Platform driver loss prediction management system and method based on multivariate data analysis

Abstract

The invention discloses a platform driver loss prediction management system and method based on multivariate data analysis, relates to the technical field of internet and data analysis, and aims to get rid of a traditional classification method of rough division according to a single index in the aspect of driver classification, realize accurate classification of drivers by integrating multiple factors, and improve the classification efficiency of the drivers. The platform can clearly insight the real condition and the potential value of the driver, and a solid foundation is provided for subsequent management; for loss management, the driver loss condition is effectively predicted by means of historical data, early warning is performed in advance, so that the platform has enough time to take retention measures, and the stability of a driver team is effectively guaranteed; in the aspect of driver number prediction, the number change of various types of drivers can be accurately predicted, and key support is provided for a platform to reasonably plan human resources and deploy transport capacity; in addition, the system also reduces the number of lost drivers by optimizing mechanisms such as order allocation, thereby reducing the loss rate and improving the competitiveness of the platform in the market.

Description

Technical Field

[0001] The present invention relates to the field of Internet and data analysis technology, and in particular to a platform driver loss prediction management system and method based on multivariate data analysis. Background Art

[0002] In today's booming online freight industry, platform operators face the key challenge of how to efficiently manage and accurately control all drivers. As an important bridge connecting drivers and shippers, the stability and activity of the driver team of the online freight platform plays a decisive role in the sustainable development of the platform.

[0003] At present, the traditional online freight platform management model often lacks refined classification and in-depth analysis of drivers. On the one hand, the classification method for drivers is relatively simple, and most of them rely on a single indicator such as the number of completed orders for rough classification, which makes it difficult to fully and accurately reflect the real status and potential value of drivers. On the other hand, in order to attract and retain users, existing online platforms have launched various advertising strategies. This competitive situation has greatly broadened the choice space for drivers and increased the mobility of drivers. Therefore, how to predict the number of drivers on the platform in the future and take timely measures to intervene and retain them has become an urgent problem to be solved by the current online freight platform. Summary of the invention

[0004] The purpose of the present invention is to provide a platform driver loss prediction management system and method based on multivariate data analysis to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solution, a platform driver loss prediction management method based on multivariate data analysis, the method comprising: Step S1: collecting historical data of the online freight platform through the Internet, wherein the historical data includes the number of platform logins, registration time, order completion volume, and user ratings corresponding to each completed order of each driver; Step S2: classifying each driver in the historical data based on the order completion volume, the platform login times and the registration time, and counting the number of drivers of each category after classification; Step S3: Classify the drivers of the online freight platform within the first preset time according to the classification standard of each driver in the historical data, and obtain the initial number of each type of drivers and the number of platform logins, order completion volume and user score corresponding to each completed order of each driver within the first preset time, and calculate the comprehensive driver score of each driver in each type of driver according to the user score and order completion volume; Step S4: Preliminarily predicting the number of drivers who will be lost within the second preset time by analyzing the comprehensive scores of the drivers, and calculating a first loss rate based on the number of drivers; Step S5: predicting the number of drivers of each type within the second preset time according to the initial number of drivers of each type within the first preset time, the number of platform logins of each driver, the number of completed orders, and the comprehensive driver score of each driver, and calculating the second churn rate according to the number of drivers; Step S6: obtaining the driver turnover rate within a second preset time through the first turnover rate and the second turnover rate, judging the turnover rate, and issuing a warning prompt according to the judgment result.

