High-performance multi-factor commodity value evaluation sorting method and system

Abstract

The invention relates to the technical field of value evaluation, and discloses a high-performance multi-factor commodity value evaluation sorting method and system, and the method comprises the steps: obtaining transaction data recorded during commodity circulation; the commodities are classified, after first-level classification is obtained, re-classification is carried out, and second-level classification is obtained; respectively carrying out single-thread processing on each category in the first-level classification and the second-level classification; after each thread processes the respective batch, results are summarized to a central processing unit, final data fusion is carried out, and the commodity value is comprehensively evaluated; and generating a value sequence of the commodities according to a comprehensive evaluation result. And the performance bottleneck of traditional single-thread processing is overcome. By introducing a comprehensive scoring model and combining multi-dimensional factors such as sales volume, gross profit contribution and sales promotion activities, more comprehensive and accurate commodity sorting is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of value assessment, and in particular to a high-performance multi-factor commodity value assessment ranking method and system. Background Art

[0002] In modern retail, product display and promotion are crucial to attracting customers. With the rapid development of e-commerce platforms, consumers not only rely on product sales and user reviews when choosing products, but also hope to quickly find products that suit their needs. This places higher demands on product sorting algorithms. Traditional product sorting algorithms are usually based on a single or a few indicators, such as sales, user reviews, click-through rates, etc. Although these systems can provide basic product recommendation functions, they often seem powerless when faced with multi-dimensional data. Summary of the invention

[0003] In view of the above-mentioned problems, the present invention is proposed.

[0004] Therefore, the technical problem solved by the present invention is that the existing commodity value evaluation and ranking method relies on a single indicator and other problems.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a high-performance multi-factor commodity value evaluation and ranking method, comprising:

[0006] Obtain transaction data recorded when goods are circulated;

[0007] Classify the goods, and after obtaining the primary classification, reclassify them to obtain the secondary classification;

[0008] Performing single-thread processing on each category in the first-level classification and the second-level classification respectively; constructing multi-threaded calculation under the first-level classification and multi-threaded calculation under the second-level classification, and dividing the multi-threaded calculation into batches according to the computational load of each thread in the multi-threads, and each thread performs calculation according to the batches;

[0009] After each thread has processed its own batch, it aggregates the results to the central processor for final data fusion and comprehensive evaluation of the product value;

[0010] Based on the results of the comprehensive evaluation, a value ranking of the commodities is generated, and based on the multi-threaded calculation under the primary classification and the multi-threaded calculation under the secondary classification, a sub-ranking of the value of the commodities is generated.

[0011] As a preferred solution of the high-performance multi-factor commodity value evaluation and ranking method of the present invention, the transaction data recorded when the commodity is circulated includes recording the transaction time and amount of all the input categories in the historical records by inputting the category of the commodity;

[0012] Count the transaction time and amount in the past thirty days and the past year respectively, and generate the number of transactions and the amount of each transaction under each first-level category in the past thirty days and the past year respectively; at the same time, generate the number of transactions and the amount of each transaction under each second-level category in the past thirty days and the past year respectively;

[0013] By selecting a region, only the transaction data of the commodities in the selected region is evaluated for value, and a ranking of the commodity value evaluation in the selected region is generated.

[0014] As a preferred solution of the high-performance multi-factor commodity value evaluation and ranking method of the present invention, wherein: the first-level classification includes the classification results obtained by dividing the commodities according to commodity categories;

[0015] The secondary classification includes the classification results obtained by dividing the commodities in each primary classification according to the commodity models;

[0016] According to the classification results of the first-level classification, the value assessment of each commodity category is calculated as an independent thread, which is defined as the first-level thread;

[0017] According to the classification results of the secondary classification, the value evaluation of different models of the same category of goods is calculated as an independent thread, which is defined as a secondary thread.

[0018] As a preferred solution of the high-performance multi-factor commodity value evaluation and ranking method of the present invention, wherein: the multi-threaded calculation under the primary classification and the multi-threaded calculation under the secondary classification are constructed, for each of the primary thread and the secondary thread, a multi-factor characteristic index is generated according to the transaction data;

[0019] Calculate the gross profit contribution score for the past 30 days:

[0020]

[0021] Among them, MLF represents the gross profit contribution score of the current commodity category in the past 30 days, accounting for 25% of the total score of 100; XL 30 Indicates the sales volume of the current commodity category in the past 30 days; SJ0 indicates the current average selling price of the current commodity category; CB0 indicates the current average cost of the current commodity category; MAX(...) indicates the maximum value of the input content; i indicates the i-th day in the historical transaction record of the current category, i∈(31, I), I indicates the maximum number of record days minus 30; XL i Indicates the sales volume of the current category in the 30 days before the i-th day in the historical transaction records; SJ i Indicates the average selling price of the current category in the historical transaction records of the 30 days before the i-th day; CB iIndicates the average cost of the historical transaction records of the current category in the 30 days before the i-th day;

[0022] Calculate the sales score in the past 30 days:

[0023]

[0024] Calculate the annual cumulative sales score:

[0025]

