E-commerce platform order analysis system and method based on data mining
By classifying e-commerce platform orders in both forward and reverse directions and mining their features and influencing parameters, the problem of existing systems failing to uncover information on unsold orders has been solved, enabling multi-dimensional and dynamic product sales forecasting and strategy formulation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JIANGSU OPEN UNIVERSITY (THE CITY VOCATIONAL COLLEGE OF JIANGSU)
- Filing Date
- 2026-01-30
- Publication Date
- 2026-05-15
AI Technical Summary
Existing e-commerce platform order analysis systems fail to effectively uncover potential information in unsold orders, resulting in highly limited analysis results that cannot adapt to the rapid changes in the e-commerce market and affect merchants' ability to formulate future product sales strategies.
By classifying historical orders into positive and negative orders through the order classification module, their characteristics are extracted, and combined with cyclical characteristics and demand impact parameters, future commodity demand is predicted, and adjustment strategies are formulated.
It enables cross-temporal and multi-dimensional order information mining, provides dynamic product sales forecasts, and helps merchants adapt to market changes and formulate effective sales strategies on e-commerce platforms.
Smart Images
Figure CN121616379B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of business analytics, specifically to an e-commerce platform order analysis system and method based on data mining. Background Technology
[0002] E-commerce platforms store a large number of orders, which contain users' shopping behavior. By analyzing the shopping behavior in these orders, merchants can uncover potential sales information and use it as a data foundation for making sales and profits in the future, thereby ensuring that merchants can operate healthily and efficiently within the e-commerce platform.
[0003] Patent application CN119090542A discloses a data mining-based order analysis method and system for e-commerce platforms. Through a multi-layered data lake architecture combined with Lambda, it can extract structured and semi-structured data from the raw data area for large-scale analysis. The system's ability to efficiently identify and process abnormal order data is achieved through the training of an autoencoder anomaly detection model and precise threshold settings. The combination of a product recommendation system and an inventory management system significantly improves sales and inventory turnover, maximizing cost-effectiveness. Continuous acquisition and analysis of order data enables rapid adjustment of product prices, ensuring that pricing strategies always align with the current market environment.
[0004] Existing order analysis systems typically focus only on and analyze completed orders, and perform product sales trend and demand forecasting based on these orders. This approach ignores the potential information contained in uncompleted orders, resulting in a relatively singular dimension for the mining and analysis of potential information. Consequently, both the analyzed potential information and the forecast results are limited. Furthermore, it fails to effectively combine diverse features from the past and future, making the analysis and forecast results relatively static. This makes it difficult for merchants to adapt to the rapidly changing dynamic environment of the e-commerce market and hinders them from accurately formulating product sales strategies for future moments. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides an e-commerce platform order analysis system and method based on data mining.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] An e-commerce platform order analysis system based on data mining, comprising:
[0008] The order classification module is used to extract the product names of goods in historical orders within the previous analysis period, and import historical orders with the same product name into the empty spaces of the blank queue in order to generate an order queue;
[0009] The feature mining module is used to identify the order attributes of historical orders in the order queue, divide historical orders into forward orders and reverse orders according to the order attributes, and mine the forward features of forward orders and the reverse features of reverse orders respectively.
[0010] The demand forecasting module is used to determine the cyclical characteristics of the next analysis period, integrate the cyclical characteristics, positive characteristics, and negative characteristics into comprehensive order data, and input the comprehensive order data into the commodity demand model to predict the commodity demand value for the next analysis period.
[0011] The demand calculation module is used to extract demand impact parameters from forward and reverse orders, and calculate the actual demand value for the next analysis period based on the demand impact parameters and the product demand value.
[0012] The demand analysis module compares the inventory balance and planned increase values for the next analysis period with the actual demand values to determine the demand status of goods in the next analysis period and formulate corresponding product adjustment strategies.
[0013] To optimize the above technical solution, the specific measures also include:
[0014] Furthermore, the generated order queue specifically refers to:
[0015] Establish A basic queues with two parallel queue lines, and set B consecutive empty spaces between the two queue lines to obtain A empty queues;
[0016] Query the generation time of all historical orders one by one, and import historical orders with the same product name into the empty space of the blank queue in chronological order, and assign a unique number to each of the A product names.
[0017] Create blank queue frames at the head of each of the A blank queues, and add the product name number to each queue frame to convert the blank queues into order queues, thus obtaining A order queues.
[0018] Furthermore, the order attributes include profit from sales and profit from unsold items;
[0019] The specific steps for classifying historical orders into forward and reverse orders based on order attributes are as follows:
[0020] According to the order number in ascending order, the order management system is used to query the item names of historical orders in the A order queue one by one, and the order items with the item names of sales status and profit status are recorded as sales items and profit items respectively.
