Resource information processing method and device

Abstract

The invention discloses a resource information processing method and device, and relates to the technical field of computers. A specific embodiment of the method comprises the steps of obtaining attribute information of sample resources and demand data in different historical time periods, extracting attribute features of the attribute information, and determining demand change difference features of the sample resources in adjacent historical time periods; inputting the attribute features into an original logistic regression model corresponding to each demand mode, and constructing an adaptation probability expression of the sample resources and each demand mode; constructing a probability expression of all the demand change difference characteristics of the sample resources, and constructing a joint probability expression of all the demand change difference characteristics of the sample resources under the condition that the sample resources are matched with each demand mode; and based on the adaptive probability expression, the joint probability expression and the probability expression, calculating a value of a regression coefficient in each original logistic regression model, and obtaining a logistic regression model after training of each demand mode. According to the embodiment, the accuracy of the model can be improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a resource information processing method and device. Background Art

[0002] Different resources are adapted to different demand patterns, so appropriate forecasting strategies and inventory strategies can be adapted according to the demand patterns of resources to better provide replenishment suggestions, assist operations personnel in making replenishment decisions, improve fulfillment rates and reduce inventory costs.

[0003] Due to the large number of resource types, currently, manual experience, unified algorithms, and model building methods are mainly used to determine resource demand patterns. However, manual experience has a coarse identification dimension and can only provide the demand pattern of resources in the category dimension, which is difficult to refine to a single resource. In addition, manual experience considers limited factors, has low accuracy, and takes a long time. Neither unified algorithms nor model building methods can provide the factors that affect changes in resource demand, resulting in a lack of interpretability of the results, and it is impossible to adapt appropriate forecasting strategies and inventory strategies to resources with different data characteristics. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a resource information processing method and device, which can at least solve the problem in the prior art that the granularity of manual experience recognition is coarse, the unified algorithm and model construction method lack interpretability, and different replenishment strategies cannot be adapted to different resources.

[0005] To achieve the above object, according to one aspect of an embodiment of the present invention, a resource information processing method is provided, comprising:

[0006] Acquire attribute information of sample resources and demand data in different historical time periods, extract attribute features of the attribute information, and determine, based on the demand data, demand change difference features of the sample resources in adjacent historical time periods;

[0007] Inputting the attribute features into the original logistic regression model corresponding to each demand pattern to construct an adaptation probability expression between the sample resource and each demand pattern;

[0008] Constructing a probability expression of all demand change difference characteristics of the sample resources, and constructing a joint probability expression of all demand change difference characteristics of the sample resources under the condition of being adapted to each of the demand patterns;

[0009] Based on the adaptation probability expression, the joint probability expression and the probability expression, the value of the regression coefficient in each original logistic regression model is calculated, so as to obtain the trained logistic regression model of each demand pattern.

[0010] Optionally, the attribute information includes category attribute information and value attribute information, and the extracting attribute features of the attribute information includes:

[0011] Determine the valid code corresponding to each category attribute information, and obtain the category attribute feature of each category attribute information;

[0012] The value of each numerical attribute information is quantified to obtain the numerical attribute characteristics of each numerical attribute information.

[0013] Optionally, obtaining demand data of sample resources in different historical time periods includes: obtaining demand data of sample resources in a historical preset time period, dividing the historical preset time period into an early historical time period, a mid-term historical time period, and a late historical time period;

[0014] Determining the difference characteristics of demand changes of sample resources in adjacent historical time periods according to the demand data includes:

[0015] Use the hypothesis verification method to determine whether the mean of the sample resource demand data in the early historical period is equal to the mean of the demand data in the mid-term historical period, and obtain the difference characteristics of the early and mid-term demand changes; and

[0016] Determine whether the mean of the demand data of the sample resources in the mid-term historical period is equal to the mean of the demand data in the late historical period, and obtain the difference characteristics of the mid-term and late-term demand changes.

[0017] Optionally, constructing a joint probability expression of all demand change difference characteristics of sample resources under the condition of being adapted to each demand pattern includes:

[0018] A probability expression of each demand change difference feature of the sample resource under the condition of being adapted to each demand pattern is constructed, and the probability expressions of all demand change difference features are multiplied to obtain the joint probability expression.

[0019] Optionally, after obtaining the trained logistic regression model for each demand pattern, the method further includes:

[0020] Input the attribute characteristics of the sample resources into the logistic regression model corresponding to each demand pattern to obtain the matching probability between the sample resources and each demand pattern, and use the demand pattern with the largest matching probability as the target demand pattern that matches the sample resources;

[0021] Receive an adaptive replenishment strategy for each sample resource tag;

[0022] The current inventory data of the sample resources is obtained, and the attribute characteristics of the sample resources, the demand data in different historical time periods, the inventory data, the adapted target demand pattern, and the adapted replenishment strategy are input into the original replenishment model for training to obtain the trained replenishment model.

[0023] Optionally, after obtaining the trained replenishment model, the method further includes:

[0024] Obtain attribute information of new resources, demand data in different historical time periods, and current inventory data, and extract attribute features of the attribute information;

[0025] Input the attribute information of the new resource into the logistic regression model corresponding to each demand pattern to obtain the adaptation probability of the new resource to each demand pattern, and take the demand pattern with the largest adaptation probability as the target demand pattern adapted to the new resource;

[0026] The attribute characteristics of the new resource, the demand data in different historical time periods, the inventory data, and the adapted target demand pattern are input into the replenishment model to obtain the target replenishment strategy.

[0027] Optionally, the replenishment strategy includes a forecasting strategy and an inventory strategy;

[0028] The attribute characteristics of the new resource, demand data in different historical time periods, inventory data, and the adapted target demand pattern are input into the replenishment model to obtain the target replenishment strategy, including:

[0029] Input the attribute characteristics of the new resource, the demand data in different historical time periods, the inventory data, and the adapted target demand pattern into the replenishment model to obtain a forecasting strategy set and an inventory strategy set;

[0030] Combine the forecast strategies in the forecast strategy set and the inventory strategies in the inventory strategy set to obtain multiple combination schemes, apply each combination scheme to the new resources, obtain application result data, and use the combination scheme with the largest application result data as the target replenishment strategy.

