Commodity loading recommendation method and system based on matching degree calculation

Abstract

The invention discloses a commodity sampling recommendation method and system based on matching degree calculation. The method comprises the following steps: 1, obtaining basic data, wherein the basic data comprises a store commodity transaction number, a commodity unit price, accumulated sampling time, commodity category information, store feature attributes, commodity feature attributes and a commodity application scene; 2, calculating a commodity-store matching degree; 3, calculating commodity-commodity similarity: constructing a commodity-commodity similarity graph between the commodities, and 4, recommending the commodities: comprehensively considering the commodity-store real matching degree and the commodity-commodity similarity, and carrying out commodity sampling recommendation. According to the invention, by collecting and analyzing the sales data, the product information data and other related information of the store, accurate recommendation of the sample-loaded commodities of the store is realized, and decision support is provided for a store principal. Meanwhile, the method can accurately capture the feature change of the commodities sold in the store, achieves the dynamic commodity recommendation, adapts to the market change, and improves the sales efficiency.

Description

Technical Field

[0001] The present invention belongs to the field of information technology, and in particular relates to a method and system for recommending product samples based on matching degree calculation. Background Art

[0002] At present, enterprises are facing increasingly fierce market competition, and consumers have more diverse choices. As a channel directly facing consumers, offline stores have become particularly important in recommending samples. With the accumulation of data and the development of artificial intelligence technology, in-depth mining of massive sales data, personalized recommendations for sampled products in different stores, and improved accuracy and effectiveness of sampled products are particularly important for the development of enterprises.

[0003] In the traditional product sampling process, the experience and subjective judgment of the store manager are the most important influencing factors, while there is less research on the matching degree between products and stores, and the matching degree between products, and lack of mining of implicit information in a large amount of data. Therefore, the present invention mines and analyzes massive product sales data and store information, calculates the matching degree between products and between products and stores, and develops a more accurate product sampling recommendation method, so that companies can more accurately understand consumer needs, provide personalized recommendation services, and improve consumers' shopping experience and satisfaction. At the same time, it brings more sales opportunities and market share to companies. Summary of the invention

[0004] The purpose of the present invention is to propose and develop a product sampling recommendation method based on matching degree calculation to address the shortcomings of existing store product sampling that relies on subjective judgment.

[0005] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0006] Step 1: Obtain basic data: including the number of store product transactions, product unit price, cumulative sample time, product category information, store characteristic attributes, product characteristic attributes and product application scenarios, and obtain data from your own enterprise information system through SQL;

[0007] Step 2: Calculate product-store matching;

[0008] Step 3: Calculate product-product similarity:

[0009] Build a product-product similarity graph between products.

[0010] Step 4: Product recommendation: Comprehensively consider the product-store real match and product-product similarity to make product recommendations.

[0011] Furthermore, the method supports the following test scenarios: offline store product sampling and recommendation for consumer goods companies with multiple stores in multiple regions, where the company has collected basic store information, products, and sales data or has the ability to collect these data.

[0012] Furthermore, the calculation of product-store matching degree described in step 2 is specifically implemented as follows:

[0013] 2-1. Traverse the time of single product sample loading in each store, calculate the cumulative sales of each product in the store single_sales, count the total sample loading time of the product in the store single_days_count, and define the average daily output value of sampled products sales_per_day;

[0014] 2-2. Calculate the total average daily output value of products of the same category and positioning, that is, add the average daily output value sales_per_day of the product category information. The total average daily output value of products of the same category and positioning is stored in the form of a dictionary.

[0015] 2-3. Calculate the average daily output value of all stores and in-store sample products and the total average daily output value of products of the same category and positioning;

[0016] 2-4. After the calculation of the real matching degree between products and stores is completed, the store characteristic attributes, product characteristic attributes, and the real matching degree between products and stores are used to construct a product-store data structure diagram.

[0017] Furthermore, the calculation of the average daily output value of goods sales_per_day is as follows:

[0018]

[0019] The cumulative sales single_sales is calculated by multiplying the transaction quantity "ORDER_QTY" of the product in this store by the unit price "SINGLE_PRICE".

