TWI615789B - Commodity recommending method - Google Patents

Abstract

The invention discloses a product recommendation method, which comprises the following steps: inputting transaction information. Generate feature time series and potential time series. A bridging pattern is generated based on the feature time series and the potential time series. The feature time series are merged according to the bridging pattern to produce a performance time series. The item code following the item code corresponding to the purchase habit sequence in the time series is provided.

Description

Product recommendation method

The present invention relates to a product recommendation method, and more particularly to a product recommendation method for generating a performance time series according to a bridging pattern, and correspondingly recommending related products in accordance with a purchase habit sequence.

Nowadays, Internet technology is developed, and the public has become accustomed to purchasing goods through the Internet without having to go shopping.

However, in the past, most of the methods for recommending products on shopping websites adopted the association rule, which used group behavior to predict individual behavior.

However, because it predicts individuals by group behavior, the results obtained by statistical analysis of group behavior are not necessarily suitable for some individuals; therefore, how to solve the above problems will be urgently needed and discussed. of.

In view of the above-mentioned problems, it is an object of the present invention to provide a product recommendation method for solving the problems faced by the prior art.

Based on the above object, the present invention provides a product recommendation method comprising the steps of setting a continuous sales threshold for each product category. During a plurality of calculation times, the commodity codes of each commodity type that reach the corresponding continuous sales threshold in each calculation time are sequentially recorded to obtain a plurality of characteristic time series. Between each characteristic time series, each commercial product corresponding to the consumption behavior is recorded sequentially The commodity code of the class to obtain a plurality of potential time series. Searching for a bridge pattern of at least one potential time series connecting the boundaries of the previous feature time series and the latter feature time series. Combining the previous feature time series with the latter feature time sequence according to at least one bridge pattern to obtain a performance time series. When the plurality of commodity codes included in the purchase habit sequence meets at least one partial sequence of the performance time series, at least one commodity code located in the at least one performance time series after the partial sequence is sequentially provided.

Preferably, before obtaining the plurality of feature time series, the following steps may be included: inputting a plurality of transaction information, each transaction information including a sales time and an item code. The relative cycle time is generated based on the sales time. The transaction time series is generated based on the commodity code and the relative cycle time.

Preferably, the relative cycle time may include a relative number of years, a week of weeks, a number of days per week, and an hour.

Preferably, the bridging pattern may include a starting bridge pattern in which one potential time series is from the first Ni-1 commodity codes and the last commodity code of the previous feature time series, and one of the potential time series is from the last number Nj-1 The end bridge style of the first item code of the item code and the last feature time series.

Preferably, when the purchase habit sequence conforms to a part of the sequence of the plurality of performance time series, the number of at least one commodity code after the partial sequence may be determined according to the plurality of performance time series, and the performance time is determined by at least a sorting manner. The sequence provides the order of at least one commodity code after the partial sequence.

Preferably, when the purchase habit sequence conforms to a part of the sequence of the plurality of performance time series, the performance time may be determined by the order of the at least one commodity code after the partial sequence according to the plurality of performance time series. The sequence provides the order of at least one commodity code after the partial sequence.

Preferably, the feature time series, the late time series and the purchase habit sequence may be composed of a plurality of commodity codes.

As described above, the product recommendation method of the present invention can find the feature time series and the potential time series by inputting the transaction information, and further find the bridge pattern between the feature time series and the potential time series, and then merge into the performance time. The sequence, in order to make the consumer's purchase habit sequence conform to the partial sequence of the performance time series, may provide the commodity corresponding to the product type code after the partial sequence is recommended, so as to more accurately recommend the product according to the consumer's consumption pattern.

Steps S11 to S16, S21 to S23, S301 to S313‧‧

Fig. 1 is a first flow chart of the product recommendation method of the present invention.

Fig. 2 is a second flow chart of the product recommendation method of the present invention.

Figure 3 is a third flow chart of the product recommendation method of the present invention.

In order to understand the features, contents, and advantages of the present invention, and the advantages thereof, the present invention will be described in conjunction with the drawings, and the description of the embodiments will be described in detail below. The use of the present invention and the accompanying drawings are not necessarily the true proportions and precise configurations of the present invention. Therefore, the scope and configuration of the attached drawings should not be construed as limiting the scope of the invention.