[0031] To achieve the above object, according to another aspect of an embodiment of the present invention, a resource information processing device is provided, comprising:

[0032] An acquisition module, used to acquire attribute information of sample resources and demand data in different historical time periods, extract attribute features of the attribute information, and determine, based on the demand data, demand change difference features of sample resources in adjacent historical time periods;

[0033] A first construction module is used to input the attribute features into the original logistic regression model corresponding to each demand pattern to construct an adaptation probability expression between the sample resource and each demand pattern;

[0034] The second construction module is used to construct a probability expression of all demand change difference characteristics of the sample resources, and to construct a joint probability expression of all demand change difference characteristics of the sample resources under the condition of being adapted to each demand pattern;

[0035] The training module is used to calculate the value of the regression coefficient in each original logistic regression model based on the adaptation probability expression, the joint probability expression and the probability expression, so as to obtain the trained logistic regression model of each demand pattern.

[0036] Optionally, the attribute information includes category attribute information and value attribute information, and the acquisition module is used to:

[0037] Determine the valid code corresponding to each category attribute information, and obtain the category attribute feature of each category attribute information;

[0038] The value of each numerical attribute information is quantified to obtain the numerical attribute characteristics of each numerical attribute information.

[0039] Optionally, the acquisition module is used to:

[0040] Obtaining demand data of sample resources in a historical preset time period, dividing the historical preset time period into an early historical time period, a mid-term historical time period, and a late historical time period; and

[0041] Use the hypothesis verification method to determine whether the mean of the sample resource demand data in the early historical period is equal to the mean of the demand data in the mid-term historical period, and obtain the difference characteristics of the early and mid-term demand changes; and

[0042] Determine whether the mean of the demand data of the sample resources in the mid-term historical period is equal to the mean of the demand data in the late historical period, and obtain the difference characteristics of the mid-term and late-term demand changes.

[0043] Optionally, the second building module is used to:

[0044] A probability expression of each demand change difference feature of the sample resource under the condition of being adapted to each demand pattern is constructed, and the probability expressions of all demand change difference features are multiplied to obtain the joint probability expression.

[0045] Optionally, the device further includes a replenishment model training module, which is used to:

[0046] Input the attribute characteristics of the sample resources into the logistic regression model corresponding to each demand pattern to obtain the matching probability between the sample resources and each demand pattern, and use the demand pattern with the largest matching probability as the target demand pattern that matches the sample resources;

[0047] Receive an adaptive replenishment strategy for each sample resource tag;

[0048] The current inventory data of the sample resources is obtained, and the attribute characteristics of the sample resources, the demand data in different historical time periods, the inventory data, the adapted target demand pattern, and the adapted replenishment strategy are input into the original replenishment model for training to obtain the trained replenishment model.

[0049] Optionally, the device further comprises a replenishment model application module, which is used to:

[0050] Obtain attribute information of new resources, demand data in different historical time periods, and current inventory data, and extract attribute features of the attribute information;

[0051] Input the attribute information of the new resource into the logistic regression model corresponding to each demand pattern to obtain the adaptation probability of the new resource to each demand pattern, and take the demand pattern with the largest adaptation probability as the target demand pattern adapted to the new resource;

[0052] The attribute characteristics of the new resource, the demand data in different historical time periods, the inventory data, and the adapted target demand pattern are input into the replenishment model to obtain the target replenishment strategy.

[0053] Optionally, the replenishment strategy includes a forecasting strategy and an inventory strategy;

[0054] The replenishment model application module is used to:

[0055] Input the attribute characteristics of the new resource, the demand data in different historical time periods, the inventory data, and the adapted target demand pattern into the replenishment model to obtain a forecasting strategy set and an inventory strategy set;

[0056] Combine the forecast strategies in the forecast strategy set and the inventory strategies in the inventory strategy set to obtain multiple combination schemes, apply each combination scheme to the new resources, obtain application result data, and use the combination scheme with the largest application result data as the target replenishment strategy.

[0057] To achieve the above objective, according to another aspect of an embodiment of the present invention, a resource information processing electronic device is provided.

[0058] The electronic device of an embodiment of the present invention includes: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement any of the resource information processing methods described above.

[0059] To achieve the above objective, according to another aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, any of the resource information processing methods described above is implemented.

[0060] According to the scheme provided by the present invention, an embodiment of the above invention has the following advantages or beneficial effects: Compared with the existing manual labeling training model method, this scheme uses the resource's own attribute information for effective modeling to construct a priori demand pattern recognition result, and then uses the change difference characteristics of the resource historical demand data in a specific time period as observations to correct the priori demand pattern recognition result to obtain the final resource demand pattern recognition result. This process does not require human participation, thereby reducing the degree of human subjective influence; it can be performed under different dimensional data, such as item dimension, category dimension, etc., and the demand pattern recognition results of the required dimensions can be given according to the actual needs of the business. It can give both coarse-grained category dimension demand pattern recognition results and refined demand pattern recognition results to a single resource dimension, which solves the problem that the existing identification dimension based on manual experience is difficult to refine to a single resource, and can better assist operators in making replenishment decisions, improve fulfillment rates and reduce inventory costs.

[0061] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0063] Figure 1 is a schematic diagram of the main flow of a resource information processing method according to an embodiment of the present invention;

[0064] Figure 2 is a flow chart of an optional resource information processing method according to an embodiment of the present invention;

[0065] Figure 3 is a flowchart of another optional resource information processing method according to an embodiment of the present invention;

[0066] Figure 4 is a flowchart of another optional resource information processing method according to an embodiment of the present invention;

[0067] Figure 5 It is a schematic diagram of the overall process of the resource information processing method according to an embodiment of the present invention;

[0068] Figure 6 is a schematic diagram of main modules of a resource information processing device according to an embodiment of the present invention;

[0069] Figure 7 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0070] Figure 8 It is a schematic diagram of the structure of a computer system of a mobile device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0072] It should be pointed out that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of the user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Take necessary measures for user personal information, such as de-identification or anonymization, to protect user privacy, prevent illegal access to user personal information data, and maintain user personal information security, network security and national security. Once these user personal information data are no longer needed, the risks should be minimized by restricting or even prohibiting data collection and / or deleting data.