[0020] Furthermore, steps 2-3 are implemented as follows:

[0021] For each sampled product, re-traverse all stores to find the store where the product is being sampled. If the average daily output value of the product in a sampled store is 0, the real match between the product and the store is 0; if its average daily output value is not 0, the real match between the product and the store real_match_valu is calculated as follows:

[0022]

[0023] Furthermore, steps 2-4 are specifically implemented as follows: stores and products are taken as nodes, and stores and products respectively contain corresponding store feature attributes and product feature attributes, and product-store real matching degrees are taken as edges. Stores and products are connected to form an undirected graph as a product-store data structure diagram.

[0024] Furthermore, the calculation of product-product similarity described in step 3 is specifically implemented as follows:

[0025] 3-1. Group all products of the same category and positioning together, list the product-store real matching degree of each product of the same category and positioning in each store where the products are sold, and sort them from large to small, and remove products with a product-store real matching degree of 0 in the set.

[0026] 3-2. Take the top 60% as the hot-selling products of the same category and positioning products set, and the product-store real matching degree of the last product in the top 60% is used as the judgment threshold τ of the hot-selling products of the same category and positioning products.

[0027] 3-3. Traverse all sampled products in all stores again, and determine in which stores the sampled product is a hot-selling product based on the hot-selling threshold τ of the same category and positioning products. That is, if the real product-store match between a product A and its corresponding store S is greater than τ, that is, real_match(store S, product A)>τ, then the hot-selling relationship value "MATCH_STANDARD" between product A and store S is recorded as 1, otherwise it is 0.

[0028] 3-4. According to the hot-selling relationship value "MATCH_STANDARD" between each product and its sample store, record the hot-selling store to which the product belongs; introduce product similarity Similarity, for any two products G i ,G j The similarity formula is:

[0029]

[0030] Among them, S i It is product G i Collection of hot-selling stores, S j It is product G j The set of hot-selling stores, Similarity() represents the similarity solving function; length() represents the set length solving function.

[0031] Furthermore, the product-store matching degree is specifically implemented as follows:

[0032] Suppose there are several similar products in store S, among which the average daily output value of high-cost-effective product A is A1 yuan, and the average daily output value of high-cost-effective product B is B1 yuan. There are several similar products in store S', among which the average daily output value of high-cost-effective product A is A2 yuan, and the average daily output value of high-cost-effective product C is C1 yuan. Then the average daily total output value of high-cost-effective sofas is: A1+B1+A2+C1; the product-store matching degree of product A and store S is:

[0033]

[0034] Furthermore, the hot-selling store set is specifically solved as follows:

[0035] Suppose product A is available in a store group G = {store a, store b, store c, store d, store e, store f, store g}. By calculating the product-store matching degree, the set of stores that are higher than the hot-selling product judgment threshold τ is G' = {store a, store b, store d, store f}, then G' is the set of hot-selling stores for sofa A. Similarly, calculate the set of hot-selling stores for another similar product D, and assume that H' = {store b, store c} is obtained.

[0036] Furthermore, the present invention also provides a product sampling recommendation system based on matching degree calculation, which includes four modules: a data reading module, a product-store matching degree calculation module, a product-product matching degree calculation module, and a log module.

[0037] The data reading module will obtain all required data from its own enterprise information system by executing SQL statements;

[0038] The product-store matching calculation module constructs a data structure of all stores and sampled products in the store by calculating the average daily output value of sampled products and the total average daily output value of products of the same category and positioning, and calculates the actual matching degree of products-stores, and arranges them in order from large to small.

[0039] The product-product matching degree calculation module determines the threshold of hot-selling products in the same scene and price positioning product set, and records the matching relationship between products and stores; finally, based on the matching relationship between each product and its sample store, the product-product matching degree is calculated, and the products are arranged in descending order to obtain a pre-recommended product list.

[0040] The log module writes all calculation results into the log file of the corresponding date. Users can view it by themselves, and can display the executed SQL, product-store matching and product-product matching execution information.

[0041] Beneficial effects of the present invention:

[0042] The present invention calculates the matching degree between goods and stores and goods and goods through a large amount of goods sales data, and makes sample recommendations based on the calculation results of the hot-selling goods that best match the target store, which is conducive to the accurate delivery of goods, improves customer satisfaction and sales performance, and reduces the sales cost of the enterprise.

