An inventory optimization method based on data-driven return prediction

Abstract

The application discloses an inventory optimization method based on data-driven return prediction, which comprises the following steps: 1. analyzing and processing data in an information system of an enterprise; 2. calculating a predicted total return quantity of an ordering period; 3. constructing a multi-period inventory optimization model considering return prediction and calculating an optimal inventory level of each ordering period; and 4. deciding an optimal replenishment quantity. The application can combine return prediction and inventory optimization, considers product return as a kind of inventory source to decide the optimal replenishment quantity, thereby reducing enterprise inventory and lowering total loss of the enterprise in the whole sales period.

Description

Technical Field

[0001] This invention belongs to the field of e-commerce technology, specifically a data-driven inventory optimization method for return prediction. Background Technology

[0002] With the rapid development of information technology and logistics, e-commerce has become an integral part of people's lives. However, in the e-commerce environment, customers cannot experience physical goods and rely on express delivery to receive them, which easily leads to problems such as incorrect sizing, color differences from online pictures, delivery errors, and damage during delivery, resulting in a high rate of returns. Some returned online shopping products do not have quality issues; the main reason for returns is a mismatch with customer needs, such as clothing and accessories due to color differences or incorrect sizes. Products without quality problems can be resold after simple cleaning and repackaging. Product returns are also an important source of replenishment for businesses' inventory. However, most companies make inventory decisions primarily based on consumer demand, determining replenishment quantities based on inventory flow, while neglecting consumer returns. In the era of increasingly popular online shopping, the role of product returns as a source of inventory is becoming increasingly undeniable. Good inventory management is related to a company's losses and profits. Well-known methods for optimizing corporate inventory all start from the perspective of consumer demand, aiming to minimize inventory losses when deciding on replenishment quantities. However, this method and apparatus ignore the impact of consumer returns, leading to excessive replenishment. Therefore, existing enterprise inventory optimization methods and apparatus cannot achieve the goal of minimizing inventory losses when considering consumer returns.

[0003] Existing enterprise inventory optimization methods and devices do not consider the impact of consumer returns on inventory. Given the increasingly serious problem of consumer returns, existing methods are insufficient to meet the needs of effective inventory management. Summary of the Invention

[0004] The present invention aims to address the shortcomings of the existing technology by providing a data-driven return forecasting-based inventory optimization method. This method combines return forecasting with inventory optimization, considering product returns as a source of inventory to determine the optimal replenishment quantity, thereby reducing enterprise inventory and lowering the enterprise's total losses throughout the sales period.

[0005] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0006] The data-driven return forecasting and inventory optimization method of this invention is characterized by the following steps:

[0007] Step 1: Analyze and process the data in the company's information system:

[0008] Step 1.1: Obtain the total product sales volume S, total returned quantity R, and the quantity of returned goods resold RS from the enterprise's information system, and then use equations (1) and (2) to calculate the average return rate P of the product. r The resale ratio β of returned products:

[0009]

[0010]

[0011] Step 1.2: Obtain the return time and corresponding purchase time of all returned products from the enterprise's information system, and subtract the purchase time from the return time to obtain the lag time set d = {d1, ...,d2} of the returned products. r ,…,d R}, where d r This represents the lag time for the r-th returned product, where R is the total number of returns.

[0012] Step 1.3: Fit the density plot and distribution plot of the lag time set d to obtain the cumulative distribution function F that the lag time of returned products follows. d ;

[0013] Step 1.4: Obtain the sales volume set for each ordering cycle from the enterprise's information system, denoted as D = {D1, ..., D2}. j , ..., D |D|}, where D j The sales volume refers to the sales volume in the j-th ordering cycle, where |D| is the total number of ordering cycles.

