TWI626609B - Production method and system for personal product and vaccine purchase combination - Google Patents

Abstract

A production method and system for personal product and expense purchase combination, mainly comprising a product combination access component, a product combination analysis component, a product combination filter component, a product combination component and a product combination feedback component It extracts the customer record with the product combination access component, and filters the product portfolio and calculates the product combination score through the product combination analysis component, the product combination filter component and the product combination component, outputs the product combination, and then produces the product combination. The component calculates the product mix price and extracts the product portfolio weight, and finally the customer portfolio feedback component extracts the customer feedback data and calculates the product recommendation list.

Description

Production method and system for personal product and expense purchase combination

The invention relates to a product purchasing recommendation method and system, in particular to a production method capable of producing a personalized product and a cost purchasing combination suitable for a customer product combination recommendation according to a product recommendation that is repeatedly selected according to a plurality of recommendation results and a customer's product combination recommendation. And its system.

In order to cope with the current information diversification era, the company's product portfolio needs to meet customer needs, and its product portfolio changes are endless, resulting in the diversified product portfolio can be adapted to customer needs, and the common market portfolio recommendation methods are mostly Analyze the nature of the product to link with the characteristics of the customer. For example, the Patent Application No. 201227564 of the Republic of China on July 1, 2012, "Online Shopping Method and Website Server for Hot Products" is mainly based on consumers. The total number of clicks is recommended as the sort, but it is not considered that the total number of clicks on the new products will be less, and the number of total clicks in the morning products is higher, which makes it easier to recommend products to consumers. Most of the products that are put on the shelf earlier are not the most suitable and most popular products for consumers. In addition, the case uses the list provided by consumers as the source of recommended products, and it is easy to lead to more limited or too many recommendations. Product type of personal preference factor.

The main object of the present invention is to provide a production method for a personalized product and a cost-purchasing combination which can be used as a source of recommended product combinations according to customer needs, product portfolio association, product sales association, and customer acceptable price and cost performance. system.

The secondary object of the present invention is to provide a production method and system for a personalized product and cost purchase combination that can be adjusted according to product shelf time and product popularity.

Another object of the present invention is to provide a production method for a personalized product and a cost-purchasing combination which can be used as a reference value by referring to a product combination recommended by a customer and a product combination actually purchased by the customer, so as to improve the sorting accuracy of the recommended list. And its system.

A further object of the present invention is to provide a production method and system for a personalized product and a fee purchase combination that obtain a recommended list from a product combination list according to a combination of products and a cost performance, and a product combination that is actually purchased by the customer.

110‧‧‧Receive records

120‧‧‧Computed product portfolio

121‧‧‧Screening product portfolio

122‧‧‧Compute product portfolio score

123‧‧‧Output product mix sort

130‧‧‧Output recommendation list

131‧‧‧Computed product portfolio price/performance ratio

132‧‧‧Extract product portfolio weights

133‧‧‧Extract customer feedback information

134‧‧‧Computed product recommendation list

210‧‧‧Product Combination Access Components

220‧‧‧Product portfolio analysis component

230‧‧‧Product Combination Filter Element

240‧‧‧Product combination components

250‧‧‧Product portfolio feedback component

1 is a system architecture diagram of the present invention; FIG. 2 is a main flowchart of the present invention; FIG. 3 is a flowchart of a processing product processing procedure of the present invention; and FIG. 4 is a processing recommendation list processing procedure of the present invention. flow chart.

Please refer to FIG. 1 , a production method and system for the personalized product and expense purchase combination of the present invention, the system mainly comprising a product combination access component 210 , a product combination analysis component 220 , and a product combination filter component 230 . a product combination production component 240 and a product combination feedback component 250. The product combination access component 210 is mainly responsible for extracting information such as customer usage records and product shelves, including the time, purchase quantity, and purchase price of the customer to subscribe to a certain product. Combination and customer characteristics and demand conditions, product shelf time, sales amount, preferential items, product content, the product portfolio analysis component 220 is mainly responsible for analyzing the correlation between products and commodity sales analysis, and losing The input data is obtained by the product combination access component 210, and includes the time when the customer subscribes to a certain product, the purchase quantity, the purchase product combination, the product shelf time, the sales amount, the preferential item, and the product content, and the product combination filter component 230 is mainly responsible for processing The customer needs to perform screening and filtering, and the input data is obtained by the product combination access component 210, and includes the customer demand condition. The product combination production component 240 is mainly responsible for calculating the analyzed and filtered results through the calculation production recommendation list, and inputting the input. The data is obtained by the product combination analysis component 220 and the product combination filter component 230, and includes product inter-related degree, product sales analysis, and screening and filtering according to customer requirements, and the product portfolio feedback component 250 is mainly responsible for collecting customer feedback data for analysis. And the input data is obtained by the product combination production component 240, and includes a calculated production recommendation list.

