An intelligent data analysis recommendation engine and system

Abstract

The present invention belongs to the field of intelligent recommendation technology, and specifically relates to an intelligent data analysis recommendation engine and system, which includes: a data collection and preprocessing module, a feature selection module and a model recommendation module. When clustering all enterprises, the ordered queue and the result queue are iteratively updated, including: dividing the sorted to-be-processed enterprises into two groups according to the group length, determining the feasibility of grouping with different group lengths according to the distribution concentration of the first group and the distribution difference of the two groups, adding the to-be-processed enterprises in the first group with the largest group feasibility as core enterprises to the result queue, and deleting them from the ordered queue until the ordered queue is empty, obtaining the clustering results according to the result queue, recording the enterprises of the same cluster as the same type of enterprises, determining the interest degree of enterprise customers in each type of enterprise, so as to determine the number of enterprises of each type recommended to enterprise customers. The present invention realizes accurate recommendation for enterprise customers.

Description

Technical Field

[0001] The present invention relates to the field of intelligent recommendation technology, and more specifically, to an intelligent data analysis recommendation engine and system. Background Art

[0002] The Online Financing Service Platform is a financing service portal for corporate customers. By directly connecting with financial institutions, it breaks through the information barrier between financial institutions and corporate customers and realizes a fast online financing channel for corporate customers.

[0003] When recommending applicant companies to corporate clients, in order to ensure that the applicant companies recommended to financial institutions match the needs and characteristics of corporate clients, it is necessary to divide all applicant companies into multiple different types and make accurate recommendations based on the types that corporate clients are interested in.

[0004] In the prior art, a clustering algorithm is usually used to divide all applicant companies into multiple different types. During the clustering process, there is a large number of iterative updates required to complete the entire clustering process, resulting in a large consumption of computing time and computing costs, which in turn affects the efficiency and practicality of the intelligent data analysis recommendation engine. Summary of the invention

[0005] In order to solve the technical problem that when all applicant companies are divided into multiple different types through conventional clustering algorithms, the computing time and computing cost are consumed greatly, thereby affecting the efficiency and practicality of the intelligent data analysis recommendation engine, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides an intelligent data analysis recommendation engine and system, the system comprising the following modules: a data collection and preprocessing module for collecting business information of each enterprise; a feature selection module for clustering all enterprises according to the business information by a density-based clustering algorithm, including: iteratively updating the ordered queue and the result queue until the ordered queue is empty, obtaining the clustering result according to the order of the core enterprises in the result queue and their reachable distances, and recording all enterprises belonging to the same cluster as enterprises of the same type; the iterative update process is: sorting all the to-be-processed enterprises in the ordered queue in the order of their reachable distances from small to large; sorting the queued enterprises according to the group length All the enterprises to be processed after sorting are divided into two groups, and the number of all the enterprises to be processed in the first group is equal to the group length; the distribution concentration of the first group corresponding to the group length is calculated; the distribution difference of the two groups corresponding to the group length is calculated; the feasibility of grouping with different group lengths is determined, and the grouping feasibility is positively correlated with the distribution concentration and the distribution difference; the target grouping length is determined according to the size of the grouping feasibility; all the enterprises to be processed in the first group corresponding to the target grouping length are added to the result queue as core enterprises, and are deleted from the ordered queue; a model recommendation module is used to recommend different types of enterprises to corporate customers; according to the degree of interest of corporate customers in each type of enterprise, the enterprises recommended to corporate customers in the future are determined.

[0007] The present invention improves the process of iteratively updating the ordered queue and the result queue in the conventional OPTICS clustering algorithm. Each time the ordered queue and the result queue are iteratively updated, the reachable distances of all the to-be-processed enterprises in the ordered queue are sorted and analyzed, and at least one to-be-processed enterprise is selected as a core enterprise to be added to the result queue according to the analysis result, and is deleted from the ordered queue. Compared with the conventional OPTICS clustering algorithm, the present invention requires fewer iterative updates to complete the entire clustering process. By reducing the number of iterative updates required to complete the entire clustering process, the consumption of computing time and computing cost is reduced, thereby improving the efficiency and practicality of the intelligent data analysis recommendation engine of the present invention, and realizing accurate recommendations for corporate customers.

