Product recommendation method based on data pre-screening and intermediate variable introduction

Abstract

The invention discloses a product recommendation method based on data pre-screening and intermediate variable introduction, and the method comprises the steps: taking a condition that each customer has a similar common purchase feature for a product combination which is purchased historically and comprises a to-be-recommended product, and screening a basis customer list as a basis for determining whether a to-be-recommended customer is a target recommendation customer or not in advance; whether the to-be-recommended customer is the target recommended customer or not is subsequently determined, and the historical purchase data of each customer in the customer list is taken as an analysis basis, so that the data analysis amount is greatly reduced, and the recommendation efficiency is improved under the scene that complex recommendation analysis calculation needs to be performed by considering the mutual recommendation influence relationship between products. And the analysis speed of simultaneously carrying out product recommendation on a huge customer group is improved. By introducing the intermediate variable and constructing the mutual recommendation influence relationship between the products, the target recommendation product capable of meeting the brand pursuit and consumption ability of the customers to be recommended can be accurately screened out from a large number of the products to be recommended, and the accuracy of building material product recommendation is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of product recommendation, and in particular to a product recommendation method based on data pre-screening and introduction of intermediate variables. Background Art

[0002] The most core issue that needs to be solved in product recommendation is how to improve the accuracy of product recommendations. In addition, when the crawled customer base is large and it is necessary to quickly calculate the recommended products for each customer in this large group in a short period of time, how to improve the analysis speed is another important issue that needs to be solved to ensure the user experience of the product recommendation analysis system.

[0003] At present, the existing technologies for improving the accuracy of product recommendations focus on improving the recommendation algorithm, but do not consider the mutual recommendation influence relationship between different products purchased by the same customer. For example, assuming that customer U has purchased product A, the existing technology uses a set of algorithms to extract the customer purchase features of customer U when purchasing product A, and then matches product B with product features similar to product A, and finally uses customer U as the proposed recommendation object for product B. When product C needs to be recommended, similarly, the product feature similarity between product C and product A is calculated, and product C with similarity is recommended to customers who have purchased product A. Under such a product recommendation algorithm principle, the current improvement of the algorithm focuses on how to improve the accuracy of the analysis of customer purchase features of product A, or improve the accuracy of the analysis of the similarity of product features between product B or product C and product A. However, when a customer has purchased product A and product B or product C at the same time in the past, the common purchase characteristics between the customer's purchase of product A, product B, and product C are not considered. Especially when product A, product B, and product C belong to the same category, accurately analyzing these common purchase characteristics can greatly improve the accuracy of product recommendations. However, how to accurately analyze these common purchase characteristics is the first technical problem that this application needs to solve.

[0004] As exemplified above, the existing product recommendation algorithm focuses on the similarity of product features between a recommendable product, such as product B, and a previously purchased product, such as product A, that the customer has purchased. The feature data required for the product feature similarity matching process is acquired and screened during the algorithm analysis process. When the acquisition or screening process is complex, and the number of recommendable products or customers to be recommended that need to perform product feature similarity calculations is large, this in-process data acquisition or screening process will seriously affect the speed of calculation and analysis, and is not conducive to quickly calculating the recommended products for each customer in a large customer group in a short period of time. In particular, in scenarios where the mutual recommendation influence relationship between the above-mentioned products needs to be considered, in-process data acquisition or screening will be more time-consuming, affecting the user's experience of the system. Therefore, how to change in-process data screening to pre-process data screening, especially in scenarios where the mutual recommendation influence relationship between products needs to be considered, how to change data screening to pre-process screening becomes the second technical problem that needs to be solved in this application. Summary of the invention

[0005] The present invention takes into account the mutual recommendation influence relationship between products to improve the accuracy of product recommendations, and in the scenario of considering the mutual recommendation influence relationship between products, changes the data screening required for recommendation to pre-screening, in order to increase the analysis speed of product recommendations for a large customer group at the same time. A product recommendation method based on data pre-screening and introduction of intermediate variables is provided.

[0006] To achieve this object, the present invention adopts the following technical solutions:

[0007] A product recommendation method based on data pre-screening and introduction of intermediate variables is provided, comprising the steps of:

[0008] S1, using the fact that each customer of the product combination that the customer has purchased in the past, including the product to be recommended, has similar common purchase characteristics for the product combination as a screening condition, and screening out a list of base customers as a basis for determining whether the customer to be recommended is a target recommended customer;

[0009] S2, filtering out products that have been purchased by the customer to be recommended and by each customer in the customer list, to form a third recommendation basis list;

[0010] S3, calculating the first variable value of the intermediate variable between each product in the third recommendation basis list purchased by the proposed customer in history, and extracting the maximum first variable value, and then calculating the positive influence factor f and the reverse influence factor f′ of the proposed product and each building material product in the third recommendation basis list in sequence according to the maximum first variable value, and further constructing the mutual recommendation influence relationship between the proposed product and each building material product in the third recommendation basis list;

[0011] S4, solving the fifth deviation between the mutual recommendation influence relationship between the intended recommended customers calculated for the same product combination and the mutual recommendation history influence relationship between the historical purchasing customers referred to in the third recommendation basis list, and taking the intended recommended customers who meet the condition that the fifth deviation is less than the preset fifth deviation threshold as the target recommended customers for recommending the intended recommended products.

[0012] Preferably, in step S1, the rule for determining that each customer has similar common purchasing characteristics for the product combination is: different customers purchase the same brand classification level of the first building material product in the product combination, and the same brand classification level of the second building material product purchased by different customers, and the deviation between the first average values ​​of the common characteristics of brand pursuit of the product combination is less than a preset first common deviation threshold, and the deviation between the second average values ​​of the common characteristics of consumption capacity of the product combination is less than a preset second common deviation threshold;

[0013] The intermediate variable includes the common characteristics pursued by the brands and / or the common characteristics of consumption capabilities associated with the same product combination; the variable value of the intermediate variable is the first average value or the second average value, or the weighted sum of the first average value and the second average value.

[0014] Preferably, the method for calculating the first average value representing the common characteristics of the brand pursuit of the first building material product and the second building material product in the product portfolio comprises the steps of:

[0015] A1, classifying the first building material product and the second building material product constituting the product combination by brand, and then assigning the same brand classification grade value to each brand under each category;

[0016] A2, calculating a first proportion of the first brand classification grade value of the first building material product of the corresponding brand that the customer has purchased in the past among the brand classification grade values ​​corresponding to each brand of the first building material product, and calculating a second proportion of the second brand classification grade value of the second building material product of the corresponding brand that the same customer has purchased in the past among the brand classification grade values ​​corresponding to each brand of the second building material product;

[0017] A3, determining whether a first deviation between the first proportion and the second proportion is less than a preset first deviation threshold,

[0018] If yes, then calculating a first average value of the first brand classification grade value and the second brand classification grade value as the common characteristics of the brand pursuit of the first building material product and the second building material product by the customer;

[0019] If not, it is determined that the customer does not have the common characteristics of brand pursuit for the first building material product and the second building material product.

