into a trained clue scoring model to obtain the clue score of the driver, including: inputting the combined features, periodic features and city features of the driver into a trained clue scoring model to obtain the clue score of the driver.

[0011] In one embodiment of the present invention, after the step of cross-combining the clue features of the driver according to preset rules to obtain the combined features of the truck driver, the step also includes: inputting the driver's historical conversations in the multi-channel data into the trained NLP model to obtain the driver's willingness to join score; inputting the driver's combined features and periodic features into the trained clue scoring model to obtain the driver's clue score, including: inputting the driver's combined features, periodic features and willingness to join score into the trained clue scoring model to obtain the driver's clue score.

[0012] In one embodiment of the present invention, the combined features and periodic features of the driver are input into a trained clue scoring model to obtain the clue score of the driver, and then the step further includes: performing operational optimization based on the clue score of the driver and the inviter information of the city where the driver is located to obtain the clue level of the driver.

[0013] To achieve the above-mentioned purpose and other related purposes, the present invention also provides a driver clue scoring device based on data fusion, comprising: a data acquisition unit, used to acquire multi-channel data of the driver to be scored; a data cleaning unit, used to obtain the clue characteristics of the driver based on the multi-channel data; a feature mining unit, used to obtain the periodic characteristics of the driver based on the multi-channel data, and used to cross-combine the clue characteristics of the driver according to preset rules to obtain the combined characteristics of the driver; a scoring unit, used to input the combined characteristics and periodic characteristics of the driver into a trained clue scoring model to obtain the clue score of the driver.

[0014] To achieve the above-mentioned purpose and other related purposes, the present invention also provides an electronic device, including a processor, a memory and a communication bus; the communication bus is used to connect the processor and the memory; the processor is used to execute a computer program stored in the memory to implement a method provided in any one of the above embodiments.

[0015] To achieve the above object and other related objects, the present invention further provides a computer-readable storage medium on which a computer program is stored, wherein the computer program is used to enable a computer to execute the method provided in any one of the above embodiments.

[0016] Beneficial effects of the present invention: The present invention proposes a driver lead scoring method and device based on data fusion, which is based on the existing data of the platform, cleans and processes multi-channel data, integrates data from different sources, generates driver lead features, and enriches driver feature data through feature combination, periodic feature design and other methods. Finally, based on the model, the driver feature data is processed into driver leads that are converted into platform capacity possibility scores. The score can intuitively and accurately show the quality of driver leads. When inviters invite driver leads with higher scores, they can be more purposeful and can improve conversion rates while reducing workload. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 A flowchart of a first driver clue scoring method provided by an embodiment of the present invention;

[0019] Figure 2 A flowchart of obtaining preset rules provided by an embodiment of the present invention;

[0020] Figure 3 A flow chart of a second driver clue scoring method provided by an embodiment of the present invention;

[0021] Figure 4 A flowchart of a third driver clue scoring method provided by an embodiment of the present invention;

[0022] Figure 5 A flowchart of a fourth driver clue scoring method provided by an embodiment of the present invention;

[0023] Figure 6 A flowchart of a fifth driver clue scoring method provided by an embodiment of the present invention;

[0024] Figure 7 A flowchart of a sixth driver clue scoring method provided by an embodiment of the present invention;

[0025] Figure 8 A schematic diagram of a first driver clue scoring device provided by an embodiment of the present invention;

[0026] Fig. 9 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention.

[0027] Explanation of the reference numerals: 101, data acquisition unit; 102, data cleaning unit; 103, feature mining unit; 104, scoring unit; 201, processor; 202, memory. DETAILED DESCRIPTION

[0028] The following describes the embodiments of the present invention through specific specific embodiments, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. It should be noted that the following embodiments and the features in the embodiments can be combined with each other without conflict. In addition to the specific methods, equipment, and materials used in the embodiments, according to the technical personnel in this technical field's grasp of the prior art and the records of the present invention, any methods, equipment, and materials of the prior art that are similar or equivalent to the methods, equipment, and materials described in the embodiments of the present invention can also be used to implement the present invention.