[0021] When the text of the sales status in the sales item is "sold" and the number of the profit status in the profit item is positive, the order attribute is "sales profit" and the historical order is recorded as a positive order;
[0022] When the text of the sales status in the sales item is "sold" and the number of the profit status in the profit item is a non-positive number, the order attribute is "unsold profit" and the historical order is recorded as a reverse order;
[0023] When the sales status text in a sales project is "not sold", the historical order will be recorded as a reverse order.
[0024] Furthermore, the positive characteristics of the positive orders include the unit price fluctuation range, discount level, and promotional exposure frequency;
[0025] The reverse characteristics of the reverse orders include the percentage increase in freight costs and the duration of delays.
[0026] Furthermore, the periodic characteristics include time period type value and market activity; the method for determining the time period type value is as follows: multiply the average daily sales peaks in the next analysis period by the corresponding preset expansion factor to generate holiday values, and then sum up all the holiday values in the next analysis period to calculate the time period type value.
[0027] Furthermore, the commodity demand model is based on LSTM and is obtained by training on comprehensive order data and corresponding commodity demand values.
[0028] Furthermore, the demand-influencing parameters include the equilibrium score and the return / exchange rate;
[0029] The specific process of extracting the equilibrium score value from positive and negative orders is as follows:
[0030] Mark the generation time of all forward and reverse orders one by one, and record the first generation time and the last generation time as the start time and end time in chronological order;
[0031] Retrieve the logistics score, quality score, and service score for product A at the start and end times respectively;
[0032] The logistics score, quality score, and service score at the end time are subtracted from the logistics score, quality score, and service score at the beginning time, respectively. The differences are then assigned to the corresponding weighting factors and summed to obtain the balanced score.
[0033] Furthermore, the calculation of the actual demand value for the next analysis cycle is specifically as follows:
[0034] Multiply the balanced score by the calibrated unit score to obtain the first influence value, and multiply the return / exchange rate by the calibrated unit return / exchange volume to obtain the second influence value.
[0035] The demand impact value is calculated by adding the first impact value and the second impact value, and the actual demand value is calculated by subtracting the commodity demand value from the demand impact value.
[0036] Furthermore, determining the commodity demand status for the next analysis period specifically involves:
[0037] Add the remaining inventory value and planned increase value of product A in the next analysis period, and then calculate the difference value of product A by subtracting it from the actual demand value of product A.
[0038] When the difference between the goods and the inventory limit is greater than or equal to the limit, the demand for the goods is in a state of abundance.
[0039] When the difference between the goods is greater than or equal to 0 and less than the upper limit of inventory, the demand for goods is in a balanced state.
[0040] When the difference in commodity demand is less than 0, the commodity demand is in a state of shortage.
[0041] The specific steps for formulating the corresponding product adjustment strategy are as follows:
[0042] When the demand for goods is abundant, formulate a strategy to reduce inventory.
[0043] When the demand for goods is in equilibrium, a strategy of maintaining the status quo is adopted.
[0044] When the demand for a product is in a state of shortage, a strategy to increase inventory is formulated.
[0045] This invention also proposes a data mining-based method for analyzing e-commerce platform orders, implemented through the system described above, comprising the following steps:
[0046] S1. Extract the product names from historical orders in the previous analysis period, and import historical orders with the same product names into the empty spaces of the blank queue in order to generate an order queue;
[0047] S2. Identify the order attributes of historical orders in the order queue, divide historical orders into forward orders and reverse orders according to the order attributes, and mine the forward features of forward orders and the reverse features of reverse orders respectively.
[0048] S3. Determine the cycle characteristics of the next analysis period, integrate the cycle characteristics, positive characteristics, and negative characteristics into comprehensive order data, and input the comprehensive order data into the commodity demand model to predict the commodity demand value for the next analysis period.
[0049] S4. Extract demand impact parameters from positive and negative orders, and calculate the actual demand value for the next analysis period based on the demand impact parameters and the product demand value.
[0050] S5. Compare the inventory balance and planned increase value for the next analysis period with the actual demand value to determine the commodity demand status for the next analysis period and formulate corresponding commodity adjustment strategies.
[0051] The beneficial effects of this invention are:
[0052] (1) By dividing historical orders into positive orders and negative orders and mining positive and negative features, this invention can effectively distinguish between the successful sales profits and unsuccessful sales profits of merchants in the past time period. At the same time, it can effectively mine and analyze the potential sales information of goods in historical orders from the positive and negative dimensions. This achieves the effect of order mining and analysis in two parallel dimensions, effectively avoiding the limitations of single-dimensional order mining and analysis methods, and improving the comprehensiveness of mining and analysis of potential information in orders.