[0026] Among them, XL 365 Indicates the cumulative sales volume of the current product category in the past 365 days; XL 365，u Indicates the cumulative sales volume of the current product category in the 365 days before the uth day in the historical transaction records, u∈(366, U), where U represents the maximum number of recorded days minus 365;

[0027] Calculate the current discount strength score:

[0028]

[0029] Among them, YJ represents the average original price of the current commodity category;

[0030] Calculate the scores of product labels and promotion policies in the past thirty days. For each first-level thread, count the scores of different types of products:

[0031]

[0032] Where j represents the index of different models of the same category of goods; J represents the number of models; CX j Indicates the points of the j-th model. If there is a promotion, it is recorded as 1, otherwise it is recorded as 0;

[0033] For each secondary thread, if the calculated model has a promotion, the product label and promotion policy score for the past thirty days are recorded as 35; if the calculated model does not have a promotion, the product label and promotion policy score for the past thirty days are recorded as 0.

[0034] As a preferred solution of the high-performance multi-factor commodity value evaluation and ranking method of the present invention, wherein: the batching of the calculation of multiple threads includes: preferentially calculating the primary thread, and after completing the calculation of the primary thread, calculating the secondary thread;

[0035] Before calculation, the batches are divided according to the amount of calculations generated by each thread in the historical calculation process and the maximum amount of calculations of the computer;

[0036] When batching, the first-level threads are batched first, and then the second-level threads are batched;

[0037] During calculation, in each level of threads, calculations are performed sequentially in the order in which the batch results are generated;

[0038] Assume: there are A threads of calculation, YSL a is the mode of the amount of computation data generated by the thread selected for the ath time during the historical computation process, MAX YSL Indicates the maximum computing capacity of the computer;

[0039] The batch partitioning constraint is expressed as:

[0040]

[0041] Among them, a represents the ath thread selected in this division process, and among threads of the same level, the selection is random; n represents the number of threads in the batch.

[0042] As a preferred solution of the high-performance multi-factor commodity value evaluation and ranking method described in the present invention, the comprehensive evaluation of the commodity value includes: for the first-level thread, through the neural network, analyzing XLF and NXLF to determine whether there are seasonal factors in the commodity, if there are seasonal factors, then paying more attention to NXLF; analyzing ZKF, CXF, XLF and NXLF to determine whether the commodity has a high premium, if there is a high premium, then reducing the attention of MLF and increasing the attention of ZKF and CXF;

[0043] The neural network includes a first neural network and a second neural network;

[0044] The first neural network comprises, an input layer: input feature vector [XLF, NXLF];

[0045] Hidden layer: 2 hidden layers, 64 neurons in each layer, using ReLU activation function;

[0046] Output layer: Output the weight factor α used to adjust the seasonal factors;

[0047] α=f MLP (XLF,NXLF)

[0048] Among them, f MLP Represents a neural network model of a multilayer perceptron; if the product is affected by seasonal factors through the neural network analysis, a weight factor α greater than 1 is generated according to the interference intensity of the seasonal factors; if the product is not affected by seasonal factors through the neural network analysis, a weight factor α equal to 1 is generated;

[0049] The second neural network includes, an input layer: input feature vector [XLF, NXLF, ZKF, CXF];

[0050] Hidden layer: 2 hidden layers, 64 neurons in each layer, using ReLU activation function;

[0051] Output layer: three output nodes β, γ and δ, used to adjust the weight factors of premium factors;

[0052] [β, γ, δ] = f MLP (XLF,NXLF,ZKF,CXF)

[0053] If the neural network analysis shows that the product has a high premium, then according to the strength of the premium interference, weight factors β and γ greater than 1 and a weight factor δ less than 1 are generated; if the neural network analysis shows that the product is not affected by the premium, β = γ = δ = 1 is generated;

[0054] For the secondary thread, there is no interference from seasonal factors, and a weight factor α=1 is directly generated; through the second neural network, nodes β, γ and δ are output to adjust the weight factor of the premium factor;

[0055] Final score for each thread:

[0056] ZZ t =MLF×δ+XLF+NXLF×α+ZKF×β+CXF×γ

[0057] Among them, ZZ t Indicates the comprehensive evaluation result of the commodity value calculated by thread t.

[0058] As a preferred solution of the high-performance multi-factor commodity value evaluation and ranking method of the present invention, wherein: the commodity value ranking includes, without any restrictions on regions, arranging the final scores of each thread in order of size, and using the arrangement order as the value ranking of the commodity category;

[0059] In the case of region restrictions, for the same product category or model q, in the restricted region, the final score of the calculation thread is ZZ q ; When there is no region restriction, the final score of the calculation thread is ZZ q 0;

[0060] Adjust the final score under regional restrictions:

[0061]

[0062] Using the adjusted final scores, arrange the items in order of size, and use the order of arrangement as the value ranking of the product category;

[0063] In the arrangement sequence, if threads with the same score appear, the threads with the same score are arranged in the same sorting position, and the sorting position of the next thread is the number of all previous threads plus one;

[0064] The value sub-ranking of the commodities includes giving priority to outputting the value ranking of the primary classification, and by selecting the category, being able to pull down and generate the value ranking of the secondary classification under the primary classification.