[0006] Furthermore, in step S2, each driver in the historical data is classified based on the order completion volume, the number of platform logins, and the registration time, and the classification criteria are as follows: If the driver's registration time on the online freight platform is less than a first time threshold, the order completion volume is less than a first quantity threshold, and the number of platform logins is less than a first frequency threshold, the driver is determined to be a new driver; If the driver's registration time on the online freight platform is greater than or equal to the first time threshold and less than the second time threshold, the order completion volume is less than the first quantity threshold, and the number of platform logins is less than the first frequency threshold, the driver is determined to be a silent driver; If the driver's registration time on the online freight platform is greater than or equal to the first time threshold and less than the second time threshold, the order completion volume is greater than or equal to the first quantity threshold and less than the second quantity threshold, and the number of platform logins is greater than or equal to the first frequency threshold and less than the second frequency threshold, then the driver is determined to be an ordinary driver; If the driver's registration time on the online freight platform is greater than or equal to the first time threshold and less than the second time threshold, the order completion volume is greater than or equal to the second quantity threshold, and the number of platform logins is greater than or equal to the second frequency threshold, the driver is determined to be an active driver; If the driver's registration time on the online freight platform is greater than or equal to a second time threshold, the number of completed orders is greater than or equal to a second quantity threshold, and the number of platform logins is greater than or equal to a second frequency threshold, the driver is determined to be a loyal driver; If the driver's registration time on the online freight platform is greater than or equal to the first time threshold, the order completion volume is less than the first quantity threshold, and the number of platform logins is less than the first frequency threshold, the driver is determined to be a lost driver; Among them, the first time threshold is smaller than the second time threshold, the first quantity threshold is smaller than the second quantity threshold, and the first frequency threshold is smaller than the second frequency threshold.

[0007] Further, step S3 includes: Step S3-1: Count the user scores of each driver for each order completed within the first preset time, and obtain the user comprehensive score by weighted average of the user scores; The user ratings in the above steps directly reflect the satisfaction of the shippers with the driver's service. By statistically averaging the ratings of a large number of orders, the randomness and bias of individual ratings can be eliminated, thereby obtaining a relatively objective user comprehensive rating that can represent the driver's overall service level. For example, if a driver completes multiple orders within a period of time, the ratings given by different shippers may be high or low. Through weighted average calculation, a score that comprehensively reflects the stability of his service quality can be obtained, avoiding excessive influence on the driver's evaluation due to individual extreme evaluations. Step S3-2: Calculate the driver's comprehensive score based on the user's comprehensive score and the order completion volume. The comprehensive score calculation method is as follows: ; Among them, G ij A represents the comprehensive score of the jth driver in the i-th category of drivers. ij represents the comprehensive user rating of the jth driver among the ith drivers, B ij represents the order completion volume of the jth driver among the ith drivers, α1 represents the weight corresponding to the user's comprehensive score, and α2 represents the weight corresponding to the order completion volume; The calculation method in the above steps fully considers the two core dimensions of service quality and business volume. The comprehensive user score reflects the quality of the driver's service, and the order completion volume reflects the diligence of his work and the size of his business contribution. The setting of weights α1 and α2 is crucial, and they can be flexibly adjusted according to the platform's operating priorities and actual business needs. If the platform currently pays more attention to improving service quality to enhance user stickiness, α1 can be appropriately increased. If it is in the business expansion stage and hopes to motivate drivers to increase the order volume, α2 can be set relatively high. For example, for drivers with outstanding service quality but a slightly smaller order volume, if α1 is large, they can still obtain a higher comprehensive score, thereby obtaining a reasonable tilt in platform resource allocation, reward mechanism, etc., to encourage them to continue to maintain high-quality service. For drivers with a high order completion volume but whose service score needs to be improved, they can also clarify their own improvement direction through comprehensive scores, promote the improvement of the overall service level, and achieve the healthy development and management optimization of the platform driver team.

[0008] Further, step S4 includes: Step S4-1: Calculate the user comprehensive score of each lost driver by using the user score corresponding to each completed order of each lost driver in the historical data, and calculate the score threshold according to the order completion volume of each lost driver and the user comprehensive score. The calculation method is as follows: ; Among them, Q represents the scoring threshold, C irepresents the order completion volume of the i-th lost driver in the historical data, D i represents the user comprehensive score of the i-th lost driver in the historical data, β1 represents the weight corresponding to the order completion volume, β2 represents the weight corresponding to the user comprehensive score, and N represents the number of lost drivers in the historical data; In the above steps, the score threshold is calculated by processing the relevant data of each lost driver in the historical data. The score threshold is determined by combining the two factors of order completion volume and user comprehensive score through a specific weighted calculation method. Under the effect of the weight, the calculated score threshold will be more accurate, thereby providing a reasonable reference standard for subsequent judgments.