[0020] Preferably, the method for calculating the second average value representing the common characteristics of the consumption capacity of the first building material product and the second building material product in the product combination comprises the steps of:

[0021] B1, classifying the selling price ranges of each brand under each brand grade classification of the first building material product and the second building material product constituting the product combination, and assigning corresponding price range classification grade values ​​to each brand under the same brand grade classification;

[0022] B2, calculating a first ratio of a first price range classification level value corresponding to the selling price range into which the purchase price of the first building material product of the corresponding brand purchased by the customer in history falls, to the price range classification level values ​​respectively corresponding to each brand under the same brand classification level of the first building material product, and obtaining a second ratio of a second price range classification level value corresponding to the selling price range into which the purchase price of the second building material product of the corresponding brand purchased by the same customer in history falls, to the price range classification level values ​​respectively corresponding to each brand under the same brand classification level of the second building material product;

[0023] B3, determining whether a second deviation between the first ratio and the second ratio is less than a preset second deviation threshold,

[0024] If so, calculating a second average value of the first price range classification grade value and the second price range classification grade value as the common characteristics of the consumption capacity of the customer for the first building material product and the second building material product;

[0025] If not, it is determined that the customer does not have the common consumption capacity characteristics for the first building material product and the second building material product.

[0026] Preferably, in step B1, the method for assigning the price range classification level values ​​corresponding to the brands under the same brand classification level comprises the steps of:

[0027] B11, determine whether there is any overlapping price range between brands in the same brand classification level.

[0028] If so, the influence of lowering and / or raising prices on the price range classification level value is calculated for each brand with overlapping product price ranges, and then the process goes to step B12;

[0029] If not, then assign the price range classification level values ​​from high to low to the selling price ranges corresponding to the brands under the same brand classification level according to the highest price in the product price range from high to low;

[0030] B12, assigning the same price intersection interval classification level value to the price intersection intervals in the selling price ranges of the two brands for solving the influence amount, and then adding the influence amount to the price intersection interval classification level value to form the price interval classification level value corresponding to the brand with the price reduction and / or price increase.

[0031] Preferably, the influence amount in step B11 is calculated by the following method steps:

[0032] B111, segmenting the first selling price range and the second selling price range corresponding to the first brand and the second brand at the same brand classification level participating in the calculation of the influence amount respectively with a fixed step length to obtain a plurality of first price sub-ranges associated with the first selling price range and a plurality of second price sub-ranges associated with the second selling price range;

[0033] B112, counting the number of segments with the same price sub-interval as a reference number for solving the influence amount, and counting the number of first segments in which the price in each of the first price sub-intervals and / or each of the second price sub-intervals is less than the minimum price in each of the same price sub-intervals, and / or counting the number of second segments in which the price is greater than the maximum price in each of the same price sub-intervals;

[0034] B113, calculating a first ratio and a second ratio of the first segment number and / or the second segment number to the reference number respectively;

[0035] B114, calculate the first product of the first ratio and the price intersection interval classification level value as a negative downward influence amount, and / or calculate the second product of the second ratio and the price intersection interval classification level value as a positive upward influence amount.

[0036] Preferably, in step S3, the mutual recommendation influence relationship is constructed by the following method steps:

[0037] S31, constructing an influence relationship matrix of a product combination consisting of the proposed product and a building material product in the third recommendation list and including the maximum first variable value, and performing element value normalization processing, and then calculating the positive influence factor f and the reverse influence factor f′ of the proposed product and the building material product in the product combination;

[0038] S32, calculating the sixth deviation of the positive influence factor f and the negative influence factor f' associated with the product combination as the mutual recommendation influence relationship constructed for the product combination.

[0039] Preferably, in step S31, the constructed influence relationship matrix containing the maximum first variable value is expressed by the following expression (1):

[0040]

[0041] In expression (1), M represents the influence relationship matrix;

[0042] represents the jth characteristic value of the proposed recommended product under the i-th type of characteristics; represents the jth characteristic value of a building material product in the third recommendation basis list that constitutes a product combination with the proposed recommended product under the i′th type of characteristics, j=1,2,…,n, n represents the number of characteristic values ​​under the i-th or i′th type of characteristics; i or i′=1,2,…,I, I represents the number of characteristic types of the proposed recommended product or the building material product; v 1 、v 2 represents a first average value and a second average value respectively representing the common characteristics of brand pursuit and the common characteristics of consumption ability used to solve the maximum first variable value;

[0043] The positive impact factor f is calculated by the following method steps:

[0044] S311, calculating a first influence value of each element data pair generated from the proposed product in the normalized matrix after normalizing the matrix M to generate the first average value, and calculating a second influence value of each element data pair generated from the building material product that constitutes a product combination with the proposed product in the normalized matrix to generate the second average value;

[0045] S312: Calculate the weighted average of the first impact value and the second impact value as the positive impact factor f of the proposed product 历史 ;

[0046] In step S311, the calculation method of the first impact value or the second impact value is expressed by the following formula (2):

[0047]

[0048] In formula (2), x = 1, 2, c 1 represents the first impact value, c 2 represents the second impact value;

[0049] y=1,2;

[0050] v y ′ represents v y The normalized value of

[0051] b re Represents the eigenvalue of the element in the rth row and eth column of the matrix M;

[0052] w re Indicates b re In calculating c x The weight of time;

[0053] The values ​​of x and y may be the same or different.

[0054] The present invention has the following beneficial effects:

[0055] 1. Based on the screening conditions of each customer's historical purchase characteristics of the product combination including the proposed recommended product, a customer list is screened in advance as the basis for determining whether the proposed recommended customer is the target recommended customer. Subsequently, whether the proposed recommended customer is the target recommended customer is determined based on the historical purchase data of each customer in the customer list. This greatly reduces the amount of data analysis, which is conducive to improving the analysis speed of making product recommendations to a large customer group at the same time in scenarios where complex recommendation analysis and calculation are required to consider the mutual recommendation influence relationship between products.

