Method, device, computer equipment and storage medium for predicting business conditions of logistics network points

Abstract

The present application relates to a method, device, computer equipment and storage medium for predicting the business situation of a logistics network. The method includes: obtaining the network feature indicators corresponding to the logistics network to be predicted, and the index data corresponding to the network feature indicators; inputting the index data into a pre-trained complaint prediction model to determine the change trend of the index data through the complaint prediction model, obtaining the predicted index data corresponding to the index data according to the change trend, and determining the predicted complaint volume and / or predicted complaint rate of the logistics network according to the predicted index data; obtaining the predicted complaint volume and / or predicted complaint rate output by the complaint prediction model as the user complaint volume of the logistics network, realizing the advance prediction of the customer complaint situation, and moreover, by using the index data of the corresponding feature indicators of the logistics network, it can conform to the business situation of the logistics network and improve the prediction accuracy of the user complaint volume.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular, to a method, device, computer device, and storage medium for predicting the business situation of a logistics network point. Background Art

[0002] With the continuous growth of business volume, the number of users served by logistics enterprises is also increasing. In order to provide better services to users, logistics enterprises often improve their operations based on user complaints or suggestions.

[0003] In the prior art, when facing user complaints, it is often necessary to review the operations after the business is completed, find the reasons for user complaints, and avoid their recurrence, so as to control user complaints. However, this control method has obvious lag and is difficult to predict customer complaints in advance. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, device, computer device, and storage medium for predicting the business situation of a logistics network point to solve the above technical problems.

[0005] A method for predicting the business situation of a logistics network point, the method comprising:

[0006] Obtaining the network point characteristic indicators corresponding to the logistics network point to be predicted, and the indicator data corresponding to the network point characteristic indicators; the network point characteristic indicators are obtained by performing feature engineering processing on a plurality of logistics characteristic indicators of the logistics network point;

[0007] Inputting the indicator data into a pre-trained complaint prediction model to determine the change trend of the indicator data through the complaint prediction model, obtaining the predicted indicator data corresponding to the indicator data according to the change trend, and determining the predicted complaint volume and / or predicted complaint rate of the logistics network point according to the predicted indicator data;

[0008] Obtaining the predicted complaint volume and / or predicted complaint rate output by the complaint prediction model as the user complaint volume of the logistics network point.

[0009] Optionally, the method further comprises:

[0010] When the user complaint volume exceeds a preset threshold, obtaining the controllable characteristic indicators from the plurality of network point characteristic indicators;

[0011] Determining the current business characteristic period of the logistics network point, and obtaining the data fluctuation range of the controllable characteristic indicators in the business characteristic period;

[0012] Determining the indicator data adjustment range according to the data fluctuation range and the indicator data corresponding to the controllable characteristic indicators;

[0013] Generate an index data regulation suggestion based on the regulation range of the index data, where the index data regulation suggestion is used to improve the number of user complaints.

[0014] Optionally, obtaining the data fluctuation range of the controllable feature index during the business feature period includes:

[0015] When the business feature period is the business peak period, determine the data fluctuation range corresponding to the controllable feature index according to the peak period index data corresponding to the controllable feature index in the historical business peak period;

[0016] And / or,

[0017] When the business feature period is the business stable period, starting from the current time point, obtain the stable period index data corresponding to the controllable feature index within a preset time range, and determine the data fluctuation range corresponding to the controllable feature index according to the stable period index data.

[0018] Optionally, it further includes:

[0019] Obtain the network feature index corresponding to the logistics network from multiple preset logistics feature indexes, and obtain the sample index data and complaint volume label corresponding to the network feature index;

[0020] Input the sample index data into the time series model to be trained, so as to determine the change trend of the sample index data through the time series model, obtain the sample prediction index data corresponding to the sample index data according to the change trend, and determine the sample prediction complaint volume and / or sample prediction complaint rate according to the sample index data;

[0021] Determine the training error according to the sample prediction complaint volume and / or sample prediction complaint rate, and the complaint volume label, and adjust the time series model according to the training error until the training end condition is met, and obtain the trained complaint prediction model.

[0022] Optionally, obtaining the network feature index corresponding to the logistics network from multiple preset logistics feature indexes includes:

[0023] According to the extreme gradient boosting algorithm and the index data corresponding to multiple preset logistics feature indexes, determine the feature gain scores corresponding to each logistics feature index, and sum the multiple feature gain scores to obtain the total score;

[0024] Determine the ratio of each feature gain score to the total score, and determine the logistics feature index whose ratio exceeds the preset ratio threshold as the network feature index corresponding to the logistics network.

[0025] Optionally, before the step of determining the feature gain scores corresponding to each logistics feature index according to the extreme gradient boosting algorithm and the index data corresponding to a plurality of preset logistics feature indexes, the method further includes:

[0026] Obtain a plurality of original logistics feature indexes, and determine the information value corresponding to each original logistics feature index;

[0027] Determine the plurality of original logistics feature indexes whose information value exceeds the preset value threshold as the plurality of preset logistics feature indexes;

[0028] Wherein, the plurality of original logistics feature indexes include at least two of the following:

[0029] Logistics pickup and delivery link feature indexes, logistics transfer link feature indexes, logistics transportation link feature indexes, logistics process customer complaint feature indexes.

[0030] Optionally, the sample index data includes multiple groups of sample index data, and the obtaining of the sample index data corresponding to the network point feature index includes:

[0031] Obtain multiple groups of candidate index data corresponding to the network point feature index, and the complaint volume label corresponding to each group of candidate index data; the multiple groups of candidate index data are index data within time periods of different time lengths;

[0032] For each group of candidate index data, input the candidate index data into the time series model to be trained, obtain the predicted complaint volume and / or predicted complaint rate output by the time series model, and determine and / or predict the complaint rate according to the predicted complaint volume, and the complaint volume label corresponding to the candidate index data, and determine the training error;

[0033] Select the candidate index data with the smallest corresponding training error from the multiple groups of candidate index data as the sample index data corresponding to the network point feature index.