[0053] (2) By collecting the periodic features of the next analysis period and combining them with positive and negative features to predict the demand value of goods, this invention can collect features that affect the actual demand quantity of goods in past and future time periods across time periods, realizing the dynamic feature mining effect along the time line. At the same time, by combining the demand influence parameters to calculate the actual demand value, it can provide a diversified data foundation for setting the actual demand quantity of goods in future time periods, thereby realizing the cross-time and diversified analysis effect of potential information in orders, which makes it easier for merchants to formulate plans and strategies for the sale of goods in future time periods, and meets the sales needs of merchants on e-commerce platforms. Attached Figure Description
[0054] Figure 1 This is a schematic diagram of the modules of the e-commerce platform order analysis system based on data mining provided in Embodiment 1 of the present invention;
[0055] Figure 2 This is a flowchart illustrating the data mining-based order analysis method for e-commerce platforms provided in Embodiment 2 of the present invention. Detailed Implementation
[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0057] Example 1
[0058] This invention proposes an e-commerce platform order analysis system based on data mining, such as... Figure 1 As shown, it includes:
[0059] The order classification module extracts the product names of historical orders from the previous analysis period one by one, establishes an empty queue corresponding to the product names, and imports the historical orders into the empty queue in order to generate the order queue.
[0060] The analysis period is the time frame for analyzing and mining the data contained in orders within an e-commerce platform. This allows the analysis period to serve as a time limit for order collection in the past and also as a time limit for the specific order demand and changes in the future of the e-commerce platform.
[0061] In this embodiment, the specific duration of the analysis period is not a fixed value. The specific duration of the analysis period can be customized according to the user's actual needs, or it can be set according to the interval between two adjacent order analyses within the e-commerce platform.
[0062] It should be noted that in order to ensure that there are an appropriate number of orders within the analysis period to support subsequent analysis needs, the specific duration of the analysis period should not be too short or too long. For example, the specific duration of the analysis period is 15 days, 30 days, 45 days, and 60 days.
[0063] Once the specific duration of the analysis cycle is determined, the end point of the time period corresponding to the previous analysis cycle is the current moment, and the start point of the time period is the moment after counting back one analysis cycle along the timeline.
[0064] Historical orders are orders for goods purchased by users that were generated and collected in the order database by the e-commerce platform during the previous analysis period. This means that the purchase transaction information generated by the e-commerce platform and users during the previous analysis period can be recorded and stored.
[0065] Product names are used to represent the names of specific products purchased by users in historical orders. They can serve as a direct basis for classifying historical orders. Since there are a wide variety of products sold on e-commerce platforms, by determining the product names in historical orders, the products sold and traded on the e-commerce platform during the analysis period can be directly represented.
[0066] Specifically, when extracting product names from historical orders, the order management system is used to query all historical orders for which the item name is the product name, and the text in the order item is recorded as the product name, thus obtaining the product names of A products.
[0067] It should be noted that a single historical order may contain multiple different product names, and multiple historical orders may also contain the same product name. Therefore, the number of product names may be greater than, equal to, or less than the number of historical orders.
[0068] After obtaining the product names of historical orders, the historical orders can be classified based on the product names, and historical orders with the same product name can be aggregated to form a corresponding order queue.
[0069] Specifically, the method for generating the order queue is as follows:
[0070] Create a basic queue A with two parallel queue lines, and set B consecutive empty spaces between the two queue lines to obtain A blank queues; the blank queue is a queue that has not imported any historical orders, and the empty space is the basic unit that makes up the blank queue and provides a unique position constraint for the import of subsequent historical orders;
[0071] Query the generation time of all historical orders one by one, and import historical orders with the same product name into the empty space of the blank queue in chronological order, and assign a unique number to each of the A product names.
[0072] Create blank queue frames at the head of each of the A blank queues, and annotate the queue frames with the product name number. This will convert the blank queues into order queues, resulting in A order queues. The head of the queue refers to the end of the blank queue containing the first empty space, and serves as the starting position for importing historical orders into that space.
[0073] It should be noted that converting product names into numbers allows each order queue to be assigned a unique number, facilitating subsequent management and sorting of the A order queues.
[0074] The feature mining module identifies the order attributes of historical orders in the order queue, divides historical orders into forward orders and reverse orders, and mines the forward features of forward orders and the reverse features of reverse orders respectively.
[0075] Order attributes are used to indicate whether historical orders were successfully sold and generated profits in actual sales on e-commerce platforms; that is, they can be used to represent the actual sales and profit situation of historical orders.
[0076] Because e-commerce platforms have certain rules regarding returns and exchanges, historical orders may result in cancellations, returns with refunds, or refunds only after they are generated. In order to accurately distinguish between historical orders that were successfully sold and those that were not, it is necessary to identify the order attributes of historical orders and, based on the identification results, classify historical orders into positive orders and negative orders.
[0077] Order attributes include profit from sales and profit from unsold orders; where profit from sales indicates that the merchant successfully sold the goods and made a profit in this historical order; profit from unsold orders indicates that the merchant did not successfully sell the goods and made a profit in this historical order.