[0073] This solution can be applied to a variety of scenarios, such as replenishment of items in retail scenarios, resource scheduling in computer scenarios, etc. Taking items as an example, some items (such as moon cakes, glutinous rice balls, etc.) are strongly related to holidays. These items will only have more demand during specific holidays, and less demand at other times; some items (such as masks, disinfectants, etc.) are strongly related to epidemic prevention; some items are strongly related to promotional activities, such as more demand during promotional activities such as 618, Double 11 and Double 12, and less demand at other times.

[0074] There are many types of resources. Currently, the commonly used replenishment methods are mainly the following: 1) Make replenishment decisions based on manual experience, 2) Make replenishment suggestions through the same set of forecasting strategies and inventory strategies, such as moving weighted average forecasting strategy and RSS inventory strategy, 3) Build a model from raw data such as sales and inventory to replenishment suggestions in an end-to-end manner.

[0075] The main disadvantages of these methods are as follows: 1) Manual experience has a coarse dimension for identifying resource demand patterns, and can only provide demand patterns for resources in the category dimension, such as holiday-related and promotion-related, which is difficult to refine to the identification of a single resource demand pattern. In addition, manual experience considers limited factors, has low accuracy, and is time-consuming; 2) The unified algorithm and model construction method only provide a result, and cannot provide the factors that affect changes in resource demand, that is, it is not interpretable, and cannot adapt appropriate forecasting strategies and inventory strategies for resources with different data characteristics. Among them, interpretability refers to which factors determine the output results of the model, and providing influencing factors means trying to explain the composition of the output results as much as possible through some specific information, rather than a black box result (which cannot be intuitively explained).

[0076] See also Figure 1 , which shows a main flow chart of a resource information processing method provided by an embodiment of the present invention, comprising the following steps:

[0077] S101: Acquire attribute information of sample resources and demand data in different historical time periods, extract attribute features of the attribute information, and determine demand change difference features of sample resources in adjacent historical time periods based on the demand data;

[0078] S102: Inputting the attribute features into the original logistic regression model corresponding to each demand pattern to construct an adaptation probability expression between the sample resources and each demand pattern;

[0079] S103: constructing probability expressions of all demand change difference characteristics of sample resources, and constructing joint probability expressions of all demand change difference characteristics of sample resources under the condition of being adapted to each demand pattern;

[0080] S104: Based on the adaptation probability expression, the joint probability expression and the probability expression, the value of the regression coefficient in each original logistic regression model is calculated, so as to obtain the trained logistic regression model of each demand pattern.

[0081] In the above implementation, for step S101, this solution obtains the attribute information of the sample resource, and classifies the resource attribute information according to the attribute characteristics, so as to calculate the corresponding attribute features in different ways. For example, enumerable item category attribute features such as name, brand, category, and color can be One-Hot encoded (the same value is 1, and different values ​​are 0), and directly quantifiable item numerical attribute features such as specifications, volume, and weight can directly use the actual value for subsequent calculations.

[0082] Obtain the demand data of each sample resource in the historical preset time period, divide the historical preset time period into early, middle and late periods, or divide it into 4, 5 or other small time periods, so as to obtain the demand data of resources in different historical time periods. The division method includes but is not limited to dividing the time period evenly, or dividing it according to a period before the start of the activity, during the activity, and a period after the end of the activity.

[0083] This scheme is mainly described using the early, middle and late stages as examples. The difference in the change of historical demand data of each sample resource in the early, middle and late stages of a specific time period is taken as the observation value, and the hypothesis verification method is used to calculate the difference characteristics of the demand change in the early and middle stages and the difference characteristics of the demand change in the middle and late stages.

[0084] For step S102, this solution sets different demand patterns for different resources, and for different demand patterns, the original logistic regression logistic models constructed are also different, that is, the regression coefficients in the model are different. Taking items as an example, this solution pre-sets 9 demand patterns, including demand rising all the time in the early, middle and late stages of a specific time period, rising first and then unchanged, rising first and then falling; unchanged first and then rising, unchanged all the time, unchanged first and then falling, falling first and then rising, falling first and then unchanged, and falling all the time, so as to construct 9 original logistic regression logistic models.

[0085] In this scheme, i is used to represent the i-th sample resource and k is used to represent the k-th demand pattern. For items, k = 1, 2, ..., 9. The regression coefficient of the original logistic regression model of the k-th demand pattern is set to β k0 and β kj ,j=1,...,n, where n is the number of attribute features. This solution uses logistic regression to model the prior recognition of demand patterns based on resource attribute features. The attribute features of the i-th sample resource are input into the original logistic regression model of the k-th demand pattern to construct the matching probability P(z i =k) ​​expression:

[0086]

[0087] In the formula, m represents the number of sample resources, z i It represents the prior recognition result of the resource demand pattern of the ith sample. It is an unknown quantity, but it is assumed to be known. Taking items as an example, intuitively, items of the same brand, the same category, and similar names may have similar demand patterns, and the demand patterns should be consistent with a high probability. i =k) ​​indicates the probability that the prior recognition result of the i-th sample resource demand pattern is the k-th demand pattern, x ijRepresents the attribute characteristics of the i-th sample resource itself. Formula (1) means that m resource attributes are used as variables to perform logistic regression on the prior recognition results of the demand pattern. Therefore, for the same item, 9 adaptation probability expressions for the demand pattern will be obtained.