[0008] Preferably, the method for obtaining the reachable distance of the enterprise to be processed includes: recording the core enterprise newly added to the result queue as the target core enterprise; the core enterprise newly added to the result queue refers to the core enterprise added to the result queue when the result queue was iteratively updated last time; for any enterprise to be processed in the ordered queue, calculating the distance between the operating status information of the enterprise to be processed and the operating status information of each target core enterprise, and in response to the minimum value of the distance being less than the reachable distance of the enterprise to be processed, taking the minimum value of the distance as the reachable distance of the enterprise to be processed.

[0009] Preferably, the distribution concentration of the first group corresponding to the packet length satisfies the expression: ; In the formula, Indicates the length of the packet The corresponding distribution concentration of the first group is Indicates the packet length, represents the reachable distance of the i-th enterprise to be processed in the first group, Represents the mean of the reachable distances of all processed enterprises in the first group.

[0010] Preferably, the distribution concentration of the first group corresponding to the packet length satisfies the expression: ; In the formula, Indicates the length of the packet The corresponding distribution concentration of the first group is Indicates the packet length, , They represent the maximum and minimum reachable distances of all enterprises to be processed in the first group, respectively. , They represent the lower and upper quartiles of the reachable distances of all processed enterprises in the first group, respectively.

[0011] The present invention expects that the reachable distances of the enterprises to be processed added to the result queue are as close as possible. Therefore, the distribution characteristics of the reachable distances of all the enterprises to be processed in the first group are used to calculate the distribution concentration of the first group corresponding to the group length, thereby ensuring that the reachable distances of the enterprises to be processed added to the result queue meet expectations.

[0012] Preferably, the distribution difference of the two groups corresponding to the group lengths satisfies the expression: ; In the formula, Indicates the length of the packet The distribution difference between the two groups is Indicates the packet length, represents the mean of the reachable distances of all the enterprises to be processed in the first group, It represents the mean of the reachable distances of all the processed enterprises in the second group.

[0013] Preferably, the distribution difference of the two groups corresponding to the group lengths satisfies the expression: ; In the formula, Indicates the length of the packet The distribution difference between the two groups is Indicates the packet length, , They represent the maximum and minimum reachable distances of all enterprises to be processed in the first group, respectively. , They represent the maximum and minimum reachable distances of all enterprises to be processed in the second group, respectively. , , They represent the lower quartile, median and upper quartile of the reachable distances of all the enterprises to be processed in the first group, respectively. , , They represent the lower quartile, median and upper quartile of the reachable distances of all the processed enterprises in the second group, respectively.

[0014] The present invention expects that the reachable distances of the pending enterprises added to the result queue and the remaining pending enterprises not added to the result queue are separated as much as possible. Therefore, the distribution difference of the two groups corresponding to the grouping length is calculated by the difference between the distribution characteristics of the reachable distances of all pending enterprises in the first group and the distribution characteristics of the reachable distances of all pending enterprises in the second group, thereby ensuring that the reachable distances of the pending enterprises added to the result queue meet expectations.

[0015] Preferably, the group feasibility of the group length satisfies the expression: ; In the formula, Indicates the length of the packet The feasibility of grouping Indicates the packet length, Indicates the length of the packet The corresponding distribution concentration of the first group is Indicates the length of the packet The distribution difference between the two corresponding groups.

[0016] Preferably, the group feasibility of the group length satisfies the expression: ; In the formula, Indicates the length of the packet The feasibility of grouping Indicates the packet length, Indicates the length of the packet The corresponding distribution concentration of the first group is Indicates the length of the packet The distribution difference of the two corresponding groups, A represents the number of enterprises to be processed in the ordered queue, represents an exponential function with a natural constant as base, represents the distance between the operating status information of the jth enterprise to be processed in the first group and the operating status information of the first enterprise to be processed, represents the reachable distance of the first enterprise to be processed in the first group, represents the reachable distance of the jth enterprise to be processed in the first group, Represents the minimum function.

[0017] When calculating the feasibility of grouping length, the present invention calculates the grouping length by the average of the reachable distances of all the enterprises to be processed in the first group. The distribution concentration of the corresponding first group is weighted so that the reachable distance of the to-be-processed enterprises added to the result queue is smaller, thereby ensuring that the reachable distance of the to-be-processed enterprises added to the result queue meets expectations.

