Warehousing intelligent safety inventory identification and prediction method

Abstract

The invention belongs to the technical field of intelligent warehousing, and particularly relates to a warehousing intelligent safe inventory identification and prediction method, which comprises the following steps: step 1, collecting data required by inventory identification and prediction; step 2, demand type classification: supporting demand prediction for different categories by accurately classifying demand types of inventory goods; 3, performing time sequence model prediction on the classified high-importance goods; 4, performing Bayesian model prediction on the classified low-importance cargos; and the step 4 and the step 3 can be carried out without any order, after a prediction result is obtained, an inventory replenishment plan is generated based on the prediction result, and a corresponding person in charge completes warehousing replenishment operation through a purchasing system. According to the method, accurate prediction is realized through two modes of time sequence and exponential distribution Bayesian prediction, and the extra additional cost generated by stock overstock is reduced on the basis of meeting the delivery demand.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent warehousing, and in particular relates to a method for identifying and predicting intelligent safety inventory in warehousing. Background Art

[0002] In the daily warehouse management of logistics enterprises, the management of safety stock is a key link. It involves maintaining appropriate inventory to prevent insufficient inventory, while not increasing capital and management costs due to inventory occupation. However, the current management of safety stock is generally through setting a fixed threshold. Once it is lower than the specified inventory, replenishment is initiated. However, the actual inventory demand is often different. The general safety stock strategy cannot meet the actual safety stock management requirements. The following problems usually exist in the management of safety stock: 1. Uncertainty in demand forecasting. The volatility of demand makes it complicated to maintain a suitable safety stock level. If the demand forecast is inaccurate, it may lead to insufficient or excessive inventory. 2. The instability of the supply chain, delayed delivery by suppliers, shortage of raw materials or unstable quality affect the setting of safety stock. 3. The balance of cost and risk. Safety stock needs to weigh the storage occupation cost, capital occupation cost and out-of-stock risk. Too much safety stock will increase costs, while too little will lead to out-of-stock risks. 4. Inaccurate inventory data. Incomplete inventory data records lead to inaccurate inventory data, which affects the calculation of safety stock and leads to inaccurate decision-making. 5. Changes in periodic activities. Due to temporary promotions of different categories, inventory is tight, making the calculation of safety stock more complicated. The above problems make it difficult to predict and control the management of safety stock, and the actual pipeline effect is poor. Summary of the invention

[0003] The purpose of the present invention is to provide a method for intelligent warehouse safety stock identification and prediction, which is used to solve the technical problems in the prior art that safety stock management is difficult to predict and control and the actual management effect is poor.

[0004] The method for identifying and predicting intelligent warehouse safety inventory includes the following steps.

[0005] Step 1: Collect the data required for inventory identification and forecasting.

[0006] Step 2: Classify demand types. Accurately classify the demand types of inventory goods to support demand forecasts for different categories.

[0007] Step 3: Use time series model to predict the classified high-importance goods.

[0008] Step 4: Use the Bayesian model to predict the classified low-importance goods.

[0009] Step 4 and step 3 can be performed in no particular order. After obtaining the forecast results, a replenishment inventory plan is generated based on the forecast results, and the relevant person in charge completes the warehousing and replenishment operations through the procurement system.

[0010] Preferably, in step one, the required data include inventory data, purchase cost, order data and transportation time, the processed inventory data comes from the intelligent warehousing system, and the purchase cost, order data, transportation time and other data come from the order center, transportation center and upstream ERP data center respectively.

[0011] Preferably, in the step 2, classification is first performed according to the first dimension, and the three types of ABC are classified from high to low according to the purchase amount ranking of the key factor affecting the inventory; then classification is performed according to the second dimension, and three types of XYZ are set according to the speed of inventory consumption from fast to slow; then the classifications of the two dimensions are combined to obtain nine classifications: ABC and XYZ are combined into nine classifications: AX, AY, AZ, BX, BY, BZ, CX, CY, CZ; the 9 classifications are ranked in importance, the categories with high purchase amount ranking are of high importance, and the categories with large inventory consumption and easy to increase the difficulty of inventory management are of higher importance. After sorting, the order of importance of the 9 categories of goods is AX, BX, CX, AY, BY, CY, AZ, BZ, CZ.

