A data fusion global user industry portrait implicit demand mining method

Abstract

This invention relates to the field of demand marketing, addressing the problem that traditional technologies lack the ability to systematically mine users' implicit needs based on inter-industry relationships. Specifically, it is a data fusion-based method for mining implicit needs from comprehensive user industry profiles. This method includes sequentially executing steps such as industry user profile classification, determination of explicit demand intensity and quantity, acquisition of inter-industry correlation quantity and strength, implicit demand mining, and attenuation parameter binding and output. Based on a set of user profiles from all involved industries, it extracts transaction behavior sequences and search behavior sequences according to a set statistical period, quantifies the intensity and quantity of explicit needs, constructs an industry correlation scattering graph based on the correlation quantity and strength between different industries, uses edge weights to reflect the tightness of inter-industry correlations, and outputs the mining results with threshold filtering. Finally, it corrects the mining results for timeliness through attenuation periods, achieving a quantitative assessment of implicit needs.

Description

Technical Field

[0001] This invention relates to the field of demand marketing, specifically a method for mining implicit needs of users across all industries through data fusion. Background Technology

[0002] With the rapid development of the internet economy and the increasing maturity of data collection technology, precise marketing and demand forecasting based on user profiles have become core means for various industries to improve operational efficiency and optimize resource allocation. User profiling technology has evolved from single-dimensional statistics to multi-source data fusion. In the early stages, user profiling mainly relied on structured data such as registration information and historical transaction records actively filled in by users for simple tagging processing. For example, users were classified into corresponding industry categories based on the categories of goods they purchased, and their consumption preferences were judged based on the frequency of purchases.

[0003] With the rise of big data technology, the industry has begun to introduce multi-source behavioral data to supplement and enhance user profiles. Researchers have incorporated unstructured data such as users' browsing history, search logs, social relationships, and geographical location into the profile system. They use machine learning algorithms to classify users with multiple tags and achieve a certain degree of behavioral intent recognition. On this basis, recommendation systems based on collaborative filtering and association rule mining are widely used in e-commerce, content distribution and other fields, and can push products or services that users may be interested in based on their historical behavior.

[0004] However, traditional technologies generally focus on explicit user behaviors that have already occurred, namely analyzing and predicting explicit needs that users have already searched, browsed, and purchased. They lack systematic means to mine implicit needs that users have not yet actively expressed but that objectively exist. Implicit needs refer to consumer demands that users have not yet made purchases or actively searched for, but which are likely to arise in the future due to their industry attributes, the functional relevance of purchased goods, and the natural extension of the consumption chain.

[0005] For example, users who purchase new home renovation services implicitly have a continuous demand for home appliances, home furnishings, and smart home devices in the subsequent stages. These implicit demands have cross-industry transmission and time decay characteristics, and are a key breakthrough for improving user lifetime value and cross-selling conversion rates. However, in existing technologies, industry segmentation and demand mining are disconnected, and there is a lack of a framework for systematic implicit demand reasoning based on the macro attributes of the industry and the inter-industry relationships. As a result, a large number of potential cross-industry consumption opportunities have not been identified. Therefore, a solution is proposed. Summary of the Invention

[0006] This invention is based on a user profile set covering all industries. It extracts transaction behavior sequences and search behavior sequences according to a set statistical period, quantifies the intensity and quantity of explicit needs, and then constructs an industry association scattering diagram based on the number and strength of associations between different industries. The edge weights reflect the tightness of associations between industries, and the mining results are output with threshold filtering to realize the quantification of implicit needs. This solves the problem that traditional technologies lack systematic mining of users' implicit needs based on inter-industry associations, and proposes a data fusion-based method for mining implicit needs of users across industry profiles.

[0007] The objective of this invention can be achieved through the following technical solution: a method for mining implicit needs of users across the entire user industry profile through data fusion, comprising the following steps:

[0008] Step 1: Construct an implicit demand mining architecture that includes user industry segmentation units, explicit demand statistics units, industry correlation analysis units, implicit demand mining units, and demand decay correction units.

[0009] Step 2: The user industry segmentation unit performs statistics on the relevant industries and categorizes the user profiles of each industry to obtain an industry user profile set;

[0010] Step 3: The explicit demand statistics unit determines the intensity and quantity of explicit demand within each industry based on the user profile set of each industry;

[0011] Step 4: The industry association analysis unit analyzes the associations between the industries involved to obtain the number and strength of associations;

[0012] Step 5: After obtaining the number and strength of associations, the implicit demand mining unit mines the implicit demands attached to the explicit demands based on the explicit demands sent by the explicit demand statistics unit, and obtains the implicit demands of the user profile.