[0065] A high-performance multi-factor commodity value evaluation and ranking system using the method of the present invention, wherein:

[0066] The collection unit obtains the transaction data recorded when the goods are circulated;

[0067] Classification unit, classifies the goods, and after obtaining the first-level classification, it is further classified to obtain the second-level classification;

[0068] The computing unit performs single-thread processing on each category in the first-level classification and the second-level classification respectively; constructs multi-thread computing under the first-level classification and multi-thread computing under the second-level classification, and divides the multi-thread computing into batches according to the computational load of each thread in the multi-thread, and each thread performs computing respectively according to the batches;

[0069] Comprehensive evaluation unit, after each thread has processed its own batch, the results are aggregated to the central processor for final data fusion and comprehensive evaluation of the commodity value;

[0070] The ranking generating unit generates a value ranking of commodities according to the result of the comprehensive evaluation, and generates a value sub-ranking of commodities according to the multi-threaded calculation under the primary classification and the multi-threaded calculation under the secondary classification.

[0071] A computer device comprises: a memory and a processor; the memory stores a computer program, wherein: the processor implements the steps of any one of the methods of the present invention when executing the computer program.

[0072] A computer-readable storage medium stores a computer program, wherein: when the computer program is executed by a processor, the steps of any one of the methods of the present invention are implemented.

[0073] Beneficial effects of the present invention: The high-performance multi-factor commodity value assessment and ranking method provided by the present invention greatly improves the processing speed of large-scale commodity data and overcomes the performance bottleneck of traditional single-threaded processing. By introducing a comprehensive scoring model and combining multi-dimensional factors such as sales volume, gross profit contribution, and promotional activities, a more comprehensive and accurate commodity ranking is achieved. In addition, the dynamic weight adjustment mechanism can flexibly optimize the ranking results according to market changes to ensure the timeliness and rationality of the ranking. The personalized sorting function supports customized sorting by regional characteristics, which helps to formulate accurate marketing strategies. The overall solution improves the response speed and stability of the system and enhances the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0075] Figure 1 An overall flow chart of a high-performance multi-factor commodity value evaluation and ranking method provided for the first embodiment of the present invention. DETAILED DESCRIPTION

[0076] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0077] Example 1, reference Figure 1 , is an embodiment of the present invention, and provides a high-performance multi-factor commodity value evaluation and ranking method, comprising:

[0078] S1: Obtain transaction data recorded during commodity circulation.

[0079] Furthermore, the transaction data recorded during the circulation of the goods includes recording the transaction time and amount of all the input categories in the historical records by inputting the category of the goods. The transaction time and amount in the past thirty days and the past year are counted respectively, and the number of transactions and the amount of each transaction under each first-level category in the past thirty days and the past year are generated respectively; at the same time, the number of transactions and the amount of each transaction under each second-level category in the past thirty days and the past year are generated respectively.

[0080] By selecting a region, only the transaction data of the commodities in the selected region is evaluated for value, and a ranking of the commodity value evaluation in the selected region is generated.

[0081] It is worth mentioning that by counting the transaction time and amount in the past thirty days and the past year, the sales performance of the product in the short and long term can be captured. This time dimension analysis helps to identify the sales trend of the product, seasonal changes, and consumer purchasing behavior. For example, some products may show obvious sales growth during a specific season or holiday. Counting the data of the past thirty days can help capture this short-term trend, while counting the data of the past year can provide a longer-term performance reference.

[0082] S2: Classify the goods, and after obtaining the first-level classification, reclassify them to obtain the second-level classification.

[0083] The first-level classification includes classification results obtained by dividing commodities according to commodity categories; the second-level classification includes classification results obtained by dividing commodities in each first-level classification according to commodity models.

[0084] According to the classification results of the primary classification, the value evaluation of each commodity category is calculated as an independent thread, which is defined as the primary thread. According to the classification results of the secondary classification, the value evaluation of different models of commodities in the same category is calculated as an independent thread, which is defined as the secondary thread.

[0085] By classifying products into primary and secondary categories and creating independent threads for processing, computing efficiency can be greatly improved. The primary category processes a larger data set of product categories, while the secondary category performs more detailed calculations for specific models. Through this hierarchical processing method, the system can more effectively utilize computing resources in parallel processing, reduce the computing load of each thread, and thus speed up the overall data processing speed.

[0086] The first-level classification conducts an overall value assessment based on the product category, which helps to grasp the market performance of different categories of products from a macro perspective. The second-level classification goes deeper into the specific model level to conduct a more fine-grained value assessment. This two-tier classification and evaluation method can ensure a more comprehensive and accurate assessment of the product value, not only considering the overall performance at the category level, but also paying attention to the subtle differences between each model in the market.

[0087] This hierarchical classification and multi-threaded processing design makes the system highly flexible and scalable. The introduction of new products or new models only requires adding corresponding threads for processing, which will not affect the operation of other threads. As the types and quantities of goods increase, the system can continue to maintain efficient computing performance by dynamically allocating and expanding threads. The first-level thread is responsible for the value assessment of large categories of goods to ensure the accuracy of data at the macro level; the second-level thread is responsible for the micro-analysis of specific models to ensure the integrity of the classified data. Through this multi-level and multi-dimensional evaluation method, the system can more accurately reflect the actual market value of the goods.