[0088] For step S103, when calculating the historical demand data of resources and obtaining multiple demand change difference characteristics, this solution assumes that these characteristics are independent of each other, for example, assuming that the early and mid-term demand change difference characteristics and the mid- and late-term demand change difference characteristics are independent of each other. Under the condition that the prior recognition result mapped to the i-th sample resource demand is the k-th demand pattern category, the joint probability of multiple demand difference characteristics can be expressed as follows using the data formula:

[0089]

[0090] In the formula, Q represents the number of demand change difference characteristics, q represents the qth demand change difference characteristic, P(f iq |z i = k) indicates that the prior recognition result of the resource demand pattern of the i-th sample is the feature f under the condition of the k-th demand pattern category. iq The probability, P(f i |z i = k) indicates that the prior recognition result of the resource demand pattern of the i-th sample is multiple features f under the condition of the k-th demand pattern category. i =(f i1 ,f i2 ,...,f iq ,...f iQ ). Here f i =(f i1 ,f i2 ,...,f iq ,...f iQ ) gives an extended form for the case of multiple features, but it can actually be just one. The meaning of joint probability refers to the probability of the statistical distribution of multiple variables, and this value is calculated to be used as the prior probability.

[0091] For this scheme, if we only consider the difference characteristics of demand changes in the early and middle stages and the difference characteristics of demand changes in the middle and late stages, then Q=2 here, q takes values ​​of 1 and 2, and the probability expression of the difference characteristics of demand changes in the early and middle stages and the probability expression of the difference characteristics of demand changes in the middle and late stages are multiplied to obtain the joint probability expression of these two characteristics.

[0092] P(f iq |z i=k) ​​can be modeled through common statistical distributions (including normal distribution, etc.). Parameter θ will be introduced in the modeling process. For example, when using normal distribution for modeling, the parameters mean and standard deviation need to be introduced. Parameter θ can be solved by maximizing the conditional probability P(θ|f,x). After obtaining parameter θ, the delayed identification result of resource demand pattern can be inferred by maximizing P(z|θ,f).

[0093] Through the above steps, we can obtain: the prior recognition result of the i-th resource demand pattern is the adaptation probability expression P(z i = k), the prior recognition result of the i-th resource demand pattern is the joint probability expression P(f i |z i = k), and multiplying the two, the a priori recognition result of the i-th resource demand pattern is the expression P(f i |z i = k)P(z i =k).

[0094] The prior recognition result of the i-th resource demand pattern is the accumulation of expressions of all demand change difference characteristics under each demand pattern condition. The resource demand pattern posterior result is inferred based on the demand pattern recognition model (such as the hierarchical Bayesian model), which can be expressed as follows using a mathematical formula:

[0095]

[0096] Among them, f i represents the demand change difference characteristics of the historical demand data of the i-th sample resource in different historical time periods, P(f i ) represents the feature f of the i-th sample resource i The probability expression, P(f i |z i = k) indicates that the prior recognition result of the resource demand pattern of the i-th sample is the feature f under the k-th demand pattern condition i Therefore, formula (3) is a full probability formula, which represents the characteristic f of the i-th sample resource i The probability of can be obtained by the probability of the prior recognition result of the ith resource demand pattern under different demand pattern categories and the sum of the products of the probabilities of the prior recognition result of the ith item demand pattern under different demand pattern categories.

[0097] The known quantities in the whole process are the attribute characteristics and the demand change difference characteristics, and the remaining parameters are all unknown quantities. The whole process can be regarded as, after constructing formula (1) and formula (2), formula (1) and formula (2) are input into formula (3), and the unknown quantities related to probability can be eliminated through formula (3). In addition, the number of sample resources in this scheme is multiple. Through formula (1) and formula (2) of these sample resources, the regression coefficient of each demand pattern logistic model can be calculated. For items, 9 sets of regression coefficients need to be calculated to obtain 9 demand pattern logistic models.

[0098] The method provided in the above embodiment abandons the existing idea of ​​manually labeling sample resources to adapt to demand patterns. This solution uses the resource's own attribute information for effective modeling to construct a priori demand pattern recognition result, and then uses the change difference characteristics of the resource's historical demand data in a specific time period as observation to correct the prior demand pattern recognition result to obtain the final resource demand pattern recognition result. This process does not require human participation, thereby reducing the degree of human subjective influence and improving the accuracy of the model in identifying demand patterns.

[0099] See also Figure 2 , shows a schematic flow chart of an optional resource information processing method according to an embodiment of the present invention, comprising the following steps:

[0100] S201: Acquire attribute information of sample resources and demand data in a historical preset time period, and divide the historical preset time period into an early historical time period, a mid-term historical time period, and a late historical time period;

[0101] S202: determining a valid code corresponding to each category attribute information to obtain a category attribute feature of each category attribute information; and quantizing a value of each numerical attribute information to obtain a numerical attribute feature of each numerical attribute information;

[0102] S203: using a hypothesis verification method, determining whether the mean of the demand data of the sample resources in the early historical period is equal to the mean of the demand data of the mid-term historical period, and obtaining the difference characteristics of the demand changes between the early and mid-term periods; and determining whether the mean of the demand data of the sample resources in the mid-term historical period is equal to the mean of the demand data of the late historical period, and obtaining the difference characteristics of the demand changes between the mid-term and late periods;

[0103] S204: inputting the category attribute features and the numerical attribute features into the original logistic regression model corresponding to each demand pattern to construct an adaptation probability expression between the sample resources and each demand pattern;

[0104] S205: constructing probability expressions of all demand change difference characteristics of sample resources, and constructing joint probability expressions of all demand change difference characteristics of sample resources under the condition of being adapted to each demand pattern;

[0105] S206: Based on the adaptation probability expression, the joint probability expression and the probability expression, the value of the regression coefficient in each original logistic regression model is calculated, so as to obtain the trained logistic regression model of each demand pattern.

[0106] In the above embodiment, for steps S201, S204 to S206, see Figure 1 The description shown is not given here.

[0107] For step S202, this solution obtains the attribute information of the sample resource, and classifies the resource attribute information according to the attribute characteristics, so as to calculate the corresponding attribute features in different ways. For example, the enumerable item category attribute features such as name, brand, category, and color can be One-Hot encoded (the same value is 1, and the different values ​​are 0), and the directly quantifiable item numerical attribute features such as specifications, volume, and weight can directly use the actual value for subsequent calculations.