The advantages and features of the present invention, as well as the technical methods of the present invention, are described in more detail with reference to the exemplary embodiments and the accompanying drawings, and the present invention may be implemented in various forms and should not be construed as limited thereby. The embodiments of the present invention, and the embodiments of the present invention are intended to provide a more complete and complete and complete disclosure of the scope of the present invention, and The scope of the patent application is defined.

Please refer to FIG. 1 , which is a first flowchart of the product recommendation method of the present invention. As shown in the figure, the product recommendation method of the present invention comprises the following steps:

In step S11: the continuous sales threshold of each product category is set.

In step S12, the commodity codes of the commodity types that reach the corresponding continuous sales thresholds in each calculation time are sequentially recorded in a plurality of calculation times to obtain a plurality of feature time series.

In step S13, the commodity codes of the commodity types corresponding to the consumption behavior are sequentially recorded between the feature time series to obtain a plurality of potential time series.

In step S14: searching for a bridge pattern of at least one potential time series connecting the boundary of the previous feature time series and the latter feature time series.

In step S15, the previous feature time sequence and the subsequent feature time sequence are merged according to the at least one bridge pattern to obtain a performance time series.

In step S16, when the plurality of commodity codes included in the purchase habit sequence meets a partial sequence of at least one performance time series, at least one commodity code located after the partial sequence in the at least one performance time series is sequentially provided. Wherein, providing at least one commodity code is an item that recommends at least one commodity code.

Please refer to FIG. 2, which is a second flow chart of the product recommendation method of the present invention. As shown in the figure, the following steps can be included before the plurality of feature time series are obtained:

In step S21, a plurality of transaction information is input, and each transaction information includes a sales time and an item code.

In step S22: a relative cycle time is generated based on the sales time.

In step S23: a transaction time sequence is generated based on the commodity code and the relative cycle time.

For example, there are three transaction time series, which are 0105110, 14, 0006716, 6, 0007218, 7. Among them, 0105110, 14 series indicates that the commodity code 14 was sold out at 10 am on the first day of the fifth week of 1 year ago; 0006,716, 6 indicates that the goods code 6 was received at 4 pm on the 7th day of the 6th week before this year. Sales out; 0007218, 7 series said that the goods code 7 was sold out at 6 pm on the second day of the 7th week before this year.

In other words, the above-mentioned continuous sales threshold corresponds to the sale of up to 5 groups on the same day as the commodity code 14, or the sale of the commodity code 6 on the same day to more than 10 groups; among them, the continuous sales threshold can be different The products are adjusted and set, and the continuous sales threshold for multiple products is not necessarily the same.

The calculation time mentioned above begins when the continuous sales volume of one of the commodity codes reaches the continuous sales threshold and ends below the continuous sales threshold setting.

Further, if there is a commodity code that reaches the continuous sales threshold in the calculation time, it is arranged in chronological order. Taking the above-mentioned 0105110, 14, 0006716, 6, 0007218, 7 as an example, the commodity code 14 first reaches the continuous sales threshold, and the commodity The code 6 is ranked, and the commodity code 7 is finalized, and the characteristic time series 14-6-7 composed of the oil commodity code is obtained.

It is worth mentioning that, for example, 0105110,14, which means that the commodity code 14 is sold out at 10 am on the first day of the fifth week of 1 year ago, it is known that "0105110" is the relative cycle time. The relative cycle time includes the relative number of years (01), the number of weeks of the week (05), the number of days per week (1), and the number of hours (10), wherein the number of hours is 24 hours, to compare the number 5 years ago. The sales time at 10 am on the first day of the week.

The potential time series mentioned above is constituted by the consumer's consumption behavior, and it is preferred here that the consumer puts the goods into the shopping cart or checkout, but not limited thereto. The potential time series is A sequence generated at a time other than the characteristic time series, and a joint point of the potential time series and the feature time series is defined as a bridge pattern.