[0043] Advantages of the present invention: By collecting and analyzing the sales data, product information data and other relevant information of the store, the present invention can achieve accurate recommendations for the sample goods in the store and provide decision support for the store manager. At the same time, the present invention can accurately capture the characteristic changes of the goods sold in the store, realize dynamic product recommendations, adapt to market changes, and improve sales efficiency. From the implementation level, this method is developed and implemented through Python, and the calculation complexity of the model is not high. On the premise that the enterprise provides relevant data, it does not require excessive computing resources to perform calculations, and has strong feasibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of the recommendation algorithm of the present invention. DETAILED DESCRIPTION

[0045] The present invention will be described in detail below with reference to the accompanying drawings, and the purpose and results of the present invention will become more apparent. This example is a type embodiment of the present invention, and the specific operation process is as follows Figure 1 shown.

[0046] A method for recommending product samples based on matching degree calculation, the technical solution adopted is as follows:

[0047] The test scenarios supported by the present invention are: offline store product sampling and recommendation of consumer goods companies with multiple stores in multiple regions, and the company has collected basic information, products, and sales data of the stores or has the ability to collect these data.

[0048] Based on the supported scenarios, the implementation of the present invention is as follows:

[0049] Step 1: Basic data acquisition:

[0050] The basic data required by the present invention includes the number of store commodity transactions, commodity unit price, cumulative sample loading time, commodity category information, store characteristic attributes, commodity characteristic attributes and commodity application scenarios, and the data is obtained from the enterprise information system through SQL;

[0051] Step 2: Calculate product-store matching;

[0052] 2-1. Traverse the time of single product loading in each store, and calculate the cumulative sales of each product in the store, single_sales. The calculation method is to multiply the number of transactions "ORDER_QTY" of the product in this store by the unit price "SINGLE_PRICE". Count the total loading time of the product in the store, single_days_count, that is, the cumulative number of days for loading if there are multiple loading records. Define the average daily output value of the sampled product, sales_per_day, as follows:

[0053]

[0054] For example, if product A is sampled twice, each time for 10 days, and its cumulative sales are 30,000 yuan, then its average daily output value is:

[0055]

[0056] 2-2. Calculate the total daily average output value of products of the same category and positioning

[0057] goods_same_location_type_per_day_income_count, the total daily average output value of goods of the same category and location is stored in dictionary form.

[0058] The calculation of the total average daily output value of the same category and the same positioning products is to add the average daily output value sales_per_day of the product category information.

[0059] 2-3. Calculate the average daily output value of all stores and in-store sampled goods and the total average daily output value of goods of the same category and positioning; for each sampled product, re-traverse all stores to find the store where the product is being sampled. If the average daily output value of the product in a sampled store is 0, the real match between the product and the store is 0; if its average daily output value is not 0, the real match between the product and the store real_match_valu is calculated as follows:

[0060]

[0061] 2-4. After the calculation of the real matching degree between products and stores is completed, the store characteristic attributes, product characteristic attributes, and the real matching degree between products and stores are used to construct a product-store data structure diagram.

[0062] Specifically: stores and products are taken as nodes, and stores and products contain corresponding store characteristic attributes and product characteristic attributes respectively, and the product-store real matching degree is taken as the edge. Stores and products are connected to form an undirected graph as the product-store data structure diagram

[0063] Step 3: Calculate product-product similarity:

[0064] A product-product similarity graph is constructed between products. The specific calculation method is as follows:

[0065] 3-1. Group all products of the same category and positioning together, list the product-store real matching degree of each product of the same category and positioning in each store where the products are sold, and sort them from large to small, and remove products with a product-store real matching degree of 0 in the set.

[0066] 3-2. Take the top 60% as the hot-selling products of the same category and positioning products set, and the product-store real matching degree of the last product in the top 60% is used as the judgment threshold τ of the hot-selling products of the same category and positioning products.

[0067] 3-3. Traverse all sampled products in all stores again, and determine which stores the sampled product is a hot product based on the threshold value τ of the hot-selling products of the same category and positioning. For example: if the real product-store match between a product A and its corresponding store S is greater than τ, that is, real_match(store S, product A)>τ, then the hot-selling relationship value "MATCH_STANDARD" between product A and store S is recorded as 1, otherwise it is 0.

[0068] 3-4. According to the hot-selling relationship value "MATCH_STANDARD" between each product and its sample store, record the hot-selling store to which the product belongs. Introduce product similarity, for any two products G i ,G j The similarity formula is:

[0069]

[0070] Among them, S i It is product G i Collection of hot-selling stores, S j It is product G j Similarity() represents the similarity solving function; length() represents the set length solving function. This allows the similarity between products to be calculated across categories and locations.