[0009] Step S4-2: Compare the comprehensive driver score of each driver in each category of drivers within the first preset time with the score threshold, count the number of drivers in each category of drivers whose comprehensive driver scores are lower than the score threshold, and the ratio of the weighted sum of the number of drivers to the weighted sum of the initial number of drivers in each category within the first preset time is the first churn rate; The above steps can identify which drivers' comprehensive performance is close to or lower than the level of previous lost drivers through comparison, count the number of drivers whose comprehensive scores are lower than the threshold, and further calculate the ratio of the weighted sum of the initial number of each type of drivers as the first churn rate. This churn rate can intuitively reflect the proportion of each type of driver group that may face the risk of churn at the current stage.

[0010] Furthermore, the prediction method of the number of drivers in step S5 is as follows: ; Among them, K i represents the number of drivers of the i-th category predicted within the second preset time, K0 represents the initial number of drivers of the i-th category within the first preset time, and L ij represents the number of orders completed by the jth driver in the i-th category within the first preset time, w1 is the weight coefficient corresponding to the order completion amount, L max represents the maximum number of orders completed by the i-th driver, O ij represents the number of platform logins of the jth driver in the i-th category of drivers within the first preset time, w2 is the weight coefficient corresponding to the number of platform logins, O max represents the maximum number of platform login times among the i-th type of drivers, S ij represents the comprehensive score of the jth driver in the i-th category of drivers within the first preset time, w3 is the weight coefficient corresponding to the comprehensive score of the driver, S max represents the maximum value of the comprehensive score of drivers in the i-th category; The above steps comprehensively consider the impact of multiple important factors on the number of drivers. The first is the order completion volume. By comparing it with the maximum order completion volume and combining it with the corresponding weight coefficient, it can reflect the impact of the order completion volume on the trend of driver quantity changes. For example, if a certain type of driver has generally completed a high number of orders and is close to the maximum value in the past period of time, then when predicting the future, based on this factor, their number may have a certain growth trend; the number of platform logins is also an important factor. Drivers who frequently log in to the platform tend to be more actively involved in the business. If a certain type of driver has a high overall login frequency, the number may remain stable or increase in the future under the condition that other conditions remain stable. If the number of logins is small, there may be loss, resulting in a decrease in the number; in addition, drivers with high comprehensive scores usually perform well in service quality and business capabilities, and are more likely to continue to operate on the platform, which has a positive impact on the number of such drivers. Conversely, drivers with low scores may gradually leave the platform; A second churn rate is calculated based on the number of drivers, where the second churn rate is the ratio of the number of drivers who are predicted to be lost within a second preset time to the sum of the predicted numbers of drivers of each type.

[0011] Further, the first churn rate and the second churn rate are weighted averaged to obtain the churn rate of drivers within a second preset time; the churn rate is compared with a preset churn rate, and when the churn rate is greater than the preset churn rate, an early warning is issued, and the order allocation mechanism of the platform is adjusted according to the early warning to reduce the churn rate; The above steps are a scientific and comprehensive evaluation method for determining the final churn rate by weighted averaging the first churn rate and the second churn rate. The first churn rate is based on a comparison between the characteristics of churned drivers in historical data and the comprehensive scores of current drivers, reflecting the correlation between the performance of existing drivers and past churned drivers; the second churn rate is obtained by predicting the number of future drivers and estimating the corresponding number of churned drivers, reflecting the possibility of future churn based on current trends.

[0012] Furthermore, in order to better implement the above method, a platform driver loss prediction management system based on multivariate data analysis is also provided, the system includes a data collection module, a driver classification module, a prediction module, a loss rate calculation module, and an early warning prompt module; A data collection module collects historical data of the online freight platform through the Internet, wherein the historical data includes the number of platform logins, registration time, order completion volume, and user ratings corresponding to each completed order of each driver; A driver classification module, which classifies each driver in the historical data and each driver in the first preset time by analyzing the order completion volume, the number of platform logins, and the registration time, and counts the number of each type of drivers classified in the historical data and the first preset time respectively; A prediction module, which predicts the number of drivers who will be lost within the second preset time based on the initial number of each type of drivers within the first preset time and the number of platform logins, order completion volume and user ratings corresponding to each completed order of each driver; A churn rate calculation module, which calculates the churn rate of drivers within a second preset time according to the prediction result of the prediction module; The early warning prompt module compares the churn rate calculated by the churn rate calculation module with the system preset churn rate, judges the churn rate, and issues an early warning prompt according to the judgment result.