In other words, the bridging pattern detects the randomness range by detecting the correctness of the positive positive near 50% positive and negative 10%. The potential time series is continuous from the beginning of Ni-1 and from the tail to the continuous Nj. - 1 commodity code, plus the last commodity code of the previous feature time series and the first commodity code of the latter feature time series, respectively, the start bridge style and the end bridge style, wherein Ni, Nj can be set up.

In other words, the bridging pattern may include a starting bridge pattern in which one potential time series consists of the first Ni-1 commodity codes and the last commodity code of the previous feature time series, and one of the potential time series is represented by the last number Nj-1 The end bridge style of the first item code of the item code and the last feature time series.

For example, the current feature time series is 14-6-7, the late time series is 14-2-1-5, and the latter feature time series is 22-2-6, when Ni=2, Nj=2 Because Ni-1=1, Nj-1=1, the potential time series is 14 from the first product code, 5 from the last product code, and the last product code in the previous feature time series is 7, and the latter feature time series, the first commodity code is 22, the start bridge style is 7-14, and the end bridge style is 5-22.

On the other hand, if Ni=3 and Nj=3, since Ni-1=2 and Nj-1=2, the potential time series is from the first two commodity codes to 14-2, and the last two commodity codes. For 1-5, the last item code of the previous feature time series is 7, and the first item code of the latter feature time series is 22, which is 7-14-2 at the beginning of the bridge mode, and the bridge mode is ended. For 1-5-22, and so on.

For another example, Ni=2 and Nj=2, so Ni-1=1 and Nj-1=1.

As can be seen from the above example, the start bridging pattern is 5-12, and the ending bridging pattern is 8-2. In addition, the above two sets of pre- and post-feature time series can be combined (and the potential time series are not merged), and two sets of performance time series are generated. as follows:

Df1:14-6-4-2-8-5-2-5-6-3-4-5

Df2: 14-6-4-2-8-5-2-14-7-2-5

Further, when the commodity code on a consumer C1 shopping cart (or at checkout) is measured as 4-2-8, the purchase habit sequence H1: 4-2-8 called C1 is the same consumer and the same Trading, so do not consider the purchase order problem, so this purchase habit sequence and its combination of combinations S1 = {H1|4, 2, 8} (such as 2-8-4, 8-2-4, etc.), found by comparison The partial sequence (4-2-8) in the time series Df1 is expressed, so that the electronic system actively recommends the product with the commodity code 5 after the partial sequence 4-2-8 (the commodity code after the partial sequence). When C1 is not satisfied or not adopted, the products with the product codes of 2, 6, 3, and 4 are recommended in order. On the other hand, since S1 simultaneously coincides with a partial sequence of another performance time series Df2, the commodity codes 14, 7, 2 are also products that can be recommended to consumers.

It is worth mentioning that when focusing on the economic benefit priority recommendation strategy, after the purchase habit sequence conforms to a part of the sequence of the plurality of performance time series, the number of at least one commodity code after the partial sequence is based on the plurality of performance time series. The order of at least one commodity code after the partial sequence is provided by each performance time series is determined by at least a sorting manner.

That is to say, with the above example, the product code 5-2-6-3-4 can be recommended because of the partial sequence of the time series Df1, and the recommended quantity is more than the product code 5 recommended by the other performance time series Df2. -2-14-7, so the product that represents the time series Df1 in the partial code after the partial code will be recommended first.

On the other hand, if the purchase habit sequence conforms to a part of the sequence of the plurality of performance time series, the number of at least one commodity code after the partial sequence is determined according to the plurality of performance time series, At most, the ordering manner determines the order of at least one commodity code after each of the performance time series provides a partial sequence.

That is to say, with the above example, the product code 5-2-14-7 can be recommended because of the partial sequence of the time series Df2, and the recommended quantity is less than the product code 5-2 recommended by the other performance time series Df1. -6-3-4, therefore, it is recommended to first display the commodity code of the commodity code after the partial sequence Df2.

In addition, the feature time series, potential time series and purchase habit sequence mentioned above may be composed of a plurality of commodity codes.

Please refer to FIG. 3, which is a third flow chart of the product recommendation method of the present invention. As shown in the figure, the detailed process of the product recommendation method of the present invention is to first input the sales time and its product code (S301), and then convert the sales time to generate a relative cycle time (S302), and then generate the type code to generate The transaction time series (S303), and the number of commodity types is calculated (S304).