[0071] Step 4: Product Recommendation:

[0072] Through steps 2 and 3, the real matching degree between products and stores and the similarity between products are comprehensively considered to make product recommendations;

[0073] Take the recommendation of home furnishing products as an example: suppose there are several sofas in store S, among which the average daily output value of high-cost-effective sofa A is 3,000 yuan, and the average daily output value of high-cost-effective sofa B is 2,000 yuan. There are several sofas in store S', among which the average daily output value of high-cost-effective sofa A is 1,000 yuan, and the average daily output value of high-cost-effective sofa C is 1,500 yuan. The total daily output value of high-cost-effective sofas is:

[0074] goods_same_location_type_per_day_income_count=7500

[0075] Calculate the product-store matching degree of product A and store S:

[0076]

[0077] Next, calculate the product-product matching. Assume that sofa A is available in a certain store group G = {store a, store b, store c, store d, store e, store f, store g}. By calculating the product-store matching, assume that the store set above the hot-selling product judgment threshold τ is G' = {store a, store b, store d, store f}, then G' is the hot-selling store set of sofa A. Similarly, calculate the hot-selling store set of another product D, and assume that H' = {store b, store c}. According to the product-product matching formula:

[0078]

[0079] get:

[0080]

[0081] Finally, we make sampling recommendations by comprehensively considering the product-store matching degree of 0.4 and the product-product matching degree of 0.2.

[0082] Furthermore, the present invention provides a technical architecture of the enterprise offline store product sampling recommendation method, which includes four modules: a data reading module, a product-store matching degree calculation module, a product-product matching degree calculation module, and a log module.

[0083] The data reading module will obtain all the data required by the recommendation algorithm from the enterprise's information system by executing SQL statements;

[0084] The product-store matching calculation module constructs a data structure of all stores and sampled products in the store by calculating the average daily output value of sampled products and the total average daily output value of products of the same category and positioning, and calculates the actual matching degree of products-stores, and arranges them in order from large to small.

[0085] The product-product matching degree calculation module determines the threshold of hot-selling products in the same scene and price positioning product set, and records the matching relationship between products and stores. Finally, based on the matching relationship between each product and its sample store, the product-product matching degree is calculated, and the products are arranged in order from large to small to obtain a pre-recommended product list.

[0086] The log module writes all calculation results into the log file of the corresponding date. Users can view it by themselves, and can display the execution information such as the executed SQL, product-store matching and product-product matching.

[0087] The present invention is not limited to the above specific implementations. A person skilled in the art can implement the present invention by using other various specific implementations based on the content disclosed in the present invention. Therefore, any design that adopts the design structure and ideas of the present invention and makes some simple changes or modifications falls within the protection scope of the present invention.

Claims

1. A method for recommending product samples based on matching degree calculation, characterized in that: The steps include: Step 1: Obtain basic data: including the number of store product transactions, product unit price, cumulative sample time, product category information, store characteristic attributes, product characteristic attributes and product application scenarios, and obtain data from your own enterprise information system through SQL; Step 2: Calculate product-store matching; Step 3: Calculate product-product similarity: Build a product-product similarity graph between products. Step 4: Product recommendation: Comprehensively consider the product-store real match and product-product similarity to make product recommendations.

2. The method for recommending product samples based on matching degree calculation according to claim 1, characterized in that: The test scenarios supported by this method are: offline store product sampling recommendations for consumer goods companies with multiple stores in multiple regions. The company has collected basic store information, products, and sales data or has the ability to collect these data.

3. A method for recommending product samples based on matching degree calculation according to claim 1 or 2, characterized in that: The calculation of product-store matching degree described in step 2 is specifically implemented as follows: 2-1. Traverse the time of single product sample loading in each store, calculate the cumulative sales of each product in the store single_sales, count the total sample loading time of the product in the store single_days_count, and define the average daily output value of sampled products sales_per_day; 2-2. Calculate the total average daily output value of products of the same category and positioning, that is, add the average daily output value sales_per_day of the product category information, and store the total average daily output value of products of the same category and positioning in the form of a dictionary; 2-3. Calculate the average daily output value of all stores and in-store sample products and the total average daily output value of products of the same category and positioning; 2-4. After the calculation of the real matching degree between products and stores is completed, the store characteristic attributes, product characteristic attributes, and the real matching degree between products and stores are used to construct a product-store data structure diagram.