[0013] Further, the prediction module includes a first prediction unit and a second prediction unit; The first prediction unit obtains the initial number of each type of drivers and the number of platform logins, order completion volume and user scores corresponding to each completed order of each driver within the first preset time through the data collection module, calculates the comprehensive driver score of each driver in each type of driver according to the user score and order completion volume, and makes a preliminary prediction on the number of drivers who will become lost drivers within the second preset time by analyzing the comprehensive driver scores; A second prediction unit predicts the number of drivers of each type within the second preset time according to the initial number of drivers of each type within the first preset time, the number of platform logins of each driver, the number of completed orders, and the comprehensive driver score of each driver; Further, the churn rate calculation module includes a first churn rate calculation unit and a second churn rate calculation unit; A first churn rate calculation unit, which calculates a first churn rate according to the number of drivers who are lost predicted by the first prediction unit and the initial number of each type of drivers within a first preset time; The second churn rate unit calculates a second churn rate according to the number of drivers of each type predicted by the second prediction unit, and obtains the churn rate of drivers within a second preset time by taking a weighted average of the first churn rate and the second churn rate.

[0014] Compared with the prior art, the beneficial effects of the present invention are: in terms of driver classification, it gets rid of the traditional classification method of rough classification based on a single indicator, and realizes accurate classification of drivers by integrating multiple factors, so that the platform can have a clear insight into the real situation and potential value of drivers, providing a solid foundation for subsequent management; in terms of churn management, it uses historical data to effectively predict driver churn, and early warning allows the platform to have enough time to take retention measures, effectively ensuring the stability of the driver team; in terms of driver quantity prediction, it can accurately predict the changes in the number of various types of drivers, providing key support for the platform to rationally plan human resources and capacity allocation; in addition, the system also reduces the number of lost drivers by optimizing order distribution and other mechanisms, thereby reducing the churn rate and enhancing the platform's competitiveness in the market. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a method flow diagram of a platform driver loss prediction management system and method based on multivariate data analysis of the present invention; Figure 2 The present invention is a schematic diagram of the system structure of the platform driver loss prediction management system and method based on multivariate data analysis. DETAILED DESCRIPTION

[0016] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making any creative work shall fall within the scope of protection of the present invention.