[0056] 2. By introducing intermediate variables and calculating positive and reverse influencing factors, a mutual recommendation influence relationship is established between the first building material product and the second building material product, and the impact of differences in product characteristics between different products on the accuracy of product recommendations is taken into account. This makes it possible to subsequently find the common purchasing characteristics of the proposed recommended customers and historical purchasing customers, and based on the maximum first variable value of the intermediate variable of the proposed customer, accurately screen out target recommended products that can meet the brand pursuit and consumption capacity of the proposed recommended customers from a large number of proposed recommended products, thereby greatly improving the accuracy of building material product recommendations.

[0057] 3. By identifying and extracting the common purchase features of the same customer for the first building material product and the second building material product, and introducing the common purchase features as intermediate variables, it becomes possible to establish a mutual recommendation history influence relationship between the first building material product and the second building material product.

[0058] 4. In calculating the impact value c as the basis for calculating the historical positive impact factor or the historical negative impact factor x When the normalized element value b re and the normalized common purchase features v y The ratio of ′ quantifies the influence of each characteristic value of each type of characteristics of the first building material product and the second building material product on the common purchasing characteristics of customers, and is expressed by bre The corresponding weight w re To correct this degree of influence, the solved historical positive influence factor or historical negative influence factor can reflect the common purchasing characteristics of customers who purchase the first building materials product or the second building materials product.

[0059] 5. When calculating the historical positive impact factor, the common characteristics of brand pursuit of the first building material product (i.e., the first average value) is used as the basis for calculating the first impact value, and the common characteristics of consumption capacity of the second building material product (i.e., the second average value) is used as the basis for calculating the second impact value. The historical positive impact factor is obtained by weighted summing the first impact value and the second impact value, which characterizes the influence of the same customer's historical consumption capacity for the second building material product on the brand pursuit when purchasing the first building material product. Similarly, when calculating the historical reverse impact factor, the common characteristics of brand pursuit of the second building material product is used as the basis for calculating the first impact value, and the common characteristics of consumption capacity of the first building material product is used as the basis for calculating the second impact value. The historical reverse impact factor is obtained by weighted summing the first impact value and the second impact value, which characterizes the influence of the same customer's historical consumption capacity for the first building material product on the brand pursuit when purchasing the second building material product, making it possible to accurately recommend the proposed products to the proposed customers in the future. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0061] Figure 1 This is a diagram of implementation steps of a product recommendation method based on data pre-screening and introduction of intermediate variables provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0062] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific implementation methods.

[0063] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0064] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "inside", "outside", etc. appear to indicate the orientation or position relationship, it is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as a limitation on the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0065] In the description of the present invention, unless otherwise clearly specified and limited, if the term "connection" or the like appears to indicate the connection relationship between components, the term should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two components or the interaction relationship between two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0066] In this embodiment, taking the recommendation of building materials products as an example, the provided product recommendation method based on data pre-screening and introduction of intermediate variables is explained.

[0067] The embodiment of the present invention provides a product recommendation method based on data pre-screening and introduction of intermediate variables, such as Figure 1 As shown, the steps include:

[0068] S1, using the fact that each customer of the product portfolio that the customer has purchased in the past, including the product to be recommended, has similar common purchase characteristics for the product portfolio as a screening condition, and screening out a list of historical customers as a basis for determining whether the customer to be recommended is a target recommended customer;

[0069] Assume that the product combination is composed of the first building material product and the second building material product, the first building material product is assumed to be floor heating floor, and the second building material product is assumed to be central air conditioning. Assume that the product to be recommended is building material product a, and the historical purchasing customers include cuu1-cuu100, a total of 100 customers. Among these 100 customers, assume that each of 30 customers has purchased product a in the past, and each of them has also purchased at least other types of building materials products, and at least one other building material product purchased by each of them has formed a product combination with building material product a. After calculating the mutual recommendation historical influence relationship between the two building materials products in the product combination, these 30 historical purchasing customers are determined as: the historical purchasing customers of the product combination that has been purchased in the past include the historical purchasing customers of the product to be recommended.

[0070] Then, the 30 historical purchase customers are screened for similar common purchase characteristics of the product combination, and a list of customers is screened out as the basis for determining whether the proposed recommended customer is the target recommended customer. For example, taking customers cuu1, cuu2, and cuu3 among the 30 customers as examples, it is assumed that the product combination that customer cuu1 has purchased and has built a historical influence relationship of mutual recommendation includes building materials products a and building materials products b; the product combination that customer cuu2 has purchased and has built a historical influence relationship of mutual recommendation includes building materials products a and building materials products c; the product combination that customer cuu3 has purchased and has built a historical influence relationship of mutual recommendation includes building materials products a and building materials products d. The rule for judging which customers among these three customers have similar common purchase characteristics of the product combination is: the brand classification of the first building material product purchased by different customers in the product combination is the same, and the brand classification level of the brand purchased by the second building material product is the same, and the deviation between the first average values ​​of the common characteristics of brand pursuit of the product combination is less than the preset first common deviation threshold, and the deviation between the second average values ​​of the common characteristics of consumption capacity of the product combination is less than the preset second common deviation threshold. For example, assuming that the product to be recommended is building material product a and it is also the first building material product in the product portfolio, then building material product b purchased by customer cuu1 is the second building material product in the first product portfolio constructed by building material products a and building material products b; similarly, building material product c purchased by customer cuu2 is the second building material product in the second product portfolio constructed by building material products a and building material products c; building material product d purchased by customer cuu3 is the second building material product in the third product portfolio constructed by building material products a and building material products d. Assume that the brand classification level of the brand of building material product a purchased by customer cuu1 is the first level, and the brand classification level of the brand of building material product b is the first level; the brand classification level of the brand of building material product c purchased by customer cuu2 is also the first level; the brand classification level of the brand of building material product d purchased by customer cuu3 is the second level. Since the brand classification level of the brand of the first building material product (building material product a) purchased by customer cuu1 and customer cuu2 is the same, and the brand classification level of the brand of the first building material product (the brand classification level of building material product b purchased by customer cuu1 is the first level, and the brand classification level of the brand of building material product c purchased by customer cuu2 is also the first level), the first condition for determining that customer cuu1 and customer cuu2 have similar common purchasing characteristics is met, that is: the brand classification level of the brand of the first building material product in the product combination purchased by different customers is the same, and the brand classification level of the brand of the second building material product purchased is the same. The brand classification level of the brand of the second building material product purchased by customer cuu3, namely building material product d, is the second level, which is different from the brand classification level of the brand of the second building material product purchased by cuu1 and cuu2, so cuu3 is filtered out from the 30 customers.