4. The method for recommending product samples based on matching degree calculation according to claim 3, characterized in that: The calculation of the average daily output value of goods sales_per_day is as follows: The cumulative sales single_sales is calculated by multiplying the transaction quantity "ORDER_QTY" of the product in this store by the unit price "SINGLE_PRICE".

5. A method for recommending product samples based on matching degree calculation according to claim 3 or 4, characterized in that: Steps 2-3 are implemented as follows: For each sampled product, re-traverse all stores to find the store where the product is being sampled. If the average daily output value of the product in a sampled store is 0, the actual matching degree between the product and the store is 0; If the average daily output value is not 0, the real match degree between the product and the store, real_match_valu, is calculated as follows:

6. The method for recommending product samples based on matching degree calculation according to claim 5, characterized in that: The specific implementation of steps 2-4 is as follows: stores and products are taken as nodes, and stores and products contain corresponding store characteristic attributes and product characteristic attributes respectively, and the product-store real matching degree is taken as the edge. After connecting stores and products, an undirected graph is formed as the product-store data structure diagram.

7. The method for recommending product samples based on matching degree calculation according to claim 6, characterized in that: The calculation of product-product similarity described in step 3 is specifically implemented as follows: 3-1. Group all products of the same category and positioning together, list the product-store real matching degree of each product of the same category and positioning in each store where the products are sampled, and sort them from large to small, and remove products with a product-store real matching degree of 0 in the set; 3-2. Take the top 60% as the hot-selling products of the same category and positioning products set, and the product-store real matching degree of the last product in the top 60% as the judgment threshold τ of the hot-selling products of the same category and positioning products; 3-3. Traverse all sampled products in all stores again, and determine which stores the sampled product is a hot-selling product based on the hot-selling threshold τ of the same category and positioning products. That is, if the real product-store matching degree of a product A and its corresponding store S is greater than τ, that is, real_match(store S, product A)>τ, then the hot-selling relationship value "MATCH_STANDARD" between product A and store S is recorded as 1, otherwise it is 0; 3-4. According to the hot-selling relationship value "MATCH_STANDARD" between each product and its sample store, record the hot-selling store to which the product belongs; introduce product similarity Similarity, for any two products G i ,G j The similarity formula is: Among them, S i It is product G i Collection of hot-selling stores, S j It is product G j The set of hot-selling stores, Similarity() represents the similarity solving function; length() represents the set length solving function.

8. The method for recommending product samples based on matching degree calculation according to claim 5, characterized in that: The product-store matching degree is specifically implemented as follows: Suppose there are several similar products in store S, among which the average daily output value of high-cost-effective product A is A1 yuan, and the average daily output value of high-cost-effective product B is B1 yuan. There are several similar products in store S', among which the average daily output value of high-cost-effective product A is A2 yuan, and the average daily output value of high-cost-effective product C is C1 yuan. Then the average daily total output value of high-cost-effective sofas is: A1+B1+A2+C1; the product-store matching degree of product A and store S is:

9. The method for recommending product samples based on matching degree calculation according to claim 7, characterized in that: The hot-selling store set is specifically solved as follows: Suppose product A is available in a store group G = {store a, store b, store c, store d, store e, store f, store g}. By calculating the product-store matching degree, the set of stores that are higher than the hot-selling product judgment threshold τ is G' = {store a, store b, store d, store f}, then G' is the set of hot-selling stores for sofa A. Similarly, calculate the set of hot-selling stores for another similar product D, and assume that H' = {store b, store c} is obtained.

10. A product sample recommendation system based on matching degree calculation, characterized in that: The system includes 4 modules: data reading module, product-store matching calculation module, product-product matching calculation module, and log module; The data reading module will obtain all required data from its own enterprise information system by executing SQL statements; The product-store matching calculation module constructs a data structure of all stores and sampled products in the store by calculating the average daily output value of sampled products and the total average daily output value of products of the same category and positioning, and calculates the real matching degree of products-stores, and arranges them in order from largest to smallest; The product-product matching calculation module determines the threshold of hot-selling products in the same scene and same price positioning product set, and records the matching relationship between products and stores; finally, according to the matching relationship between each product and its sample store, the product-product matching degree is calculated, and the products are arranged in descending order to obtain a pre-recommended product list; The log module writes all calculation results into the log file of the corresponding date; Users can view it by themselves, and it can display the executed SQL, product-store matching and product-product matching execution information.

Images