Continuing, after calculating the number of product types, it is determined whether the feature time series (S305) is started, and if so, it is monitored by the feature monitoring module (S306), and after checking, it is determined whether the feature time series is ended (S307), and if so, Then, the connection point is identified by the bridge pattern module (S308), and if not, the process returns to step S301.

After determining whether the feature time series (S305) is started, if not, proceeding to step S308 to identify the connection point by the bridge pattern module, and then determining whether the potential time series is started (S309), and if so, by the connection point detection module The detection (S312), after checking, determines whether the potential time series is over (S313), and if yes, returns to step S304, and if not, returns to step S312 and detects by the connection point detection module.

However, the above mentioned determining whether the potential time series starts, if not, searching by the feature time series search module (S310), and merging the feature time series (S311) by the feature time series, and checking after checking Whether the feature time series ends (S307). However, the above is only an example and should not be limited to this.

As described above, the product recommendation method of the present invention can find the feature time series and the potential time series by inputting the transaction information, and further find the bridge pattern between the feature time series and the potential time series, and then merge into the performance time. The sequence, in order to make the consumer's purchase habit sequence conform to the partial sequence of the performance time series, may provide the commodity corresponding to the product type code after the partial sequence is recommended, so as to more accurately recommend the product according to the consumer's consumption pattern.

The embodiments described above are merely illustrative of the technical spirit and the features of the present invention, and the objects of the present invention can be understood by those skilled in the art, and the scope of the present invention cannot be limited thereto. That is, the equivalent variations or modifications made by the spirit of the present invention should still be included in the scope of the present invention.

S11 to S16‧‧‧ steps

Claims (7)

A product recommendation method, comprising the steps of: setting a continuous sales threshold of one of the product categories; and recording, in the plurality of calculation times, one of each of the commodity types that reach the corresponding continuous sales threshold within the calculation time a commodity code to obtain a plurality of feature time series; each of the feature time series sequentially records the commodity code corresponding to each commodity category of a consumer behavior to obtain a plurality of potential time series; searching for at least one potential The time series is connected to a bridge pattern of the boundary between the feature time sequence and the latter feature time sequence; and the previous one of the feature time series and the subsequent one of the feature time series are combined according to at least one of the bridge patterns to obtain a performance time. a sequence; and when a plurality of the commodity codes included in a purchase habit sequence meet at least one partial sequence of the performance time series, at least one of the commodity codes in the performance time series subsequent to the partial sequence is sequentially provided. The method for recommending goods according to claim 1, wherein the obtaining the plurality of characteristic time series comprises the steps of: inputting a plurality of transaction information, each transaction information including a sales time and the commodity code; The sales time generates a relative cycle time; and a transaction time series is generated based on the commodity code and the relative cycle time. The method for recommending goods as described in claim 2, wherein the relative cycle time comprises a relative number of years, a week of weeks, a number of days per week, and an hour. The method for recommending goods according to claim 1, wherein the bridging pattern comprises one of the potential time series starting from a plurality of consecutive commodity codes and one of the last one of the commodity time codes of the previous feature time series. The style, and one of the potential time series ends the bridging pattern from one of the plurality of commodity codes and one of the first one of the commodity time codes of the subsequent feature time series. The method for recommending merchandise according to claim 1, wherein when the purchase habit sequence conforms to the partial sequence in the plurality of performance time series, the plurality of performance time series are at least after the partial sequence A quantity of the commodity code, in which the order of at least one of the commodity codes after the partial sequence is provided in each of the performance time series is determined by at least a sorting manner. The method for recommending merchandise according to claim 1, wherein when the purchase habit sequence conforms to the partial sequence in the plurality of performance time series, the plurality of performance time series are at least after the partial sequence A quantity of the commodity code, in an order of as little as possible, determining the order in which the performance time series provides at least one of the commodity codes after the partial sequence. The product recommendation method of claim 1, wherein the feature time series, the potential time series, and the purchase habit sequence are composed of a plurality of the commodity codes.