[0034] A device for predicting the business situation of a logistics network point, the device includes:

[0035] An index data acquisition module, configured to acquire the network point feature index corresponding to the logistics network point to be predicted, and the index data corresponding to the network point feature index; the network point feature index is obtained by performing feature engineering processing on multiple logistics feature indexes of the logistics network point;

[0036] An index data input module for inputting the index data into a pre-trained complaint prediction model to determine the change trend of the index data through the complaint prediction model, obtaining predicted index data corresponding to the index data according to the change trend, and determining the predicted complaint volume and / or predicted complaint rate of the logistics network point according to the predicted index data;

[0037] A user complaint volume acquisition module for acquiring the predicted complaint volume and / or predicted complaint rate output by the complaint prediction model as the user complaint volume of the logistics network point.

[0038] A computer device comprising a memory and a processor, where the memory stores a computer program, and the processor implements the steps of the method as described in any one of the above when executing the computer program.

[0039] A computer-readable storage medium having a computer program stored thereon, and the computer program implements the steps of the method as described in any one of the above when executed by a processor.

[0040] The above-mentioned method, device, computer device and storage medium for predicting the business situation of a logistics network point obtain the network point characteristic indexes corresponding to the logistics network point to be predicted and the index data corresponding to the network point characteristic indexes, input the index data into a pre-trained complaint prediction model to determine the change trend of the index data through the complaint prediction model, obtain the predicted index data corresponding to the index data according to the change trend, determine the predicted complaint volume or predicted complaint rate of the logistics network point according to the predicted index data, and use it as the user complaint volume of the logistics network point, realizing the advance prediction of customer complaints. Moreover, by using the index data of the characteristic indexes corresponding to the logistics network point, it can fit the business situation of the logistics network point and improve the prediction accuracy of the user complaint volume. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is an application environment diagram of a method for predicting the business situation of a logistics network point in an embodiment;

[0042] Figure 2 It is a flowchart showing the process of a method for predicting the business situation of a logistics network point in an embodiment;

[0043] Figure 3 It is a flowchart showing the process of steps for improving the user complaint volume in an embodiment;

[0044] Figure 4 It is a flowchart showing the process of steps for determining the network point characteristic indexes in an embodiment;

[0045] Figure 5 It is a flowchart showing the process of a method for managing the user complaint volume in an embodiment;

[0046] Figure 6 It is a structural block diagram of a business situation prediction device for a logistics network point in an embodiment;

[0047] Figure 7 It is an internal structure diagram of a computer device in an embodiment. Specific implementation manners

[0048] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0049] A method for predicting the business situation of a logistics network point provided by the present application can be applied to, for example, Figure 1 the application environment shown. Among them, the terminal 102 can communicate with the server 104 through the network. (Described in combination with the overall solution of claim 1). Among them, the terminal 102 can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers and portable wearable devices, and the server 104 can be implemented by an independent server or a server cluster composed of multiple servers.

[0050] In one embodiment, as Figure 2 shown, a method for predicting the business situation of a logistics network point is provided. Taking the application of this method to the Figure 1 server as an example for illustration. It should be understood that this method can also be used independently on the terminal. Specifically, this method can include the following steps:

[0051] Step 201, obtain the network point characteristic indicators corresponding to the logistics network point to be predicted, and the index data corresponding to the network point characteristic indicators; the network point characteristic indicators are obtained after performing feature engineering processing on multiple logistics characteristic indicators of the logistics network point.

[0052] As an example, a logistics network point can refer to a network point that provides one or more logistics services. The network point can be a real service network point, such as a network point with a fixed business address that picks up express deliveries for users and provides receiving and sending services offline; the network point can also be a virtual service network point, such as a network point that receives user requests online and responds in each region. The logistics network point can also be a network point at different levels. For example, a parent logistics network point has one or more sub-logistics network points. The parent logistics network point and / or the sub-logistics network points can be the logistics network points referred to in this embodiment; it can also be a network point divided by region, such as a network point distinguished according to administrative divisions.

[0053] The logistics feature indicators can refer to the indicators reflecting the characteristics of logistics operations. The network point feature indicators can be used to reflect the characteristics of the operations of a specified logistics network point, and the network point feature indicators can be one or more of the multiple logistics feature indicators.

[0054] Feature engineering can refer to the process of processing multiple logistics feature indicators to obtain the indicators for training the model. Among them, feature engineering can include any one or more of the following processing processes: data preprocessing, feature selection, and feature dimensionality reduction. Among them, data preprocessing can refer to optimizing and improving the data corresponding to the incomplete, inconsistent, and directly unusable logistics feature indicators; feature selection can refer to the process of screening the logistics feature indicators to screen out the feature indicators with a correlation lower than the threshold with model training; feature dimensionality reduction can refer to reducing the feature dimension when the scale of the logistics feature indicators is huge (for example, the type of logistics feature indicators exceeds the threshold).

[0055] In specific implementation, with the increase in business volume and the expansion of the user scope, logistics network points often receive complaints or suggestions from users, and different logistics network points can have different business characteristics. In order to predict the user complaint volume, one or more network point feature indicators corresponding to the logistics network point to be predicted can be obtained, and the index data corresponding to the network point feature indicators can be obtained. Among them, the network point feature indicators corresponding to the logistics network point can be obtained by processing multiple logistics feature indicators through feature engineering for this network point.

[0056] Step 202: Input the index data into a pre-trained complaint prediction model to determine the change trend of the index data through the complaint prediction model, obtain the predicted index data corresponding to the index data according to the change trend, and determine the predicted complaint volume and / or predicted complaint rate of the logistics network point according to the predicted index data.

[0057] After obtaining the index data corresponding to the network point feature indicators, the index data can be input into a pre-trained complaint prediction model, the change trend of the index data within a preset time can be determined through the complaint prediction model, and the future index data can be predicted according to the change trend to obtain the predicted index data. Furthermore, the predicted complaint volume and / or predicted complaint rate of the logistics network point can be determined according to the predicted index data.

[0058] Step 203: Obtain the predicted complaint volume and / or predicted complaint rate output by the complaint prediction model as the user complaint volume of the logistics network point.

[0059] As an example, the user complaint volume of a logistics network point can include the predicted absolute complaint volume and the predicted relative complaint volume. Among them, the predicted absolute complaint volume can be a specific prediction of the number of complaints, that is, the predicted complaint volume, and the predicted complaint volume can determine the specific number of complaints received by the logistics network point; the predicted relative complaint volume can be the ratio of the number of complaints to the number of services provided, that is, the predicted complaint rate, and the predicted complaint rate has high comparability, and different network points can be compared through the predicted complaint rate.