[0078] In this embodiment, a positive order is a historical order in which the merchant has successfully sold and profited, while a negative order is a historical order in which the merchant has not successfully sold and profited.
[0079] The method for distinguishing between forward and reverse orders is as follows:
[0080] According to the order number in ascending order, the order management system is used to query the item names of historical orders in the A order queue one by one, and the order items with the item names of sales status and profit status are recorded as sales items and profit items respectively.
[0081] When the text for the sales status in the sales item is "sold" and the number for the profit status in the profit item is positive, it means that the goods in the historical order have been successfully sold and generated a profit. The order attribute is recorded as "sales profit" and the historical order is recorded as a positive order.
[0082] When the text for the sales status in the sales item is "sold" and the number for the profit status in the profit item is a non-positive number, it means that the goods in the historical order have been successfully sold but no profit has been made. The order attribute is recorded as "unsold but profitable" and the historical order is recorded as a reverse order.
[0083] When the sales status text in a sales item is "not sold," it means that the goods in that historical order were not successfully sold. The order attribute is recorded as "unsold and unprofitable," and the historical order is recorded as a reverse order.
[0084] By dividing orders into forward and reverse orders as described above, all historical orders in order queue A can be accurately and quickly classified into forward and reverse orders. This provides two parallel dimensions of analysis for e-commerce merchants' sales profitability on the e-commerce platform and provides a basis for the final inventory and stocking status of the product.
[0085] In order to accurately understand the specific sales and profit situation of the product in the previous analysis period, it is necessary to analyze the relevant data contained in the positive and negative orders to obtain positive and negative features that can comprehensively represent the sales and profit situation in multiple dimensions.
[0086] Positive features refer to the actual sales profit of goods in positive orders. Specifically, positive features include the fluctuation range of unit price, the degree of discount, and the frequency of promotional exposure.
[0087] The unit price fluctuation range refers to the difference between the minimum and maximum values of the unit price of goods in positive orders in the order queue during the previous analysis period, which can be used to represent the magnitude of the unit price fluctuation range of goods.
[0088] In this embodiment, when mining the unit price fluctuation range, the order items with the item name "unit price" in the forward orders of A order queues are marked one by one. The numbers in the order items are recorded as the unit price of the product. The unit price fluctuation range of A products can be calculated by subtracting the maximum value and the minimum value of the unit price of the product.
[0089] Discount level refers to the maximum discount rate of the unit price of goods in the order queue during the previous analysis period; the higher the discount level, the more goods are sold and the more profit is made.
[0090] In this embodiment, when mining discount rates, the discount rates of positive orders in A order queues are marked one by one, and the maximum discount rate is recorded as the discount rate, thus obtaining A discount rates.
[0091] Promotion exposure frequency refers to the ratio of the duration of promotion exposure of the unit price of the product in the positive order in the order queue to the duration of the previous analysis period. The higher the promotion exposure frequency, the more the product is sold and the more profit is made.
[0092] In this embodiment, when mining the promotion exposure frequency, the number of days that the product name in each of the A order queues was in the promotion exposure state in the previous analysis period is queried one by one and recorded as the promotion days. The promotion days are then compared with the duration of the analysis period to obtain the A promotion exposure rates.
[0093] Reverse characteristics refer to the actual sales profit of goods in reverse orders. Specifically, reverse characteristics include the percentage increase in freight costs and the length of time the goods are held in inventory.
[0094] The freight increase ratio refers to the percentage of reverse orders in the order queue whose logistics transportation costs exceed the standard transportation costs, compared to the total number of orders. It can be used to represent the changes in logistics transportation costs. When the freight increase ratio is higher, the sales volume of the goods is lower, and the profit is also lower.
[0095] In this embodiment, when identifying the percentage increase in shipping costs, the order management system is used to query the unit weight transportation cost of reverse orders in A order queues one by one, and the number of transportation costs exceeding the standard cost threshold is counted to obtain A increase values. These A increase values are then compared with the total number of reverse orders in the A order queues to obtain A percentage increases in shipping costs. The standard cost threshold refers to the standard value of the unit weight transportation cost originally agreed upon by the merchant, e-commerce platform, and logistics provider, and can be used as the basis for subsequent judgments on whether the unit weight transportation cost has increased or decreased.
[0096] The dwell time refers to the time span between the entry and exit of goods in the reverse order in the order queue. It can be used as a representation of the time that goods remain in the warehouse. The longer the dwell time, the less the goods are sold and the less profit is made.
[0097] In this embodiment, when mining the dwell time, the order management system queries the inbound and outbound times of the reverse orders in A order queues one by one. The time between the inbound and outbound times is recorded as the dwell time. The dwell times of all A order queues are summed and averaged to obtain A dwell time.