[0018] Preferably, the method for obtaining the degree of interest of enterprise customers in enterprises of various types includes: taking the central enterprise of each cluster as the representative enterprise of each type; taking the representative enterprises of all types as the first recommended enterprises and displaying them on the recommendation page, so as to obtain the collection status and browsing time of each first recommended enterprise by the enterprise customer; in response to the enterprise customer collecting the first recommended enterprise corresponding to the r-th type, the interest degree of the enterprise customer in the enterprise of the r-th type satisfies the expression: ; Otherwise, the interest level of the enterprise customer in the rth type of enterprise satisfies the expression: ;in, It indicates the browsing time of the first recommended enterprise corresponding to the rth type by the enterprise customers. It indicates the maximum browsing time of enterprise customers for the first recommended enterprise of all types. , represent the preset first parameter and second parameter respectively, and .

[0019] The present invention analyzes the collection status and browsing time of corporate customers to obtain the degree of interest of corporate customers in various types of companies, which can more accurately reflect the needs and interests of corporate customers, thereby improving the accuracy of recommendations to corporate customers.

[0020] In a second aspect, the present invention provides an intelligent data analysis recommendation engine, which adopts the following technical solution:

[0021] An intelligent data analysis recommendation engine includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned intelligent data analysis recommendation system is implemented.

[0022] By adopting the above technical solution, the above-mentioned intelligent data analysis recommendation system is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby making a terminal device based on the memory and the processor for easy use.

[0023] The beneficial effects of the present invention are:

[0024] The present invention improves the process of iteratively updating the ordered queue and the result queue in the conventional OPTICS clustering algorithm. Each time the ordered queue and the result queue are iteratively updated, the reachable distances of all the to-be-processed enterprises in the ordered queue are sorted and analyzed, and at least one to-be-processed enterprise is selected as a core enterprise to be added to the result queue according to the analysis result, and is deleted from the ordered queue. Compared with the conventional OPTICS clustering algorithm, the present invention requires fewer iterative updates to complete the entire clustering process. By reducing the number of iterative updates required to complete the entire clustering process, the consumption of computing time and computing cost is reduced, thereby improving the efficiency and practicality of the intelligent data analysis recommendation engine of the present invention, and realizing accurate recommendations for corporate customers. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0026] Figure 1 A system block diagram of an intelligent data analysis recommendation engine and system in the present invention is schematically shown;

[0027] Figure 2 The process of iteratively updating the ordered queue and the result queue by the conventional OPTICS clustering algorithm is schematically shown;

[0028] Figure 3 The process of iteratively updating the ordered queue and the result queue according to the present invention is schematically illustrated. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0030] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0031] The present invention provides an intelligent data analysis recommendation engine and system. Figure 1 As shown, an intelligent data analysis recommendation engine and system includes a data collection and preprocessing module 100, a feature selection module 200 and a model recommendation module 300, which are described in detail below.

[0032] The Online Financing Service Platform is a financing service portal for corporate customers. By directly connecting with financial institutions, it breaks through the information barrier between financial institutions and corporate customers and realizes a fast online financing channel for corporate customers.

[0033] When recommending applicant companies to corporate clients, in order to ensure that the applicant companies recommended to financial institutions match the needs and characteristics of corporate clients, it is necessary to divide all applicant companies into multiple different types and make accurate recommendations based on the types that corporate clients are interested in.

[0034] The data collection and preprocessing module 100 is used to collect the operating status information of each enterprise.

[0035] Specifically, the business status information of an enterprise includes but is not limited to financial information, business operation information, enterprise scale information, enterprise credit information and enterprise growth information.

[0036] The financial information includes the company's revenue and profit data, asset and liability data, cash flow data, etc.; among them, the revenue and profit data include financial indicators such as the company's revenue, profit and profit margin, which can help identify the company's scale and profitability; the asset and liability data include the company's total assets, total liabilities and other financial data, which can reflect the company's financial health and risk level; the cash flow data includes the company's cash inflow and outflow data, which helps to evaluate its financial stability and operational capabilities.

[0037] The business operation information includes market share data and customer data of the main business; the market share data of the main business refers to the share of the company's main business in its market, which can reflect its market influence and competitive position; the customer data includes the distribution data of the company's main customer groups, which helps to understand the company's market demand and business development.

[0038] The enterprise scale information includes the number of employees, years of enterprise establishment, etc.; the number of employees can reflect the operating scale and organizational structure of the enterprise; the years of enterprise establishment refer to the time when the enterprise was established, which helps to evaluate its market experience and stability.