[0060] After inputting the index data into the complaint prediction model, the predicted complaint volume and / or the predicted complaint rate output by the complaint prediction model can be obtained and used as the user complaint volume of the logistics network point. After obtaining the user complaint volume, the user complaint volume can be fed back to the corresponding terminal for display.

[0061] In this embodiment, by obtaining the network point characteristic indexes corresponding to the logistics network point to be predicted and the index data corresponding to the network point characteristic indexes, the index data is input into the pre-trained complaint prediction model to determine the change trend of the index data through the complaint prediction model, obtain the predicted index data corresponding to the index data according to the change trend, determine the predicted complaint volume or the predicted complaint rate of the logistics network point according to the predicted index data, and use it as the user complaint volume of the logistics network point, realizing the advance prediction of the customer complaint situation. Moreover, by using the index data of the characteristic indexes corresponding to the logistics network point, it can fit the business situation of the logistics network point and improve the prediction accuracy of the user complaint volume.

[0062] In one embodiment, as Figure 3 shown, the method may further include the following steps:

[0063] Step 301, when the user complaint volume exceeds the preset threshold, obtain the controllable characteristic indexes from the multiple network point characteristic indexes.

[0064] As an example, the network point characteristic indexes can include non-controllable characteristic indexes and controllable characteristic indexes. Among them, the non-controllable characteristic indexes can refer to the characteristic indexes whose index data is difficult to be regulated through decisions or measures in the logistics business; the controllable characteristic indexes can refer to the characteristic indexes whose index data can be influenced through decisions or measures and the index data can be regulated in the logistics business.

[0065] Further, the controllable characteristic indicators can be divided into resource - type controllable characteristic indicators and exception - type controllable characteristic indicators. The resource - type controllable characteristic indicators can be characteristic indicators related to logistics resources, and can include any one or more of the following: the number of front - line personnel, the on - the - job rate of front - line personnel, the high - speed rail carrying capacity, the railway shipment volume, the weight of goods shipped by scheduled airlines; The exception - type controllable characteristic indicators can be controllable characteristic indicators in case of abnormal situations, and can include any one or more of the following: the initial unloaded and unshipped quantity, the number of unloaded and unshipped invoices, the collection rate, the delivery success rate, the on - time arrival rate at the transfer point, the on - time completion rate of delivery time, the on - time departure rate of branch lines, the on - time arrival rate of branch lines at the destination, the on - time departure rate of main lines, the on - time arrival rate of main lines at the destination, the proportion of tasks with a loading rate greater than the preset threshold, the normal rate of scheduled airlines. After determining the controllable characteristic indicators, the network characteristic indicators other than the controllable characteristic indicators can be determined as non - controllable characteristic indicators.

[0066] In practical applications, the predicted user complaint volume can be monitored. Specifically, a threshold for monitoring the user complaint volume, that is, the preset threshold, can be set in advance. When setting the threshold, it can be set according to the logistics network situation and / or user management requirements. After obtaining the user complaint volume output by the complaint prediction model, it can be determined whether the user complaint volume exceeds the preset threshold. If the user complaint volume does not exceed the preset threshold, no further control measures may be taken temporarily; when the user complaint volume exceeds the preset threshold, it can be determined that there may be a deterioration in the user complaint situation within the preset future time. Therefore, controllable characteristic indicators can be obtained from multiple network characteristic indicators, and by regulating the controllable characteristic indicators, an increase in the future user complaint volume can be prevented.

[0067] Step 302: Determine the current business characteristic period of the logistics network, and obtain the data fluctuation range of the controllable characteristic indicators during the business characteristic period.

[0068] In practical applications, the indicator data corresponding to the controllable characteristic indicators has a fluctuation range, and since the business volume undertaken by the logistics network at different times can be different, the corresponding fluctuation range can also be different. Therefore, after determining the controllable characteristic indicators, the current business characteristic period of the logistics network can be determined, and the data fluctuation range of the controllable characteristic indicators during this business characteristic period can be obtained.

[0069] Step 303: Determine the indicator data regulation range according to the data fluctuation range and the indicator data corresponding to the controllable characteristic indicators.

[0070] After obtaining the data fluctuation range, the indicator data regulation range can be determined according to the data fluctuation range and the indicator data currently corresponding to the controllable characteristic indicators.

[0071] Specifically, the target for regulating the index data of the controllable feature index can be determined from the data fluctuation range, and the regulation range of the index data can be determined according to the current index data corresponding to the controllable index and the target for regulating the index data. For example, if the data fluctuation range is [m, n], the current index data corresponding to the controllable feature index is x, and x is less than n, then n - x can be determined as the regulation range of the index data.

[0072] Step 304: Generate an index data regulation suggestion based on the regulation range of the index data, where the index data regulation suggestion is used to improve the user complaint volume.

[0073] After obtaining the regulation range of the index data, an index data regulation suggestion can be generated based on this regulation range to improve the user complaint volume. In a specific implementation, an index data regulation suggestion including the regulation range of the index data and the predicted user complaint volume can be generated and fed back to the terminal. After viewing the index data regulation suggestion, the user can determine whether it is necessary to pre-control the user complaints based on the predicted user complaint volume. If necessary, relevant measures can be further determined in combination with the regulation range of the index data in the index data regulation suggestion.

[0074] In this embodiment, by determining the current business characteristic period of the logistics network point and obtaining the data fluctuation range of the controllable feature index during the business characteristic period, according to the data fluctuation range and the index data corresponding to the controllable feature index, the regulation range of the index data is determined, and an index data regulation suggestion is generated based on the regulation range of the index data, realizing the pre-control of the user complaint volume. By regulating the business associated with the controllable feature index, the logistics business can be optimized before user complaints, reducing the future user complaint volume and improving the customer experience. Moreover, by determining the regulation range of the index data based on the index fluctuation range, a reasonable regulation amplitude or regulation intensity can be determined during regulation, and practical regulation suggestions can be put forward, which is beneficial to improving the user complaint volume subsequently.