[0088] S3: Perform single-thread processing on each category in the first-level classification and the second-level classification respectively; construct multi-thread calculation under the first-level classification and multi-thread calculation under the second-level classification, and at the same time, divide the multi-thread calculation into batches according to the computing amount of each thread in the multi-thread, and each thread performs calculation according to the batch.

[0089] The constructing of multi-threaded calculation under the primary classification and multi-threaded calculation under the secondary classification includes generating multi-factor characteristic indicators according to the transaction data for each of the primary thread and the secondary thread.

[0090] Calculate the gross profit contribution score for the past 30 days:

[0091]

[0092] Among them, MLF represents the gross profit contribution score of the current commodity category in the past 30 days, accounting for 25% of the total score of 100; XL 30 Indicates the sales volume of the current commodity category in the past 30 days; SJ0 indicates the current average selling price of the current commodity category; CB0 indicates the current average cost of the current commodity category; MAX(...) indicates the maximum value of the input content; i indicates the i-th day in the historical transaction record of the current category, i∈(31, I), I indicates the maximum number of record days minus 30; XL i Indicates the sales volume of the current category in the 30 days before the i-th day in the historical transaction records; SJ i Indicates the average selling price of the current category in the historical transaction records of the 30 days before the i-th day; CB i Indicates the average cost of the historical transaction records of the current category in the 30 days before the i-th day.

[0093] Calculate the sales score in the past 30 days:

[0094]

[0095] Calculate the annual cumulative sales score:

[0096]

[0097] Among them, XL 365 Indicates the cumulative sales volume of the current product category in the past 365 days; XL 365，u It represents the cumulative sales volume of the current product category in the 365 days before the u-th day in the historical transaction records, u∈(366,U), where U represents the maximum number of recorded days minus 365.

[0098] Calculate the current discount strength score:

[0099]

[0100] Among them, YJ represents the average original price of the current commodity category.

[0101] Calculate the scores of product labels and promotion policies in the past thirty days. For each first-level thread, count the scores of different types of products:

[0102]

[0103] Where j represents the index of different models of the same category of goods; J represents the number of models; CX j Represents the points of the j-th model. If there is a promotion, it is recorded as 1, otherwise it is recorded as 0.

[0104] For each secondary thread, if the calculated model has a promotion, the product label and promotion policy score for the past thirty days are recorded as 35; if the calculated model does not have a promotion, the product label and promotion policy score for the past thirty days are recorded as 0.

[0105] Furthermore, the batching of the multi-threaded calculations includes: firstly calculating the primary thread, and then calculating the secondary thread after the primary thread has completed the calculation. Before the calculation, the batches are divided according to the amount of calculation generated by each thread in the historical calculation process and the maximum amount of calculation of the computer.

[0106] When batching, the batching of the first-level threads is completed first, and then the second-level threads are batched. When calculating, in each level of threads, calculations are performed in sequence according to the order in which the batch results are generated.

[0107] Assume: there are A threads of calculation, YSL a is the mode of the amount of computation data generated by the thread selected for the ath time during the historical computation process, MAX YSL Indicates the maximum computing power of the computer.

[0108] The batch partitioning constraint is expressed as:

[0109]

[0110] Among them, a represents the ath thread selected in this division process, and among threads of the same level, the selection is random; n represents the number of threads in the batch.

[0111] Processing the computational load of multiple threads in batches and performing calculations in order of batches can effectively control resource competition between threads and avoid thread conflicts or deadlocks. This design helps to ensure the stability and reliability of the system under high concurrency conditions. By rationally dividing and prioritizing computing resources, the system's hardware resources, especially the parallel computing capabilities of multi-core processors, can be maximized while ensuring computing accuracy, further improving the overall performance of the system.

[0112] S4: After each thread has processed its own batch, it aggregates the results to the central processor for final data fusion and comprehensive evaluation of the product value.

[0113] The central processing unit integrates an algorithm unit that can perform large-scale operations on input data. The comprehensive evaluation of the value of the product includes, for the first-level thread, analyzing XLF and NXLF through a neural network to determine whether there are seasonal factors in the product. If there are seasonal factors, more attention will be paid to NXLF; analyzing ZKF, CXF, XLF and NXLF to determine whether the product has a high premium. If there is a high premium, the attention of MLF will be reduced and the attention of ZKF and CXF will be increased.

[0114] The neural network includes a first neural network and a second neural network.

[0115] The first neural network includes: an input layer: an input feature vector [XLF, NXLF].

[0116] Hidden layers: 2 hidden layers, 64 neurons in each layer, using ReLU activation function.

[0117] Output layer: Outputs the weight factor α used to adjust the seasonal factor.

[0118] α=f MLP (XLF,NXLF)

[0119] Among them, f MLP Represents a neural network model of a multilayer perceptron; if the product is affected by seasonal factors through the neural network analysis, a weight factor α greater than 1 is generated according to the interference intensity of the seasonal factors (here the intensity can be understood as the ratio of the two); if the product is not affected by seasonal factors through the neural network analysis, a weight factor α equal to 1 is generated.