[0071] Furthermore, it is necessary to determine whether the second condition of having similar common purchasing characteristics is met for cuu1 customers and cuu2 customers, that is, whether the deviation between the first average values ​​of the common characteristics of brand pursuit as a product combination is less than the preset first common deviation threshold, and whether the second condition of the deviation between the second average values ​​of the common characteristics of consumption ability as a product combination is less than the preset second common deviation threshold is met. For example, the first average value of the common characteristics of brand pursuit of building materials products a and building materials products b that cuu1 customers have purchased in the past is assumed to be The second average value representing the common characteristics of consumption capacity is The first average value of the common characteristics of the brand pursuit of building materials products a and c that customers have purchased in the past is assumed to be The second average value representing the common characteristics of consumption capacity is Then when and The absolute value of the difference (deviation) is less than the preset first commonality deviation threshold, and and When the absolute value of the difference (deviation) is less than the preset second common deviation threshold, the second condition with similar common purchase characteristics is determined to be achieved, and customers cuu1 and cuu2 are included in the customer list; otherwise, cuu1 or cuu2 or both cuu1 and cuu2 are filtered out from the 30 customers (when there is no other historical purchase with cuu1 that meets the above-mentioned second achievement condition, cuu1 is filtered out from the 30 customers, and the filtering condition for cuu2 is the same. When both cuu1 and cuu2 have no other historical purchasing customers that meet the second achievement condition, cuu1 and cuu2 are filtered out together).

[0072] The following is a detailed explanation of how to calculate the first average value of the common characteristics of the first building material product and the second building material product in the product portfolio, which represents the brand pursuit, and the second average value of the common characteristics of the consumption capacity:

[0073] The first average value representing the common characteristics pursued by brands is calculated by the following method steps:

[0074] A1, classify the first building material product and the second building material product that constitute the product portfolio into brands, and then assign the same brand classification grade value to each brand under each category;

[0075] For example, suppose that the brands of floor heating flooring as the first building material product include BP1, BP2, BP3, and BP4. Assuming that the brand level of brands BP1 and BP2 is the first level, and the brand level of brands BP3 and BP4 is the second level, brands BP1 and BP2 are assigned the same brand classification level value, and brands BP3 and BP4 are assigned another same brand classification level value. Brand classification level values ​​at different classification levels are assigned accordingly from high to low according to the classification level from high to low, that is, the brand classification level value of each brand at the first level is greater than the brand classification level value of each brand at the second level.

[0076] A2, calculating a first proportion of the first brand classification grade value of the first building material product of the corresponding brand purchased by the customer in the brand classification grade values ​​corresponding to each brand of the first building material product, and calculating a second proportion of the second brand classification grade value of the second building material product of the corresponding brand purchased by the same customer in the brand classification grade values ​​corresponding to each brand of the second building material product;

[0077] For example, the brand of floor heating floor that the customer has purchased in the past is BP1. The first brand classification level value corresponding to brand BP1 is assumed to be "1". The brand classification level value of brand BP2 of floor heating floor is also "1". The brand classification level values ​​corresponding to brands BP3 and BP4 are both 0.8. Then the first proportion is The calculation method of the second proportion is the same as the first proportion and will not be repeated here.

[0078] A3, determining whether a first deviation between the first proportion and the second proportion is less than a preset first deviation threshold,

[0079] If so, a first average value of the first brand classification grade value and the second brand classification grade value is calculated as the common characteristics of the brand pursuit of the first building material product and the second building material product by the customer;

[0080] If not, it is determined that the customer does not have common characteristics in brand pursuit for the first building material product and the second building material product.

[0081] For example, assuming the second proportion is The first deviation of the second proportion for the first proportion is the absolute value of the difference between the two, which is 0.035. Assuming that the first deviation threshold is 0.05, the "if" condition of step A3 is met, and the average value of the first brand classification level value and the second brand classification level value (defined as the first average value) is calculated as the common characteristics of customers' pursuit of the brand of the first building material product and the second building material product that constitute the product portfolio.

[0082] Two aspects need to be explained here. The first aspect is that the setting of the first deviation threshold is related to the number of brands of the first building material product and the second building material product, and the number of brand grade classifications. The first deviation threshold can be adjusted to capture the size difference in the common characteristics of brand pursuit between the first building material product and the second building material product. Since how to set the first deviation threshold is not within the scope of the rights protection required by this application, it will not be explained in detail. The second aspect is that the brand classification grade values ​​of each brand of the first building material product and the second building material product at the same classification level are preferably the same.

[0083] The second average value representing the common characteristics of consumption capacity is obtained by the following method steps:

[0084] B1, classify the selling price ranges of each brand under each brand level classification of the first building material product and the second building material product constituting the product combination, and assign corresponding price range classification level values ​​to each brand under the same brand level classification;

[0085] For example, for the above-mentioned floor heating floor as the first building material product, the selling price ranges of brands BP1 and BP2 with the first brand classification level are assumed to be RMB 200-400 and RMB 250-450 respectively. Then, the brand BP1 of the first level is assigned a price range classification level value corresponding to the selling price range of RMB 200-400, and the brand BP2 is assigned a price range classification level value corresponding to the selling price range of RMB 250-450.

[0086] The method for assigning the price range classification level values ​​corresponding to each brand under the same brand classification level comprises the following steps:

[0087] B11, determine whether there is any overlapping price range between brands in the same brand classification level.

[0088] If yes, then the influence of lowering the price and / or raising the price on the price range classification level value is solved for each brand with overlapping product price ranges, and then the process goes to step B12;

[0089] If not, then assign the price range classification level values ​​from high to low to the corresponding selling price ranges of each brand under the same brand classification level according to the highest price in the product price range;

[0090] For example, the selling price range of brand BP1 is 200-400, and the selling price range of brand BP2 is 250-450. It is determined that there is an overlapping area of ​​brand product prices between brands BP1 and BP2 with the same brand classification level, and the overlapping area is 250-400, then the "if yes" condition of step B11 is met.

[0091] When the "if" condition of step B11 is met, the influence amount in step B11 is calculated by the following method steps:

[0092] B111, segmenting the first selling price interval and the second selling price interval corresponding to the first brand and the second brand at the same brand classification level participating in the calculation of the influence amount respectively with a fixed step length, and obtaining a plurality of first price sub-intervals associated with the first selling price interval and a plurality of second price sub-intervals associated with the second selling price;

[0093] For example, the first selling price range (200-400) and the second selling price range (250-450) corresponding to the above-mentioned brand BP1 (first brand) and brand BP2 (second brand) are segmented with a fixed step size of "50" to obtain several first price sub-ranges associated with the first selling price range, which are: 200-250, 250-300, 300-350, 350-400, and several second price sub-ranges associated with the second selling price range, which are: 250-300, 300-350, 350-400, 400-450, and then proceed to step B112.