[0017] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, a platform driver loss prediction management method based on multivariate data analysis, the method comprising: Step S1: collecting historical data of the online freight platform through the Internet, wherein the historical data includes the number of platform logins, registration time, order completion volume, and user ratings corresponding to each completed order of each driver; Step S2: classifying each driver in the historical data based on the order completion volume, the platform login times and the registration time, and counting the number of drivers of each category after classification; The classification criteria for drivers are as follows: If the driver's registration time on the online freight platform is less than a first time threshold, the order completion volume is less than a first quantity threshold, and the number of platform logins is less than a first frequency threshold, the driver is determined to be a new driver; If the driver's registration time on the online freight platform is greater than or equal to the first time threshold and less than the second time threshold, the order completion volume is less than the first quantity threshold, and the number of platform logins is less than the first frequency threshold, the driver is determined to be a silent driver; If the driver's registration time on the online freight platform is greater than or equal to the first time threshold and less than the second time threshold, the order completion volume is greater than or equal to the first quantity threshold and less than the second quantity threshold, and the number of platform logins is greater than or equal to the first frequency threshold and less than the second frequency threshold, then the driver is determined to be an ordinary driver; If the driver's registration time on the online freight platform is greater than or equal to the first time threshold and less than the second time threshold, the order completion volume is greater than or equal to the second quantity threshold, and the number of platform logins is greater than or equal to the second frequency threshold, the driver is determined to be an active driver; If the driver's registration time on the online freight platform is greater than or equal to a second time threshold, the number of completed orders is greater than or equal to a second quantity threshold, and the number of platform logins is greater than or equal to a second frequency threshold, the driver is determined to be a loyal driver; If the driver's registration time on the online freight platform is greater than or equal to the first time threshold, the order completion volume is less than the first quantity threshold, and the number of platform logins is less than the first frequency threshold, the driver is determined to be a lost driver; Step S3: Classify the drivers of the online freight platform within the first preset time according to the classification standard of each driver in the historical data, and obtain the initial number of each type of drivers and the number of platform logins, order completion volume and user score corresponding to each completed order of each driver within the first preset time, and calculate the comprehensive driver score of each driver in each type of driver according to the user score and order completion volume; Wherein, step S3 comprises: Step S3-1: Count the user scores of each driver for each order completed within the first preset time, and obtain the user comprehensive score by weighted average of the user scores; Step S3-2: Calculate the driver's comprehensive score based on the user's comprehensive score and the order completion volume. The comprehensive score calculation method is as follows: ; Among them, G ij A represents the comprehensive score of the jth driver in the i-th category of drivers. ij represents the comprehensive user rating of the jth driver among the ith drivers, B ij represents the order completion volume of the jth driver among the ith drivers, α1 represents the weight corresponding to the user's comprehensive score, and α2 represents the weight corresponding to the order completion volume; In the embodiment of the present invention, the order completion volume of the jth driver in the i-th category is obtained to be 5, and the user scores corresponding to the 5 orders are 4, 3, 4.5, 3.5, and 4 respectively. The calculated user comprehensive score is 3.8. The values ​​of α1 and α2 are both set to 0.5, then the comprehensive score of this driver is G ij =0.5×3.8+0.5×5=4.4; Step S4: Preliminarily predicting the number of drivers who will be lost within the second preset time by analyzing the comprehensive scores of the drivers, and calculating a first loss rate based on the number of drivers; Wherein, step S4 comprises: Step S4-1: Calculate the user comprehensive score of each lost driver by using the user score corresponding to each completed order of each lost driver in the historical data, and calculate the score threshold according to the order completion volume of each lost driver and the user comprehensive score. The calculation method is as follows: ; Among them, Q represents the scoring threshold, C i represents the order completion volume of the i-th lost driver in the historical data, D irepresents the user comprehensive score of the i-th lost driver in the historical data, β1 represents the weight corresponding to the order completion volume, β2 represents the weight corresponding to the user comprehensive score, and N represents the number of drivers who lost drivers in the historical data; In the embodiment of the present invention, three lost drivers are selected. The first lost driver has a completed order volume of 2 and a user comprehensive score of 3; the second lost driver has a completed order volume of 3 and a user comprehensive score of 2.5; the third lost driver has a completed order volume of 1 and a user comprehensive score of 2; β1=β2=0.5 is set, then the score threshold Q=[0.5×(2+3+1)+0.5×(3+2.5+2)] / 3=2.25; Step S4-2: Compare the comprehensive driver score of each driver in each category of drivers within the first preset time with the score threshold, count the number of drivers in each category of drivers whose comprehensive driver scores are lower than the score threshold, and the ratio of the weighted sum of the number of drivers to the weighted sum of the initial number of drivers in each category within the first preset time is the first churn rate; Step S5: predicting the number of drivers of each type within the second preset time according to the initial number of drivers of each type within the first preset time, the number of platform logins of each driver, the number of completed orders, and the comprehensive driver score of each driver, and calculating the second churn rate according to the number of drivers; The method for predicting the number of drivers in step S5 is as follows: ; Among them, K i represents the number of drivers of the i-th category predicted within the second preset time, K0 represents the initial number of drivers of the i-th category within the first preset time, and L ij represents the number of orders completed by the jth driver in the i-th category within the first preset time, w1 is the weight coefficient corresponding to the order completion amount, L max represents the maximum number of orders completed by the i-th driver, O ij represents the number of platform logins of the jth driver in the i-th category of drivers within the first preset time, w2 is the weight coefficient corresponding to the number of platform logins, O max represents the maximum number of platform login times among the i-th type of drivers, S ij represents the comprehensive score of the jth driver in the i-th category of drivers within the first preset time, w3 is the weight coefficient corresponding to the comprehensive score of the driver, S max represents the maximum value of the comprehensive score of drivers in the i-th category; In an embodiment of the present invention, the initial number of drivers of the i-th category is 10, and there are three drivers of this category. The order completion quantities of these three drivers in the first preset time are 4, 6, and 5 respectively, then the maximum order completion quantity is 6, and the weight coefficient of the order completion quantity is set to 0.4; the platform login times are 8, 10, and 7 respectively, then the maximum platform login times are 10, and the weight coefficient of the platform login times is set to 0.3; the comprehensive scores of the drivers are 4.2, 4.5, and 4 respectively, then the maximum comprehensive score of the drivers is 4.5, and the weight coefficient of the comprehensive score of the drivers is set to 0.3; the number of drivers of this category in the second preset time is: K i =10×(1+0.267+0.4+0.333+0.24+0.3+0.21+0.28+0.3+0.267)=11(round up); Calculating a second churn rate based on the number of drivers, the second churn rate being a ratio of the number of drivers predicted to be lost within a second preset time to the sum of the number of drivers of each type predicted; Step S6: obtaining the driver turnover rate within a second preset time through the first turnover rate and the second turnover rate, judging the turnover rate, and issuing an early warning prompt according to the judgment result; The first churn rate and the second churn rate are weighted averaged to obtain the churn rate of drivers within a second preset time; the churn rate is compared with a preset churn rate, and when the churn rate is greater than the preset churn rate, an early warning is issued, and the order allocation mechanism of the platform is adjusted according to the early warning to reduce the churn rate; Among them, in order to better implement the above method, a platform driver loss prediction management system based on multivariate data analysis is also provided, the system includes a data collection module, a driver classification module, a prediction module, a loss rate calculation module, and an early warning prompt module; A data collection module collects historical data of the online freight platform through the Internet, wherein the historical data includes the number of platform logins, registration time, order completion volume, and user ratings corresponding to each completed order of each driver; A driver classification module, which classifies each driver in the historical data and each driver in the first preset time by analyzing the order completion volume, the number of platform logins, and the registration time, and counts the number of each type of drivers classified in the historical data and the first preset time respectively; A prediction module, which predicts the number of drivers who will be lost within the second preset time based on the initial number of each type of drivers within the first preset time and the number of platform logins, order completion volume and user ratings corresponding to each completed order of each driver; Wherein, the prediction module includes a first prediction unit and a second prediction unit; The first prediction unit obtains the initial number of each type of drivers and the number of platform logins, order completion volume and user scores corresponding to each completed order of each driver within the first preset time through the data collection module, calculates the comprehensive driver score of each driver in each type of driver according to the user score and order completion volume, and makes a preliminary prediction on the number of drivers who will become lost drivers within the second preset time by analyzing the comprehensive driver scores; A second prediction unit predicts the number of drivers of each type within the second preset time according to the initial number of drivers of each type within the first preset time, the number of platform logins of each driver, the number of completed orders, and the comprehensive driver score of each driver; A churn rate calculation module, which calculates the churn rate of drivers within a second preset time according to the prediction result of the prediction module; Wherein, the churn rate calculation module includes a first churn rate calculation unit and a second churn rate calculation unit; A first churn rate calculation unit, which calculates a first churn rate according to the number of drivers who are lost predicted by the first prediction unit and the initial number of each type of drivers within a first preset time; A second churn rate unit, which calculates a second churn rate according to the number of drivers of each type predicted by the second prediction unit, and obtains a churn rate of drivers within a second preset time by weighted average of the first churn rate and the second churn rate; The early warning prompt module compares the churn rate calculated by the churn rate calculation module with the system preset churn rate, judges the churn rate, and issues an early warning prompt according to the judgment result.