[0075] In one embodiment, the obtaining of the data fluctuation range of the controllable feature index during the business characteristic period may include the following steps:

[0076] When the business characteristic period is the business peak period, determine the data fluctuation range corresponding to the controllable feature index according to the peak period index data corresponding to the controllable feature index in the historical business peak period, and / or when the business characteristic period is the business stable period, starting from the current time point, obtain the stable period index data corresponding to the controllable feature index within a preset time range, and determine the data fluctuation range corresponding to the controllable feature index according to the stable period index data.

[0077] As an example, the peak business period can be a time period when the volume of logistics business exceeds a preset threshold, such as during commercial activity promotions, holidays, etc.; the stable business period can be a time period when the difference in the volume of logistics business is less than the difference threshold within multiple consecutive business cycles.

[0078] In practical applications, when the business characteristic period is the peak business period, since there is a significant difference in the volume of logistics business between the stable business period and the peak business period, in order to improve the reliability of the data fluctuation range, the data fluctuation range of the controllable characteristic index can be determined according to the peak period index data corresponding to the controllable index in the historical peak business period. When determining the data fluctuation range, the peak period index data corresponding to the historical peak business period in the same period can be used, and the data fluctuation range can be determined according to this peak period index data. For example, referring to the controllable index data during the same festival in the past, the data fluctuation range can fluctuate up and down by 10% based on this index data, that is, the lower limit of the data fluctuation range is obtained according to the past peak period index data during the same festival * (1 - 10%), and the upper limit of the data fluctuation range is obtained according to the past peak period index data during the same festival * (1 + 10%).

[0079] When the business characteristic period is the stable business period, since the difference in the volume of logistics business is small within multiple consecutive business cycles, when determining the data fluctuation range, starting from the current time point, the index data corresponding to the controllable characteristic index within the preset time range can be obtained as the stable period index data, and the data fluctuation range can be determined according to the stable period index data. When determining, multiple index data corresponding to the controllable index within the preset time range can be determined, and their average value can be calculated, and the data fluctuation range can be determined based on this average value. For example, by calculating the average value of the controllable index data in the past three months as the benchmark, the lower limit of the data fluctuation range is obtained according to the average value * (1 - 10%), and the upper limit of the data fluctuation range is obtained according to the average value * (1 + 10%).

[0080] In this embodiment, by determining the index data corresponding to the business characteristic period, a reasonable data fluctuation range can be determined based on this index data, which is beneficial to putting forward practical control suggestions subsequently.

[0081] In one embodiment, the method may further include the following steps:

[0082] Step 401, obtain the network characteristic index corresponding to the logistics network from multiple preset logistics characteristic indexes, and obtain the sample index data and the complaint volume label corresponding to the network characteristic index.

[0083] In specific implementation, the system construction of feature indicators can be carried out in advance, that is, obtaining multiple preset logistics feature indicators. When constructing the feature indicator system, various types of feature indicators related to user complaints can be obtained, such as business operation feature indicators, timeliness feature indicators, safety feature indicators, etc. After analyzing various types of feature indicators related to user complaints, at least the following two types of logistics feature indicators can be obtained: logistics pick-up and delivery link feature indicators, logistics transfer link feature indicators, logistics transportation link feature indicators, and logistics process customer complaint feature indicators.

[0084] In practical applications, for different outlets, the correlation degree and / or importance between different feature indicators and the user complaint volume can be different. For example, outlet A is more dependent on feature indicator A when predicting the user complaint volume, while outlet B is more dependent on feature indicator B, and feature indicator A plays a smaller role. Based on this, corresponding outlet feature indicators can be obtained from multiple preset logistics features, and the sample indicator data and complaint volume label corresponding to the outlet feature indicators can be obtained, where the complaint volume label is the real complaint volume and / or complaint rate.

[0085] Step 402: Input the sample indicator data into the time series model to be trained, so as to determine the change trend of the sample indicator data through the time series model, obtain the sample prediction indicator data corresponding to the sample indicator data according to the change trend, and determine the sample prediction complaint volume and / or sample prediction complaint rate according to the sample indicator data.

[0086] After obtaining the sample indicator data, it can be input into the time series model to be trained. The time series model is used to determine the change trend of the sample indicator data, and the sample prediction indicator data corresponding to the sample indicator data is obtained according to the change trend. The sample prediction complaint volume or sample prediction complaint rate is determined according to the sample indicator data.

[0087] In one example, since in the logistics industry, the generation time of user complaints can have obvious periodicity. For example, during traditional holidays or e-commerce promotion activities, the logistics volume will increase significantly. And for subsequent control of predicting the user complaint volume, the time series model can be a model that is sensitive to time periodicity and has interpretability, such as the prophet model. The prophet model has the characteristics of easy adjustment of the period, no need to process missing values, fast fitting speed, strong parameter interpretability, etc., and can effectively improve the prediction accuracy of the model.

[0088] Step 403: Determine the training error according to the sample prediction complaint volume and / or sample prediction complaint rate, and the complaint volume label, and adjust the time series model according to the training error until the training end condition is met, and obtain the trained complaint prediction model.

[0089] After obtaining the sample predicted complaint volume and / or sample predicted complaint rate, it can be compared with the complaint volume label to obtain the training error, and the time series model can be adjusted according to the training error until the training end condition is met. For example, when the number of iterations or the training error is less than the threshold, the current time series model can be used as the trained complaint prediction model. Among them, the training error can be calculated by the following formula:

[0090]

[0091] Among them, n is the number of sample index data, y 真实 is the complaint volume label corresponding to the sample index data, and y 预测 is the sample predicted complaint volume and / or sample predicted complaint rate corresponding to the sample index data.

[0092] In this embodiment, by training the time series model, the periodic change trend of the sample index data can be determined, which can make the model have strong interpretability and improve the accuracy of the model prediction result.

[0093] In one embodiment, as Figure 4 shown, the obtaining of the network characteristic index corresponding to the logistics network from multiple preset logistics characteristic indexes may include the following steps:

[0094] Step 501, according to the extreme gradient boosting algorithm and the index data corresponding to multiple preset logistics characteristic indexes, determine the characteristic gain scores corresponding to each logistics characteristic index, and sum the multiple characteristic gain scores to obtain the total score.

[0095] As an example, the characteristic gain score can be used to measure the change brought to the model prediction result when a logistics characteristic index is added to the model training, such as the impact on the accuracy of the prediction result. By calculating the characteristic gain score corresponding to the logistics characteristic index, the relative importance of the specified logistics characteristic index among multiple logistics characteristic indexes can be determined.