[0094] B112, counting the number of segments with the same price sub-interval as the reference number for solving the influence amount, and counting the number of first segments in which the price in each first price sub-interval and / or second price sub-interval is less than the minimum price in each same sub-interval, and / or counting the number of second segments in which the price is greater than the maximum price in each same price sub-interval;

[0095] Continuing with the above example, the same price sub-intervals of 200-250, 250-300, 300-350, 350-400 and 250-300, 300-350, 350-400, 400-450 are 250-300, 300-350, 350-400, and the number of segments with the same price sub-interval is "3", which is used as the reference number for solving the influence quantity. The minimum price in these three same sub-intervals is 250 and the maximum price is 400. The first price sub-interval of 200-250 satisfies the condition that the price is less than the minimum price in each same sub-interval, and the first segment number is "1". According to the same statistical method, the second segment number is also "1", and then the process goes to step B113.

[0096] B113, calculating a first ratio and a second ratio of the first segment number and / or the second segment number to the reference number respectively;

[0097] Continuing with the above example, the first ratio and the second ratio are both one-third.

[0098] B114, calculate the product of the first ratio and the price intersection interval classification level value (defined as the first product) as a negative downward influence amount, and / or calculate the product of the second ratio and the price intersection interval classification level value (defined as the second product) as a positive upward influence amount.

[0099] Continuing with the above example, 250-400 is the price intersection range of the first selling price range and the second selling price range. This price intersection range has the same price intersection range classification level value in the first selling price and the second selling price. Assuming it is defined as V1, the product of the first ratio and V1 is calculated as the downward influence amount, and the value of the downward influence amount is negative; the product of the second ratio and V1 is calculated as the upward influence amount, and the value of the upward influence amount is positive.

[0100] The method of assigning price intersection interval classification level values ​​is not within the scope of the rights protection claimed in this application, and has no direct impact on the technical problem solved by the present invention, so it will not be explained in detail.

[0101] When the "if no" condition in step B11 is met, the method of assigning the price range classification level value is given as follows:

[0102] Assume that the first selling price range is 200-400, and the second selling price range is 450-600, and the first selling price range and the second selling price range do not have a price overlapping range, and the highest prices of the first selling price range and the second selling price range are 400 and 600 respectively, then the highest price "600" has a larger price range classification level value than "400".

[0103] After the determination in step B11 is completed, the method of assigning the price range classification level values ​​corresponding to each brand under the same brand classification level is transferred to step:

[0104] B12, assigning the same price intersection interval classification level value to the price intersection interval in the selling price interval of the two brands whose influence amount is solved in step B11, and then adding the influence amount to the price intersection interval classification level value to form the price interval classification level value corresponding to the brand with lowering price and / or raising price. It should be emphasized here that the influence amount corresponding to lowering price is negative, and the influence amount corresponding to raising price is positive.

[0105] After completing the assignment of the price range classification level values ​​corresponding to each brand under the same brand classification level through steps B11-B12, the common characteristics of consumption capacity are analyzed and extracted and transferred to the following steps:

[0106] B2, calculating a first ratio of a first price range classification level value corresponding to a selling price range in which a purchase price of a first building material product of a corresponding brand purchased by a customer falls among price range classification level values ​​corresponding to each brand at the same brand classification level of the first building material product, and obtaining a second ratio of a second price range classification level value corresponding to a selling price range in which a purchase price of a second building material product of a corresponding brand purchased by the same customer falls among price range classification level values ​​corresponding to each brand at the same brand classification level of the second building material product;

[0107] For example, the first building material product purchased by customer U in the past was floor heating floor of brand BP1, and the purchase price was 230 yuan / square meter. The floor heating floor of brand BP1 has three models, and the selling price ranges are 150-250 yuan, 250-350, and 350-500 respectively. The price of 230 yuan / square meter falls into the selling price range of 150-250, and the first price range classification level value is the price range classification level value corresponding to the selling price range of 150-250.

[0108] Assume that brands with the same brand classification level as BP1 include BP2 and BP3. Under BP1, BP2 and BP3, there are 8 price range classification level values, namely v 11 -v 18 , Assume that the price range of 150-250 corresponds to the price range classification level value v 13 , then the first ratio is The calculation principle of the second ratio is the same as that of the first ratio and will not be repeated here.

[0109] B3, determining whether a second deviation between the first ratio and the second ratio (the absolute value of the difference between the first ratio and the second ratio) is less than a preset second deviation threshold,

[0110] If so, then the average value of the first price range classification level value and the second price range classification level value (defined as the second average value) is calculated as the common characteristics of the customer's consumption capacity for the first building material product and the second building material product constituting the product combination;

[0111] If not, it is determined that the customer does not have common consumption capacity characteristics for the first building material product and the second building material product.

[0112] It should be noted here that, similar to the setting principle of the first deviation threshold, the specific setting of the second deviation threshold is related to the types of the first building material product and the second building material product, the number of brands under the same brand classification level, the number of selling price ranges of each brand, the width of the selling price range, whether there are overlapping ranges between the selling price ranges, etc. By considering these factors and adjusting the second deviation threshold, the degree of tolerance of the difference in the consumption capacity characteristics of the customer when purchasing the first building material product and the second building material product is characterized as "commonality". The setting of the second deviation threshold and the first deviation threshold is a complicated process. Since the specific setting is not within the scope of the rights protection required by this application, no specific explanation is given.

[0113] After the customer list is screened out in step S1, the product recommendation method based on data pre-screening and introduction of intermediate variables provided in this embodiment proceeds to step:

[0114] S2, screening out the products that have been purchased by each customer in the customer list to be recommended, to form a third recommendation basis list;

[0115] For example, assuming that the customer list includes customer cuu1, customer cuu2 and customer cuu3, and cuu1 has historically purchased building materials a and building materials b, cuu2 has historically purchased building materials a and building materials c, and cuu3 has historically purchased building materials a and building materials d, then building materials a, building materials b, building materials c and building materials d are formed into the third recommendation list.