[0013] Step Six: Bind the implicit needs of the user profile with the decay parameters sent by the demand decay correction unit, and output the final implicit needs and corresponding time utility.

[0014] In a preferred embodiment of the present invention, the method by which the user industry classification unit categorizes the relevant industries is as follows:

[0015] The acquired user behavior data is input into a pre-trained industry classifier. Based on the user's historical browsing history, transaction product categories, and service subscription information, the probability of the user belonging to a preset industry tag is calculated.

[0016] Obtain a preset affiliation probability threshold, identify industry tags with affiliation probabilities higher than the threshold as the user's primary industry, and assign the user profile to the corresponding primary industry set.

[0017] In a preferred embodiment of the present invention, the method by which the explicit demand statistics unit determines the intensity of explicit demand is as follows:

[0018] Based on the set statistical period, extract the transaction behavior sequence and search behavior sequence within the statistical period, and assign different weights to different behavior types in the obtained transaction behavior sequence and search behavior sequence;

[0019] The explicit demand intensity is obtained by weighting and summing the frequency of each behavior in each behavior sequence and its corresponding weight.

[0020] In a preferred embodiment of the present invention, the method by which the explicit demand statistics unit determines the quantity of explicit demand is as follows:

[0021] The explicit demand statistics unit obtains the number of purchases in the transaction behavior sequence and calculates the arithmetic average of the purchase quantity in each purchase to obtain the explicit demand quantity.

[0022] In a preferred embodiment of the present invention, the industry association analysis unit obtains the product categories contained in industry A and compares them with the product categories in industry B, counts the total number of common product categories or functionally complementary product categories between the two industries, and records them as related items.

[0023] Compare all product categories in the two industries being compared, and select all product categories in the smaller industry as the base number.

[0024] The associated quantity is obtained by calculating the ratio between the associated item and the quantity base.

[0025] In a preferred embodiment of the present invention, the method for the industry correlation analysis unit to obtain the correlation strength is as follows:

[0026] The industry association analysis unit marks the purchase events of related projects in industry A in the historical data, and records the events of purchasing goods in industry B within a preset time window after the purchase record in industry A occurs;

[0027] Next, mark the purchase events of related projects in industry B in the historical data, and record the events of purchasing goods in industry A within a preset time window after the purchase record in industry B occurs;

[0028] The frequency of the above events is summed and recorded, and then the ratio is calculated with the sum of the total purchase events in industry A and industry B to obtain the correlation strength.

[0029] In a preferred embodiment of the present invention, the method by which the latent demand mining unit mines latent demands is as follows:

[0030] Obtain the number and strength of associations output by the industry association analysis unit, and construct an industry association scattering graph based on the number and strength of associations. Each node in each direction represents an industry, and the edge weights are determined by a weighted average of the association strength and number of associations.

[0031] The implicit demand mining unit selects industry A as the base point, traverses all industries involved, and selects the related industries with the top i edge weights as candidate implicit demand industries.

[0032] Then obtain the explicit demand intensity and demand quantity output by the explicit demand statistics unit;

[0033] The formula generates the implicit demand intensity by combining explicit demand intensity, the number of associations, and the strength of associations, and generates the implicit demand quantity by combining explicit demand quantity, the number of associations, and the strength of associations.

[0034] The implicit demand mining unit takes the set of goods whose implicit demand intensity and implicit demand quantity are both greater than the corresponding threshold as the implicit demand mining result.

[0035] In a preferred embodiment of the present invention, the demand decay correction unit is preset with decay periods corresponding to different categories of goods. The latent demand mining unit obtains the corresponding decay period according to the category of goods corresponding to the latent demand mining results, and calculates the decay of the latent demand mining results by formula based on the generation time of the latent demand mining results.

[0036] Compared with the prior art, the beneficial effects of the present invention are:

[0037] 1. This invention statistically analyzes all relevant industries and calculates the probability of belonging to each industry based on user behavior data. Industry tags with a probability of belonging higher than a threshold are identified as the user's primary industry and categorized into the corresponding industry user profile set. This provides an accurate industry attribute positioning basis for uncovering latent needs. Simultaneously, transaction behavior sequences and search behavior sequences are extracted according to a set statistical period. Differentiated weights are assigned to different behavior types, and a weighted sum is performed to quantify the intensity of explicit needs. Furthermore, the number of purchases in the transaction behavior sequence is obtained, and the quantity of each purchase is arithmetically averaged to calculate the quantity of explicit needs. This achieves a dual precise measurement of explicit needs from both the intensity and quantity dimensions, providing data support for subsequent cross-industry transmission inference of latent needs.