[0108] One-Hot encoding, also known as one-bit effective encoding, is the representation of categorical variables as binary vectors. It mainly uses N-bit state registers to encode N states. Each state has its own independent register bit, and only one bit is valid at any time. Take the name as an example, which belongs to the category attribute. Assuming there are three names in total, namely X, Y, and Z, the encoding is X=>100, Y=>010, and Z=>001. Take the brand as an example, which is also a category attribute. Assuming there are four brands in total, namely Q, W, E, and R, the encodings are Q=>1000, W=>0100, E=>0010, and R=>0001 respectively. Take the category as an example, assuming there are two types in total, namely T and U, the encodings are T=>10 and U=>01 respectively. Therefore, when a resource attribute information is Y, W, or T, the complete feature digitization result of the resource category attribute is [0,1,0], [0,1,0,0], [1,0].

[0109] Quantization refers to the process of approximating the continuous value of a signal to a finite number of discrete values. It is now widely used in life and refers to something that can be clearly and specifically measured. This solution quantifies the numerical values ​​corresponding to each attribute information. For example, items weighing more than 2kg are heavy objects, and items weighing less than 2kg are light objects. The exact indicator of 2kg eliminates the uncertainty of the two fuzzy concepts of light and heavy. Quantization can reduce certainty and is an effective means of optimizing problems. This solution preferably uses actual values. For example, if the weight of an item is 1kg, 1kg is used as the weight attribute feature of the item.

[0110] For step S203, the demand data of each sample resource in the historical preset time period is obtained, and the historical preset time period is divided into early, middle and late periods, or divided into 4, 5 or other small time periods, so as to obtain the demand data of the resource in different historical time periods. The division method includes but is not limited to dividing the time period evenly, or dividing according to a period before the start of the activity, during the activity, and a period after the end of the activity.

[0111] This scheme is mainly described using the early, middle and late stages as examples. The difference in the change of historical demand data of sample resources in the early, middle and late stages of a specific time period is taken as the observation value, and the hypothesis verification method is used to calculate the difference characteristics of the demand change in the early and middle stages and the difference characteristics of the demand change in the middle and late stages.

[0112] Taking the previous demand data and the mid-term demand data as an example, assuming that the demand data for the previous 10 days is 1000, and the demand data for the mid-term 10 days is 10000, the mean of the previous demand data is calculated to be 100, and the mean of the mid-term demand data is 1000. Using the hypothesis verification method, in the hypothesis test of whether the means of two sets of data are equal, the t statistic or p value can usually be used to determine whether the means of the two sets of data are equal, that is, whether the two sets of data are significantly different. Assuming that the mean of the previous demand data is equal to the mean of the mid-term demand data, that is, 100=1000, but the actual 100<1000, the difference feature of the previous and mid-term demand changes is set to N. However, if the actual value meets the assumed value, the difference feature of the previous and mid-term demand changes is set to Y.

[0113] In some cases, it is necessary to set the demand change. For example, if the mean of the previous demand data is less than the mean of the mid-term demand data, the difference between the previous and mid-term demand changes is set to +, indicating that the demand has increased. However, if the mean of the previous demand data is equal to the mean of the mid-term demand data, the difference between the previous and mid-term demand changes is set to =, indicating that the demand is stable. If the mean of the previous demand data is greater than the mean of the mid-term demand data, the difference between the previous and mid-term demand changes is set to -, indicating that the demand has decreased.

[0114] The method provided in the above embodiment adopts different extraction methods for attribute features with different characteristics, so that the extracted attribute features reduce uncertainty and are more in line with the actual situation. The hypothesis verification method is used to determine whether there are significant differences in the demand data. The entire process is purely computer processing without human participation, which improves the accuracy of the model.

[0115] See also Figure 3 , shows a schematic flow chart of another optional resource information processing method according to an embodiment of the present invention, comprising the following steps:

[0116] S301: Inputting the attribute characteristics of the sample resource into the logistic regression model corresponding to each demand pattern, obtaining the matching probability between the sample resource and each demand pattern, and taking the demand pattern with the largest matching probability as the target demand pattern that matches the sample resource;

[0117] S302: receiving an adaptive replenishment strategy marked for each sample resource;

[0118] S303: Obtain the current inventory data of the sample resources, input the attribute characteristics of the sample resources, the demand data in different historical time periods, the inventory data, the adapted target demand pattern, and the adapted replenishment strategy into the original replenishment model for training, and obtain the trained replenishment model.

[0119] In the above implementation, for step S301, after training the logistic model for each demand pattern, the solution can use these models to calculate the adaptation probability of each sample resource to each demand pattern. Taking an item as an example, after calculating its adaptation probability to the nine demand patterns, the demand pattern with the largest adaptation probability can be used as the target demand pattern for the item.

[0120] For step S302, this solution manually labels each sample resource in advance for which replenishment strategy it is suitable for, for example, sample resource 1-replenishment strategy B, sample resource 2-replenishment strategy C. Considering that some sample resources may be suitable for multiple replenishment strategies, it may be necessary to label them with multiple replenishment strategies, such as sample resource 3-replenishment strategy A, replenishment strategy C. Combined with the above description, assuming there are m sample resources, at least m labeling is required.

[0121] For step S303, the current inventory data of each sample resource is obtained, and the inventory data is used when selecting the inventory strategy and outputting the replenishment suggestion. This solution uses the demand data, inventory data, attribute characteristics and demand pattern recognition results of each sample resource as independent variables, and the marked replenishment strategy as the result, to train a model of resource demand pattern adaptation prediction strategy and inventory strategy, such as a tree model, which is called a replenishment model in this solution.

[0122] The method provided in the above embodiment marks the replenishment strategy adapted to each sample. When the accuracy of determining the demand pattern adapted to the sample is high, the accuracy of the trained replenishment model is also high.

[0123] See also Figure 4 , shows a schematic flow chart of another optional resource information processing method according to an embodiment of the present invention, comprising the following steps:

[0124] S401: Acquire attribute information of new resources, demand data in different historical time periods, and current inventory data, and extract attribute features of the attribute information;

[0125] S402: Inputting the attribute information of the new resource into the logistic regression model corresponding to each demand pattern, obtaining the matching probability between the new resource and each demand pattern, and taking the demand pattern with the highest matching probability as the target demand pattern that matches the new resource;

[0126] S403: Input the attribute characteristics of the new resource, the demand data in different historical time periods, the inventory data, and the adapted target demand pattern into the replenishment model to obtain the target replenishment strategy.