[0012] Preferably, in step 2, the characteristics of the 9 categories are as follows:

[0013] AX: The amount accounts for a high proportion, the inventory is consumed quickly, the storage is close to the exit, and the picking operation is convenient;

[0014] AY: The amount accounts for a high proportion, the inventory consumption rate is average, and it is stored in the general operation area to ensure inventory control;

[0015] AZ: high proportion of amount, slow inventory consumption, stored inside the warehouse, regular inspection;

[0016] BX: The amount accounts for a normal proportion, the inventory is consumed quickly, it is stored close to the exit, and it is checked frequently;

[0017] BY: The amount accounts for a normal proportion, the inventory consumption rate is normal, it is stored in a normal operation area, and it is routinely inspected;

[0018] BZ: The amount accounts for a normal proportion, the inventory is consumed slowly, stored inside the warehouse, and undergoes regular inspections;

[0019] CX: Low amount, fast inventory consumption, storage close to the exit, frequent inspection;

[0020] CY: low proportion of amount, average inventory consumption rate, stored in general operation area, routine inspection;

[0021] CZ: The amount accounts for a low proportion, the inventory is consumed slowly, it is stored inside the warehouse, and it is inspected regularly.

[0022] Preferably, in step three, high-importance goods have the characteristics of high value or high frequency, including six types of goods: AX, AY, BX, BY, CX, and CY. The step of predicting safety stock through the time series model includes: 1) calculating the average value of historical demand in the past n periods, 2) calculating the time distance, 3) considering the impact of time distance on the average value of historical demand, and combining the two formulas to calculate the predicted value of safety stock demand.

[0023] Preferably, the steps of predicting safety stock by using a time series model specifically include: 1) calculating the average value of the historical demand in the recent n periods using the formula: M = (X 1 +X 2 +...+X n ) / n, where X n Indicates the most recent historical demand for goods, X 1 It represents the first historical demand of the calculation application, and n is the number of historical demands required for calculation based on the actual outbound demand, also known as the number of cycles involved in the prediction. The greater the impact of the corresponding trend, the longer the time interval, that is, the larger n is, and vice versa.

[0024] 2) The formula for calculating time distance is: X = (X n -X 1 ) / (n-1), this formula calculates the time distance based on n historical demands, where X n represents the historical demand for goods in the most recent period, X 1 represents the historical demand for period 1 to which the calculation applies;

[0025] 3) The calculation formula for the forecast value of safety stock demand is as follows: M+X=(X 2 +X 3 +....+X n+1 ) / n,X n +1=(nX n -X 1 ) / (n-1), where X n+1 represents the forecast demand value of the forecast period, n is the number of periods involved in the forecast, X 2 , X 3 They are the historical demands for the 2nd and 3rd periods required for forecasting.

[0026] Preferably, in step 4, for the three types of goods of low importance, AZ, BZ, and CZ, historical data fitting analysis is performed, and a distribution function that conforms to the fitting result is selected in combination with the Bayesian analysis method for prediction. The specific steps are as follows:

[0027] 1) Determine and verify the distribution probability function by fitting historical data;

[0028] 2) Determine the prior prediction distribution according to the conjugate distribution selection principle of the Bayesian method;

[0029] 3) Obtain the critical inventory level and predict the inventory demand.

[0030] The present invention has the following advantages: The present invention improves the prediction accuracy of warehouse safety inventory by classifying the warehouse goods information and predicting separately through two prediction models, effectively collects goods information through multiple channels and multiple systems and pre-processes it, thereby enriching the accuracy of the prediction input data. Effective classification of goods information is achieved through demand classification method, and accurate prediction is achieved through two methods, time series and exponential distribution Bayesian prediction. It has been verified that the accuracy of the prediction by fitting and comparing with historical data is as high as 98%, thereby reducing the additional costs caused by inventory backlog on the basis of meeting the outbound demand. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 The present invention is a flowchart of a method for intelligent warehouse safety inventory identification and prediction. DETAILED DESCRIPTION

[0032] The specific implementation modes of the present invention are further explained in detail below by describing the embodiments with reference to the accompanying drawings, so as to help those skilled in the art to have a more complete, accurate and in-depth understanding of the inventive concept and technical solution of the present invention.