[0018] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A platform driver loss prediction management method based on multivariate data analysis, characterized by: The method comprises: Step S1: collecting historical data of the online freight platform through the Internet, wherein the historical data includes the number of platform logins, registration time, order completion volume, and user ratings corresponding to each completed order of each driver; Step S2: classifying each driver in the historical data based on the order completion volume, the platform login times and the registration time, and counting the number of drivers of each category after classification; Step S3: Classify the drivers of the online freight platform within the first preset time according to the classification standard of each driver in the historical data, and obtain the initial number of each type of drivers and the number of platform logins, order completion volume and user score corresponding to each completed order of each driver within the first preset time, and calculate the comprehensive driver score of each driver in each type of driver according to the user score and order completion volume; Step S4: Preliminarily predicting the number of drivers who will be lost within the second preset time by analyzing the comprehensive scores of the drivers, and calculating a first loss rate based on the number of drivers; Step S5: predicting the number of drivers of each type within the second preset time according to the initial number of drivers of each type within the first preset time, the number of platform logins of each driver, the number of completed orders, and the comprehensive driver score of each driver, and calculating the second churn rate according to the number of drivers; Step S6: obtaining the driver turnover rate within a second preset time through the first turnover rate and the second turnover rate, judging the turnover rate, and issuing an early warning prompt according to the judgment result.