[0039] The enterprise credit information includes credit score data and debt data; the credit score data can reflect its credit risk and repayment ability; the debt data includes the enterprise's historical debt data and repayment data, which is helpful to evaluate its debt management ability and financial stability.

[0040] The enterprise growth information includes growth rate, investment and R&D data, etc.; the growth rate includes the growth rate of the enterprise's revenue, profit and market share, which can reflect its growth potential and market competitiveness; the investment and R&D data includes the funds invested by the enterprise in investment and R&D, which can demonstrate its innovation ability and future development direction.

[0041] It should be noted that by integrating this information, it can help corporate customers better understand and classify companies, and also help corporate customers choose suitable partners based on their needs and interests.

[0042] The feature selection module 200 is used to cluster all enterprises according to the business status information of the enterprises through a density-based clustering algorithm, and record all enterprises belonging to the same cluster as enterprises of the same type.

[0043] It should be noted that the DBSCAN (Density-based spatial clustering of applicationswith noise) clustering algorithm is a density-based clustering algorithm. This clustering algorithm does not require the pre-specification of the number of clusters, can automatically identify the number of clusters, and obtain clusters of any shape. At the same time, it is insensitive to outliers. Therefore, it is very suitable for classifying enterprises into types. The DBSCAN clustering algorithm has two important parameters. Selecting different parameters will lead to different final clustering results. For high-dimensional business information of enterprises, it is difficult to jointly adjust these two parameters. The OPTICS (Ordering points to identify the clustering structure) clustering algorithm is proposed to help the DBSCAN clustering algorithm select appropriate parameters. Therefore, the present invention clusters all enterprises according to the business information of the enterprises through the OPTICS clustering algorithm.

[0044] Specifically, the ordered queue and the result queue are iteratively updated until the ordered queue is empty. The neighborhood radius parameter is determined according to the order of the core enterprises in the result queue and their reachable distances. The clustering results of all enterprises are obtained according to the relationship between the reachable distance of the core object and the neighborhood radius parameter. All enterprises belonging to the same cluster are recorded as enterprises of the same type. Therefore, according to the clustering results of all enterprises, all enterprises can be divided into multiple types of enterprises. For all enterprises of any type, the central enterprise of the cluster corresponding to the type is taken as the representative enterprise of the type.

[0045] It should be noted that, initially, the first enterprise among all enterprises is added to the result queue as the core enterprise, and all remaining enterprises are added to the ordered queue as pending enterprises. The distance between the operating status information of each pending enterprise in the ordered queue and the operating status information of the core enterprise is calculated as the reachable distance of each pending enterprise, and there is no reachable distance when the first enterprise is used as the core enterprise.

[0046] It should be specially noted that the distances in the present invention all refer to Euclidean distances.

[0047] It should be noted that when the conventional OPTICS clustering algorithm iteratively updates the ordered queue and the result queue, each iterative update will only select one to-be-processed enterprise with the smallest reachable distance from the ordered queue as the core enterprise to add to the result queue, and delete it from the ordered queue until the ordered queue is empty, and the entire clustering process can be completed; illustratively, when the number of all enterprises is 100, 98 iterative updates are required to complete the entire clustering process; each iterative update requires calculating the distance between the operating status information of each to-be-processed enterprise in the ordered queue and the operating status information of the newly added core enterprise in the result queue, and updating the reachable distance of each to-be-processed enterprise in the ordered queue based on the distance. The more iterative updates are performed, the greater the consumption of computing time and computing cost, which seriously affects the efficiency and practicality of the intelligent data analysis recommendation engine of the present invention.

[0048] It should be further explained that the present invention improves the process of iteratively updating the ordered queue and the result queue in the conventional OPTICS clustering algorithm. Each time the ordered queue and the result queue are iteratively updated, the reachable distances of all the enterprises to be processed in the ordered queue are sorted and analyzed, and at least one enterprise to be processed is selected as a core enterprise to be added to the result queue according to the analysis results, and is deleted from the ordered queue. Compared with the conventional OPTICS clustering algorithm, each iterative update only selects one enterprise to be processed from the ordered queue as a core enterprise to be added to the result queue and is deleted from the ordered queue. The present invention requires fewer iterative updates to complete the entire clustering process. By reducing the number of iterative updates required to complete the entire clustering process, the consumption of computing time and computing cost is reduced, thereby improving the efficiency and practicality of the intelligent data analysis recommendation engine of the present invention, and realizing accurate recommendations for corporate customers.