[0033] In the intelligent warehousing system, it is important to realize the identification and prediction of safety stock and the ability to drive replenishment tasks. The platform obtains logistics data of the entire chain, including sales order data from the source control tower or ERP system, historical warehouse operation in and out data, and indirect logistics transportation execution data obtained from the order-free execution system that cooperates with the warehousing system. In order to achieve the optimal operation process of the safety stock prediction platform, the steps are as follows.

[0034] 1. Obtain sales orders through ERP or control tower, including SKU information such as product name, code, unit price, supplier, etc., so as to obtain single product classification and price data.

[0035] 2. Based on the issuance of sales orders, the logistics order center generates outbound order tasks according to the shipping tasks and pushes them to the intelligent warehousing system to determine the warehousing outbound tasks. The warehousing outbound tasks can analyze the in-stock data and recommend commodity storage locations based on the SKU categories, support normal picking, compounding, and outbound operations, and generate warehousing outbound orders.

[0036] 3. Based on the warehouse import tasks and pre-warehousing delivery data, the order execution center realizes procurement. The procurement portal provides timeliness data such as order transportation timeliness, carriers, supply volume, vehicle maintenance and delivery.

[0037] 4. Warehouse inventory management includes warehouse outbound and inbound order data, real-time inventory data, location inventory change logs, etc. Warehouse inventory management ensures the accuracy of warehouse inventory data.

[0038] 5. After each picking task is completed, the intelligent warehousing system will perform a review and scan code to confirm the outbound operation and push a message.

[0039] Based on the outbound confirmation action, the system drives the analysis and identification of safety stock and generates asynchronous safety stock messages. After receiving the safety stock task, the message consumer calls the inventory identification and analysis algorithm, combines each storage location, inventory, SKU, supply delivery time, unit price of goods, and historical inventory change data to perform FSN product classification, high-frequency time series inventory forecasting, and low-frequency Bayesian inventory forecasting. Based on the analysis results, it determines whether to generate a safety stock replenishment task.

[0040] 6. Use the SKU forecast safety inventory data generated in step 5 to drive the generation of automated replenishment tasks, and combine the replenishment category, inventory manager, and inventory replenishment warning notification.

[0041] 7. The person in charge of inventory replenishment will add safety stock based on replenishment notifications to achieve optimal inventory and avoid cost problems caused by insufficient or excessive inventory.

[0042] like Figure 1 As shown, the present invention provides a method for intelligent warehouse safety stock identification and prediction, which implements the identification and prediction of safety stock in step 5, including the following steps.

[0043] Step 1: Collect the data required for inventory identification and forecasting.

[0044] The inventory data processed by this method comes from the intelligent warehousing system, and the procurement cost, order data, transportation time and other data come from the order center, transportation center and upstream ERP data center respectively.

[0045] Step 2: Classify demand types. Accurately classify the demand types of inventory goods to support demand forecasts for different categories.

[0046] This step adopts the fuzzy XYZ classification method combined with the ABC classification method. First, the key factors affecting inventory are classified into three types according to the purchase amount ranking, among which Class A goods are the top 20% of the goods purchase amount ranking, Class B goods are the bottom 30% of the goods purchase amount ranking, and Class C goods are the remaining part of the goods purchase amount ranking.

[0047] This step sets three types of XYZ from the second dimension according to the speed of inventory consumption. Category X goods are those whose cumulative inventory consumption has reached 90% of the total, that is, the cumulative inventory consumption ratio is not less than 90%; category Y goods are those whose cumulative inventory consumption has reached between 70% and 90% of the total, that is, the cumulative inventory consumption ratio is between 70% and 90%; category Z goods are those whose cumulative inventory consumption has not reached 70% of the total, that is, the cumulative inventory consumption ratio is below 70%.

[0048] Then, the two-dimensional classifications ABC-XYZ are combined to obtain nine classifications: AX, AY, AZ, BX, BY, BZ, CX, CY, and CZ. The specific characteristics and descriptions of the corresponding goods are shown in Table 1.