[0120] It should be noted that for seasonal factors, if the neural network analysis shows an obvious nonlinear relationship between XLF and NXLF (such as high sales in summer), and the value of XLF / NXLF is too small or too large, it means that it is greatly affected by the season. At this time, attention should be paid to the annual sales volume. (For example, normally the value of XLF / NXLF is 1 / 12, which is a relatively stable value. If there is a seasonal factor, assuming that the high sales in summer can reach 1 / 2, it means that this is a high-selling product in summer, and the sales in other seasons are bound to be very low. At this time, value analysis must be conducted from the perspective of the whole year. At this time, because it is too affected by seasonal factors, the value of α will be higher.)

[0121] The second neural network includes: an input layer: an input feature vector [XLF, NXLF, ZKF, CXF].

[0122] Hidden layers: 2 hidden layers, 64 neurons in each layer, using ReLU activation function.

[0123] Output layer: Three output nodes β, γ and δ are used to adjust the weight factors of the premium factors.

[0124] [β, γ, δ] = f MLP (XLF,NXLF,ZKF,CXF)

[0125] If the neural network analysis shows that the product has a high premium, then weight factors β and γ greater than 1 and a weight factor δ less than 1 are generated based on the intensity of the premium interference; if the neural network analysis shows that the product is not affected by the premium, then β=γ=δ=1 is generated.

[0126] It should be noted that the excessive premium mentioned here does not mean that the gross profit is too high, but that the price-performance ratio is not high in the eyes of buyers. In this case, various promotional activities can effectively increase sales and net profits (the reflection in the characteristics is that the scores of ZKF and CXF are high, and the annual sales volume is not high, but the sales volume in the past thirty days is high due to the promotion plan. At this time, the values ​​of XLF / NXLF will generally increase, but not as high as those of seasonal products. By identifying the above characteristics, it can be determined that the premium of the product is too high). At this time, it means that the profit characteristics are inflated in the evaluation of the value of the product (in the historical transactions, they are inflated. Even if they participate in promotions, there are still premium factors that interfere with trading behavior for the overall analysis). Lowering the score of MLF can overcome this. At the same time, raising the scores of discounts and promotions can partially make up for the weakening of MLF, and can reflect the value gain of promotions in the transaction of this product. The strength of interference here refers to ω×ZKF+η×CXF, where ω>η, generally ω is selected as 80% and η is selected as 20%.

[0127] For the secondary thread, there is no interference from seasonal factors, and the weight factor α=1 is directly generated; through the second neural network, nodes β, γ and δ are output to adjust the weight factor of the premium factor.

[0128] Final score for each thread:

[0129] ZZ t =MLF×δ+XLF+NXLF×α+ZKF×β+CXF×γ

[0130] Among them, ZZ t Indicates the comprehensive evaluation result of the commodity value calculated by thread t.

[0131] The results of commodity value assessment are optimized by dynamically adjusting weight factors. Through the analysis of seasonal factors and premium phenomena by neural networks, the system can adaptively adjust the weights of various characteristic indicators, so that the final commodity value score is more in line with the actual market environment. This dynamic adjustment mechanism ensures that the system can respond quickly to market changes and maintain the accuracy and timeliness of commodity sorting. Through the classification and processing of primary and secondary threads, the system can perform detailed analysis at different levels. For primary threads, considering seasonal factors and premium phenomena can better evaluate the value of the entire category; while in secondary threads, the impact of seasonal factors is small, and fixed weight factors are directly used to ensure the simplification and efficiency of the evaluation process. This design ensures the meticulousness of the evaluation without increasing unnecessary computational complexity.

[0132] S5: Generate a value ranking of commodities according to the result of the comprehensive evaluation, and generate a value sub-ranking of commodities according to the multi-threaded calculation under the primary classification and the multi-threaded calculation under the secondary classification.

[0133] The value ranking of the commodities includes arranging the final scores of each thread in order of size without any restrictions on regions, and using the arrangement order as the value ranking of the commodity category.

[0134] In the case of region restrictions, for the same product category or model q, in the restricted region, the final score of the calculation thread is ZZ q ; When there is no region restriction, the final score of the calculation thread is ZZ q 0.

[0135] Adjust the final score under regional restrictions:

[0136]

[0137] Using the adjusted final score, the products are arranged in order of size, and the order of arrangement is used as the value ranking of the product category. (Usually, there are certain differences in the use of products or models in different regions, but this difference does not represent the actual value of the product. By taking the average of the two, the difference between the two can be balanced, reflecting both regionality and the overall concept.) Through a flexible sorting mechanism, the system can dynamically respond to market changes. For example, some products may perform well in a certain region, but perform generally in other regions. This difference is reasonably reflected through the regional adjustment mechanism. The design also uses the sub-sorting function to achieve detailed sorting of product categories at different levels, helping retailers better understand the specific performance of various types of products in the market.

[0138] In the arrangement sequence, if threads with the same score appear, the threads with the same score are listed in the same sorting position, and the sorting position of the next thread is the number of all previous threads plus one.