[0038] 2. In this invention, while uncovering latent needs, the decay period corresponding to different product categories is preset in the demand decay correction unit. Based on the generation time of the latent demand mining results, the intensity and quantity of latent needs are corrected for timeliness using the decay calculation formula. The final latent demand result with time utility tag is output, realizing the dynamic tracking of the decay of latent needs over time. This makes the output latent needs have hierarchical information of timeliness level, providing a basis for judgment of subsequent marketing strategies.

[0039] 3. This invention also constructs an industry association scattering diagram based on the number and strength of associations, and uses edge weights to reflect the closeness of associations between industries. Candidate industries with high edge weights are selected as candidate directions for implicit demand. The explicit demand strength is multiplied by the association strength and the number of associations to generate the implicit demand strength, and the explicit demand quantity is multiplied by the association quantity and the association strength to generate the implicit demand quantity. Combined with threshold screening, the mining results are output, realizing a complete reasoning scheme for implicit demand from direction screening to strength quantification to threshold confirmation. Attached Figure Description

[0040] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0041] Figure 1 This is a system block diagram of the present invention;

[0042] Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0043] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0044] Example 1: Please refer to Figure 1 - Figure 2 As shown, a method for mining implicit needs of users across all industries through data fusion includes the following steps:

[0045] Step 1: Conduct statistics on all relevant industries, determine the industry scope, and categorize the user profiles corresponding to each industry to obtain a set of user profiles for each industry;

[0046] Step 2: Based on the user profile set of each industry, extract the transaction behavior sequence and search behavior sequence of users in that industry, assign differentiated weights to different behavior types in the sequence and perform weighted summation to calculate the explicit demand intensity of the industry;

[0047] Simultaneously, the number of purchases in the transaction behavior sequence is obtained and the quantity of each purchase is arithmetically averaged to calculate the explicit demand quantity of the industry;

[0048] Step 3: Analyze the relationships between industries. Compare the product categories of any two industries and count the total number of common or complementary product categories between the two industries as related items. Calculate the ratio of the related items to the smaller of the total number of product categories in the two industries to obtain the number of related items.

[0049] Then, the historical purchase data is statistically analyzed to mark events where a purchase of a related item in one industry is followed by a purchase of goods in another industry within a preset time window, as well as reverse purchase events. The frequency of these events is then compared with the total number of purchase events in the two industries to calculate the correlation strength.

[0050] Step 4: Construct an industry association scattering graph based on the number and strength of the associations obtained. The weighted average of the edge weights reflects the closeness of the associations between industries. Select the current user's industry as the base point and traverse all related industries. Select the related industries with the top i edge weights as candidate latent demand industries.

[0051] Obtain the explicit demand intensity and explicit demand quantity calculated in step two. Multiply the explicit demand intensity by the association intensity and the association quantity to generate the implicit demand intensity. Multiply the explicit demand quantity by the association quantity and the association intensity to generate the implicit demand quantity. The set of goods whose implicit demand intensity and implicit demand quantity are both greater than the corresponding threshold is taken as the implicit demand mining result.

[0052] Step 5: Preset the decay period corresponding to different product categories. Obtain the corresponding decay period based on the product category corresponding to the implicit demand mining results output in Step 4. Based on the generation time of the mining results, adjust the implicit demand intensity and quantity according to the timeliness using the decay calculation formula. Use the adjusted implicit demand intensity and quantity as the final output.

[0053] Example 2: Please refer to Figure 1 - Figure 2 As shown, a data fusion-based system for mining implicit needs of users across all industries includes a user industry segmentation unit, an explicit needs statistics unit, an industry correlation analysis unit, an implicit needs mining unit, and a needs decay correction unit.

[0054] The user industry segmentation unit performs statistical analysis on the relevant industries and categorizes the user profiles for each industry to obtain an industry user profile set. The method for the user industry segmentation unit to categorize the relevant industries is as follows:

[0055] The acquired user behavior data is input into a pre-trained industry classifier. Based on the user's historical browsing history, transaction product categories, and service subscription information, the probability of the user belonging to a preset industry tag is calculated.

[0056] Obtain a preset affiliation probability threshold, identify industry tags with affiliation probabilities higher than the affiliation probability threshold as the user's primary industry, and assign the user profile to the corresponding primary industry set.