[0127] In the above implementation, for steps S401 and S402, after constructing the logistic regression model and replenishment model for each demand pattern, this solution needs to apply them to the actual scenario. For new resources, three pieces of information also need to be obtained: attribute information, demand data in different historical time periods, and current inventory data. For attribute feature extraction, see Figure 1 and Figure 2 Input its attribute characteristics into the logistic regression model of each demand pattern to calculate the adaptation probability of the new resource to each demand pattern, and take the demand pattern with the largest adaptation probability as the target demand pattern for the new resource to adapt.

[0128] For step S403, the demand data, inventory data, attribute characteristics and demand pattern recognition results of the new resource are input into the trained replenishment model as independent variables, so that the replenishment model outputs replenishment suggestions, that is, replenishment strategies.

[0129] The replenishment strategy in this solution includes forecasting strategy and inventory strategy. Because it is uncertain which replenishment strategy is most suitable for new resources, the actual output is a forecasting strategy set and an inventory strategy set. For example, the forecasting strategies include A1, A2, ..., Aa, and the inventory strategies include B1, B2, ..., Bb. The forecasting strategy set {A2, A4, A7} output by the replenishment model for new resources, and the inventory strategy set is {B1, B3, B9}. Combining these two sets in pairs, we get multiple combination schemes: A2B1, A2B3, A2B9, A4B1, A4B3, A4B9, A7B1, A7B3, A7B9.

[0130] Apply various combinations of forecasting strategies and inventory strategies to new resources, analyze the replenishment effect of each new resource, that is, apply the result data, such as order fulfillment rate and inventory turnover, and select a forecasting strategy and inventory strategy combination with the best replenishment effect as the best solution. In actual operation, the best solution can be the best forecasting strategy and inventory strategy combination, or a weighted solution of several combinations. Therefore, the number of the best replenishment effect here can be one or more.

[0131] Example 1: The demand for resources has changed significantly before, during, and after a specific period of time in history. The demand pattern recognition results show that the demand for resources has a certain trend of change. At this time, the promotion forecast strategy and RS inventory strategy are adapted. Among them, the RS inventory strategy is: check the inventory status at any time, and when the inventory is found to be reduced to the order point level R, start replenishment, and keep the maximum inventory unchanged after replenishment, which is the constant S.

[0132] For example, the demand for items changes significantly before, during, and after a promotion. The result of demand pattern recognition is that demand increases from before the promotion to during the promotion, and then decreases from during the promotion to after the promotion. Through the replenishment model, it is adapted to the promotion increment model and RS inventory strategy, which constitute the replenishment strategy.

[0133] Example 2: If the demand for resources does not change significantly before, during, or after a specific period of time, the demand pattern recognition result shows that the demand for items has a stable and unchanged trend. At this time, the conventional forecasting strategy and the Ss inventory strategy are adapted. Among them, the Ss inventory strategy is: check the inventory status at any time, and when the inventory is found to be reduced to the safety inventory level s, start replenishment, and keep the maximum inventory unchanged after replenishment, which is the constant S.

[0134] For example, the demand for items did not change significantly before, during, and after a promotion. The result of demand pattern recognition was that demand was stable from before the promotion to during the promotion, and then stable from during the promotion to after the promotion. Through the replenishment model, it was adapted to the exponential smoothing model and the Ss inventory strategy, which constituted the replenishment strategy.

[0135] The method provided in the above embodiment uses a replenishment model to adapt the forecasting strategy and inventory strategy based on the historical demand data, inventory data, attribute characteristics and demand pattern recognition results of the resource, and finally gives a replenishment suggestion.

[0136] See also Figure 5 , shows a schematic diagram of the overall process of the resource information processing method of an embodiment of the present invention, including a logistic regression model training stage and a replenishment model training stage:

[0137] The replenishment problem of resources is actually to make replenishment decisions based on the existing resource-related information (including resource attribute information, demand data, inventory data, etc.) to meet the demand for resources in the future as much as possible under controllable inventory costs. In retail scenarios, it is usually possible to obtain the attribute characteristics of the item itself and the sales (demand) data and inventory data of the item over a period of time.

[0138] 1. During the training phase of the logistic regression model, multiple sample resources are required. The input data includes resource attribute information and historical demand data. The resource attribute information is used to perform a priori recognition modeling of the demand pattern. The historical demand data is then used as observations for a hierarchical Bayesian model. Combined with the prior recognition results, the posterior recognition results of the resource demand pattern are inferred.

[0139] Specifically, attribute features are extracted from the attribute information of each sample resource, including categorical attribute features and numerical attribute features. Category attribute features include name, brand, category, color, etc., and numerical attribute features include specification, volume, weight, etc. From the demand data of each sample resource in different historical time periods, such as the early, middle and late stages of the activity, calculate the difference characteristics of demand changes in the early and middle stages and the difference characteristics of demand changes in the middle and late stages. Based on the attribute features and demand change difference features, they are input into a demand pattern recognition model (such as a hierarchical Bayesian model) for training to obtain the value of each regression coefficient of the logistic regression model. Subsequently, the demand pattern adapted to each sample resource can be calculated.

[0140] 2. During the replenishment model training phase, multiple sample resources are also required. The input data includes resource attribute information, historical demand data, adapted demand patterns, and inventory data. Based on this information, the adapted forecasting strategy and inventory strategy are obtained to better provide replenishment strategies and assist operators in making replenishment decisions. Specifically, a sample resource may be adapted to one or more demand patterns, and the replenishment strategy is a combination of the forecasting strategy and the inventory strategy.