[0098] It should be noted that positive features refer to the relevant features of positive orders in the order queue, while negative features refer to the relevant features of negative orders in the order queue. This allows for comprehensive collection of both positive and negative feature data from historical orders across two dimensions, achieving the effect of analyzing historical orders from two parallel dimensions.
[0099] The demand forecasting module determines the cyclical characteristics of the next analysis period, integrates the cyclical characteristics, positive characteristics, and negative characteristics into comprehensive order data, and inputs the comprehensive order data into the commodity demand model to predict the commodity demand value for the next analysis period.
[0100] After identifying the positive and negative features of historical orders from the previous analysis period, these features can be used as two bases for determining the sales volume of goods on the e-commerce platform in the next analysis period, and provide an analytical foundation for predicting the sales results of goods in the next analysis period.
[0101] Cyclical characteristics refer to the relevant features that will affect the sales volume of goods when the time corresponding to the next analysis period is reached. In other words, they can be used to comprehensively represent factors such as the number of days and dates that actually correspond to the next analysis period.
[0102] Specifically, cyclical characteristics include time period type values and market activity;
[0103] The time period type value refers to the numerical result of the peak sales days included in the product sales time range within the next analysis period. The more peak sales days there are, the more the product is sold and the more profit is generated.
[0104] In this embodiment, peak sales days include, but are not limited to, various statutory holidays, non-statutory holidays, and e-commerce platform activity days; for example, peak sales days are May Day, National Day, Qixi Festival, and Double Eleven shopping festival.
[0105] After determining the peak sales days for the next analysis period, it is also necessary to represent the sales activity level of the products corresponding to each peak sales day. When representing the sales activity level, the number of holiday days for each peak sales day is multiplied by the corresponding preset multiple based on the specific date of each peak sales day.
[0106] Specifically, when the peak sales day is the National Day holiday, the holiday value for the National Day holiday is obtained by multiplying the number of holiday days for the National Day holiday by the corresponding preset expansion multiple.
[0107] For example, when the National Day holiday is 7 days long and the preset expansion multiple for National Day is 4, the holiday value for National Day is 28.
[0108] After obtaining all holiday values for the next analysis period, all holiday values can be summed up to calculate the time period type value.
[0109] Market activity refers to the level of activity of e-commerce platforms in promoting products through marketing methods in the next analysis period. It can be used to represent the sales performance of products in the next analysis period due to policy benefits. The higher the market activity, the more products are sold and the more profits are generated.
[0110] In this embodiment, when determining market activity, the sales incentive policies formulated by the e-commerce platform in the next analysis period are queried out. Sales incentive policies containing product names are recorded as valid policies. The number of valid policies is compared with the number of sales incentive policies to obtain A market activity levels.
[0111] After determining the time period type value and market activity level for the next analysis period, the time period type value and market activity level can be merged and summarized with the positive and negative features mined in the previous analysis period. This will result in the time period type value, market activity level, unit price fluctuation range, discount level, promotion exposure frequency, freight increase rate, and dwell time being merged and summarized into A comprehensive order data.
[0112] It should be noted that a single comprehensive order data set contains only one value for time period type, market activity, unit price fluctuation range, discount, promotion exposure frequency, shipping cost increase rate, and dwell time. Therefore, by integrating the time period type value with A values for market activity, unit price fluctuation range, discount, promotion exposure frequency, shipping cost increase rate, and dwell time, the required A comprehensive order data sets can be obtained.
[0113] After obtaining the comprehensive order data, it can be imported into the pre-trained product demand model. Based on the comprehensive order data, the product demand value of product A in the next analysis period can be predicted.
[0114] The commodity demand model is based on a machine learning model. It is obtained by massive training and optimization based on a large amount of comprehensive order data and corresponding commodity demand values. It can predict the commodity demand value in the next analysis period based on the positive and negative features of the previous analysis period and the periodic features of the next analysis period. The machine learning model can be LSTM (Long Short-Term Memory), MLP (Multilayer Perceptron), GRU (Gated Recurrent Unit), Random Forest, etc., and can be selected as needed.
[0115] In this embodiment, the commodity demand value refers to the quantity of commodities that meet the e-commerce platform's external sales demand within the analysis period. The commodity demand value is obtained by querying the actual sales volume of commodities in all historical orders within an analysis period through the order management system.
[0116] When training a product demand model, it is necessary to collect multiple sets of comprehensive order data and product demand values in advance.
[0117] Each set of comprehensive order data is labeled as a training feature, and the corresponding product demand value for each set of training features is labeled.
[0118] The labeled training features are divided into a training set and a test set. 70% of the training features are used as the training set and 30% of the training features are used as the test set. The training set is used to train the product demand model, and the test set is used to test the product demand model.
[0119] A preset error threshold is set. When the mean of the prediction errors of all training features in the test set is less than the error threshold, a commodity demand model is obtained.