[0049] Table 1: Description of ABC-XYZ classification

[0050] category Features illustrate <![CDATA[ AX ]]> <![CDATA[ The amount accounts for a high proportion and the inventory is consumed quickly ]]> <![CDATA[ Close to the exit, convenient picking operation ]]> <![CDATA[ AY ]]> <![CDATA[ The amount accounts for a high proportion, and the inventory consumption is average ]]> <![CDATA[ General operation area, ensuring inventory control ]]> <![CDATA[ AZ ]]> <![CDATA[ The amount accounts for a high proportion, and the inventory consumption is slow ]]> <![CDATA[ Inside the warehouse, routine inspection ]]> <![CDATA[ BX ]]> <![CDATA[ The amount accounts for a general proportion, and the inventory is consumed quickly ]]> <![CDATA[ Close to the exit, frequent inspection ]]> <![CDATA[ BY ]]> <![CDATA[ The amount accounts for an average proportion, and the inventory consumption is average ]]> <![CDATA[ General work area, routine inspection ]]> <![CDATA[ BZ ]]> <![CDATA[ The amount accounts for an average proportion, and inventory consumption is slow ]]> <![CDATA[ Inside the warehouse, routine inspection ]]> <![CDATA[ CX ]]> <![CDATA[ The amount accounts for a low proportion, and the inventory is consumed quickly ]]> <![CDATA[ Close to the exit, frequent inspection ]]> <![CDATA[ CY ]]> <![CDATA[ The amount accounts for a low proportion, and the inventory consumption is average ]]> <![CDATA[ General work area, routine inspection ]]> <![CDATA[ CZ ]]> <![CDATA[ The amount accounts for a low proportion and the inventory consumption is slow ]]> <![CDATA[ Inside the warehouse, routine inspection ]]>

[0051] The goods are classified by the above dimensions, and the 9 categories are ranked by importance to support the next step of forecasting model parameter input. Based on the ABC-XYZ classification method, the two categories of goods BX and CX need to be upgraded in importance because these two categories of inventory have large consumption and are likely to increase the difficulty of inventory management. Therefore, this solution classifies these two categories into the same category as AX. The corresponding three categories of goods AZ, BZ, and CZ have the smallest consumption, so these three categories of goods should be stored in the inventory. Therefore, the order of importance of the nine categories of goods determined comprehensively is AX, BX, CX, AY, BY, CY, AZ, BZ, and CZ.

[0052] Step 3: Use time series model to predict the classified high-importance goods.

[0053] High-importance goods have the characteristics of high value or high frequency, including AX, AY, BX, BY, CX, CY, and are also the top six types of goods in terms of importance. For these six types of goods, this step uses a time series model for prediction and verifies it with historical data of high-value and high-frequency goods. The steps for prediction through a time series model include:

[0054] 1) Calculate the average value of the historical demand in the recent n periods. The formula is: M = (X 1 +X 2 +...+X n ) / n, where X n Indicates the most recent historical demand for goods, X 1It represents the first historical demand of the calculation application, and n is the number of historical demands required for calculation based on the actual outbound demand, also known as the number of cycles involved in the prediction. The greater the corresponding trend impact, the longer the time interval, that is, the larger n is, and vice versa.

[0055] 2) Calculate the time distance, the formula is: X = (X n -X 1 ) / (n-1), this formula calculates the time distance based on n historical demands, where X n represents the historical demand for goods in the most recent period, X 1 Represents the historical demand for period 1 to which the calculation applies.

[0056] 3) Considering the impact of time distance on the average value of historical demand, we have X = (X n+1 -X 1 ) / n, combine the two formulas to calculate X n+1 The value of is calculated as follows: M+X=(X 2 +X 3 +....+X n+1 ) / n,X n +1=(nX n -X 1 ) / (n-1), where X n+1 represents the forecast demand value of the forecast period, n is the number of periods involved in the forecast, X 2 , X 3 They are the historical demands for the 2nd and 3rd periods required for forecasting.