[0029] It should be understood that the terms used in the embodiments of the present invention are for describing specific embodiments rather than for limiting the scope of protection of the present invention. Unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those generally understood by those skilled in the art.

[0030] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In some of the embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0031] The flowchart and block diagram in the accompanying drawings illustrate the system architecture, functions and operations that may be implemented according to the methods and computer program products of various embodiments disclosed in the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, the program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of the boxes in the block diagram and / or the flowchart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0032] See also Figure 1 , Figure 1 A driver clue scoring method based on data fusion is provided for one embodiment of the present invention, including step S100, step S200, step S300 and step S600. These steps take a driver to be scored as an example, and describe in detail how to score based on the clue data of the driver. In actual application, all unconverted driver data can be processed simultaneously.

[0033] Step S100: Obtain multi-channel data of the driver to be rated.

[0034] In a specific embodiment of the present invention, multi-channel data includes manual outbound call data, artificial intelligence outbound call (AI outbound call) data, and historical behavior event data of the driver on the freight platform. For the platform, all the data about the driver accumulated from different channels is the clue data of the driver. The clue data is a comprehensive information of the driver, such as the channel source of the data and the data content. The data content can be, for example, the registration device, registration time, registration city, registered vehicle model, submitted information, etc., or the time of the last outbound call, the time of the last information update, the number of historical follow-up times, etc.

[0035] Step S200: Obtain the clue features and periodic features of the driver based on the multi-channel data. The clue features of the driver are obtained by cleaning the multi-channel data. The clue features can be understood as primary features, which are obtained by simple processing of the data, such as the driver's registered device name, registration time, registration city, last outbound call time, etc. These data can be obtained by simple data cleaning.

[0036] The periodic characteristics of drivers need to be mined from multi-channel data. In a specific embodiment of the present invention, the dimension characteristics of the preset period can be extracted from the historical behavior event data of the driver on the freight platform to obtain the periodic characteristics. The preset period here can be, for example, 7 days, 30 days, 90 days, etc. These three periods correspond to weeks, months, and quarters. Of course, other preset periods can also be set as needed.

[0037] For example, the manual outbound call data of multi-channel data includes the time record of each follow-up. By mining these time records, we can obtain periodic characteristics such as the number of follow-ups in the past 7 days, the number of follow-ups in the past 30 days, and the number of follow-ups in the past 90 days.

[0038] Step S300, cross-combining the clue features of the driver according to the preset rules to obtain the combined features of the driver. Cross-combination is a conventional operation for fusing features, and the clue features of the driver after cross-combination are the combined features. In this embodiment, by setting the preset rules, only the combined features that meet the preset rules are selected, so as to selectively fuse the clue features of the driver, and specifically select the fused features that have a greater impact on the driver's score to input into the subsequent model, so that the driver's clues can be scored more accurately.

[0039] See also Figure 2 In a specific embodiment of the present invention, the preset rule in step S300 can be obtained according to steps S301 to S303.

[0040] Step S301, cross-combining clue features of all drivers to obtain multiple first features. For example, the clue features of the driver include the two features of "registration device" and "registration channel", and by combining them, a "registration device-registration channel" combined feature can be obtained. "Registration device" has values ​​such as ios and android, and "registration channel" has values ​​such as offline referral, information flow, mini program, and application market. Then, by crossing the feature values ​​in pairs, the value of the "registration device-registration channel" feature obtained by combination can be ios_mini program, ios_information flow, android_information flow, android_application market, etc.

[0041] Step S302: Using multiple first features of the driver as input features and whether the driver is converted into platform capacity as a label, the importance of each first feature is calculated using a logistic regression model. The logistic regression model can rank the features by importance by analyzing the estimated values ​​of the model parameters. In logistic regression, the importance (or coefficient, weight) of each first feature represents the degree of influence of the feature on the prediction result. The greater the importance, the greater the influence of the feature on the model prediction.