[0014] Estimate the probability distribution function Φ of the sales volume for each ordering cycle based on the sales volume set D;

[0015] Step 2: Calculate the predicted total return quantity r for the j-th ordering cycle. j :

[0016] Step 2.1, define the period during which consumers are allowed to return products after purchase;

[0017] Let the start time and end time of the j-th ordering cycle be respectively

[0018] Let D be the sales volume of the m-th ordering cycle preceding the j-th ordering cycle. m , m < j;

[0019] Let t be the time when the i-th product is purchased within the m-th ordering cycle. m (i), i = 1, ..., D m ;

[0020] Use equation (3) to calculate the probability that a consumer who purchases the i-th product in the m-th ordering cycle will hold it until the j-th ordering cycle.

[0021]

[0022] Step 2.2, use equation (4) to calculate the probability P that a consumer purchases the i-th product in the m-th ordering cycle and decides to return it after holding it until the j-th ordering cycle. mj (i):

[0023]

[0024] Step 2.3, use equation (5) to calculate the sales volume D for the m-th ordering cycle. m The number of returns r in the j-th ordering cycle mj :

[0025]

[0026] Step 2.4, use equation (6) to calculate the expected total return quantity r in the j-th ordering cycle. j :

[0027]

[0028] Step 3: Construct a multi-period inventory optimization model that takes into account return forecasts to calculate the optimal inventory level for each ordering cycle;

[0029] Step 3.1, assume c is the replenishment loss per unit product, and h and p are the holding loss per unit product and the stockout loss per unit product, respectively;

[0030] Let x be the number of times x is used. j Let y be the beginning inventory for the j-th ordering cycle. j Let be the inventory level of the online retailer after replenishment in the j-th ordering cycle; then the replenishment loss in the j-th ordering cycle is c(y). j -x j Using equation (7), calculate the beginning inventory x of the enterprise at the end of the j+1 ordering cycle. j+1 :

[0031] x j+1 =y j +β(r j +∈ j )-D j (7)

[0032] In equation (7), ∈ j The difference between the actual return quantity and the predicted return quantity in the j-th ordering cycle;

[0033] Calculate the inventory loss ω(y) for the j-th ordering cycle using equation (8). j ):

[0034] ω(y j )=h[y j +β(r j +∈ j )-d j ] + +p[d j -y j -β(r j +∈ j )] + (8)

[0035] In equation (8), [] + This indicates that the element in brackets [] is the maximum value between itself and "0";

[0036] Step 3.2, establish a multi-period inventory optimization model:

[0037] Assuming the product continues to be sold until the Tth ordering cycle, the total loss function V in the multi-period inventory optimization model constructed using equation (9) from the jth ordering cycle to the Tth ordering cycle is... j (x j ):

[0038]

[0039] Let the boundary condition of the multi-period inventory optimization model be V. T+1 (x T+1 )=-cx T+1 Where γ is the discount factor; x T+1 This represents the beginning inventory for the T+1th ordering period;

[0040] Step 3.3: Solve the multi-period inventory optimization model using backward induction, and then use equation (10) to obtain the optimal inventory level S for the j-th ordering cycle. j :

[0041]

[0042] Step 4: Determine the optimal replenishment quantity:

[0043] x is the beginning inventory for the j-th ordering cycle. j With optimal inventory level S j For comparison, if the initial inventory is x j Below the optimal inventory level S j Then the replenishment quantity for the j-th ordering cycle is S. j -x j Otherwise, it means that no replenishment is needed in the j-th ordering cycle.

[0044] The present invention provides an electronic device, including a memory and a processor, wherein the memory is used to store a program that supports the processor in executing the inventory optimization method, and the processor is configured to execute the program stored in the memory.

[0045] The present invention discloses a computer-readable storage medium on which a computer program is stored, wherein the computer program is executed by a processor to perform the steps of the inventory optimization method.

[0046] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0047] 1. This invention proposes a data-driven return prediction method, which predicts the number of product returns within a certain period of time based on the enterprise information system. It is simple to operate and easy to implement, overcoming the shortcomings of existing technologies that require a large amount of data and are complex to operate, and achieving a balance between solution speed and accuracy.