[0049] Specifically, the process of iteratively updating the ordered queue and the result queue includes steps 1 to 5, which are as follows:

[0050] Step 1: Sort all pending enterprises in the ordered queue in ascending order of their reachable distances; divide all sorted pending enterprises into two groups according to the group length, and the number of all pending enterprises in the first group is equal to the group length.

[0051] Among them, the method for obtaining the reachable distance of the enterprise to be processed includes: recording the core enterprise newly added to the result queue as the target core enterprise; the core enterprise newly added to the result queue refers to the core enterprise added to the result queue when the result queue was iteratively updated last time; for any enterprise to be processed in the ordered queue, calculating the distance between the operating status information of the enterprise to be processed and the operating status information of each target core enterprise, and in response to the minimum value of the distance being less than the reachable distance of the enterprise to be processed, taking the minimum value of the distance as the reachable distance of the enterprise to be processed.

[0052] The packet length is taken All integers in the range, A represents the number of enterprises to be processed in the ordered queue, Indicates rounding down.

[0053] It should be noted that after each iterative update, the to-be-processed enterprises with a smaller reachable distance will be added to the result queue as core enterprises and deleted from the ordered queue. After the above-mentioned grouping operation, the present invention will add all the to-be-processed enterprises in the first group as core enterprises to the result queue and delete them from the ordered queue. Therefore, the number of core enterprises added to the result queue is different for different grouping lengths; and for the core enterprises added to the result queue, that is, all the to-be-processed enterprises in the first group, it is expected that the reachable distances of these to-be-processed enterprises are as close as possible, that is, the distribution concentration of the first group corresponding to the grouping length is as large as possible. At the same time, it is expected that the reachable distances of the to-be-processed enterprises added to the result queue and the remaining to-be-processed enterprises not added to the result queue, that is, all the to-be-processed enterprises in the first group and all the to-be-processed enterprises in the second group are as separated as possible, that is, the distribution difference of the two groups corresponding to the grouping length is as large as possible.

[0054] Step 2: Calculate the distribution concentration of the first group corresponding to the group length.

[0055] In one embodiment, the distribution concentration of the first group corresponding to the packet length satisfies the expression:

[0056] ;

[0057] In the formula, Indicates the length of the packet The corresponding distribution concentration of the first group is Indicates the packet length, represents the reachable distance of the i-th enterprise to be processed in the first group, Represents the mean of the reachable distances of all processed enterprises in the first group.

[0058] It should be noted that the mean of the reachable distances of all the enterprises to be processed in the first group represents the central tendency of the reachable distances of all the enterprises to be processed in the first group. The difference between the reachable distance of the ith enterprise to be processed in the first group and the mean of the reachable distances of all the enterprises to be processed in the first group is The smaller it is, the closer the reachable distance of the i-th enterprise to be processed in the first group is to the central trend of the reachable distance of the first group, which further indicates that the closer the reachable distances of all enterprises to be processed in the first group are to the central trend, the smaller the group length is. The corresponding first group has a greater distribution concentration.

[0059] In another embodiment, the distribution concentration of the first group corresponding to the packet length satisfies the expression:

[0060] ;

[0061] In the formula, Indicates the length of the packet The corresponding distribution concentration of the first group is Indicates the packet length, , They represent the maximum and minimum reachable distances of all enterprises to be processed in the first group, respectively. , They represent the lower and upper quartiles of the reachable distances of all processed enterprises in the first group, respectively.

[0062] It should be noted that the maximum value, minimum value, lower quartile and upper quartile describe the important quantitative indicators of the distribution characteristics of the reachable distances of all the enterprises to be processed in the first group. The difference between the maximum value and the minimum value Represents the distribution range of the reachable distance. The smaller the distribution range, the longer the group length. The greater the concentration of the distribution of the first group, the greater the difference between the lower quartile and the upper quartile. Represents the concentrated distribution range of the reachable distance. The smaller the concentrated distribution range, the longer the group length. The corresponding first group has a greater distribution concentration.

[0063] It should be further explained that the present invention expects that the reachable distance of the enterprises to be processed added to the result queue is as close as possible. Therefore, the distribution characteristics of the reachable distances of all the enterprises to be processed in the first group are used to calculate the distribution concentration of the first group corresponding to the group length, thereby ensuring that the reachable distances of the enterprises to be processed added to the result queue meet expectations.