[0096] In practical applications, for each logistics network, according to the extreme gradient boosting algorithm (xgboost) and the index data of each preset logistics characteristic index, determine the characteristic gain score corresponding to each logistics characteristic index, and sum the multiple characteristic gain scores to obtain the total score. Among them, in the extreme gradient boosting algorithm, multiple decision trees in the algorithm can be used to perform attribute tests on the index data respectively to obtain the final prediction result.

[0097] Specifically, the characteristic gain scores corresponding to each logistics characteristic index can be calculated by the following formula:

[0098]

[0099] Among them, is the score of the left subtree, is the score of the right subtree, is the score when not splitting, δ is the complexity cost introduced by adding a leaf node in the decision tree, and Gain is the feature gain score.

[0100] Step 502: Determine the ratio of each feature gain score to the total score, and determine the logistics feature indicators whose ratio exceeds the preset ratio threshold as the network feature indicators corresponding to the logistics network points.

[0101] After obtaining the total score, the ratio of each gain feature score to the total score can be calculated, and this ratio can be determined as the relative importance of the logistics feature indicators. Furthermore, the logistics feature indicators whose ratio exceeds the preset ratio threshold can be determined as the network feature indicators corresponding to the logistics network points. In another example, according to the ratios corresponding to each logistics feature indicator, multiple logistics feature indicators can be sorted in descending order, and the preset number of logistics feature indicators ranked at the top can be determined as the network feature indicators.

[0102] In this embodiment, by determining the ratio of each feature gain score to the total score, and determining the logistics feature indicators whose ratio exceeds the preset ratio threshold as the network feature indicators corresponding to the logistics network points, the importance of each logistics feature indicator to the logistics network point can be accurately measured. Furthermore, the logistics feature indicators with importance within the preset range can be selected as the network feature indicators. At the same time, by reducing the feature indicators used for training the model, training resources and training time can be avoided.

[0103] In one embodiment, before the step of determining the feature gain scores corresponding to each logistics feature indicator according to the extreme gradient boosting algorithm and the indicator data corresponding to a preset multiple of logistics feature indicators, the method may further include the following steps:

[0104] Obtain a plurality of original logistics feature indicators, and determine the information value corresponding to each original logistics feature indicator; determine the plurality of original logistics feature indicators whose information value exceeds the preset value threshold as the preset multiple of logistics feature indicators.

[0105] As an example, the plurality of original logistics feature indicators may include at least two of the following:

[0106] Logistics collection and delivery link feature indicators, logistics transfer link feature indicators, logistics transportation link feature indicators, logistics process customer complaint feature indicators.

[0107] Among them, the characteristic indicators of the logistics collection and delivery link can reflect the inherent collection and delivery business characteristics of the logistics outlets, and may include any one or more of the following: the number of collections, the number of deliveries, the type of items collected and delivered, whether they are large fresh items, the number of collections and deliveries exceeding the weight threshold, the collection rate, the collection rate, the proportion of items arriving at the preset time, the delivery success rate, the number of stranded items, the efficiency of the front line, the on-the-job rate of the front line personnel, the actual number of people deployed in the front line, and the number of people temporarily deployed in the front line; the efficiency of the front line can be calculated by the following formula:

[0108] First-line efficiency = (number of parcels received + number of parcels delivered) / actual number of people on the first line

[0109] Since the operation of the transit link is complex and the amount of logistics handled is huge, a large number of user complaints are associated with the transit link. Based on this, characteristic indicators related to the transit link can be constructed, namely, characteristic indicators of the logistics transit link. The characteristic indicators of the logistics transit link can be indicators that reflect the business characteristics of logistics outlets in the logistics middle link. The characteristic indicators of the logistics transit link are highly correlated with the model prediction effect. When constructing a characteristic indicator system, it can be constructed from multiple dimensions. Specifically, the characteristic indicators of the logistics transit link can include any one or more of the following: the amount of circulating pieces, the efficiency of transit, the number of first and second-level vehicles, the number of third-level vehicles, the amount of untimely unloading during transit, the amount of untimely delivery during transit, the actual number of operators, the amount of cross-shift delivery (tickets), the proportion of cross-shift delivery during delivery, the amount of initial unloaded and unsent, the amount of initial loaded and unsent, the proportion of tasks with a loading rate greater than the preset threshold, the proportion of self-employed drivers, the proportion of self-employed vehicles, and the proportion of self-employed tasks.

[0110] In the logistics and transportation link, there is the possibility of incurring user complaints due to insufficient loading rate, improper transportation, delays and other reasons. Therefore, characteristic indicators related to the logistics and transportation link can be obtained to obtain characteristic indicators of the logistics and transportation link. The characteristic indicators of the logistics and transportation link can be indicators obtained for any one or more of the following transportation methods: all-cargo aircraft transportation, charter flight transportation, railway transportation, and traffic transportation. When constructing the characteristic indicators of the logistics and transportation link, a characteristic indicator system can be constructed from the aspects of transportation capacity and punctuality. The characteristic indicators of the logistics and transportation link may include any one or more of the following: all-cargo aircraft shipment volume, charter flight shipment weight, railway shipment volume, high-speed rail operation volume, all-cargo aircraft punctuality rate, charter flight cargo rate, charter flight normal rate, main line departure punctuality rate, main line arrival punctuality rate, branch line departure punctuality rate, branch line arrival punctuality rate, and checkpoint punctuality rate.

[0111] Since the model needs to make predictions for different logistics outlets, for example, predicting the logistics outlets corresponding to different business strategic regions or administrative divisions, and the user complaint situations in different business strategic regions or administrative divisions are not the same. For the logistics outlets with serious historical user complaints, the user complaint behavior has guiding significance for the model prediction results. By paying attention to the user complaint situations of each outlet, relevant indicators that lead to the deterioration of user complaint situations can be determined, which is beneficial for the model to predict the user complaint volume or generate targeted improvement measures. Therefore, the customer complaint characteristic indicators of the logistics process can be obtained, and this indicator can include any one or more of the following: the user complaint volume corresponding to the group assessment and / or regional assessment, the user complaint rate corresponding to the group assessment and / or regional assessment, the express security complaint volume and complaint rate, the express timeliness complaint volume and complaint rate, the service experience complaint volume and complaint rate. Among them, the complaint rate (or user complaint rate) can be calculated through the following formula:

[0112] Complaint rate = Complaint volume / (Received volume + Delivered volume) * 1000000

[0113] It should be understood that in addition to the original logistics characteristic indicators that can include the above characteristic indicators, the preset logistics characteristic indicators can also include the above characteristic indicators.