[0048] 2. This invention combines return forecasting with inventory decision-making. When making the optimal replenishment quantity decision, it is necessary to first predict the quantity of product returns. This overcomes the shortcomings of existing technologies that ignore product returns as a source of inventory, effectively reducing enterprise inventory and lowering the total loss of the enterprise throughout the sales period. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the process of the present invention;

[0050] Figure 2 This is a probability density function graph of the holding time d of the present invention;

[0051] Figure 3 This is a graph showing the cumulative distribution function of the holding time d of this invention;

[0052] Figure 4 The probability density function graph of weekly sales of this invention

[0053] Figure 5 This is the optimal ordering decision diagram for the first set of experiments in this invention, considering both return prediction and non-return prediction. Detailed Implementation

[0054] In this embodiment, as Figure 1 As shown, a data-driven return forecasting and inventory optimization method is carried out according to the following steps:

[0055] Step 1: Analyze and process the data in the company's information system:

[0056] Step 1.1: Obtain the total product sales volume S, total returned quantity R, and the quantity of returned goods resold RS from the enterprise's information system, and then use equations (1) and (2) to calculate the average return rate P of the product. r The resale ratio β of returned products:

[0057]

[0058]

[0059] In this embodiment, the data used comes from a fast fashion apparel company. Transaction and return data for a specific product of this company were obtained, resulting in a total sales volume S of 31,744, a total return quantity R of 16,641, and 13,479 of the returned items resold. Therefore, the average return rate P is calculated. r The resale rate (β) of returned products is 0.524, and the resale rate (β) is 0.81.

[0060] Step 1.2: Obtain the return time and corresponding purchase time of all returned products from the enterprise's information system, and subtract the purchase time from the return time to obtain the lag time set d = {d1, ..., d2} of the returned products. r , ..., d R}, where d r Let R represent the lag time for the r-th returned product, where R is the total number of returns. In this embodiment, if the return date is 2019 / 10 / 7 and the corresponding purchase date is 2019 / 10 / 3, then the lag time is 5 days, which is a multiple of one day.

[0061] Step 1.3: Fit the density plot and distribution plot of the lag time set d to obtain the cumulative distribution function F that the lag time of returned products follows. d In this embodiment, the density map and distribution map of the lag time set d are as follows: Figure 2 and Figure 3 As shown, the lag time of returned products follows a truncated log-normal distribution.

[0062] Step 1.4: Obtain the sales volume set for each ordering cycle from the enterprise's information system, denoted as D = {D1, ..., D2}. j , ..., D |D|}, where D j The sales volume refers to the sales volume in the j-th ordering cycle, where |D| is the total number of ordering cycles.

[0063] The probability distribution function Φ of sales volume for each ordering cycle is estimated based on the sales volume set D; in this embodiment, one week is considered one ordering cycle, such as... Figure 4 As shown, weekly sales follow a normal distribution.

[0064] Step 2: Calculate the predicted total return quantity r for the j-th ordering cycle. j :

[0065] Step 2.1, define the period during which consumers are allowed to return products after purchase;

[0066] Let the start time and end time of the j-th ordering cycle be respectively

[0067] Let D be the sales volume of the m-th ordering cycle preceding the j-th ordering cycle. m , m < j;

[0068] Let t be the time when the i-th product is purchased within the m-th ordering cycle. m (i), i = 1, ..., D m ;

[0069] Use equation (3) to calculate the probability that a consumer who purchases the i-th product in the m-th ordering cycle will hold it until the j-th ordering cycle.

[0070]

[0071] Step 2.2, use equation (4) to calculate the probability P that a consumer purchases the i-th product in the m-th ordering cycle and decides to return it after holding it until the j-th ordering cycle. mj (i):

[0072]

[0073] Step 2.3, use equation (5) to calculate the sales volume D for the m-th ordering cycle. m The number of returns r in the j-th ordering cycle mi :

[0074]

[0075] Step 2.4, use equation (6) to calculate the expected total return quantity r in the j-th ordering cycle. j :

[0076]

[0077] Step 3: Construct a multi-period inventory optimization model that takes into account return forecasts to calculate the optimal inventory level for each ordering cycle;

[0078] Step 3.1, assuming c is the replenishment loss per unit product, and h and p are the holding loss per unit product and the stockout loss per unit product, respectively; in this embodiment, c = 2, h = 1.9, and p = 12 are set.