[0057] The explicit demand statistics unit determines the intensity and quantity of explicit demand within each industry based on the user profile set of each industry. Specifically, the method used by the explicit demand statistics unit to determine the intensity of explicit demand is as follows:

[0058] Based on the set statistical period, extract the transaction behavior sequence and search behavior sequence within the statistical period, and assign different weights to different behavior types in the obtained transaction behavior sequence and search behavior sequence. Different behaviors include at least transaction completion behavior, adding to cart behavior and search behavior.

[0059] The intensity of explicit demand is obtained by weighting and summing the frequency of each behavior in each behavior sequence and its corresponding weight.

[0060] Specifically, the method for determining the quantity of explicit demand in the explicit demand statistics unit is as follows:

[0061] The explicit demand statistics unit obtains the number of purchases in the transaction behavior sequence and calculates the arithmetic average of the purchase quantity in each purchase to obtain the explicit demand quantity.

[0062] The industry association analysis unit analyzes the associations between the industries involved. First, it obtains the product categories contained in industry A through the industry association analysis unit, compares them with the product categories in industry B, and counts the total number of common product categories or complementary product categories between the two industries, which are recorded as related items.

[0063] Compare all product categories in the two industries being compared, and select all product categories in the smaller industry as the base number.

[0064] The associated quantity Rab is obtained by calculating the ratio between the associated items and the quantity base. Nab represents the related project, and Na and Nb represent the number of all product categories in industries A and B, respectively.

[0065] The method for obtaining the correlation strength in the industry correlation analysis unit is as follows:

[0066] The industry association analysis unit marks purchase events of related projects in industry A in historical data, and records the events of purchasing goods in industry B within a preset time window after the purchase record in industry A, and records them as Tab;

[0067] Next, mark the purchase events of related projects in industry B in the historical data, and record the events of purchasing goods in industry A within a preset time window after the purchase record in industry B, and record them as Tba;

[0068] The frequency of the above events is summed and recorded, and then the ratio is calculated with the sum of the total purchase events in industries A and B to obtain the association strength Cab. , where Ta+Tb is the sum of all purchase events in industry A and industry B.

[0069] After obtaining the number and strength of associations, the implicit demand mining unit mines the implicit demands attached to the explicit demands based on the explicit demands sent by the explicit demand statistics unit, and obtains the implicit demands of the user profile.

[0070] The method used by the implicit requirements mining unit to mine implicit requirements is as follows:

[0071] Obtain the number and strength of associations output by the industry association analysis unit, and construct an industry association scattering graph based on the number and strength of associations. Each node in each direction represents an industry, and the edge weights are determined by a weighted average of the association strength and number of associations.

[0072] The latent demand mining unit selects industry A as the base point, traverses all industries involved, and selects the related industries with the top i edge weights as candidate latent demand industries.

[0073] Then obtain the explicit demand intensity and demand quantity output by the explicit demand statistics unit;

[0074] The implicit demand intensity Hab is generated by combining explicit demand intensity, number of associations, and association strength using a formula. Where Sa represents the intensity of explicit demand, and the quantity of implicit demand Hq is generated by combining the quantity of explicit demand, the quantity of related demands, and the intensity of related demands. Qa represents the explicit demand quantity;

[0075] The latent demand mining unit compares the intensity of latent demand with an intensity threshold and the quantity of latent demand with a quantity threshold. If the intensity of latent demand is greater than the intensity threshold and the quantity of latent demand is greater than the quantity threshold, then the corresponding set of goods is taken as the result of latent demand mining.

[0076] The demand decay correction unit pre-sets decay periods Ti corresponding to different product categories. The latent demand mining unit obtains the corresponding decay period based on the product category corresponding to the latent demand mining results, and calculates the decay of the latent demand mining results based on the generation time of the latent demand mining results using a formula. The specific calculation formula is as follows: ,in The attenuation ratio is denoted as t, and t is the time interval between the current time and the generation time of the implicit demand mining result. The demand attenuation correction unit multiplies the attenuation ratio with the implicit demand mining result and uses the result as the final implicit demand intensity and final implicit demand quantity after time correction.

[0077] Thresholds, preset values, preset ranges, etc. are set for result comparison and analysis to determine whether they are good or bad. The value of these thresholds is determined by a combination of large-scale model analysis of sample data and human experience. They can also be adjusted appropriately based on seasonal or common-sense influences.