[0057] The above-mentioned time series model can calculate the predicted demand value of the forecast period, that is, the predicted value, through historical data. This forecasting model is suitable for six types of goods, AX, AY, BX, BY, CX, and CY, which have high capital occupation and fast consumption, thereby realizing the safety inventory demand forecast for the corresponding goods.

[0058] Step 4: Use the Bayesian model to predict the classified low-importance goods.

[0059] The three types of goods, AZ, BZ, and CZ, whose cumulative inventory consumption ratio is below 70%, have low capital occupation and can be stored for a long time. This step is based on historical data fitting analysis, and selects the distribution function that meets the fitting results from the classic distribution functions such as normal distribution and exponential distribution, combined with the Bayesian analysis method for prediction. This step can be performed in no particular order with step three. The specific steps of this step are as follows.

[0060] 1) Determine and verify the distribution probability function by fitting historical data.

[0061] Taking the historical data conforming to the exponential distribution as an example, the exponential distribution is selected as the prior distribution probability density function of the current type of goods, as follows:

[0062] f(λ)=λe (-λx) (0<x<∞,λ>0),

[0063] Where λ is the rate parameter of the exponential function and x is the amount of goods consumed. The variance of this function is 1 / λ 2 , the mean of this function is 1 / λ.

[0064] 2) Then determine the prior predictive distribution based on the conjugate distribution selection principle of the Bayesian method.

[0065] In this step, the conjugate distribution of Poisson distribution is selected as gamma distribution (Γ), and its shape parameter is set to α (α>0), and the scale parameter is set to β (β>0). The corresponding formula is as follows:

[0066]

[0067] q represents the consumption of goods, t is the time in the order cycle, and the obtained prior density function is:

[0068]

[0069] Among them, α'=α+q, β'=β+tn, n is the quantity required each time, and denote the means of the shape parameter and scale parameter, respectively.

[0070] The prior distribution of the occurrence of excessive consumption of goods during the forecast period is:

[0071]

[0072] Where d is the differential symbol and L represents the order cycle.

[0073] 3) Obtain the critical inventory level and predict the inventory demand.

[0074] Therefore, the maximum requirement to be achieved is to achieve a basic inventory level that meets the market demand satisfaction rate P. The critical inventory level S satisfies the following inequality:

[0075]

[0076] Among them, R(t) represents the market demand calculated based on time t. The results calculated by the above formula are combined with the inequality to finally obtain the critical inventory level S, and then the inventory demand for BZ and CZ spare parts is further predicted.

[0077] The above steps complete the safety stock forecast of dividing the goods into 9 categories, and this step drives the execution of the procurement task through the forecast results.

[0078] After obtaining the forecast results, a replenishment plan is generated based on the forecast results, and responsibilities are assigned according to the replenishment tasks, and the corresponding position manager is pushed. The manager receives the replenishment task, completes the warehousing replenishment operation through the procurement system, and classifies the goods in real time, repeating the previous identification and forecasting process.

[0079] The present invention is described above by way of example in conjunction with the accompanying drawings. It is obvious that the specific implementation of the present invention is not limited to the above-mentioned method. As long as various non-substantial improvements are made using the inventive concept and technical solution of the present invention, or the inventive concept and technical solution are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A method for identifying and predicting intelligent warehouse safety inventory, characterized by: The following steps are involved: Step 1: Collect data required for inventory identification and forecasting; Step 2: Classify demand types, and accurately classify the demand types of inventory goods to support demand forecasting for different categories; Step 3: Use time series model to predict the classified high-importance goods; Step 4: Use the Bayesian model to predict the classified low-importance goods; Step 4 and step 3 can be performed in no particular order. After obtaining the forecast results, a replenishment inventory plan is generated based on the forecast results, and the relevant person in charge completes the warehousing and replenishment operations through the procurement system.

2. The method for identifying and predicting intelligent warehouse safety inventory according to claim 1, characterized in that: In step 1, the required data include inventory data, purchase cost, order data and transportation time. The processed inventory data comes from the intelligent warehousing system, and the purchase cost, order data, transportation time and other data come from the order center, transportation center and upstream ERP data center respectively.