[0114] In specific implementation, after obtaining multiple original logistics characteristic indicators, the information value (i.e., Information Value, iv value) corresponding to each logistics characteristic indicator can be determined. Specifically, for each original logistics characteristic indicator, the weight of evidence (i.e., Weight of Evidence, woe value) can be calculated first, and the weight of evidence can be calculated through the following formula:

[0115]

[0116] where woe i is the woe value corresponding to each group after automatically grouping each dimension of logistics characteristics; py i is the proportion of the number of positive samples in each group to the total number of positive samples after automatically grouping each dimension of logistics characteristics; pni is the proportion of the number of negative samples in each group to the total number of negative samples after automatically grouping each dimension of logistics characteristics; #y i is the number of positive samples in each group after automatically grouping each dimension of logistics characteristics; #y T is the number of positive samples in all samples; #n i is the number of positive samples in each group after automatically grouping each dimension of logistics characteristics; #n T is the number of negative samples in all samples.

[0117] After obtaining the weight of evidence, the information value can be calculated based on the weight of evidence, and it can be calculated through the following formula:

[0118]

[0119]

[0120] Among them, each dimension of logistics feature corresponds to an iv value, and this iv value can be the sum of each group of iv values.

[0121] After obtaining the information value corresponding to each original logistics feature index, the preliminary screening of logistics features can be carried out. Specifically, the logistics feature indexes whose information value exceeds the preset value threshold can be determined as a preset number of logistics feature indexes, and then the preset number of logistics features can be subjected to secondary screening through the extreme gradient boosting algorithm.

[0122] In another example, when the information value is available, multiple original logistics feature indexes can be sorted in descending order, and the preset number of logistics feature indexes ranked at the top can be determined as the preset number of logistics feature indexes.

[0123] In this embodiment, two methods, namely the value information and the extreme gradient boosting algorithm, can be combined to screen the logistics feature indexes, improving the reliability of the final outlet feature indexes, which is beneficial to improving the accuracy of the model prediction results. At the same time, by reducing the feature indexes used for training the model, training resources and training time can be avoided.

[0124] In one embodiment, the sample index data includes multiple groups of sample index data, and the obtaining of the sample index data corresponding to the outlet feature indexes may include the following steps:

[0125] Step 601, obtain multiple groups of candidate index data corresponding to the outlet feature indexes, and the complaint volume label corresponding to each group of candidate index data; the multiple groups of candidate index data are index data within time periods of different time lengths.

[0126] As an example, the multiple groups of candidate index data may refer to index data within time periods of different time lengths. For example, for the same outlet feature index, index data within x days, y days, and z days can be collected, and then multiple groups of candidate index data can be obtained.

[0127] In specific implementation, multiple groups of candidate data of the outlet feature indexes and the complaint volume label corresponding to each group of candidate index data can be obtained.

[0128] Step 602. For each group of candidate index data, input the candidate index data into the time series model to be trained, obtain the predicted complaint volume and / or predicted complaint rate output by the time series model, and determine the training error based on the predicted complaint volume and / or predicted complaint rate and the complaint volume label corresponding to the candidate index data.

[0129] Since the time series model can perform machine learning based on the performance of historical data, training the time series model with index data of different time periods can yield different training effects: if the time period is too short, the information learned by the model is incomplete; if the time period is too long, it is easy for the model to learn more noise. Both of the above situations will reduce the accuracy of the model prediction results.

[0130] Based on this, after obtaining multiple groups of candidate index data, the multiple groups of candidate index data can be used for model training. Specifically, for each group of candidate index data, it can be input into the time series model to be trained, obtain the predicted complaint volume and / or predicted complaint rate output by the time series model, and determine the training error in combination with the complaint volume label corresponding to this group of candidate index data.

[0131] Step 603. Select the candidate index data with the smallest corresponding training error from the multiple groups of candidate index data as the sample index data corresponding to the network point feature index.

[0132] After obtaining the training error, it can be determined that the time period corresponding to the candidate index data with the smallest training error has the best effect, and the candidate index data with the smallest training error can be used as the sample index data for subsequent model training.

[0133] In this embodiment, by respectively inputting multiple groups of candidate index data with different time periods into the time series model to be trained, training, and selecting the candidate index data with the smallest training error as the sample index data corresponding to the network point feature index, it is possible to obtain index data with a suitable time period as the sample index data for training the model, effectively improving the model training effect and the accuracy of the model prediction results.

[0134] To enable those skilled in the art to better understand the above steps, the following provides an exemplary illustration of the embodiments of the present application through an example, but it should be understood that the embodiments of the present application are not limited thereto.

[0135] Such as Figure 5As shown, feature indicators can be extracted for the pick-up and delivery links, transfer links, transportation links, and customer complaint information of the logistics business respectively to obtain pick-up and delivery features, transfer features, transportation features, and customer complaint features. Furthermore, feature screening can be performed through feature engineering, the xgboost algorithm, and the iv value to obtain the network feature indicators corresponding to the logistics network points, and the corresponding indicator data can be obtained as a training set to train the prophet model, and the model effect can be tested through the test set. After passing the test, the model can be used to predict the number of user complaints, and corresponding measures can be taken to control the site management process according to the prediction results to adjust the indicator data and reduce the customer complaint rate.

[0136] It should be understood that although Figures 1-5 the steps in the flowchart of Figures 1-5 are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover,

[0137] In one embodiment, as Figure 6 shown, a device for predicting the business situation of a logistics network point is provided. The device may include:

[0138] An index data acquisition module 601, configured to acquire the network feature indicators corresponding to the logistics network point to be predicted, and the index data corresponding to the network feature indicators; the network feature indicators are obtained by performing feature engineering processing on multiple logistics feature indicators of the logistics network point;

[0139] An index data input module 602, configured to input the index data into a pre-trained complaint prediction model, so as to determine the change trend of the index data through the complaint prediction model, obtain the predicted index data corresponding to the index data according to the change trend, and determine the predicted complaint volume and / or predicted complaint rate of the logistics network point according to the predicted index data;

[0140] A user complaint volume acquisition module 603, configured to acquire the predicted complaint volume and / or predicted complaint rate output by the complaint prediction model as the user complaint volume of the logistics network point.