[0079] Let x be the number of times x is used. j Let y be the beginning inventory for the j-th ordering cycle. j Let be the inventory level of the online retailer after replenishment in the j-th ordering cycle; then the replenishment loss in the j-th ordering cycle is c(y). j -x j Using equation (7), calculate the beginning inventory x of the enterprise at the end of the j+1 ordering cycle. j+1 :

[0080] x j+1 =y j +β(r j +∈ j )-D j (7)

[0081] In equation (7), ∈ j The difference between the actual return quantity and the predicted return quantity in the j-th ordering cycle;

[0082] Calculate the inventory loss ω(y) for the j-th ordering cycle using equation (8). j ):

[0083] ω(y j )=h[y j +β(r j +∈ j )-d j ] + +p[d j -y j -β(r j +∈ j )] + (8)

[0084] In equation (8), [] + This indicates that the element in brackets [] is the maximum value between itself and "0";

[0085] Step 3.2, establish a multi-period inventory optimization model:

[0086] Assuming the product continues to be sold until the Tth ordering cycle, the total loss function V in the multi-period inventory optimization model constructed using equation (9) from the jth ordering cycle to the Tth ordering cycle is... j (x j ):

[0087]

[0088] Let the boundary condition of the multi-period inventory optimization model be V. T+1 (x T+1 )=-cx T+1 Where γ is the discount factor; x T+1This represents the beginning inventory for the (T+1)th ordering cycle; in this embodiment, T = 20 is set.

[0089] Step 3.3: Solve the multi-period inventory optimization model using backward induction, and then use equation (10) to obtain the optimal inventory level S for the j-th ordering cycle. j :

[0090]

[0091] Step 4: Determine the optimal replenishment quantity:

[0092] x is the beginning inventory for the j-th ordering cycle. j With the optimal inventory level S j For comparison, if the initial inventory is x j Below the optimal inventory level S j Then the replenishment quantity for the j-th ordering cycle is S. j -x j Otherwise, it means that no replenishment is needed in the j-th ordering cycle.

[0093] Experimental demonstrations were conducted on the method of this invention: a total of 5 sets of experiments were carried out. By comparing and analyzing the optimal inventory decision and total loss under the two cases of considering return forecasting and not considering return forecasting, the value of return forecasting for inventory management was analyzed. Figure 5 Taking the first group of experiments as an example, this paper presents a comparison of the optimal order quantity under the two conditions of not considering return forecasting and considering return forecasting. Table 1 presents the total losses under the two conditions of not considering return forecasting and considering return forecasting in the five groups of experiments.

[0094] Table 1

[0095]

[0096] Note:

[0097] The results in Table 1 show that return forecasting can reduce order quantities, thereby reducing order losses and holding losses. The method of this invention has achieved good results in effectively reducing enterprise losses.

[0098] In this embodiment, an electronic device includes a memory and a processor. The memory stores a program that supports the processor in executing the above-described method, and the processor is configured to execute the program stored in the memory.

[0099] In this embodiment, a computer-readable storage medium stores a computer program, which is executed by a processor to perform the steps of the above method.