3. The method for identifying and predicting intelligent warehouse safety inventory according to claim 1, characterized in that: In the step 2, classification is first performed based on the first dimension, and the key factors affecting inventory are classified into three types from high to low according to the purchase amount ranking; Then, we classify according to the second dimension, and set three types, XYZ, according to the speed of inventory consumption from fast to slow; Then combine the classifications of the two dimensions to get nine categories: ABC and XYZ are combined into nine categories: AX, AY, AZ, BX, BY, BZ, CX, CY, CZ; rank the 9 categories in order of importance, categories with high purchase amount ranking are of high importance, and categories with large inventory consumption and easy to increase inventory management difficulty are of high importance. After sorting, the order of importance of the 9 categories of goods is AX, BX, CX, AY, BY, CY, AZ, BZ, CZ.

4. The method for identifying and predicting intelligent warehouse safety inventory according to claim 3, characterized in that: In step 2, the characteristics of the 9 categories are as follows: AX: The amount accounts for a high proportion, the inventory is consumed quickly, the storage is close to the exit, and the picking operation is convenient; AY: The amount accounts for a high proportion, the inventory consumption rate is average, and it is stored in the general operation area to ensure inventory control; AZ: high proportion of amount, slow inventory consumption, stored inside the warehouse, regular inspection; BX: The amount accounts for a normal proportion, the inventory is consumed quickly, it is stored close to the exit, and it is checked frequently; BY: The amount accounts for a normal proportion, the inventory consumption rate is normal, it is stored in a normal operation area, and it is routinely inspected; BZ: The amount accounts for a normal proportion, the inventory is consumed slowly, stored inside the warehouse, and undergoes regular inspections; CX: Low amount, fast inventory consumption, storage close to the exit, frequent inspection; CY: low proportion of amount, average inventory consumption rate, stored in general operation area, routine inspection; CZ: The amount accounts for a low proportion, the inventory is consumed slowly, it is stored inside the warehouse, and it is inspected regularly.

5. The method for identifying and predicting intelligent warehouse safety inventory according to claim 1, characterized in that: In step three, high-importance goods have the characteristics of high value or high frequency, including six types of goods: AX, AY, BX, BY, CX, and CY. The steps of predicting safety stock through the time series model include: 1) calculating the average value of historical demand in the past n periods, 2) calculating the time distance, and 3) considering the impact of time distance on the average value of historical demand, and combining the two formulas to calculate the predicted value of safety stock demand.

6. A method for identifying and predicting intelligent warehouse safety inventory according to claim 5, characterized in that: The steps of predicting safety stock through time series model include: 1) The formula for calculating the average value of the historical demand in the recent n periods is: M = (X1 + X2 + ... + X n ) / n, where X n Indicates the most recent historical demand for goods, X1 indicates the first historical demand for calculation application, and n is the number of historical demands required for calculation based on actual outbound demand, also known as the number of cycles involved in the forecast. The greater the impact of the corresponding trend, the longer the time interval, that is, the larger n is, and vice versa; 2) The formula for calculating time distance is: X = (X n -X1) / (n-1), this formula calculates the time distance based on n historical demands, where X n represents the historical demand for the goods in the most recent period, and X1 represents the historical demand in the first period to which the calculation is applied; 3) The calculation formula for the forecast value of safety stock demand is as follows: M+X=(X2+X3+....+X n+1 ) / n,X n +1=(nX n -X1) / (n-1), where X n+1 It represents the forecast demand value of the forecast period, n is the number of periods involved in the forecast, X2 and X3 are the historical demands of the second and third periods required for the forecast respectively.

7. The method for identifying and predicting intelligent warehouse safety inventory according to claim 5, characterized in that: In step 4, for the three types of goods of low importance, AZ, BZ, and CZ, after historical data fitting analysis, a distribution function that meets the fitting results is selected and predicted in combination with the Bayesian analysis method. The specific steps are as follows: 1) Determine and verify the distribution probability function by fitting historical data; 2) Determine the prior prediction distribution according to the conjugate distribution selection principle of the Bayesian method; 3) Obtain the critical inventory level and predict the inventory demand.

Images