2. The platform driver loss prediction management method based on multivariate data analysis according to claim 1 is characterized by: Each driver in the historical data is classified based on the order completion volume, the number of platform logins and the registration time. The classification criteria are as follows: If the driver's registration time on the online freight platform is less than a first time threshold, the order completion volume is less than a first quantity threshold, and the number of platform logins is less than a first frequency threshold, the driver is determined to be a new driver; If the driver's registration time on the online freight platform is greater than or equal to the first time threshold and less than the second time threshold, the order completion volume is less than the first quantity threshold, and the number of platform logins is less than the first frequency threshold, the driver is determined to be a silent driver; If the driver's registration time on the online freight platform is greater than or equal to the first time threshold and less than the second time threshold, the order completion volume is greater than or equal to the first quantity threshold and less than the second quantity threshold, and the number of platform logins is greater than or equal to the first frequency threshold and less than the second frequency threshold, then the driver is determined to be an ordinary driver; If the driver's registration time on the online freight platform is greater than or equal to the first time threshold and less than the second time threshold, the order completion volume is greater than or equal to the second quantity threshold, and the number of platform logins is greater than or equal to the second frequency threshold, the driver is determined to be an active driver; If the driver's registration time on the online freight platform is greater than or equal to a second time threshold, the number of completed orders is greater than or equal to a second quantity threshold, and the number of platform logins is greater than or equal to a second frequency threshold, the driver is determined to be a loyal driver; If the driver's registration time on the online freight platform is greater than or equal to a first time threshold, the order completion volume is less than a first quantity threshold, and the number of platform logins is less than a first frequency threshold, the driver will be determined as a lost driver.

3. The platform driver loss prediction management method based on multivariate data analysis according to claim 1 is characterized by: The step S3 comprises: Step S3-1: Counting the user scores of each driver for each order completed within the first preset time, and taking the weighted average of the user scores to obtain a comprehensive user score; Step S3-2: Calculate the driver's comprehensive score based on the user's comprehensive score and the order completion volume. The comprehensive score calculation method is as follows: ； Among them, G ij A represents the comprehensive score of the jth driver in the i-th category of drivers. ij represents the comprehensive user rating of the jth driver among the ith drivers, B ij represents the order completion volume of the jth driver among the ith drivers, α1 represents the weight corresponding to the user's comprehensive score, and α2 represents the weight corresponding to the order completion volume.

4. The platform driver loss prediction management method based on multivariate data analysis according to claim 1 is characterized by: Step S4 includes: Step S4-1: Calculate the user comprehensive score of each lost driver by using the user score corresponding to each completed order of each lost driver in the historical data, and calculate the score threshold according to the order completion volume of each lost driver and the user comprehensive score. The calculation method is as follows: ； Among them, Q represents the scoring threshold, C i represents the order completion volume of the i-th lost driver in the historical data, D i represents the user comprehensive score of the i-th lost driver in the historical data, β1 represents the weight corresponding to the order completion volume, β2 represents the weight corresponding to the user comprehensive score, and N represents the number of lost drivers in the historical data; Step S4-2: Compare the comprehensive driver score of each driver in each category within the first preset time with the score threshold, count the number of drivers in each category whose comprehensive driver score is lower than the score threshold, and the ratio of the weighted sum of the number of drivers to the weighted sum of the initial number of drivers in each category within the first preset time is the first churn rate.