[0064] Step 3: Calculate the distribution difference of the two groups corresponding to the grouping length.

[0065] In one embodiment, the distribution difference of the two groups corresponding to the group lengths satisfies the expression:

[0066] ;

[0067] In the formula, Indicates the length of the packet The distribution difference between the two groups is Indicates the packet length, represents the mean of the reachable distances of all the enterprises to be processed in the first group, It represents the mean of the reachable distances of all the processed enterprises in the second group.

[0068] It should be noted that the mean of the reachable distances of all pending enterprises in the first group represents the central tendency of the reachable distances of all pending enterprises in the first group, and the mean of the reachable distances of all pending enterprises in the second group represents the central tendency of the reachable distances of all pending enterprises in the second group. The larger it is, the more dissimilar the central trends of the reachable distances of the two groups are, and the greater the distribution difference between the two groups is.

[0069] In another embodiment, the distribution difference of the two groups corresponding to the group lengths satisfies the expression:

[0070] ;

[0071] In the formula, Indicates the length of the packet The distribution difference between the two groups is Indicates the packet length, , They represent the maximum and minimum reachable distances of all enterprises to be processed in the first group, respectively. , They represent the maximum and minimum reachable distances of all enterprises to be processed in the second group, respectively. , , They represent the lower quartile, median and upper quartile of the reachable distances of all the enterprises to be processed in the first group, respectively. , , They represent the lower quartile, median and upper quartile of the reachable distances of all the processed enterprises in the second group, respectively.

[0072] It should be noted that the maximum value, minimum value, lower quartile, median and upper quartile describe the important quantitative indicators of the distribution characteristics of the reachable distances of all the enterprises to be processed in the first and second groups. The difference in the important quantitative indicators of the distribution characteristics of the first and second groups is The larger it is, the more dissimilar the central trends of the reachable distances of the two groups are, and the greater the distribution difference between the two groups is.

[0073] It should be further explained that the present invention expects that the reachable distances of the pending enterprises added to the result queue and the remaining pending enterprises not added to the result queue are separated as much as possible. Therefore, the distribution characteristics of the reachable distances of all pending enterprises in the first group and the distribution characteristics of the reachable distances of all pending enterprises in the second group are used to calculate the distribution differences of the two groups corresponding to the grouping lengths, thereby ensuring that the reachable distances of the pending enterprises added to the result queue meet expectations.

[0074] Step 4: Calculate the feasibility of grouping with different group lengths based on the distribution concentration of the first group corresponding to different group lengths and the distribution difference of the two groups corresponding to different group lengths.

[0075] Specifically, the grouping feasibility is positively correlated with the distribution concentration and distribution diversity.

[0076] In one embodiment, the group feasibility of the group length satisfies the expression:

[0077] ;

[0078] In the formula, Indicates the length of the packet The feasibility of grouping Indicates the packet length, Indicates the length of the packet The corresponding distribution concentration of the first group is Indicates the length of the packet The distribution difference between the two corresponding groups.

[0079] It should be noted that the present invention expects that the reachable distances of these pending enterprises added to the result queue are as close as possible. At the same time, it is expected that the reachable distances of the pending enterprises added to the result queue and the remaining pending enterprises not added to the result queue are as far apart as possible. That is to say, the distribution concentration of the first group corresponding to the finally obtained group length is as large as possible and the distribution difference of the two groups corresponding to the group length is as large as possible, so the group length The distribution difference between the two corresponding groups The larger the group length The corresponding distribution concentration of the first group The larger the packet length, Feasibility of grouping The bigger.

[0080] In another embodiment, the group feasibility of the group length satisfies the expression:

[0081] ;

[0082] In the formula, Indicates the length of the packet The feasibility of grouping Indicates the packet length, Indicates the length of the packet The corresponding distribution concentration of the first group is Indicates the length of the packet The distribution difference of the two corresponding groups, A represents the number of enterprises to be processed in the ordered queue, represents an exponential function with a natural constant as base, represents the distance between the operating status information of the jth enterprise to be processed in the first group and the operating status information of the first enterprise to be processed, represents the reachable distance of the first enterprise to be processed in the first group, represents the reachable distance of the jth enterprise to be processed in the first group, Represents the minimum function.