[0042] Step S303, obtain the preset rules according to the combination method of the preset number of first features ranked high in importance. The preset number here can be set according to the demand, which corresponds to the first features ranked top in importance that we are about to select. The preset number can be, for example, 20, 30, 40, etc. Taking the preset number of 30 as an example, we need to select the features ranked top 30 in importance. Assuming that the importance ranking of the feature "registration device-registration channel" is 20, which belongs to the top 30, then the combination method of the first feature can be used as the preset rule. When executing step S300 later, "registration device" and "registration channel" are directly combined to obtain a combined feature.

[0043] It can be understood that if the preset number is 30, then the preset rule corresponds to the combination of 30 first features.

[0044] Of course, in addition to ranking the first feature in importance, the driver's clue feature itself can also participate in the importance ranking, so the preset number of features with the highest importance ranking finally screened out can include not only the first feature, but also the clue feature.

[0045] In this embodiment, the historical data of drivers (including the data of drivers who have been converted to platform capacity) is used to construct a training set; then the training data set is used to train a logistic regression model to obtain the coefficient estimate of each first feature; after the model training is completed, the coefficient of each first feature is checked, which can be positive or negative. A positive coefficient indicates that the probability of an event increases with the increase of the feature value, and a negative coefficient indicates that the probability of an event decreases with the increase of the feature value; then the absolute value of the coefficient of each first feature is calculated to obtain the importance. This is done because we are concerned about the influence of the first feature, not its direction of influence; then the features are sorted according to the importance, the greater the importance, the higher the importance of the feature; finally, it is only necessary to screen out a preset number of first features, and then the preset rules can be obtained based on the combination of these first features.

[0046] Step S600: Input the combined features and periodic features of the driver into the trained clue scoring model to obtain the driver's clue score. Through the machine learning algorithm, the driver leads (including all converted and unconverted driver leads) accumulated by the platform in the past six months (or other time) are used as training samples, that is, all driver leads are processed according to steps S100 to S300, and the obtained combined features and periodic features are used as inputs to the clue scoring model, and are labeled according to whether these drivers have been converted into platform capacity, for example, the converted driver label is 1 and the unconverted driver label is 0. The clue scoring model is trained using the training samples, and the trained model needs to be tested and verified if necessary. The trained clue scoring model can directly score the driver characteristics to be scored, and obtain the probability of the driver lead being converted into platform capacity, and its value range is [0,1].

[0047] It should be noted that, when actually scoring, it may be necessary to calculate N-in-one intention scores, which are divided into different categories, such as one-in-one, two-in-one, three-in-one, four-in-one, etc. For example, in a certain embodiment, we only want to obtain the driver's one-in-one willingness score, that is, the willingness to convert to platform capacity, then when constructing the above training samples, the labels of those drivers who have been converted to platform capacity are all 1. If you want to obtain the driver's two-in-one willingness (certified and paid a deposit), then when constructing the above training samples, the labels of those drivers who have been converted to platform capacity but have not paid a deposit should be 0, and only those drivers who have been converted to platform capacity and have paid a deposit should have a label of 1. The score obtained by the clue scoring model trained with this training sample is the two-in-one willingness score.

[0048] The same is true for other all-in-one willingness scores. Therefore, the above-mentioned lead scoring model can not only calculate the driver's willingness score to be converted into platform capacity, but also calculate a more complex N-in-one intention score.

[0049] See also Figure 3 In a specific embodiment of the present invention, step S300 also includes: S400, obtaining the city characteristics of the driver based on the order volume, order acceptance rate and driver activity characteristics of the city where the driver is located on the freight platform. The order volume, order acceptance rate and driver activity of the city will also affect the driver's willingness to convert into platform capacity, so this feature needs to be considered when scoring leads. That is, step S600 includes: S600', inputting the driver's combined characteristics, periodic characteristics and city characteristics into the trained lead scoring model to obtain the driver's lead score. On the basis of the combined characteristics and periodic characteristics, the driver's city characteristics are integrated to further improve the accuracy of the driver's lead scoring results.