[0141] In one embodiment, the device may further include:

[0142] A controllable feature index acquisition module, configured to obtain a controllable feature index from the multiple network point feature indexes when the user complaint volume exceeds a preset threshold;

[0143] A data fluctuation range acquisition module, configured to determine the current business characteristic period of the logistics network point and obtain the data fluctuation range of the controllable feature index during the business characteristic period;

[0144] A regulation range determination module, configured to determine an index data regulation range according to the data fluctuation range and the index data corresponding to the controllable feature index;

[0145] A regulation suggestion generation module, configured to generate an index data regulation suggestion based on the index data regulation range, where the index data regulation suggestion is used to improve the user complaint volume.

[0146] In one embodiment, the data fluctuation range acquisition module includes:

[0147] A peak period data fluctuation range acquisition sub-module, configured to determine the data fluctuation range corresponding to the controllable feature index according to the peak period index data corresponding to the controllable feature index in the historical business peak period when the business characteristic period is the business peak period;

[0148] And / or,

[0149] A stable period data fluctuation range acquisition sub-module, configured to, when the business characteristic period is the business stable period, take the current time point as the starting point, obtain the stable period index data corresponding to the controllable feature index within a preset time range, and determine the data fluctuation range corresponding to the controllable feature index according to the stable period index data.

[0150] In one embodiment, the device may further include:

[0151] A sample index data acquisition module, configured to obtain the network point feature index corresponding to the logistics network point from multiple preset logistics feature indexes, and obtain the sample index data and complaint volume label corresponding to the network point feature index;

[0152] A sample index data input module, configured to input the sample index data into a time series model to be trained, so as to determine the change trend of the sample index data through the time series model, obtain the sample prediction index data corresponding to the sample index data according to the change trend, and determine the sample prediction complaint volume and / or sample prediction complaint rate according to the sample index data;

[0153] A complaint prediction model acquisition module, which is used to determine a training error according to the predicted complaint volume and / or predicted complaint rate of the samples and the complaint volume label, and adjust the time series model according to the training error until the training end condition is met, so as to obtain a trained complaint prediction model.

[0154] In one embodiment, the sample index data acquisition module includes:

[0155] A feature gain score acquisition sub-module, which is used to determine the feature gain scores corresponding to each logistics feature index according to the extreme gradient boosting algorithm and the index data corresponding to multiple preset logistics feature indexes, and sum the multiple feature gain scores to obtain a total score;

[0156] A first logistics feature index screening sub-module, which is used to determine the ratio of each feature gain score to the total score, and determine the logistics feature index corresponding to the logistics network point as the network point feature index corresponding to the logistics network point if the ratio exceeds a preset ratio threshold.

[0157] In one embodiment, the device may further include:

[0158] An information value acquisition sub-module, which is used to acquire multiple original logistics feature indexes and determine the information values corresponding to each original logistics feature index;

[0159] A second logistics feature index screening module, which is used to determine multiple original logistics feature indexes with information values exceeding a preset value threshold as the preset multiple logistics feature indexes;

[0160] Wherein, the multiple original logistics feature indexes include at least two of the following:

[0161] Logistics pickup and delivery link feature indexes, logistics transfer link feature indexes, logistics transportation link feature indexes, and logistics process customer complaint feature indexes.

[0162] In one embodiment, the sample index data includes multiple groups of sample index data, and the sample index data acquisition module includes:

[0163] A candidate index data acquisition sub-module, which is used to acquire multiple groups of candidate index data corresponding to the network point feature index and the complaint volume label corresponding to each group of candidate index data; the multiple groups of candidate index data are index data within time periods of different time lengths;

[0164] A training sub-module, which is used to input each group of candidate index data into the time series model to be trained, obtain the predicted complaint volume and / or predicted complaint rate output by the time series model, and determine the training error according to the predicted complaint volume and / or predicted complaint rate and the complaint volume label corresponding to the candidate index data;

[0165] A sample index data screening sub-module, configured to select, from multiple groups of candidate index data, the candidate index data with the minimum corresponding training error as the sample index data corresponding to the network point feature index.

[0166] For the specific limitations of a device for predicting the business conditions of a logistics network point, reference may be made to the limitations of a method for predicting the business conditions of a logistics network point in the foregoing text, which will not be elaborated herein. Each module in the foregoing device for predicting the business conditions of a logistics network point can be implemented in whole or in part by software, hardware, and their combination. The foregoing modules can be embedded in or independent of a processor in a computer device in the form of hardware, or stored in a memory in the computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to the foregoing modules.

[0167] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as Figure 7 shown. The computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store logistics network point feature index data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a method for predicting the business conditions of a logistics network point.

[0168] Those skilled in the art can understand that Figure 7 the structure shown in

[0169] is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have a different component layout.

[0170] Obtain the network point feature index corresponding to the logistics network point to be predicted, and the index data corresponding to the network point feature index; the network point feature index is obtained by performing feature engineering processing on multiple logistics feature indexes of the logistics network point;

[0171] Input the index data into a pre-trained complaint prediction model to determine the change trend of the index data through the complaint prediction model, obtain the predicted index data corresponding to the index data according to the change trend, and determine the predicted complaint volume and / or predicted complaint rate of the logistics network point according to the predicted index data;

[0172] Obtain the predicted complaint volume and / or predicted complaint rate output by the complaint prediction model as the user complaint volume of the logistics network point.

[0173] In one embodiment, when the processor executes the computer program, it also implements the steps in the above-mentioned other embodiments.

[0174] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0175] Obtain the network point characteristic index corresponding to the logistics network point to be predicted, and the index data corresponding to the network point characteristic index; the network point characteristic index is obtained after performing feature engineering processing on multiple logistics characteristic indexes of the logistics network point;

[0176] Input the index data into a pre-trained complaint prediction model to determine the change trend of the index data through the complaint prediction model, obtain the predicted index data corresponding to the index data according to the change trend, and determine the predicted complaint volume and / or predicted complaint rate of the logistics network point according to the predicted index data;

[0177] Obtain the predicted complaint volume and / or predicted complaint rate output by the complaint prediction model as the user complaint volume of the logistics network point.