Referring to FIG. 2, the production method of the personalized product and the expense purchase combination of the present invention comprises the following steps: Step 1: Extracting the record 110, including extracting the customer record, which is responsible for collecting the time and purchase of the customer to subscribe to a certain product. Quantity, purchase product combination, customer characteristics and demand conditions; extract product catalogue, which is responsible for recording product shelf time, sales amount, preferential items, product content; Step 2, calculating product portfolio 120, which is responsible for linking products to products The degree, the degree of association between the product and the customer is weighted, and the product portfolio is produced; and the third step, the output recommendation list 130, is responsible for numerically sorting and combining the product combinations to produce a product recommendation list.

Referring to FIG. 3, a production method of the personalized product and the expense purchase combination of the present invention, wherein the calculation product combination 120 of the second step further comprises the following steps: Step 1: screening the product combination 121, calculating and screening the product items. To find a product portfolio with a support greater than or equal to the minimum support, the support represents the ratio of the variables in this rule in the same record in the database. For example, if the total number of transactions is 10 million, and the number of occurrences of the product portfolio S ₁ "communication time 100 minutes + unlimited Internet access" is 2.5 million, the support degree is 2.5 million / 10 million = 0.25, which is greater than the minimum support of 0.2. Then it is included in the next stage analysis; the number of occurrences of product portfolio S ₂ "communication time 30 minutes + 1G Internet traffic" is 500,000, and its support is 500,000 / 10 million = 0.05, not greater than the minimum support of 0.2, then not Including the next stage analysis; Step 2, calculating the product combination score 122, which is responsible for the product combination S _j , 1 obtained by the screening product combination 121 j n , the sum score is calculated according to the weight of each factor, and the factor includes product sales popularity, product shelf time, customer characteristics/purchase product relationship. (Transaction number / total number of transactions comprising the portfolio S _J) Hot sales of the product of the combination of _J S Support (S _j), for a total of transactions such as a list of 100,000, wherein the product-containing composition has _an S transactions Five thousand pens, then Support ( S ₁ ) is 0.05. The shelf time factor OST ( S _j ) of the product combination S _j is the standardized shelf time (month), that is, ,0 OST ( S _j ) 1, where Max ( i ) is the maximum time (month) for all product shelves, Min ( i ) is the minimum time (month) for all product shelves, and x is the shelf time (months) for this product combination. Because the longer the product is put on the shelf, its popularity will be higher in theory. Therefore, it is necessary to use the time to offset the impact of the sales popularity factor to balance the chances of the old product and the new product being recommended. Therefore, the longer the shelf time is. For product items, the value of the shelf time factor will be smaller. If the maximum and minimum time for all product items are 60 months and 2 months respectively, and the shelf time of product combination S ₁ is 10 months, then OST ( S ₁ ) is (60-10)/(60-2 ) = 0.86, the shelf time of product portfolio S ₃ is 45 months, then OST ( S ₃ ) is (60-45) / (60-2) = 0.26. The customer relationship/purchase product relationship factor Relation ( S _j ) is . If the customer characteristics can be, gender For women, age 24 years old, professional For the financial industry, the customer information is listed as N customer characteristics C _i ,1 i N , where There are v _i values for the i- th customer feature. And for each customer feature , with each product combination S _j , can get customer feature relevance based on past sales records , that is, in the sales record (the number of customers who subscribe to S _j / all the same customer characteristics The number of customers), and finally add the customer feature relevance to the customer characteristics / purchase product relationship factor ,Such as That is, the relationship between the customer characteristics of "female + 20-29 years old + financial industry" and the purchase product combination S ₁ is 0.76. Finally, the product portfolio score is the sum of the above three factors (popularity, shelf time, customer characteristics and product relevance), W ( S _j )= w 1* Support ( S _j )+ w 2* OST ( S _j )+ w 3* Relation ( S _j ), because each factor has a different proportion, so each is multiplied by a weight value w _i ,1 i 3, the purpose of this step is to find out the customer's suitability product, so customer characteristics and product relevance are more important than other factors, so w 3 is set to 0.5; w 1 is set to 0.3; w 2 is set to 0.2 . The weight value of each factor can be adjusted by subsequent feedback analysis to obtain an optimized value. The consideration factor of this step includes not only the popularity of the product, but also the popularity of the product is usually too long, which may cause the popularity of the product to be missed. Therefore, the present invention also refers to the shelf time factor. To increase the chances of new shelves being recommended, thereby balancing the bias of product sales that are too popular or unpopular; and step three, output product mix ranking 123, according to the calculation of product portfolio score 122 step two to calculate the score of each product portfolio, And obtain the cost of the product portfolio, and then filter according to the customer's demand conditions or the maximum acceptable cost of self-input, filter out the product portfolio whose cost exceeds the customer's budget, and finally output the product portfolio list according to the score, as shown in Figure 5, the output project There are: product mix, sort, score, cost.