[0083] It should be noted that after each iterative update, the enterprises with smaller reachable distances will be added to the result queue as core enterprises. Therefore, the smaller the mean reachable distance of all enterprises in the first group, the better. The group length is calculated by the mean reachable distance of all enterprises in the first group. The distribution concentration of the corresponding first group is weighted. The smaller the mean of the reachable distance of all the enterprises to be processed in the first group and the longer the group length is, the smaller the mean of the reachable distance of all the enterprises to be processed in the first group is. The distribution difference between the two corresponding groups The larger the packet length, Feasibility of grouping The larger the number of all pending enterprises in the first group is, the , the number of all pending enterprises in the second group is equal to , the product of the number of all pending enterprises in the two groups The smaller the size, the greater the difference in the number of all pending enterprises in the two groups, and the greater the difference between the two groups. The group length is calculated by multiplying the number of all pending enterprises in the two groups. The distribution difference of the two corresponding groups is weighted. The smaller the product of the number of all enterprises to be processed in the two groups and the smaller the group length is, the The greater the difference in the distribution of the two corresponding groups, the longer the group length Feasibility of grouping The bigger.

[0084] It should be further explained that, considering that the first enterprise to be processed in the first group will definitely be added to the result queue as a core enterprise, the minimum value of the distance between the operating status information of the jth enterprise to be processed in the first group and the operating status information of the first enterprise to be processed and the reachable distance of the jth enterprise to be processed in the first group is As the final result of the calculation.

[0085] Step 5: Determine the target group length according to the feasibility of the grouping; add all the enterprises to be processed in the first group corresponding to the target group length as core enterprises to the result queue and delete them from the ordered queue.

[0086] Specifically, the group length corresponding to the maximum value of the group feasibility is used as the target group length; all the unprocessed enterprises in the first group corresponding to the target group length are added to the result queue as core enterprises, and are deleted from the ordered queue, thereby realizing an iterative update of the ordered queue and the result queue.

[0087] For example, there are 7 enterprises in total, and their indexes are 1 to 7, and their operating status information is (1,2), (2,5), (8,7), (3,6), (8,8), (7,3), and (4,5), respectively. All enterprises are clustered using the conventional OPTICS clustering algorithm, and the process of iteratively updating the ordered queue and the result queue is as follows: Figure 2 As shown in the figure, it took 6 iterations to complete the entire clustering process.

[0088] When the ordered queue and the result queue are iteratively updated in the present invention, at the beginning, the first enterprise among all enterprises is added as the core enterprise to the result queue. At this time, there is only one core enterprise in the result queue, that is, the enterprise with index 1, and all the remaining enterprises are added as the to-be-processed enterprises to the ordered queue. At this time, there are 6 core enterprises in the ordered queue, which are enterprises with indexes 2 to 7 respectively. The process of iteratively updating the ordered queue and the result queue is as follows: Figure 3 As shown, the entire clustering process is completed after 5 iterative updates, which reduces the number of iterative updates required to complete the entire clustering process, reduces the consumption of computing time and computing costs, and thus improves the efficiency and practicality of the intelligent data analysis recommendation engine of the present invention, and realizes accurate recommendations for corporate customers.

[0089] Among them, after the first iterative update of the ordered queue and the result queue, the obtained ordered queue is {3,4,5,6,7}, and the obtained result queue is {1,2};

[0090] Among them, when the ordered queue and the result queue are iteratively updated for the second time, the packet length is taken All integers in the range; the reachable distances of enterprises with indexes from 3 to 7 are 6.32, 1.41, 6.7, 5.38, and 2 respectively. All the pending enterprises in the ordered queue are sorted in ascending order of their reachable distances, and the sorted result is {4, 7, 6, 3, 5}; when the group length When the first group is {4} and the second group is {7,6,3,5}, the distribution concentration of the first group corresponding to the group length , the distribution difference of the two groups corresponding to the group length , feasibility of grouping of group length ; When the packet length When the first group is {4,7}, the second group is {6,3,5}, the distribution concentration of the first group corresponding to the group length , the distribution difference of the two groups corresponding to the group length , feasibility of grouping by group length ; When the packet length When the first group is {4,7,6}, the second group is {3,5}, and the distribution concentration of the first group corresponding to the group length is , the distribution difference of the two groups corresponding to the group length , feasibility of grouping of group length According to the above calculation of the feasibility of the grouping of different grouping lengths, the grouping length corresponding to the maximum value of the feasibility of the grouping, that is, the grouping length as the target group length; all the to-be-processed enterprises in the first group {4,7} corresponding to the target group length are added to the result queue as core enterprises, and are deleted from the ordered queue, thereby realizing an iterative update of the ordered queue and the result queue.