[0139] The value sub-ranking of the commodities includes giving priority to outputting the value ranking of the primary classification, and by selecting the category, being able to pull down and generate the value ranking of the secondary classification under the primary classification.

[0140] By arranging the final scores of the products in order of size and making regional adjustments, the system can generate a very accurate product ranking. Even if the scores are the same, the system can handle them reasonably, making the product ranking results more in line with actual needs. This accuracy allows the ranking results to truly reflect the market performance of the products and provide retailers with a reliable basis for decision-making.

[0141] On the other hand, this embodiment also provides a high-performance multi-factor commodity value evaluation and ranking system, which includes:

[0142] Collection unit, obtains transaction data recorded when goods are circulated.

[0143] The classification unit classifies the goods, and after obtaining the first-level classification, it is further classified to obtain the second-level classification.

[0144] The computing unit performs single-thread processing on each category in the first-level classification and the second-level classification respectively; constructs multi-threaded computing under the first-level classification and multi-threaded computing under the second-level classification, and at the same time divides the multi-threaded computing into batches according to the amount of computation of each thread in the multi-threading, and each thread performs computing according to the batches.

[0145] The comprehensive evaluation unit, after each thread has processed its own batch, summarizes the results to the central processor for final data fusion and comprehensive evaluation of the commodity value.

[0146] The ranking generating unit generates a value ranking of commodities according to the result of the comprehensive evaluation, and generates a value sub-ranking of commodities according to the multi-threaded calculation under the primary classification and the multi-threaded calculation under the secondary classification.

[0147] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0148] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0149] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0150] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0151] Example 2 is an embodiment of the present invention, which provides a high-performance multi-factor commodity value evaluation and ranking method. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0152] We selected a category of popular products on a large e-commerce platform for the experiment. The product category covers electronic products (such as smartphones, tablets, smart watches, etc.) and includes specific products of different models. The experimental data comes from the transaction records of the platform in the past year. In order to ensure the comprehensiveness and representativeness of the data, the experiment selected data from different regions across the country, and focused on the sales volume in the past 30 days and the cumulative sales volume for the whole year.

[0153] During the experiment, the commodities were first classified into primary classification (electronic products) and secondary classification (specific models). Based on these classifications, a multi-threaded computing system with primary and secondary threads was constructed, and multi-factor feature indicators were generated for the transaction data of different commodities. The feature indicators include the gross profit contribution (MLF) in the past 30 days, the sales score (XLF) in the past 30 days, the annual cumulative sales score (NXLF), the current discount strength score (ZKF), and the commodity label and promotion policy score (CXF) in the past 30 days. The recorded data is shown in Table 1.

[0154] The specific implementation process is as follows:

[0155] Data preparation: The commodity transaction data required for the experiment was extracted from the e-commerce platform, including sales volume, sales amount, discount information, promotional tags, etc. The data covers different regions across the country, and the transaction records for the past 30 days and the whole year are counted.

[0156] Product classification and thread construction:

[0157] Classify products into categories, such as smartphones, tablets, smart watches, etc.

[0158] The specific models under each category are further classified into secondary categories, such as iPhone, Samsung Galaxy, iPad, etc.

[0159] Build a corresponding multi-threaded computing system, with the first-level classification processing larger data sets and the second-level classification processing specific model data.

[0160] Calculation of multi-factor characteristic indicators:

[0161] For each first-level category product, calculate its gross profit contribution score (MLF) in the past 30 days and include it in the total score with a weight of 25%.

[0162] Calculate the sales score of each product in the past 30 days (XLF) and the annual cumulative sales score (NXLF) to determine whether there are seasonal factors in the product.

[0163] Combine the current discount strength score (ZKF) and the product label and promotion policy score (CXF) to analyze the product premium.

[0164] If the neural network analysis finds that the product has obvious seasonal factors or is overpriced, the weight factor will be adjusted to ensure the rationality of the score.

[0165] Comprehensive evaluation and ranking:

[0166] After each thread completes the calculation, the results are aggregated to the central processor for data fusion and a comprehensive assessment of the product value.

[0167] The value ranking of the products is generated based on the comprehensive evaluation results, and the adjusted final scores are calculated respectively with and without regional restrictions for global and regional ranking.

[0168] Table 1 Data record table

[0169]

[0170]

[0171] Through the analysis of the tabular data, it can be seen that the commodity value evaluation ranking algorithm of the present invention has shown excellent performance in multiple dimensions. First, in the comprehensive score, multiple factors such as sales volume in the past 30 days (XLF), annual cumulative sales volume (NXLF), gross profit contribution (MLF), discount intensity (ZKF) and promotion policy (CXF) are comprehensively considered to avoid the limitations of traditional single indicator ranking. For example, smartphone model A performs well in various indicators, with a comprehensive score of 82.3, showing its market competitiveness. Although the sales volume of smart watch model E is relatively low in the past 30 days, through promotional activities (high ZKF and CXF scores), its comprehensive score still reached 77.5, reflecting the positive impact of promotional policies on its market performance.