[0141] The method provided by the embodiment of the present invention has at least the following beneficial effects compared with the prior art:

[0142] 1. Different from the existing idea of ​​manually labeling sample resources to adapt to demand patterns, this solution uses the resource's own attribute information for effective modeling to construct a priori demand pattern recognition result, and then uses the difference characteristics of the historical demand data of the resource in a specific time period as observations to correct the priori demand pattern recognition result to obtain the final resource demand pattern recognition result. This process does not require human participation, thereby reducing the degree of human subjective influence. Because it considers more and more complex data features and factors than manual experience, and comprehensively considers multiple types of resource-related information to give a more comprehensive recognition, the accuracy of the model in identifying demand patterns is improved.

[0143] 2. It can be performed under data of different dimensions, such as item dimension, category dimension, etc., and the demand pattern recognition results of the required dimensions can be given according to the actual needs of the business. It can give both coarse-grained demand pattern recognition results of category dimensions and demand pattern recognition results refined to a single resource dimension.

[0144] 3. According to the resource demand data, attribute characteristics, inventory data and adaptive demand patterns, appropriate forecasting strategies and inventory strategies are adapted to provide better replenishment strategies. From the perspective of explainability, it can explain what factors affect the changes in resource demand, help operators make replenishment decisions, reduce the decision-making time of operators, improve fulfillment rates and reduce inventory costs.

[0145] See also Figure 6 , which shows a schematic diagram of main modules of a resource information processing device 600 provided in an embodiment of the present invention, including:

[0146] The acquisition module 601 is used to acquire the attribute information of the sample resource and the demand data in different historical time periods, extract the attribute characteristics of the attribute information, and determine the demand change difference characteristics of the sample resource in adjacent historical time periods according to the demand data;

[0147] The first construction module 602 is used to input the attribute features into the original logistic regression model corresponding to each demand pattern to construct an adaptation probability expression between the sample resource and each demand pattern;

[0148] The second construction module 603 is used to construct a probability expression of all demand change difference characteristics of the sample resources, and to construct a joint probability expression of all demand change difference characteristics of the sample resources under the condition of being adapted to each demand pattern;

[0149] The training module 604 is used to calculate the value of the regression coefficient in each original logistic regression model based on the adaptation probability expression, the joint probability expression and the probability expression, so as to obtain the trained logistic regression model of each demand pattern.

[0150] In the implementation device of the present invention, the attribute information includes category attribute information and value attribute information, and the acquisition module 601 is used to:

[0151] Determine the valid code corresponding to each category attribute information, and obtain the category attribute feature of each category attribute information;

[0152] The value of each numerical attribute information is quantified to obtain the numerical attribute characteristics of each numerical attribute information.

[0153] In the implementation device of the present invention, the acquisition module 601 is used to:

[0154] Obtaining demand data of sample resources in a historical preset time period, dividing the historical preset time period into an early historical time period, a mid-term historical time period, and a late historical time period; and

[0155] Use the hypothesis verification method to determine whether the mean of the sample resource demand data in the early historical period is equal to the mean of the demand data in the mid-term historical period, and obtain the difference characteristics of the early and mid-term demand changes; and

[0156] Determine whether the mean of the demand data of the sample resources in the mid-term historical period is equal to the mean of the demand data in the late historical period, and obtain the difference characteristics of the mid-term and late-term demand changes.

[0157] In the implementation device of the present invention, the second building module 603 is used to:

[0158] A probability expression of each demand change difference feature of the sample resource under the condition of being adapted to each demand pattern is constructed, and the probability expressions of all demand change difference features are multiplied to obtain the joint probability expression.

[0159] The implementation device of the present invention also includes a replenishment model training module, which is used to:

[0160] Input the attribute characteristics of the sample resources into the logistic regression model corresponding to each demand pattern to obtain the matching probability between the sample resources and each demand pattern, and use the demand pattern with the largest matching probability as the target demand pattern that matches the sample resources;

[0161] Receive an adaptive replenishment strategy for each sample resource tag;

[0162] The current inventory data of the sample resources is obtained, and the attribute characteristics of the sample resources, the demand data in different historical time periods, the inventory data, the adapted target demand pattern, and the adapted replenishment strategy are input into the original replenishment model for training to obtain the trained replenishment model.

[0163] The implementation device of the present invention also includes a replenishment model application module, which is used to:

[0164] Obtain attribute information of new resources, demand data in different historical time periods, and current inventory data, and extract attribute features of the attribute information;

[0165] Input the attribute information of the new resource into the logistic regression model corresponding to each demand pattern to obtain the adaptation probability of the new resource to each demand pattern, and take the demand pattern with the largest adaptation probability as the target demand pattern adapted to the new resource;

[0166] The attribute characteristics of the new resource, the demand data in different historical time periods, the inventory data, and the adapted target demand pattern are input into the replenishment model to obtain the target replenishment strategy.

[0167] In the implementation device of the present invention, the replenishment strategy includes a forecasting strategy and an inventory strategy;

[0168] The replenishment model application module is used to:

[0169] Input the attribute characteristics of the new resource, the demand data in different historical time periods, the inventory data, and the adapted target demand pattern into the replenishment model to obtain a forecasting strategy set and an inventory strategy set;

[0170] Combine the forecast strategies in the forecast strategy set and the inventory strategies in the inventory strategy set to obtain multiple combination schemes, apply each combination scheme to the new resources, obtain application result data, and use the combination scheme with the largest application result data as the target replenishment strategy.

[0171] In addition, the specific implementation content of the device described in the embodiment of the present invention has been described in detail in the method described above, so the repeated content will not be described again here.

[0172] Figure 7 An exemplary system architecture 700 to which embodiments of the present invention may be applied is shown, including terminal devices 701 , 702 , 703 , a network 704 and a server 705 (only an example).

[0173] Terminal devices 701, 702, 703 can be various electronic devices with display screens and supporting web browsing, and various communication client applications are installed. Users can use terminal devices 701, 702, 703 to interact with server 705 through network 704 to receive or send messages, etc.

[0174] The network 704 is used to provide a medium for communication links between the terminal devices 701, 702, 703 and the server 705. The network 704 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0175] The server 705 may be a server that provides various services. It should be noted that the method provided in the embodiment of the present invention is generally executed by the server 705 , and accordingly, the device is generally set in the server 705 .