[0116] It should be noted here that when the proposed product is building material product a, the proposed customer has never purchased building material product a in history. In this application, whether building material product a, which is the proposed product, is actually recommended as the target recommended product to this proposed customer is mainly determined by two factors. First, the similarity of customer characteristics between the proposed customer and each historical purchasing customer referred to in the third recommendation list. Second, when the proposed customer purchases the proposed product based on the purchase characteristics of at least one building material product in the third recommendation list in the past, whether the mutual recommendation influence relationship between the proposed product and the building material products in the third recommendation list meets the recommendation requirements.

[0117] The first determining factor mentioned above has screened out a list of reference customers having customer characteristics similar to those of the proposed recommended customers through step S1 using "similar common purchase characteristics" as the screening condition.

[0118] In the above-mentioned second determining factor, when the customer to be recommended purchases the product to be recommended based on the purchase characteristics of at least one building material product in the third recommendation list in the past, whether the mutual recommendation influence relationship between the product to be recommended and the building material product in the third recommendation list meets the recommendation requirements is achieved by the method recorded in the following step S3:

[0119] S3, calculate the first variable value of the intermediate variable between each product in the third recommendation list of the historical purchase of the customer to be recommended, and extract the maximum first variable value, and then calculate the positive influence factor f and the reverse influence factor f′ of the product to be recommended and each product in the third recommendation basis list in sequence according to the maximum first variable value, and further calculate the mutual recommendation influence relationship between the product to be recommended and each building material product in the third recommendation basis list;

[0120] The calculation method of the variable value of the intermediate variable between two building materials products is specifically described in the above steps A1-A3 and steps B1-B3, and will not be repeated here. When there are multiple variable values ​​(defined as first variable values) of the intermediate variable between each product in the third recommendation list that the customer intends to recommend has purchased in history, that is, the customer intends to recommend has purchased more than two products in the third recommendation list, the maximum value of each first variable value is extracted, that is, the maximum first variable value, and then the recommendation influence relationship between the intended recommended product and each building materials product in the third recommendation basis list is calculated based on this maximum first variable value.

[0121] The mutual recommendation influence relationship in step S3 is constructed by the following method steps:

[0122] S31, constructing an influence relationship matrix of a product combination consisting of the proposed recommended product and a building material product in the third recommendation list and including the maximum first variable value, and performing element value normalization processing, and then calculating the positive influence factor f and the reverse influence factor f′ between the proposed recommended product in the product combination and the building material product;

[0123] The constructed influence relationship matrix containing the maximum first variable value is expressed by the following expression (1):

[0124]

[0125] In expression (1), M represents the influence relationship matrix;

[0126] It represents the jth characteristic value of the product to be recommended under the i-th type of characteristics; represents the jth characteristic value of a building material product in the third recommendation basis list that constitutes a product combination with the proposed product under the i′th type of characteristics, j=1,2,…,n, n represents the number of characteristic values ​​under the i-th or i′th type of characteristics; i or i′=1,2,…,I, I represents the number of characteristic types of the proposed product or building material product; v 1 、v 2 Represents a first average value and a second average value, which respectively represent the common characteristics of brand pursuit and the common characteristics of consumption ability, which are used to solve the value of the first variable.

[0127] For example, the characteristics of floor heating floor as a proposed product include comfort characteristics, environmental protection characteristics, energy-saving characteristics, price characteristics, etc., among which the comfort characteristics include thermal conductivity characteristics, heat resistance characteristics, constant temperature radiation characteristics, thermal stability characteristics, noise reduction and sound insulation characteristics, floor thickness characteristics, etc. The characteristics of building materials products such as central air conditioners in the third recommendation basis list that constitute a product combination with the proposed product and the specific characteristics under each characteristic category will not be given as examples.

[0128] It should be noted here that when constructing an influence relationship matrix for two building materials products (such as the first building materials product and the second building materials product) purchased by a customer in the past, the v in the above matrix 1 、v 2 It represents the first average value and the second average value of the common characteristics of brand pursuit and the common characteristics of consumption ability that respectively represent the variable values ​​used to solve the intermediate variables of the first building material product and the second building material product.

[0129] The positive impact factor f is calculated by the following method steps:

[0130] S311, calculating a first influence value of each element data pair generated from the proposed recommended product in the normalized matrix after normalizing the matrix M to generate a first average value, and calculating a second influence value of each element data pair generated from the building material product that constitutes a product combination with the proposed recommended product in the normalized matrix to generate a second average value;

[0131] The method of normalizing the matrix M is expressed by the following expression (3):

[0132]

[0133] I′ re It represents the normalized value of the element value in the rth row and the eth column in the matrix M after normalization;

[0134] I re Represents the element value of the rth row and eth column in the matrix M;

[0135] N represents the number of columns in the matrix M;

[0136] P represents the number of rows in the matrix M.

[0137] In step S311, the calculation method of the first impact value or the second impact value is expressed by the following formula (2):

[0138]

[0139] In formula (2), x = 1, 2, c 1 represents the first impact value, c 2 represents the second impact value;

[0140] y=1,2;

[0141] v y ′ represents v y The normalized value of

[0142] b re Represents the eigenvalue of the element in the rth row and eth column of the matrix M;

[0143] w re Indicates b re In calculating c x The weight of time;

[0144] The values ​​of x and y are the same or different. They are preferably the same to simplify the calculation process of positive or negative impact factors. re It is related to the importance of the feature type and each feature under the feature type in calculating the common purchase feature, and it is related to the product similarity of the first building material product (such as the proposed recommended product) and the second building material product (such as the building material product in the third recommendation basis list). re The method of granting is not within the scope of the rights protected by this application, so it will not be explained in detail.

[0145] S312, calculating the weighted average of the first impact value and the second impact value (ie, c 1 、c 2 The weighted average of ) is taken as the positive influencing factor f of the product to be recommended;

[0146] It should be noted here that the calculation method of the reverse impact factor f' is similar to the calculation method of f, the difference is that when calculating f', the weighted average of the first impact value generated by each element data pair of the second building material product in the normalized matrix to generate the first average value and the second impact value generated by each element data pair of the first building material product to generate the second average value is f'. In addition, when calculating the reverse impact factor, optionally, x≠y.

[0147] After completing the calculation of the positive influence factor f and the reverse influence factor f′ of the proposed recommended product, the method for constructing the mutual recommendation influence relationship in step S3 proceeds to step:

[0148] S32, calculating the deviation of the positive influencing factor f and the negative influencing factor f' associated with the product portfolio (defined as the sixth deviation, which is the absolute value of the difference between f and f') as the mutual recommendation influence relationship for constructing the product portfolio.