Please refer to FIG. 4, the production method of the personalized product and the expense purchase combination of the present invention, wherein the output recommendation list 130 of the third step further comprises the following steps:

Step 1: Calculate the product cost performance ratio 131. This step is to find the cost performance ( CP value) of the product portfolio, and define it as CP _k ( L , M ) to represent the service volume per unit of a certain k service. The value range is CP _k ( L , M ) (0,1], where the lower bound value does not contain L and the upper bound value contains M , and k 1~ n . Assume that if k =1 is defined as mobile communication, CP ₁ ( L , M ) is the price/performance ratio between L seconds and M seconds per call, and if k = 2 is defined as data communication, CP ₂ ( L , M ) is the price/performance ratio of L Mbps to M Mbps per element, and if k = 3 is defined as the channel, CP ₃ ( L , M ) is the channel that can be used per unit. The number is between L and M , etc., etc., and so on. Then define each CP _k ( L , M ), taking k =1 as the mobile communication, CP ₁ (0,10)=0.1, CP ₁ (10,20)=0.2, CP ₁ (20,30)= 0.3, CP ₁ (30, 40) = 0.4, CP ₁ (40, 60) = 0.5, CP ₁ (60, 80) = 0.6, CP ₁ (80, 100) = 0.7, CP ₁ (100, 120) = 0.8, CP ₁ (120,200)=0.9, CP ₁ (200,∞)=1; take k =2 as data communication as an example, CP ₂ (0,15)=0.1, CP ₂ (15,20)=0.2, CP ₂ (20 , 25) = 0.3, CP ₂ (25, 30) = 0.4, CP ₂ (30, 35) = 0.5, CP ₂ (35, 40) = 0.6, CP ₂ (40, 45) = 0.7, CP ₂ (45 , 50) = 0.8, CP ₂ (50, 100) = 0.9, CP ₂ (100, ∞) = 1, etc., and so on. Therefore, when the combination product may be mobile communications, data communications and the like ... channel combinations can be obtained the CP _k (L, M) where k 1~ n , but the number of each product combination may not be equal, so its n value will be the product quantity of the product portfolio, and finally through the formula The cost performance value of the product combination is obtained, and the result is defined as α ; Step 2, extracting the product combination weight 132, the step is to extract the weight value of the product combination, and define it as WT _k and k 1~ n , the definition of this part k is the same as step one, that is, k =1 is the mobile communication, etc., and so on, the n value will be the product quantity of the product combination. And the WT _k value comes from the [calculation product portfolio] 120 result, so the last pass through the formula The weight value of the product combination is obtained, and the result is defined as β ; Step 3: Extract customer feedback data 133, this step will obtain customer feedback data to generate a basic aggregate, and the data source is after multiple recommendation to customer product portfolio, The product portfolio purchased by the actual customer is used as customer feedback data. The data that has been recommended to the customer is defined as FC _i ( α , β ) where i 1~ m , its m value will be the number of product combinations that all customers actually choose. Then find out that the basic aggregate of these customer feedback data is defined as ω , assuming that R ( FC _i , ω ) represents the degree of association between the customer feedback data and the basic aggregate, then all customer feedback data FC _i ( α , β ) to the basic set The sum of the body ω correlations is the largest, that is, The maximum value, which is the basic aggregate ω representing the customer feedback data; and the fourth step, the calculation product recommendation list 134, which is finally produced in the product recommendation list, which will utilize the basic aggregate ω of the customer feedback data in step 3, and Step 1 calculates the product combination cost performance 131 and step 2 extracts the product combination weight 132, and obtains the α and β values by the [computation product combination] 120 product combination list data, and defines it as F _j ( α , β ) where j 1~ s , its s value will be the number of combination lists produced by [Calculate Product Portfolio] 120. Then F _j (α, β) of these products in combination with the basic information comparison Correlative assembly [omega], is defined for each product composition data F _j (α, β) to the base assembly [omega] is the correlation degree R (F _j, ω ) where j 1~ s . Therefore, it can be assumed that the greater the degree of association, that is, the larger the value of R ( F _j , ω ), the more consistent with the actual selection of the product portfolio of the customer, and the ranking of the product portfolio recommendation list that is currently recommended for the final output. However, the result of the final selection of the same customer will change the content of the basic aggregate ω of the customer feedback data in step three, and the calculated correlation difference will also change, so the method for retrieving the product recommendation list can be achieved. The retrievable product recommendation list method of the present invention will be recommended to the customer's product portfolio, and the actual purchased product combination is used as customer feedback data, and the feedback information includes the relationship between the recommended ranking of the product portfolio and the cost performance, and the recommended ranking The factor includes popularity, shelf time and customer characteristics/purchase product relationship, so it can reflect consumers' consideration of the above factors, thereby finding a product portfolio that is more suitable for consumers and improving the accuracy of recommendation list sorting.