[0178] In one embodiment, when the computer program is executed by a processor, it also implements the other steps in the above-mentioned embodiments.

[0179] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above various methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0180] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0181] The above-described embodiments merely represent several implementation manners of the present application. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for predicting the business situation of a logistics network point, characterized in that, The method includes: Obtaining the network feature indicators corresponding to the logistics network to be predicted, and the indicator data corresponding to the network feature indicators; the network feature indicators are obtained after performing feature engineering processing on multiple logistics feature indicators of the logistics network; Inputting the indicator data into a pre-trained complaint prediction model to determine the change trend of the indicator data through the complaint prediction model, obtaining the predicted indicator data corresponding to the indicator data according to the change trend, and determining the predicted complaint volume and / or predicted complaint rate of the logistics network according to the predicted indicator data; Obtaining the predicted complaint volume and / or predicted complaint rate output by the complaint prediction model as the user complaint volume of the logistics network; When the user complaint volume exceeds a preset threshold, obtaining controllable feature indicators from multiple network feature indicators; the controllable feature indicators are feature indicators in the logistics business that can affect the indicator data and regulate the indicator data through decisions or measures; Determining the indicator data regulation range according to the data fluctuation range of the controllable feature indicators in the business characteristic period of the logistics network and the indicator data corresponding to the controllable feature indicators; Generating an indicator data regulation suggestion based on the indicator data regulation range, and the indicator data regulation suggestion is used to improve the user complaint volume.

2. The method according to claim 1, characterized in that, After obtaining the controllable feature indicators from multiple network feature indicators, it further includes: Determining the current business characteristic period of the logistics network.

3. The method according to claim 1, characterized in that, The data fluctuation range of the controllable feature indicators in the business characteristic period of the logistics network is obtained through the following steps: When the business characteristic period of the logistics network is the business peak period, determining the data fluctuation range corresponding to the controllable feature indicators according to the peak period indicator data corresponding to the controllable feature indicators in the historical business peak period; And / or, When the business characteristic period of the logistics network is the business stable period, taking the current time point as the starting point, obtaining the stable period indicator data corresponding to the controllable feature indicators within a preset time range, and determining the data fluctuation range corresponding to the controllable feature indicators according to the stable period indicator data.

4. The method according to claim 1, characterized in that, It further includes: Obtaining the network feature indicators corresponding to the logistics network from multiple preset logistics feature indicators, and obtaining the sample indicator data and complaint volume label corresponding to the network feature indicators; Inputting the sample indicator data into a time series model to be trained to determine the change trend of the sample indicator data through the time series model, obtaining the sample predicted indicator data corresponding to the sample indicator data according to the change trend, and determining the sample predicted complaint volume and / or sample predicted complaint rate according to the sample indicator data; Determining the training error according to the sample predicted complaint volume and / or sample predicted complaint rate and the complaint volume label, and adjusting the time series model according to the training error until the training end condition is met to obtain a trained complaint prediction model.

5. The method according to claim 4, characterized in that, The obtaining the network feature indicators corresponding to the logistics network from multiple preset logistics feature indicators includes: According to the extreme gradient boosting algorithm and the index data corresponding to multiple preset logistics feature indexes, determine the feature gain scores corresponding to each logistics feature index, and sum the multiple feature gain scores to obtain the total score; Determine the ratio of each feature gain score to the total score, and determine the logistics feature index whose ratio exceeds the preset ratio threshold as the network point feature index corresponding to the logistics network point.

6. The method according to claim 5, characterized in that, Before the step of determining the feature gain scores corresponding to each logistics feature index according to the extreme gradient boosting algorithm and the index data corresponding to multiple preset logistics feature indexes, the method further includes: Obtain multiple original logistics feature indexes and determine the information value corresponding to each original logistics feature index; Determine the multiple original logistics feature indexes whose information value exceeds the preset value threshold as the multiple preset logistics feature indexes; Wherein, the multiple original logistics feature indexes include at least two of the following: Logistics pickup and delivery link feature index, logistics transfer link feature index, logistics transportation link feature index, logistics process customer complaint feature index.

7. The method according to claim 4, characterized in that, The sample index data includes multiple groups of sample index data. Obtaining the sample index data corresponding to the network point feature index includes: Obtain multiple groups of candidate index data corresponding to the network point feature index and the complaint volume label corresponding to each group of candidate index data; the multiple groups of candidate index data are index data within time periods of different time lengths; For each group of candidate index data, input the candidate index data into the time series model to be trained, obtain the predicted complaint volume and / or predicted complaint rate output by the time series model, and determine and / or predict the complaint rate according to the predicted complaint volume, and determine the training error according to the predicted complaint volume and the complaint volume label corresponding to the candidate index data; Select the candidate index data with the smallest corresponding training error from the multiple groups of candidate index data as the sample index data corresponding to the network point feature index.

8. A device for predicting the business situation of a logistics network point, characterized in that, The device includes: An index data acquisition module, configured to acquire the network point feature index corresponding to the logistics network point to be predicted and the index data corresponding to the network point feature index; the network point feature index is obtained by performing feature engineering processing on multiple logistics feature indexes of the logistics network point; An index data input module, configured to input the index data into a pre-trained complaint prediction model to determine the change trend of the index data through the complaint prediction model, obtain the predicted index data corresponding to the index data according to the change trend, and determine the predicted complaint volume and / or predicted complaint rate of the logistics network point according to the predicted index data; A user complaint volume acquisition module, configured to acquire the predicted complaint volume and / or predicted complaint rate output by the complaint prediction model as the user complaint volume of the logistics network point; A controllable feature index acquisition module, configured to, when the user complaint volume exceeds the preset threshold, acquire a controllable feature index from the multiple network point feature indexes; the controllable feature index is a feature index in the logistics business that can affect the index data through decisions or measures and regulate the index data; A regulation range determination module, configured to determine an index data regulation range according to the data fluctuation range of the controllable feature index during the business feature period of the logistics network point and the index data corresponding to the controllable feature index; A regulation suggestion generation module, configured to generate an index data regulation suggestion based on the index data regulation range, where the index data regulation suggestion is used to improve the user complaint volume.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images