[0149] What needs to be explained here is that the mutual recommendation influence relationship constructed for two building materials products purchased by customers in the past, such as the first building materials product and the second building materials product, is defined as the mutual recommendation historical influence relationship. The mutual recommendation historical influence relationship is the same as the construction method of the above-mentioned mutual recommendation influence relationship in principle. The difference is that when constructing the mutual recommendation historical influence relationship, the mutual recommendation historical influence relationship is the average value (defined as the third average value) of the third deviations solved for the same product combination by different customers with similar common purchase characteristics. The third deviation is the historical positive influence factor f between the first building materials product and the second building materials product that constitute the product combination. 历史 and historical reverse impact factor f′ 历史 The absolute value of the difference. 历史 With f, f′ 历史 The calculation principle is the same as f′, so it will not be repeated here. It should also be noted that having similar common purchase characteristics for the same product combination means that when different customers purchase the first building material product and the second building material product in the product combination, the brand classification level of the first building material product purchased is the same, the brand classification level of the second building material product purchased is the same, and the deviation between the first average values ​​of the common characteristics of brand pursuit as the product combination is less than the preset first common deviation threshold, and the deviation between the second average values ​​of the common characteristics of consumption capacity as the product combination is less than the preset second common deviation threshold.

[0150] After calculating the mutual recommendation influence relationship between the proposed product and each building material product in the third recommendation basis list in step S3, the product recommendation method based on data pre-screening and introduction of intermediate variables provided in this embodiment proceeds to step:

[0151] S4, solve the fifth deviation (the absolute value of the difference between the mutual recommendation influence relationship and the mutual recommendation historical influence relationship) between the mutual recommendation influence relationship of the associated proposed recommended customers calculated for the same product combination and the mutual recommendation historical influence relationship between the historical purchasing customers referred to in the associated third recommendation basis list, and use the proposed recommended customers whose fifth deviation is less than the preset fifth deviation threshold as the target recommended customers for recommending the proposed recommended products.

[0152] For example, after going through step S3, for the same proposed recommended customer cus1, the first mutual recommendation influence relationship between product a and product b is calculated, and the second mutual recommendation influence relationship between product a and product c is calculated. Assume that historical purchasing customer cuu2 has purchased product a and product b, and the first mutual recommendation historical influence relationship is calculated for product a and product b purchased by customer cuu2; historical purchasing customer cuu3 has purchased product a and product c, and the second mutual recommendation historical influence relationship is calculated for product a and product c purchased by customer cuu3. Then calculate the fifth deviation between the first mutual recommendation influence relationship and the first mutual recommendation historical influence relationship; calculate the fifth deviation between the second mutual recommendation influence relationship and the second mutual recommendation historical influence relationship. As long as at least one of the two fifth deviations is less than the preset fifth deviation threshold, the proposed recommended customer cus1 will be used as the target recommended customer for recommending the proposed recommended product a.

[0153] Finally, it should be emphasized that the building materials product in this embodiment is only an explanatory example product. As long as the common purchasing features of customers between two products can be extracted, the product recommendation method provided in this application can be used to achieve accurate and efficient recommendations for products.

[0154] In summary, the present invention uses the similar common purchase characteristics of the product combinations that each customer has purchased in the past, including the proposed recommended product, as a screening condition, and pre-screens a list of customers as a basis for determining whether the proposed recommended customer is a target recommended customer. The subsequent determination of whether the proposed recommended customer is a target recommended customer is based on the historical purchase data of each customer in the customer list, which greatly reduces the amount of data analysis and is conducive to improving the analysis speed of product recommendations for a large customer group at the same time in the scenario where complex recommendation analysis and calculation are required to consider the mutual recommendation influence relationship between products. By introducing intermediate variables and calculating positive and negative influence factors, a mutual recommendation influence relationship is established between the first building material product and the second building material product, and the influence of product feature differences between different products on the accuracy of product recommendation is considered, so that the common purchase characteristics of the proposed recommended customer and the historical purchasing customer can be found in the future, and the maximum first variable value of the intermediate variable of the proposed recommended customer is used as the basis, and the target recommended product that can meet the brand pursuit and consumption ability of the proposed recommended customer can be accurately screened from a large number of proposed recommended products, greatly improving the accuracy of building material product recommendations.

[0155] It should be noted that the above specific implementations are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art should understand that various modifications, equivalent substitutions, changes, etc. can be made to the present invention. However, as long as these changes do not deviate from the spirit of the present invention, they should be within the scope of protection of the present invention. In addition, some terms used in the specification and claims of this application are not restrictive, but are only for the convenience of description.

Claims

1. A product recommendation method based on data pre-screening and introduction of intermediate variables, characterized in that: Includes steps: S1, using the fact that each customer of the product combination that the customer has purchased in the past, including the product to be recommended, has similar common purchase characteristics for the product combination as a screening condition, and screening out a list of base customers as a basis for determining whether the customer to be recommended is a target recommended customer; S2, filtering out products that have been purchased by the customer to be recommended and by each customer in the customer list, to form a third recommendation basis list; S3, calculating the first variable value of the intermediate variable between each product in the third recommendation basis list of the historical purchase of the proposed customer, and extracting the maximum first variable value, and then calculating the positive influence factor f and the reverse influence factor f of the proposed recommended product and each building material product in the third recommendation basis list in sequence according to the maximum first variable value ′ , and further construct the mutual recommendation influence relationship between the proposed recommended product and each building material product in the third recommendation basis list; S4, solving the fifth deviation between the mutual recommendation influence relationship between the intended recommended customers calculated for the same product combination and the mutual recommendation history influence relationship between the historical purchasing customers referred to in the third recommendation basis list, and taking the intended recommended customers who meet the condition that the fifth deviation is less than the preset fifth deviation threshold as the target recommended customers for recommending the intended recommended products.

2. The product recommendation method based on data pre-screening and introduction of intermediate variables according to claim 1 is characterized in that: In step S1, the rule for determining that each customer has similar common purchasing characteristics for the product combination is: different customers purchase the same brand classification level of the first building material product in the product combination, and the same brand classification level of the second building material product purchased by different customers, and the deviation between the first average values ​​of the common characteristics of brand pursuit of the product combination is less than a preset first common deviation threshold, and the deviation between the second average values ​​of the common characteristics of consumption capacity of the product combination is less than a preset second common deviation threshold; The intermediate variables include the common characteristics of brand pursuit and / or the common characteristics of consumption capacity associated with the same product combination; The variable value of the intermediate variable is the first average value or the second average value, or a weighted sum of the first average value and the second average value.