[0050] See also Figure 4In a specific embodiment of the present invention, after step S300, the following steps are further followed: S500, inputting the driver's historical conversations in the multi-channel data into the trained NLP model to obtain the driver's willingness to join score. In the driver's historical conversation data, there is a high probability that their willingness to join score is directly reflected. At this time, the trained NLP model can be used to process the conversation content to obtain the willingness to join score. NLP (Natural Language Processing) is natural language processing, which is a branch of artificial intelligence and linguistics. It is dedicated to enabling computers to understand, interpret and generate human language content. NLP models are algorithms and statistical models for implementing these tasks, and they can process and analyze a large amount of natural language data. In this embodiment, we use the NLP model to obtain the driver's willingness to join score, and use it as one of the features to input it into the clue scoring model, that is, step S600 includes: S600 "input the driver's combined features, periodic features and willingness to join score into the trained clue scoring model to obtain the driver's clue score. On the basis of the combined features and periodic features, the driver's willingness to join score is integrated to further improve the accuracy of the driver's clue scoring results.

[0051] See also Figure 5 In a specific embodiment of the present invention, the above-mentioned step S400 and step S500 may be included simultaneously, that is, the driver's city characteristics and willingness to join score are extracted simultaneously, and input into the clue scoring model together with the combination characteristics and periodic characteristics to obtain the driver's clue score.

[0052] The above-mentioned clue scoring model can be, for example, a LightGBM model. LightGBM is a gradient boosting framework developed by Microsoft, using a tree-based learning algorithm, which is distributed and efficient and is specifically designed to process large-scale data sets.

[0053] Since the NLP model outputs the willingness to join score, it can be regarded as a scoring result rather than a feature. In a specific embodiment of the present invention, the clue scoring model includes a LightGBM model and an NLP model. The driver's combined features, periodic features, and city features are input into the trained LightGBM model to obtain the driver's first clue score; the driver's historical conversations in the multi-channel data are input into the trained NLP model to obtain the driver's second clue score; based on the first clue score and the second clue score, the driver's clue score is obtained.

[0054] See also Figure 6In a specific embodiment of the present invention, after step S600, the following further includes: S700, performing operational optimization based on the clue score of the driver and the inviter information of the city where the driver is located to obtain the clue level of the driver. For the inviter, the clue score is still not intuitive enough. Generally, the clue score needs to be processed into a clue level before being distributed to the inviter. The inviter can send an invitation based on the clue level. The clue level can be, for example, high, medium, and low, or other levels.

[0055] For different cities, even if the lead scores of drivers are the same, different inviter information will lead to different lead levels. Inviter information generally includes inviter level, number of people, and upper limit of average daily allocation per person. For example, if a driver's lead score is 0.7 points, for city A and city B, assuming that there are more inviters in city A and their levels are higher, the number of drivers that can be invited every day is naturally large. Then, after operations optimization, drivers with 0.7 points may be rated as high priority. Assuming that there are fewer inviters in city B, the number of drivers that can be invited every day is naturally small. Drivers who may reach 0.8 points are worthy of their attention. Drivers with 0.7 points may be rated as medium priority after operations optimization. Operational optimization makes it easier for inviters to carry out subsequent processing.

[0056] It should be noted that the step division of the above methods is only for clear description. When implemented, they can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they contain the same logical relationship, they are all within the scope of protection of this patent; adding insignificant modifications to the algorithm or process or introducing insignificant designs without changing the core design of the algorithm and process are all within the scope of protection of this patent.