[0172] Secondly, the comparison of the data in the table further proves the advantages of the invention in dealing with seasonal factors and premium phenomena. Taking smartphone model B as an example, although the annual cumulative sales volume is high (NXLF is 88), due to the high premium, its sales performance in the past 30 days is poor (XLF is 60). After adjustment through the neural network, the comprehensive score is 73.2. In contrast, due to the high discount strength and promotion policy support of smart watch model F (ZKF is 95, CXF is 85), its comprehensive score still reached 70.4, although its annual cumulative sales volume and gross profit contribution are low.

[0173] The present invention adopts a dynamic weight adjustment mechanism in the comprehensive evaluation, which uses a neural network to intelligently identify the seasonal factors and premium phenomena of commodities, and adjusts the weights of characteristic indicators accordingly. This innovative adjustment method makes the scoring results closer to the actual market demand, can dynamically respond to market changes, and provide more timely and reasonable ranking results. In addition, the present invention significantly improves the processing speed and stability of the system through multi-threaded processing and data fusion technology, and shows strong computing efficiency and accuracy when facing large-scale data.

[0174] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A high-performance multi-factor commodity value evaluation and ranking method, characterized in that: include: Obtain transaction data recorded when goods are circulated; Classify the goods, and after obtaining the primary classification, reclassify them to obtain the secondary classification; Performing single-thread processing on each category in the first-level classification and the second-level classification respectively; constructing multi-threaded calculation under the first-level classification and multi-threaded calculation under the second-level classification, and dividing the multi-threaded calculation into batches according to the computational load of each thread in the multi-threads, and each thread performs calculation according to the batches; After each thread has processed its own batch, it aggregates the results to the central processor for final data fusion and comprehensive evaluation of the commodity value; According to the result of the comprehensive evaluation, a value ranking of the commodities is generated, and according to the multi-threaded calculation under the primary classification and the multi-threaded calculation under the secondary classification, a sub-ranking of the value of the commodities is generated.

2. The high-performance multi-factor commodity value evaluation and ranking method according to claim 1, characterized in that: The transaction data recorded when the commodity is circulated includes, by inputting the commodity category, recording the transaction time and amount in the historical records of all the input categories; Count the transaction time and amount in the past thirty days and the past year respectively, and generate the number of transactions and the amount of each transaction under each first-level category in the past thirty days and the past year respectively; at the same time, generate the number of transactions and the amount of each transaction under each second-level category in the past thirty days and the past year respectively; By selecting a region, only the transaction data of the commodities in the selected region is evaluated for value, and a ranking of the commodity value evaluation in the selected region is generated.

3. The high-performance multi-factor commodity value evaluation and ranking method according to claim 2, characterized in that: The first-level classification includes the classification results obtained by dividing the commodities according to commodity categories; The secondary classification includes the classification results obtained by dividing the commodities in each primary classification according to the commodity models; According to the classification results of the first-level classification, the value assessment of each commodity category is calculated as an independent thread, which is defined as the first-level thread; According to the classification results of the secondary classification, the value evaluation of different models of the same category of goods is calculated as an independent thread, which is defined as a secondary thread.

4. The high-performance multi-factor commodity value evaluation and ranking method according to claim 3, characterized in that: The constructing of multi-threaded calculation under the primary classification and multi-threaded calculation under the secondary classification includes, for each of the primary thread and the secondary thread, generating multi-factor characteristic indicators according to the transaction data; Calculate the gross profit contribution score for the past 30 days: Among them, MLF represents the gross profit contribution score of the current commodity category in the past 30 days, accounting for 25% of the total score of 100; XL 30 Indicates the sales volume of the current commodity category in the past 30 days; SJ0 indicates the current average selling price of the current commodity category; CB0 indicates the current average cost of the current commodity category; MAX(...) indicates the maximum value of the input content; i indicates the i-th day in the historical transaction record of the current category, i∈(31, I), I indicates the maximum number of record days minus 30; XL i Indicates the sales volume of the current category in the 30 days before the i-th day in the historical transaction records; SJ i Indicates the average selling price of the current category in the historical transaction records of the 30 days before the i-th day; CB i Indicates the average cost of the historical transaction records of the current category in the 30 days before the i-th day; Calculate the sales score in the past 30 days: Calculate the annual cumulative sales score: Among them, XL 365 Indicates the cumulative sales volume of the current product category in the past 365 days; XL 365，u Indicates the cumulative sales volume of the current product category in the 365 days before the uth day in the historical transaction records, u∈(366,U), where U represents the maximum number of record days minus 365; Calculate the current discount strength score: Among them, YJ represents the average original price of the current commodity category; Calculate the scores of product labels and promotion policies in the past thirty days. For each first-level thread, count the scores of different types of products: Where j represents the index of different models of the same category of goods; J represents the number of models; CX j Indicates the points of the j-th model. If there is a promotion, it is recorded as 1, otherwise it is recorded as 0; For each secondary thread, if the calculated model has a promotion, the product label and promotion policy score for the past thirty days are recorded as 35; if the calculated model does not have a promotion, the product label and promotion policy score for the past thirty days are recorded as 0.