[0078] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for mining implicit needs of users across all industries through data fusion, characterized in that, Includes the following steps: Step 1: Construct an implicit demand mining architecture that includes user industry segmentation units, explicit demand statistics units, industry correlation analysis units, implicit demand mining units, and demand decay correction units. Step 2: The user industry segmentation unit performs statistics on the relevant industries and categorizes the user profiles of each industry to obtain an industry user profile set; Step 3: The explicit demand statistics unit determines the intensity and quantity of explicit demand within each industry based on the user profile set of each industry; Step 4: The industry association analysis unit analyzes the associations between the industries involved to obtain the number and strength of associations; Step 5: After obtaining the number and strength of associations, the implicit demand mining unit mines the implicit demands attached to the explicit demands based on the explicit demands sent by the explicit demand statistics unit, and obtains the implicit demands of the user profile. Step Six: Bind the implicit needs of the user profile with the decay parameters sent by the demand decay correction unit, and output the final implicit needs and corresponding time utility.

2. The method for mining implicit needs of full-domain user industry profiles through data fusion according to claim 1, characterized in that, The method by which the user industry segmentation unit categorizes the relevant industries is as follows: The acquired user behavior data is input into a pre-trained industry classifier. Based on the user's historical browsing history, transaction product categories, and service subscription information, the probability of the user belonging to a preset industry tag is calculated. Obtain a preset affiliation probability threshold, identify industry tags with affiliation probabilities higher than the threshold as the user's primary industry, and assign the user profile to the corresponding primary industry set.

3. The method for mining implicit needs of full-domain user industry profiles through data fusion according to claim 1, characterized in that, The method for determining the intensity of explicit demand by the explicit demand statistics unit is as follows: Based on the set statistical period, extract the transaction behavior sequence and search behavior sequence within the statistical period, and assign different weights to different behavior types in the obtained transaction behavior sequence and search behavior sequence; The explicit demand intensity is obtained by weighting and summing the frequency of each behavior in each behavior sequence and its corresponding weight.

4. The method for mining implicit needs of full-domain user industry profiles through data fusion according to claim 1, characterized in that, The method for determining the quantity of explicit demand in the explicit demand statistics unit is as follows: The explicit demand statistics unit obtains the number of purchases in the transaction behavior sequence and calculates the arithmetic average of the purchase quantity in each purchase to obtain the explicit demand quantity.

5. The method for mining implicit needs of full-domain user industry profiles through data fusion according to claim 1, characterized in that, The industry association analysis unit obtains the product categories contained in industry A and compares them with the product categories in industry B. It then counts the total number of common or complementary product categories between the two industries and records them as related items. Compare all product categories in the two industries being compared, and select all product categories in the smaller industry as the base number. The associated quantity is obtained by calculating the ratio between the associated item and the quantity base.

6. The method for mining implicit needs of full-domain user industry profiles through data fusion according to claim 5, characterized in that, The method for obtaining the correlation strength in the industry correlation analysis unit is as follows: The industry association analysis unit marks the purchase events of related projects in industry A in the historical data, and records the events of purchasing goods in industry B within a preset time window after the purchase record in industry A occurs; Next, mark the purchase events of related projects in industry B in the historical data, and record the events of purchasing goods in industry A within a preset time window after the purchase record in industry B occurs; The frequency of the above events is summed and recorded, and then the ratio is calculated with the sum of the total purchase events in industry A and industry B to obtain the correlation strength.

7. The method for mining implicit needs of full-domain user industry profiles through data fusion according to claim 1, characterized in that, The method by which the implicit demand mining unit mines implicit demands is as follows: Obtain the number and strength of associations output by the industry association analysis unit, and construct an industry association scattering graph based on the number and strength of associations. Each node in each direction represents an industry, and the edge weights are determined by a weighted average of the association strength and number of associations. The implicit demand mining unit selects industry A as the base point, traverses all industries involved, and selects the related industries with the top i edge weights as candidate implicit demand industries. Then obtain the explicit demand intensity and demand quantity output by the explicit demand statistics unit; The formula generates the implicit demand intensity by combining explicit demand intensity, the number of associations, and the strength of associations, and generates the implicit demand quantity by combining explicit demand quantity, the number of associations, and the strength of associations. The implicit demand mining unit takes the set of goods whose implicit demand intensity and implicit demand quantity are both greater than the corresponding threshold as the implicit demand mining result.

8. The method for mining implicit needs of full-domain user industry profiles through data fusion according to claim 1, characterized in that, The demand decay correction unit is preset with decay periods corresponding to different product categories. The latent demand mining unit obtains the corresponding decay period based on the product category corresponding to the latent demand mining results, and calculates the decay of the latent demand mining results using a formula based on the generation time of the latent demand mining results.

Images