As described above, the production method of the personalized product and the expense purchase combination of the present invention mainly utilizes the product combination access component 210 to extract the information of the customer use record and the product catalogue, and then the customer use record and the product. The information on the catalogue is analyzed by the product portfolio analysis component 220 for product correlation and merchandise sales analysis, and the product portfolio filter component 230 is selected according to the analysis result and the customer demand to find a product portfolio with a support degree greater than or equal to the minimum support degree. Then, the product combination component 240 calculates the sum score according to the factored weight of the selected product combination, and obtains the cost of the product combination. At this time, the product combination filter element 230 filters out according to the customer demand condition or the maximum acceptable fee. The product portfolio exceeds the customer's budget, and the product portfolio component 240 sorts the output product portfolio according to the sum score of each product portfolio, and Calculate the cost performance of each product combination and extract the weight value of each product combination. Finally, the product portfolio feedback component 250 will provide the customer's product portfolio list and the customer's actual product combination as the customer feedback data, and extract the customer feedback data. The basic aggregate, then the basic aggregate is compared with each product portfolio, and a recommendation list is generated according to the degree of relevance.

In summary, this case has the following features and functions compared to the recommended product combination recommendation method and system:

1. The product combination feedback method of the present invention will be recommended to the customer's product portfolio, and the actual purchased product combination is used as customer feedback data. The feedback information includes the relationship between the recommended ranking and the cost performance, and can reflect the consumer's Factor considerations and find a product portfolio that is more suitable for consumers.

2. In addition to the recommendation of the product, the invention considers the cost-effective relationship, takes the recommended product cost ratio into consideration, and comprehensively analyzes the relationship between the product recommendation result and the cost performance, so as to achieve a product that can be recommended to the customer and is worthy.

3. The consideration of the present invention may vary depending on the shelf time. It may have an early shelf time and more people have an impression. As a result, the popularity of the new products is affected. Therefore, the time of the shelves is taken into consideration to reduce the new shelves. The impact of little people knowing.