[0057] For example Figure 7The illustrated embodiment includes the following steps: (1) Obtaining old clues. The old clues are divided into a public sea pool, a private sea pool, and a recycling pool. The clues in the private sea pool refer to clues that have been currently allocated, the public sea pool refers to clues that are currently unallocated, and the recycling pool refers to clues that are currently unallocated and of poor quality. In this embodiment, allocation can be made based on clues in the public sea pool and the recycling pool. (2) Determine the current clue status. That is, after obtaining the clue, it is necessary to first confirm whether it is an unauthenticated driver clue. (3) Determine the quality of the clue. The quality judgment of leads is divided into two parts, namely daily scheduled update and event trigger prediction. The reason for daily scheduled update is that as the date changes, the characteristics of the driver will also change, and the score obtained according to the above steps will also change accordingly. Therefore, for all old leads, the algorithm estimates the quality of the leads based on offline characteristics on a daily basis (before the inviter goes to work); event trigger prediction is for those matters that have a greater impact on the score, such as when the event monitors "new customer service contact records, updated registration status, AI outbound call high willingness (algorithm), the driver has opened the app and other terminal behaviors, the algorithm updates the corresponding features in real time, and then combines the offline features to re-estimate the quality of the leads that trigger the event, that is, re-score immediately after these behaviors occur. (4) The subsequent clue feature combination, model prediction, and old clue level operation have been explained in detail in the above embodiments, and will not be repeated here.

[0058] In the above embodiments, the main point is how to score a single driver. Figure 7 The illustrated embodiment includes when to score drivers for clues and which drivers to score for clues, and the steps are more complete.

[0059] See also Figure 8 In a specific embodiment of the present invention, a driver clue scoring device based on data fusion is also provided, comprising: a data acquisition unit 101, a data cleaning unit 102, a feature mining unit 103 and a scoring unit 104. The data acquisition unit 101 is used to acquire multi-channel data of the driver to be scored; the data cleaning unit 102 is used to obtain the clue features of the driver according to the multi-channel data; the feature mining unit 103 is used to obtain the periodic features of the driver according to the multi-channel data, and to cross-combine the clue features of the driver according to a preset rule to obtain the combined features of the driver; the scoring unit 104 is used to input the combined features and periodic features of the driver into the trained clue scoring model to obtain the clue score of the driver.

[0060] It should be noted that the above-mentioned functional modules can be fully or partially integrated into one physical entity during actual implementation, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, some or all of the steps of the above method, or the above functional modules can be completed by hardware integrated logic circuits in the processor element or instructions in software form.

[0061] It should be noted that the driver clue scoring device of this embodiment is a device corresponding to the above-mentioned driver clue scoring method, and the functional modules in the driver clue scoring device may correspond to the corresponding steps in the driver clue scoring method. The driver clue scoring device of this embodiment can be implemented in conjunction with the driver clue scoring method, that is, in the absence of conflict, the relevant technical details mentioned in the driver clue scoring method of the above-mentioned embodiment can also be applied to the driver clue scoring device of this embodiment.

[0062] See also Fig. 9 , Fig. 9 An electronic device provided for one embodiment of the present invention includes a processor 201, a memory 202 and a communication bus; the communication bus is used to connect the processor 201 and the memory 202; the processor 201 is used to execute the computer program stored in the memory 202 to implement the above-mentioned driver clue scoring method.

[0063] The above-mentioned processor can be, for example, a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components; the above-mentioned memory may include random access memory (RAM), and may also include non-volatile memory (Non-volatile Memory), such as at least one disk storage.

[0064] The above-mentioned electronic device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0065] The electronic device mentioned above can be any electronic product that can perform human-computer interaction with a user, such as a personal computer, a tablet computer, a smart phone, a personal digital assistant (PDA), a game console, an interactive network television (IPTV), a smart wearable device, etc.

[0066] The electronic device may also include a network device and / or a user device, wherein the network device includes, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud consisting of a large number of hosts or network servers based on cloud computing.

[0067] The network where the above electronic devices are located includes but is not limited to the Internet, wide area network, metropolitan area network, local area network, virtual private network (VPN), etc.

[0068] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. The computer program is used to enable a computer to execute the above-mentioned driver clue scoring method.