Claims

1. A data-driven method for return forecasting and inventory optimization, characterized by: Follow these steps: Step 1: Analyze and process the data in the company's information system: Step 1.1: Obtain the total product sales volume S, total returned quantity R, and the quantity of returned goods resold RS from the enterprise's information system, and then use equations (1) and (2) to calculate the average return rate P of the product. r The resale ratio β of returned products: Step 1.2: Obtain the return time and corresponding purchase time of all returned products from the enterprise's information system, and subtract the purchase time from the return time to obtain the lag time set d = {d1, ...,d2} of the returned products. r ,…,d R }, where d r This represents the lag time for the r-th returned product, where R is the total number of returns. Step 1.3: Fit the density plot and distribution plot of the lag time set d to obtain the cumulative distribution function F that the lag time of returned products follows. d ; Step 1.4: Obtain the sales set for each ordering cycle from the enterprise's information system, denoted as D = {D1, ..., D...} j ,…,D |D| }, where D j The sales volume refers to the sales volume in the j-th ordering cycle, where |D| is the total number of ordering cycles. Estimate the probability distribution function Φ of the sales volume for each ordering cycle based on the sales volume set D; Step 2: Calculate the predicted total return quantity r for the j-th ordering cycle. j : Step 2.1, define the period during which consumers are allowed to return products after purchase; Let the start time and end time of the j-th ordering cycle be respectively Let D be the sales volume of the m-th ordering cycle preceding the j-th ordering cycle. m m <j; Let t be the time when the i-th product is purchased within the m-th ordering cycle. m (i), i = 1, ..., D m ; Use equation (3) to calculate the probability that a consumer who purchases the i-th product in the m-th ordering cycle will hold it until the j-th ordering cycle. Step 2.2, use equation (4) to calculate the probability P that a consumer purchases the i-th product in the m-th ordering cycle and decides to return it after holding it until the j-th ordering cycle. mj (i): Step 2.3, use equation (5) to calculate the sales volume D for the m-th ordering cycle. m The number of returns r in the j-th ordering cycle mj : Step 2.4, use equation (6) to calculate the expected total return quantity r in the j-th ordering cycle. j : Step 3: Construct a multi-period inventory optimization model that takes into account return forecasts to calculate the optimal inventory level for each ordering cycle; Step 3.1, assume c is the replenishment loss per unit product, and h and p are the holding loss per unit product and the stockout loss per unit product, respectively; Let x be the number of times x is used. j Let y be the beginning inventory for the j-th ordering cycle. j Let be the inventory level of the online retailer after replenishment in the j-th ordering cycle; then the replenishment loss in the j-th ordering cycle is c(y). j -x j Using equation (7), calculate the beginning inventory x of the enterprise at the end of the j+1 ordering cycle. j+1 : x j+1 =y j +β(r j +∈ j )-D j (7) In equation (7), ∈ j The difference between the actual return quantity and the predicted return quantity in the j-th ordering cycle; Calculate the inventory loss ω(y) for the j-th ordering cycle using equation (8). j ): ω(y j )=h[y j +β(r j +∈ j )-d j ] + +p[d j -y j -β(r j +∈ j )] + (8) In equation (8), [] + This indicates that the element in brackets [] is the maximum value between itself and "0"; Step 3.2, establish a multi-period inventory optimization model: Assuming the product continues to be sold until the Tth ordering cycle, the total loss function V in the multi-period inventory optimization model constructed using equation (9) from the jth ordering cycle to the Tth ordering cycle is... j (x j ): Let the boundary condition of the multi-period inventory optimization model be V. T+1 (x T+1 )=-cx T+1 Where γ is the discount factor; x T+1 This represents the beginning inventory for the T+1th ordering period; Step 3.3: Solve the multi-period inventory optimization model using backward induction, and then use equation (10) to obtain the optimal inventory level S for the j-th ordering cycle. j : Step 4: Determine the optimal replenishment quantity: x is the beginning inventory for the j-th ordering cycle. j With the optimal inventory level S j For comparison, if the initial inventory is x j Below the optimal inventory level S j Then the replenishment quantity for the j-th ordering cycle is S. j -x j Otherwise, it means that no replenishment is needed in the j-th ordering cycle.

2. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store a program that supports the processor in executing the inventory optimization method of claim 1, the processor being configured to execute the program stored in the memory.

3. A computer-readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to perform the steps of the inventory optimization method of claim 1.

Images