[0176] It should be understood that Figure 7 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0177] Reference below Figure 8, which shows a schematic diagram of the structure of a computer system 800 of a terminal device suitable for implementing an embodiment of the present invention. Figure 8 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0178] like Figure 8 As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage part 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the system 800 are also stored. The CPU 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0179] The following components are connected to the I / O interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, a modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed, so that a computer program read therefrom is installed into the storage section 808 as needed.

[0180] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the central processing unit (CPU) 801, the above-mentioned functions defined in the system of the present invention are executed.

[0181] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0182] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0183] The modules involved in the embodiments of the present invention may be implemented by software or hardware. The modules described may also be set in a processor, for example, may be described as: a processor includes an acquisition module, a first construction module, a second construction module, and a training module. The names of these modules do not constitute a limitation on the modules themselves in some cases, for example, the training module may also be described as a "logistic regression model training module".

[0184] As another aspect, the present invention further provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device executes any of the resource information processing methods described above.

[0185] The computer program product of the present invention comprises a computer program, and when the computer program is executed by a processor, the resource information processing method in the embodiment of the present invention is implemented.

[0186] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A resource information processing method, characterized in that: include: Acquire attribute information of sample resources and demand data in different historical time periods, extract attribute features of the attribute information, and determine, based on the demand data, demand change difference features of the sample resources in adjacent historical time periods; Inputting the attribute features into the original logistic regression model corresponding to each demand pattern to construct an adaptation probability expression between the sample resource and each demand pattern; Constructing a probability expression of all demand change difference characteristics of the sample resources, and constructing a joint probability expression of all demand change difference characteristics of the sample resources under the condition of being adapted to each of the demand patterns; Based on the adaptation probability expression, the joint probability expression and the probability expression, the value of the regression coefficient in each original logistic regression model is calculated, so as to obtain the logistic regression model after training for each demand pattern.

2. The method according to claim 1, characterized in that The attribute information includes category attribute information and value attribute information, and the extracting attribute features of the attribute information includes: Determine the valid code corresponding to each category attribute information, and obtain the category attribute feature of each category attribute information; The value of each numerical attribute information is quantified to obtain the numerical attribute characteristics of each numerical attribute information.

3. The method according to claim 1, characterized in that Obtaining demand data of sample resources in different historical time periods, including: obtaining demand data of sample resources in a historical preset time period, dividing the historical preset time period into an early historical time period, a mid-term historical time period, and a late historical time period; Determining the difference characteristics of demand changes of sample resources in adjacent historical time periods according to the demand data includes: Use the hypothesis verification method to determine whether the mean of the sample resource demand data in the early historical period is equal to the mean of the demand data in the mid-term historical period, and obtain the difference characteristics of the early and mid-term demand changes; and Determine whether the mean of the demand data of the sample resources in the mid-term historical period is equal to the mean of the demand data in the late historical period, and obtain the difference characteristics of the mid-term and late-term demand changes.

4. The method according to claim 1, characterized in that The constructing of a joint probability expression of all demand change difference characteristics of the sample resources under the condition of being adapted to each demand pattern includes: A probability expression of each demand change difference feature of the sample resource under the condition of being adapted to each demand pattern is constructed, and the probability expressions of all demand change difference features are multiplied to obtain the joint probability expression.

5. The method according to claim 1, characterized in that After obtaining the trained logistic regression model for each demand pattern, the method further includes: Input the attribute characteristics of the sample resources into the logistic regression model corresponding to each demand pattern to obtain the matching probability between the sample resources and each demand pattern, and use the demand pattern with the largest matching probability as the target demand pattern that matches the sample resources; Receive an adaptive replenishment strategy for each sample resource tag; The current inventory data of the sample resources is obtained, and the attribute characteristics of the sample resources, the demand data in different historical time periods, the inventory data, the adapted target demand pattern, and the adapted replenishment strategy are input into the original replenishment model for training to obtain the trained replenishment model.

6. The method according to claim 5, characterized in that After obtaining the trained replenishment model, the method further includes: Obtain attribute information of new resources, demand data in different historical time periods, and current inventory data, and extract attribute features of the attribute information; Input the attribute information of the new resource into the logistic regression model corresponding to each demand pattern to obtain the adaptation probability of the new resource to each demand pattern, and take the demand pattern with the largest adaptation probability as the target demand pattern adapted to the new resource; The attribute characteristics of the new resource, the demand data in different historical time periods, the inventory data, and the adapted target demand pattern are input into the replenishment model to obtain the target replenishment strategy.

7. The method according to claim 5 or 6, characterized in that: Replenishment strategies include forecasting strategies and inventory strategies; The attribute characteristics of the new resource, demand data in different historical time periods, inventory data, and the adapted target demand pattern are input into the replenishment model to obtain the target replenishment strategy, including: Input the attribute characteristics of the new resource, the demand data in different historical time periods, the inventory data, and the adapted target demand pattern into the replenishment model to obtain a forecasting strategy set and an inventory strategy set; Combine the forecast strategies in the forecast strategy set and the inventory strategies in the inventory strategy set to obtain multiple combination schemes, apply each combination scheme to the new resources, obtain application result data, and use the combination scheme with the largest application result data as the target replenishment strategy.

8. A resource information processing device, characterized in that: include: An acquisition module, used to acquire attribute information of sample resources and demand data in different historical time periods, extract attribute features of the attribute information, and determine demand change difference features of sample resources in adjacent historical time periods based on the demand data; A first construction module is used to input the attribute features into the original logistic regression model corresponding to each demand pattern to construct an adaptation probability expression between the sample resource and each demand pattern; The second construction module is used to construct a probability expression of all demand change difference characteristics of the sample resources, and to construct a joint probability expression of all demand change difference characteristics of the sample resources under the condition of being adapted to each demand pattern; The training module is used to calculate the value of the regression coefficient in each original logistic regression model based on the adaptation probability expression, the joint probability expression and the probability expression, so as to obtain the trained logistic regression model of each demand pattern.

9. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

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

Images