[0069] In general, the present invention provides a method for intelligently scoring registered driver leads that have not been converted into platform capacity, and evaluating their conversion into platform capacity. It utilizes AI, manual outbound calls and other data, and based on feature combination and periodic feature design methods, extracts data features of the driver lead group, enriching the group portrait of large vehicle leads; at the same time, it also introduces an NLP model to extract the willingness to join the conversation data between the driver and the platform inviter, further fuses the features, and improves the accuracy of the intelligent scoring of the leads.

[0070] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A driver clue scoring method based on data fusion, characterized in that: include: Obtain multi-channel data of drivers to be rated; Obtaining clue features and periodic features of the driver based on the multi-channel data; Cross-combining the clue features of the driver according to a preset rule to obtain a combined feature of the driver; The combined features and periodic features of the driver are input into the trained clue scoring model to obtain the clue score of the driver.

2. The driver clue scoring method based on data fusion according to claim 1 is characterized in that: In the step of cross-combining the clue features according to a preset rule to obtain the combined features of the driver, the preset rule is obtained according to the following steps: Cross-combining clue features of all drivers to obtain multiple first features; Using multiple first features of the driver as input features and whether the driver is converted into platform capacity as a label, a logistic regression model is used to calculate the importance of each of the first features; The preset rule is obtained according to a combination of a preset number of first features ranked high in importance.

3. The driver clue scoring method based on data fusion according to claim 1, characterized in that: The multi-channel data includes manual outbound call data, artificial intelligence outbound call data and the driver’s historical behavioral event data on the freight platform.

4. The driver clue scoring method based on data fusion according to claim 3 is characterized in that: According to the multi-channel data, the periodic characteristics of the driver are obtained, including: Dimensional features of a preset period are extracted from the multi-channel data to obtain the periodic features.

5. The driver clue scoring method based on data fusion according to claim 1, characterized in that: After the step of cross-combining the clue features of the driver according to a preset rule to obtain the combined features of the driver, the following step further comprises: According to the order volume, order acceptance rate and driver activity characteristics of the city where the driver is located on the freight platform, the city characteristics of the driver are obtained; The combined features and periodic features of the driver are input into the trained clue scoring model to obtain the clue score of the driver, including: The combined features, periodic features and city features of the driver are input into the trained clue scoring model to obtain the clue score of the driver.

6. The driver clue scoring method based on data fusion according to claim 1, characterized in that: After the step of cross-combining the clue features of the driver according to a preset rule to obtain the combined features of the truck driver, the following step further comprises: Input the driver's historical conversations in the multi-channel data into the trained NLP model to obtain the driver's willingness to join the test; The combined features and periodic features of the driver are input into the trained clue scoring model to obtain the clue score of the driver, including: The combined features, periodic features and willingness score of the driver are input into the trained clue scoring model to obtain the clue score of the driver.

7. The driver clue scoring method based on data fusion according to claim 1, characterized in that: After the step of inputting the combined features and periodic features of the driver into the trained clue scoring model to obtain the clue score of the driver, the method further includes: According to the lead score of the driver and the inviter information of the city where the driver is located, operations optimization is performed to obtain the lead level of the driver.

8. A driver clue scoring device based on data fusion, characterized in that: include: A data acquisition unit, used to acquire multi-channel data of the driver to be rated; A data cleaning unit, used for obtaining clue features of the driver according to the multi-channel data; A feature mining unit, used to obtain the periodic features of the driver according to the multi-channel data, and to cross-combine the clue features of the driver according to a preset rule to obtain the combined features of the driver; The scoring unit is used to input the combined features and periodic features of the driver into a trained clue scoring model to obtain a clue score of the driver.

9. An electronic device, characterized in that: The system comprises a processor, a memory and a communication bus; the communication bus is used to connect the processor and the memory; the processor is used to execute a computer program stored in the memory to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and the computer program is used to make a computer execute the method according to any one of claims 1 to 7.

Images