[0120] After obtaining the product demand model, the data of A comprehensive orders are imported into the product demand model one by one to predict the product demand value of A products in the next analysis period.
[0121] The demand calculation module extracts demand impact parameters from forward and reverse orders, calculates the demand impact value, and combines the product demand value with the demand impact value to calculate the actual demand value for the next analysis period.
[0122] Demand impact parameters refer to parameters in forward and reverse orders that may have a positive or negative impact on subsequent product sales. Thus, under the multiple influences of demand impact parameters, the product demand value of product A in the next analysis period will be affected.
[0123] Specifically, the demand-influencing parameters include the equilibrium score and the return / exchange rate.
[0124] The equilibrium score refers to the sum of the fluctuations in the system scores of logistics, quality and service for product A in the previous analysis period, and is one of the factors affecting the actual sales volume of the product in the next analysis period.
[0125] Specifically, when mining the equilibrium score value, the generation time of all positive and negative orders is marked one by one. The first generation time and the last generation time are recorded as the start time and the end time. The logistics score, quality score and service score of A products at the start time and the end time are queried respectively. The logistics score, quality score and service score at the end time are subtracted from the logistics score, quality score and service score at the start time, respectively. The difference is assigned to the corresponding weight factor and then added together to obtain A equilibrium score values.
[0126] The formula for calculating the balanced score is:
[0127] ;
[0128] In the formula, To balance the score, The logistics score at the end of the process. The quality score at the termination time. The service rating at the time of termination. The logistics score at the start time. The quality score at the start time. Rate the service at the start time. , , These are weight factors that are greater than 0, and .
[0129] The return / exchange ratio refers to the ratio between the number of returned / exchanged orders for product A in the previous analysis period and the total number of orders. When the return / exchange ratio is higher, the proportion of returns / exchanges for the product will increase in the next analysis period, and the demand for the product will decrease.
[0130] In this embodiment, when mining the return and exchange rate, the return and exchange status of all historical orders is queried through the order management system. Historical orders with the return and exchange status of return and exchange are recorded as return and exchange orders. The number of return and exchange orders for product A is counted. After comparing the number of return and exchange orders with the number of historical orders, the return and exchange rate of product A is calculated.
[0131] The demand impact value refers to the specific numerical value that affects the demand value of a commodity in the next analysis period under the influence of the demand impact parameter, thus serving as the basis for the actual demand value of the commodity in the next analysis period.
[0132] Specifically, when calculating the demand impact value, the equilibrium score value of product A is multiplied by the calibrated unit score to obtain A first impact values. The return / exchange rate is multiplied by the calibrated unit return / exchange volume to obtain A second impact values. The demand impact value of product A is then calculated by adding the A first impact values and the A second impact values. The calibrated unit score refers to the calibrated unit value of the equilibrium score's impact on the demand value, and can be used as the data basis for calculating the first impact value. The calibrated unit return / exchange volume refers to the calibrated unit value of the return / exchange rate's impact on the demand value, and can be used as the data basis for calculating the second impact value. Both the calibrated unit score and the calibrated unit return / exchange volume are less than the product's demand value.
[0133] The formula for calculating the impact of demand is:
[0134] ;
[0135] In the formula, This represents the impact of demand. To balance the score, The percentage of returns and exchanges. For the calibrated unit score, This refers to the specified unit return / exchange quantity.
[0136] Actual demand value refers to the actual quantity of goods that meet sales demand in the next analysis period, which can be used as a guide for merchants' actual inventory in the next analysis period.
[0137] Specifically, the formula for calculating the actual demand value is as follows:
[0138] ;
[0139] In the formula, This is the actual demand value. This represents the demand value for the product.
[0140] The demand analysis module determines the inventory balance and planned increment for the next analysis period. By comparing the actual demand, inventory balance, and planned increment, it determines the product demand status for the next analysis period and formulates product adjustment strategies corresponding to the product demand status.
[0141] Inventory balance refers to the amount of a product in a merchant's warehouse during the next analysis period on the e-commerce platform, and serves as one of the bases for influencing a merchant's decision to purchase that product.
[0142] The planned incremental value refers to the quantity of the product that a merchant plans to purchase and put into inventory in the next analysis period on the e-commerce platform, and serves as one of the bases for influencing the merchant's purchase of the product.
[0143] In this embodiment, both the inventory balance and the planned increase value are obtained by querying the merchant's database.
[0144] After obtaining the inventory balance and planned increase values, the inventory balance, planned increase values and actual demand values can be combined and compared, and the commodity demand status in the next analysis period can be determined based on the results of the combination comparison.