5. The high-performance multi-factor commodity value evaluation and ranking method according to claim 4, characterized in that: The batching of the multi-thread calculations includes: firstly calculating the primary thread, and after completing the calculation of the primary thread, calculating the secondary thread; Before calculation, the batches are divided according to the amount of calculations generated by each thread in the historical calculation process and the maximum amount of calculations of the computer; When batching, the first-level threads are batched first, and then the second-level threads are batched; During calculation, in each level of threads, calculations are performed sequentially in the order in which the batch results are generated; set up :There are a total of A threads of calculation, YSL a is the mode of the amount of computation data generated by the thread selected for the ath time during the historical computation process, MAX YSL Indicates the maximum computing capacity of the computer; The batch partitioning constraint is expressed as: Among them, a represents the ath thread selected in this division process, and among threads of the same level, the selection is random; n represents the number of threads in the batch.

6. The high-performance multi-factor commodity value evaluation and ranking method according to claim 5, characterized in that: The comprehensive evaluation of the commodity value includes: for the first-level thread, analyzing XLF and NXLF through a neural network to determine whether there are seasonal factors in the commodity, and if there are seasonal factors, paying more attention to NXLF; analyzing ZKF, CXF, XLF and NXLF to determine whether the commodity has an excessively high premium, and if there is an excessively high premium, reducing the attention of MLF and increasing the attention of ZKF and CXF; The neural network includes a first neural network and a second neural network; The first neural network comprises, an input layer: input feature vector [XLF, NXLF]; Hidden layer: 2 hidden layers, 64 neurons in each layer, using ReLU activation function; Output layer: Output the weight factor α used to adjust the seasonal factors; α=f MLP (XLF,NXLF) Among them, f MLP Represents a neural network model of a multilayer perceptron; if the product is affected by seasonal factors through the neural network analysis, a weight factor α greater than 1 is generated according to the interference intensity of the seasonal factors; if the product is not affected by seasonal factors through the neural network analysis, a weight factor α equal to 1 is generated; The second neural network includes, an input layer: input feature vector [XLF, NXLF, ZKF, CXF]; Hidden layer: 2 hidden layers, 64 neurons in each layer, using ReLU activation function; Output layer: three output nodes β, γ and δ, used to adjust the weight factors of premium factors; [β,γ,δ]=f MLP (XLF,NXLF,ZKF,CXF) If the neural network analysis shows that the product has a high premium, then according to the strength of the premium interference, weight factors β and γ greater than 1 and a weight factor δ less than 1 are generated; if the neural network analysis shows that the product is not affected by the premium, β = γ = δ = 1 is generated; For the secondary thread, there is no interference from seasonal factors, and a weight factor α=1 is directly generated; through the second neural network, nodes β, γ and δ are output to adjust the weight factor of the premium factor; Final score for each thread: ZZ t =MLF×δ+XLF+NXLF×α+ZKF×β+CXF×γ Among them, ZZ t Indicates the comprehensive evaluation result of the commodity value calculated by thread t.

7. The high-performance multi-factor commodity value evaluation and ranking method according to claim 6, characterized in that: The value ranking of the commodities includes, without any restrictions on regions, arranging the final scores of each thread in order of magnitude, and using the arrangement order as the value ranking of the commodity category; In the case of region restrictions, for the same product category or model q, in the restricted region, the final score of the calculation thread is ZZ q ; When there is no region restriction, the final score of the calculation thread is ZZ q 0; Adjust the final score under regional restrictions: Using the adjusted final scores, arrange the items in order of size, and use the order of arrangement as the value ranking of the product category; In the arrangement sequence, if threads with the same score appear, the threads with the same score are arranged in the same sorting position, and the sorting position of the next thread is the number of all previous threads plus one; The value sub-ranking of the commodities includes giving priority to outputting the value ranking of the primary classification, and by selecting the category, being able to pull down and generate the value ranking of the secondary classification under the primary classification.

8. A high-performance multi-factor commodity value evaluation and ranking system using the method according to any one of claims 1 to 7, characterized in that: The collection unit obtains the transaction data recorded when the goods are circulated; Classification unit, classifies the goods, and after obtaining the first-level classification, it is further classified to obtain the second-level classification; The computing unit performs single-thread processing on each category in the first-level classification and the second-level classification respectively; constructs multi-thread computing under the first-level classification and multi-thread computing under the second-level classification, and divides the multi-thread computing into batches according to the computational load of each thread in the multi-thread, and each thread performs computing respectively according to the batches; Comprehensive evaluation unit, after each thread has processed its own batch, the results are aggregated to the central processor for final data fusion and comprehensive evaluation of the commodity value; The ranking generating unit generates a value ranking of commodities according to the result of the comprehensive evaluation, and generates a value sub-ranking of commodities according to the multi-threaded calculation under the primary classification and the multi-threaded calculation under the secondary classification.

9. A computer device comprising: Memory and processor; The memory stores a computer program, which is characterized in that when the processor executes the computer program, it implements the steps of a high-performance multi-factor commodity value evaluation and ranking method.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a high-performance multi-factor commodity value evaluation and ranking method are implemented.

Images