Claims (14)

A production method of a personalized product and a cost purchase combination, the main steps include: Step 1: Using the product combination access component to extract customer usage records and product listing information; Step 2, using customer usage records and product shelves The catalogue information uses the product portfolio analysis component to analyze the correlation between the products and the merchandise sales. Step 3: The product portfolio filter component selects the product portfolio with the support degree greater than or equal to the minimum support level according to the analysis result and the customer demand; 4. The product combination production component calculates the sum score according to the factor of the selected product combination and obtains the cost of the product combination; Step 5: The product combination filter component filters out the cost according to the customer demand condition or the maximum acceptable cost. Product portfolio of customer budget; Step 5: Product component production component sorts the product portfolio list according to the sum score of each product portfolio, and calculates the cost performance of each product portfolio and extracts the weight value of each product combination; Step 6: Product portfolio feedback component Products that will be delivered to customers multiple times The list and the product combination actually purchased by the customer are used as the customer feedback data, and the basic collection according to the customer feedback data is extracted; step 7, the product combination feedback component compares the basic assembly with each product combination data, and The degree of association produces a list of recommendations. The production method of the personalized product and the fee purchase combination described in claim 1, wherein the customer usage record and the product listing information include the time when the customer subscribes to a certain product, and the number of purchases. Quantity, purchase product combination, product shelf time, sales amount, preferential items and product content. The production method of the personalized product and expense purchase combination described in claim 1, wherein the support ratio refers to a ratio of occurrence of the same transaction in the total number of transactions. The production method of the personalized product and expense purchase combination described in claim 1, wherein the product combination factor includes product sales popularity, product shelf time, customer characteristics, and purchase product relationship. The production method of the personalized product and the fee purchase combination according to claim 1, wherein the basic aggregate is the aggregate of the sum of the correlations in the customer feedback data. The production method of the personalized product and expense purchase combination according to claim 1, wherein the product combination list includes a product combination, a ranking, a score, and a fee. The production method of the personalized product and expense purchase combination according to claim 1, wherein the product combination feedback component defines the product portfolio data as F _j (α according to the basic aggregate, the price/performance ratio and the weight value of the product combination. After β , it is compared with the basic aggregate. A production system for personal product and expense purchase combination, which mainly comprises: a product combination access component, which is used for extracting customer usage records and product listing information; a product portfolio analysis Component, the product combination analysis component is based on the information extracted by the product combination access component, and analyzes the correlation between products and the sales of the product; a product combination filter component, which is analyzed by the product combination analysis component The information is filtered and filtered according to customer needs, and the product portfolio with support greater than or equal to the minimum support is found, and the product portfolio whose cost exceeds the customer's budget is filtered according to the customer's demand or the maximum acceptable fee; A product combination production component, the product combination component is a product combination of filtering and filtering the product combination filter component, outputting a product combination list according to the factor weighted calculation total score, and calculating the cost performance of each product combination in the product combination list. And extracting the weight value; a product combination feedback component, the product combination feedback component is a product combination list and a product combination actually purchased by the customer as customer feedback data, and extracting a basic aggregate according to the customer feedback data, and the basic assembly body Compare with each product portfolio and then generate a recommendation list based on the degree of relevance. The production system of the personalized product and expense purchase combination according to claim 8, wherein the product combination feedback component defines the product portfolio data as F _j (α according to the basic aggregate, the price/performance ratio and the weight value of the product combination. After β , it is compared with the basic aggregate. The production system of the personalized product and the expense purchase combination described in claim 8, wherein the customer usage record and the product catalogue information include the time when the customer subscribes to a certain product, the quantity purchased, the product combination purchased, and the product shelf time. , sales amount, preferential items and product content. The production system of the personalized product and expense purchase combination as claimed in claim 8, wherein the support ratio is the ratio of the same transaction occurring in the total number of transactions. The production system of the personalized product and expense purchase combination according to claim 8, wherein the product combination factor includes product sales popularity, product shelf time, customer characteristics, and purchase product relationship. The production system of the personalized product and the expense purchase combination according to claim 8, wherein the basic aggregate is the aggregate of the sum of the associations in the customer feedback data. The production system of the personalized product and expense purchase combination of claim 8, wherein the product portfolio list includes product combinations, rankings, scores, and fees.