[0145] Product demand status is used to indicate whether a product can meet sales demand in the next analysis period;
[0146] Specifically, the demand status of goods includes a shortage state, a balance state, and a surplus state. Among them, a shortage state means that the goods cannot meet the sales demand in the next analysis period, a balance state means that the goods can meet the sales demand in the next analysis period, and a surplus state means that the goods exceed the sales demand in the next analysis period.
[0147] The method for determining the demand status of a commodity is as follows:
[0148] Add the remaining inventory value and planned increase value of each product A in the next analysis period, and then calculate the difference between the difference value of each product A and the actual demand value corresponding to each product A.
[0149] The formula for calculating the difference in commodity value is:
[0150] ;
[0151] In the formula, The difference in value between goods. This represents the remaining inventory value. This is the planned incremental value;
[0152] When the difference between the inventory and sales volume is greater than or equal to the inventory limit, it indicates that the inventory will exceed the actual sales demand in the next analysis period. At this time, the quantity of the inventory is large, and the demand status of the inventory is full. The inventory limit refers to the maximum quantity of inventory that the warehouse can hold in the next analysis period under normal circumstances, and serves as a direct basis for judging whether the warehouse inventory exceeds the limit. The specific value of the inventory limit can be customized according to actual needs.
[0153] When the difference between the quantity and the inventory is greater than or equal to 0 and less than the upper limit of inventory, it means that the product can meet the actual sales demand in the next analysis period. At this time, the quantity of the product is relatively balanced, and the product demand status is balanced.
[0154] When the difference in product value is less than 0, it means that the product cannot meet the actual sales demand in the next analysis period. At this time, the quantity of the product is insufficient, and the product demand status is a shortage state.
[0155] Product adjustment strategies are specific solutions to address different demand states for products, ensuring that each product can meet sales demand in the next analysis cycle and avoiding product shortages or oversupply.
[0156] Specifically, commodity adjustment strategies include inventory reduction strategies, inventory maintenance strategies, and inventory increase strategies.
[0157] When formulating a product adjustment strategy, if the demand for a product is abundant, and the product is likely to have excessive inventory in the next analysis period, then an inventory reduction strategy should be formulated.
[0158] When the demand for a product is in a balanced state, the product will show a balanced inventory in the next analysis period, and a strategy of maintaining the status quo will be formulated.
[0159] When the demand for a product is in a shortage state, the product will be in short supply in the next analysis period, and an inventory increase strategy will be formulated.
[0160] After formulating the corresponding product adjustment strategy, merchants can adjust the actual inventory quantity of products in the next analysis cycle according to the specific strategy content, thereby achieving dynamic balance in the sales performance of each product and ensuring that merchants can maintain a healthy and efficient product sales status in the next analysis cycle of the e-commerce platform.
[0161] Example 2
[0162] Please see Figure 2As shown, parts not described in detail in this embodiment are described in Embodiment 1. A data mining-based e-commerce platform order analysis method is provided, implemented through a data mining-based e-commerce platform order analysis system, including:
[0163] S1. Extract the product names from historical orders in the previous analysis period, and import historical orders with the same product names into the empty spaces of the blank queue in order to generate an order queue;
[0164] S2. Identify the order attributes of historical orders in the order queue, divide historical orders into forward orders and reverse orders according to the order attributes, and mine the forward features of forward orders and the reverse features of reverse orders respectively.
[0165] S3. Determine the cycle characteristics of the next analysis period, integrate the cycle characteristics, positive characteristics, and negative characteristics into comprehensive order data, and input the comprehensive order data into the commodity demand model to predict the commodity demand value for the next analysis period.
[0166] S4. Extract demand impact parameters from positive and negative orders, and calculate the actual demand value for the next analysis period based on the demand impact parameters and the product demand value.
[0167] S5. Compare the inventory balance and planned increase value for the next analysis period with the actual demand value to determine the commodity demand status for the next analysis period and formulate corresponding commodity adjustment strategies.
[0168] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0169] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.