5. The platform driver loss prediction management method based on multivariate data analysis according to claim 1 is characterized by: The method for predicting the number of drivers in step S5 is as follows: ； Among them, K i represents the number of drivers of the i-th category predicted within the second preset time, K0 represents the initial number of drivers of the i-th category within the first preset time, and L ij represents the number of orders completed by the jth driver in the i-th category within the first preset time, w1 is the weight coefficient corresponding to the order completion amount, L max represents the maximum number of orders completed by the i-th driver, O ij represents the number of platform logins of the jth driver in the i-th category of drivers within the first preset time, w2 is the weight coefficient corresponding to the number of platform logins, O max represents the maximum number of platform login times among the i-th type of drivers, S ij represents the comprehensive score of the jth driver in the i-th category of drivers within the first preset time, w3 is the weight coefficient corresponding to the comprehensive score of the driver, S max represents the maximum value of the comprehensive score of drivers in the i-th category; A second churn rate is calculated based on the number of drivers, where the second churn rate is the ratio of the number of drivers who are predicted to be lost within a second preset time to the sum of the predicted numbers of drivers of each type.

6. The platform driver loss prediction management method based on multivariate data analysis according to claim 1 is characterized by: The first churn rate and the second churn rate are weighted averaged to obtain the churn rate of drivers within a second preset time; the churn rate is compared with the preset churn rate, and when the churn rate is greater than the preset churn rate, an early warning is issued, and the order allocation mechanism of this platform is adjusted according to the early warning to reduce the churn rate.

7. A platform driver loss prediction management system based on multivariate data analysis, used to execute the platform driver loss prediction management method based on multivariate data analysis as claimed in any one of claims 1 to 6, characterized in that: The system includes a data collection module, a driver classification module, a prediction module, a churn rate calculation module, and an early warning prompt module; The data collection module collects historical data of the online freight platform through the Internet, wherein the historical data includes the number of platform logins, registration time, order completion volume, and user ratings corresponding to each completed order of each driver; The driver classification module classifies each driver in the historical data and each driver within the first preset time by analyzing the order completion volume, the platform login times and the registration time, and counts the number of drivers of each type after classification within the historical data and the first preset time respectively; The prediction module predicts the number of drivers who will be lost within the second preset time based on the initial number of each type of drivers within the first preset time and the number of platform logins, order completion volume and user rating corresponding to each completed order of each driver; The churn rate calculation module calculates the churn rate of the driver within a second preset time according to the prediction result of the prediction module; The early warning prompt module compares the churn rate calculated by the churn rate calculation module with the system preset churn rate, judges the churn rate, and issues an early warning prompt according to the judgment result.

8. The platform driver loss prediction management system based on multivariate data analysis according to claim 7 is characterized by: The prediction module includes a first prediction unit and a second prediction unit; The first prediction unit obtains, through the data acquisition module, the initial number of each type of drivers and the number of platform logins, order completion volume and user scores corresponding to each completed order of each driver within the first preset time, calculates the comprehensive driver score of each driver in each type of driver according to the user scores and order completion volume, and makes a preliminary prediction on the number of drivers who will become lost drivers within the second preset time by analyzing the comprehensive driver scores; The second prediction unit predicts the number of drivers of each category within the second preset time based on the initial number of drivers of each category within the first preset time, the number of platform logins of each driver, the number of completed orders and the comprehensive driver score of each driver.

9. The platform driver loss prediction management system based on multivariate data analysis according to claim 7 is characterized by: The churn rate calculation module includes a first churn rate calculation unit and a second churn rate calculation unit; The first churn rate calculation unit calculates the first churn rate according to the number of drivers who are lost predicted by the first prediction unit and the initial number of each type of drivers within the first preset time; The second churn rate unit calculates a second churn rate according to the number of drivers of each type predicted by the second prediction unit, and obtains a churn rate of drivers within a second preset time by weighted averaging the first churn rate and the second churn rate.

Images