3. The product recommendation method based on data pre-screening and introduction of intermediate variables according to claim 2 is characterized in that: The method for calculating the first average value representing the common characteristics of the brand pursuit of the first building material product and the second building material product in the product portfolio comprises the steps of: A1, classifying the first building material product and the second building material product constituting the product combination by brand, and then assigning the same brand classification grade value to each brand under each category; A2, calculating a first proportion of the first brand classification grade value of the first building material product of the corresponding brand that the customer has purchased in the past among the brand classification grade values ​​corresponding to each brand of the first building material product, and calculating a second proportion of the second brand classification grade value of the second building material product of the corresponding brand that the same customer has purchased in the past among the brand classification grade values ​​corresponding to each brand of the second building material product; A3, determining whether a first deviation between the first proportion and the second proportion is less than a preset first deviation threshold, If yes, then calculating a first average value of the first brand classification grade value and the second brand classification grade value as the common characteristics of the brand pursuit of the first building material product and the second building material product by the customer; If not, it is determined that the customer does not have the common characteristics of brand pursuit for the first building material product and the second building material product.

4. The product recommendation method based on data pre-screening and introduction of intermediate variables according to claim 2 is characterized in that: The method for calculating the second average value representing the common characteristics of the consumption capacity of the first building material product and the second building material product in the product combination comprises the steps of: B1, classifying the selling price ranges of each brand under each brand grade classification of the first building material product and the second building material product constituting the product combination, and assigning corresponding price range classification grade values ​​to each brand under the same brand grade classification; B2, calculating a first ratio of a first price range classification level value corresponding to the selling price range into which the purchase price of the first building material product of the corresponding brand purchased by the customer in history falls, to the price range classification level values ​​respectively corresponding to each brand under the same brand classification level of the first building material product, and obtaining a second ratio of a second price range classification level value corresponding to the selling price range into which the purchase price of the second building material product of the corresponding brand purchased by the same customer in history falls, to the price range classification level values ​​respectively corresponding to each brand under the same brand classification level of the second building material product; B3, determining whether a second deviation between the first ratio and the second ratio is less than a preset second deviation threshold, If so, calculating a second average value of the first price range classification grade value and the second price range classification grade value as the common characteristics of the consumption capacity of the customer for the first building material product and the second building material product; If not, it is determined that the customer does not have the common consumption capacity characteristics for the first building material product and the second building material product.

5. The product recommendation method based on data pre-screening and introduction of intermediate variables according to claim 4 is characterized in that: In step B1, the method for assigning the price range classification level values ​​corresponding to the brands under the same brand classification level comprises the steps of: B11, determine whether there is any overlapping price range between brands in the same brand classification level. If so, the influence of lowering and / or raising prices on the price range classification level value is calculated for each brand with overlapping product price ranges, and then the process goes to step B12; If not, then assign the price range classification level values ​​from high to low to the selling price ranges corresponding to the brands under the same brand classification level according to the highest price in the product price range from high to low; B12, assigning the same price intersection interval classification level value to the price intersection intervals in the selling price ranges of the two brands for solving the influence amount, and then adding the influence amount to the price intersection interval classification level value to form the price interval classification level value corresponding to the brand with the price reduction and / or price increase.

6. The product recommendation method based on data pre-screening and introduction of intermediate variables according to claim 5 is characterized in that: The influence amount in step B11 is calculated by the following method steps: B111, segmenting the first selling price range and the second selling price range corresponding to the first brand and the second brand at the same brand classification level participating in the calculation of the influence amount respectively with a fixed step length to obtain a plurality of first price sub-ranges associated with the first selling price range and a plurality of second price sub-ranges associated with the second selling price range; B112, counting the number of segments with the same price sub-interval as a reference number for solving the influence amount, and counting the number of first segments in which the price in each of the first price sub-intervals and / or each of the second price sub-intervals is less than the minimum price in each of the same price sub-intervals, and / or counting the number of second segments in which the price is greater than the maximum price in each of the same price sub-intervals; B113, calculating a first ratio and a second ratio of the first segment number and / or the second segment number to the reference number respectively; B114, calculate the first product of the first ratio and the price intersection interval classification level value as a negative downward influence amount, and / or calculate the second product of the second ratio and the price intersection interval classification level value as a positive upward influence amount.

7. The product recommendation method based on data pre-screening and introduction of intermediate variables according to claim 1 is characterized in that: In step S3, the mutual recommendation influence relationship is constructed by the following method steps: S31, constructing an influence relationship matrix of a product combination consisting of the proposed recommended product and a building material product in the third recommendation list and including the maximum first variable value, and performing element value normalization processing, and then calculating the positive influence factor f and the reverse influence factor f between the proposed recommended product and the building material product in the product combination ′ ; S32, calculate the positive impact factor f and negative impact factor f associated with the product portfolio ′ The sixth deviation serves as the mutual recommendation influence relationship for constructing the product portfolio.

8. The product recommendation method based on data pre-screening and introduction of intermediate variables according to claim 7 is characterized in that: In step S31, the constructed influence relationship matrix containing the maximum first variable value is expressed by the following expression (1): In expression (1), M represents the influence relationship matrix; represents the jth characteristic value of the proposed recommended product under the i-th type of characteristics; Indicates that a building material product in the third recommendation basis list that forms a product combination with the proposed recommended product is in the i ′ The jth eigenvalue under the class characteristics, j = 1, 2, ..., n, n represents the i-th class or i-th ′ The number of eigenvalues ​​under the class feature; i or i ′ =1,2,…,I, where I represents the number of characteristic types of the proposed recommended product or the building material product; v1 and v2 represent the first average value and the second average value respectively representing the common characteristics of brand pursuit and the common characteristics of consumption capacity for solving the maximum first variable value; The positive impact factor f is calculated by the following method steps: S311, calculating a first influence value of each element data pair generated from the proposed product in the normalized matrix after normalizing the matrix M to generate the first average value, and calculating a second influence value of each element data pair generated from the building material product that constitutes a product combination with the proposed product in the normalized matrix to generate the second average value; S312, calculating a weighted average of the first influence value and the second influence value as the positive influence factor fhistory of the proposed recommended product; In step S311, the calculation method of the first impact value or the second impact value is expressed by the following formula (2): In formula (2), x=1, 2, c1 represents the first influence value, and c2 represents the second influence value; y＝1,2； v y ′ Indicates v y The normalized value of b re Represents the eigenvalue of the element in the rth row and eth column of the matrix M; w re Indicates b re In calculating c x The weight when ; the values ​​of x and y are the same or different.

Images