Claims
1. An e-commerce platform order analysis system based on data mining, characterized in that, include: The order classification module is used to extract the product names of goods in historical orders within the previous analysis period, and import historical orders with the same product name into the empty spaces of the blank queue in order to generate an order queue; The feature mining module is used to identify the order attributes of historical orders in the order queue, divide historical orders into positive orders and negative orders based on the order attributes, and mine the positive features of positive orders and the negative features of negative orders respectively; the order attributes include sales profit and unsold profit; the specific steps of dividing historical orders into positive orders and negative orders based on order attributes are as follows: According to the order number in ascending order, the order management system is used to query the item names of historical orders in the A order queue one by one, and the order items with the item names of sales status and profit status are recorded as sales items and profit items respectively. When the text of the sales status in the sales item is "sold" and the number of the profit status in the profit item is positive, the order attribute is "sales profit" and the historical order is recorded as a positive order; When the text of the sales status in the sales item is "sold" and the number of the profit status in the profit item is a non-positive number, the order attribute is "unsold profit" and the historical order is recorded as a reverse order; When the sales status text in a sales project is "not sold", the historical order will be recorded as a reverse order; The positive characteristics of positive orders include the unit price fluctuation range, discount level, and promotional exposure frequency; The reverse characteristics of the reverse orders include the percentage increase in freight costs and the duration of delays. The demand forecasting module is used to determine the cyclical characteristics of the next analysis period, integrate the cyclical characteristics, positive characteristics, and negative characteristics into comprehensive order data, and input the comprehensive order data into the commodity demand model to predict the commodity demand value for the next analysis period. The demand calculation module is used to extract demand impact parameters from positive and negative orders, and calculate the actual demand value for the next analysis period based on the demand impact parameters and the product demand value; the demand impact parameters include the equilibrium score value and the return / exchange rate. The specific process of extracting the equilibrium score value from positive and negative orders is as follows: Mark the generation time of all forward and reverse orders one by one, and record the first generation time and the last generation time as the start time and end time in chronological order; Retrieve the logistics score, quality score, and service score for product A at the start and end times respectively; The logistics score, quality score, and service score at the end time are subtracted from the logistics score, quality score, and service score at the beginning time, respectively. The differences are then assigned to the corresponding weighting factors and summed to obtain the balanced score. The specific calculation of the actual demand value for the next analysis cycle is as follows: Multiply the balanced score by the calibrated unit score to obtain the first influence value, and multiply the return / exchange rate by the calibrated unit return / exchange volume to obtain the second influence value. The demand impact value is calculated by adding the first impact value and the second impact value, and the actual demand value is calculated by subtracting the commodity demand value from the demand impact value. The demand analysis module compares the inventory balance and planned increase values for the next analysis period with the actual demand values to determine the demand status of goods in the next analysis period and formulate corresponding product adjustment strategies.
2. The e-commerce platform order analysis system based on data mining as described in claim 1, characterized in that, The order queue is specifically defined as follows: Establish A basic queues with two parallel queue lines, and set B consecutive empty spaces between the two queue lines to obtain A empty queues; Query the generation time of all historical orders one by one, and import historical orders with the same product name into the empty space of the blank queue in chronological order, and assign a unique number to each of the A product names. Create blank queue frames at the head of each of the A blank queues, and add the product name number to each queue frame to convert the blank queues into order queues, thus obtaining A order queues.
3. The e-commerce platform order analysis system based on data mining as described in claim 1, characterized in that, The periodic characteristics include time period type value and market activity; the method for determining the time period type value is as follows: multiply the average daily sales peaks in the next analysis period by the corresponding preset expansion multiple to generate holiday values, and then sum up all the holiday values in the next analysis period to calculate the time period type value.
4. The e-commerce platform order analysis system based on data mining as described in claim 1, characterized in that, The commodity demand model is based on LSTM and is obtained by training on comprehensive order data and corresponding commodity demand values.
5. The e-commerce platform order analysis system based on data mining as described in claim 1, characterized in that, The specific steps for determining the commodity demand status for the next analysis period are as follows: Add the remaining inventory value and planned increase value of product A in the next analysis period, and then calculate the difference value of product A by subtracting it from the actual demand value of product A. When the difference between the goods and the inventory limit is greater than or equal to the limit, the demand for the goods is in a state of abundance. When the difference between the goods is greater than or equal to 0 and less than the upper limit of inventory, the demand for goods is in a balanced state. When the difference in commodity demand is less than 0, the commodity demand is in a state of shortage. The specific steps for formulating the corresponding product adjustment strategy are as follows: When the demand for goods is abundant, formulate a strategy to reduce inventory. When the demand for goods is in equilibrium, a strategy of maintaining the status quo is adopted. When the demand for a product is in a state of shortage, a strategy to increase inventory is formulated.
6. A data mining-based order analysis method for e-commerce platforms, implemented using the system described in any one of claims 1-5, characterized in that... Includes the following steps: S1. Extract the product names from historical orders in the previous analysis period, and import historical orders with the same product names into the empty spaces of the blank queue in order to generate an order queue; S2. Identify the order attributes of historical orders in the order queue, divide historical orders into forward orders and reverse orders according to the order attributes, and mine the forward features of forward orders and the reverse features of reverse orders respectively. S3. Determine the cycle characteristics of the next analysis period, integrate the cycle characteristics, positive characteristics, and negative characteristics into comprehensive order data, and input the comprehensive order data into the commodity demand model to predict the commodity demand value for the next analysis period. S4. Extract demand impact parameters from positive and negative orders, and calculate the actual demand value for the next analysis period based on the demand impact parameters and the product demand value. S5. Compare the inventory balance and planned increase value for the next analysis period with the actual demand value to determine the commodity demand status for the next analysis period and formulate corresponding commodity adjustment strategies.