[0091] The model recommendation module 300 is used to determine the interest level of the enterprise customers in various types of enterprises, so as to determine the types of enterprises recommended to the enterprise customers.

[0092] Specifically, on the recommendation page of the intelligent data analysis recommendation system, all types of representative enterprises are listed as first recommended enterprises and displayed on the recommendation page; corporate customers click on the first recommended enterprises displayed on the recommendation page and browse the operating status information of each first recommended enterprise. At the same time, corporate customers will collect the first recommended enterprises that they are interested in.

[0093] Furthermore, when the corporate customer has browsed all the first recommended companies, the corporate customer's interest in each type of company is determined based on the corporate customer's collection and browsing time of each first recommended company, and the interest level is positively correlated with the collection and browsing time.

[0094] Furthermore, for the rth type of enterprise, in response to the enterprise customer collecting the first recommended enterprise corresponding to the type, the enterprise customer's interest in the enterprise of the type satisfies the expression: In response to the enterprise customer not collecting the first recommended enterprise corresponding to the type, the enterprise customer's interest in the enterprise of the type satisfies the expression: ;in, Indicates the degree of interest of corporate customers in the rth type of enterprise, It indicates the browsing time of the first recommended enterprise corresponding to the rth type by the enterprise customers. It indicates the maximum browsing time of enterprise customers for the first recommended enterprise of all types. , represent the preset first parameter and second parameter respectively, and .

[0095] It should be noted that the present invention analyzes the collection status and browsing time of corporate customers to obtain the degree of interest of corporate customers in various types of companies, which can more accurately reflect the needs and interests of corporate customers, thereby improving the accuracy of recommendations to corporate customers.

[0096] Among them, the specific values ​​of the first parameter and the second parameter can be set according to the actual application scenario and requirements, requiring that the first parameter is less than the second parameter, and the first parameter and the second parameter are both positive numbers. The present invention sets the first parameter to 0.5 and the second parameter to 1.5.

[0097] Furthermore, the sum of the interest of enterprise customers in all types of enterprises is recorded as the first sum value, and the ratio of the interest of enterprise customers in each type of enterprise to the first sum value is recorded as the recommendation ratio of each type of enterprise; among all the second recommended enterprises recommended to enterprise customers on the recommendation page, for every N second recommended enterprises, the number of enterprises of the rth type is , represents the recommended proportion of enterprises of the rth type, N represents the preset number, Indicates rounding down.

[0098] Among them, the specific value of the preset number can be set according to the actual application scenario and requirements. The preset number is required to be a positive integer greater than 10. The present invention sets the preset number to 30.

[0099] The present invention improves the process of iteratively updating the ordered queue and the result queue in the conventional OPTICS clustering algorithm. Each time the ordered queue and the result queue are iteratively updated, the reachable distances of all the to-be-processed enterprises in the ordered queue are sorted and analyzed, and at least one to-be-processed enterprise is selected as a core enterprise to be added to the result queue according to the analysis result, and is deleted from the ordered queue. Compared with the conventional OPTICS clustering algorithm, the present invention requires fewer iterative updates to complete the entire clustering process. By reducing the number of iterative updates required to complete the entire clustering process, the consumption of computing time and computing cost is reduced, thereby improving the efficiency and practicality of the intelligent data analysis recommendation engine of the present invention, and realizing accurate recommendations for corporate customers.

[0100] An embodiment of the present invention further discloses an intelligent data analysis recommendation engine, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent data analysis recommendation system according to the present invention is implemented.

[0101] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0102] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0103] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. An intelligent data analysis recommendation system, characterized in that: include: Data collection and preprocessing module, used to collect information on the operating conditions of each enterprise; The feature selection module is used to cluster all enterprises according to the business status information through a density-based clustering algorithm, including: iteratively updating the ordered queue and the result queue until the ordered queue is empty, obtaining the clustering result according to the order of the core enterprises in the result queue and their reachable distance, and recording all enterprises belonging to the same cluster as the same type of enterprises; the iterative update process is: All the enterprises to be processed in the ordered queue are sorted in ascending order of their reachable distances; all the sorted enterprises to be processed are divided into two groups according to the group length, and the number of all enterprises to be processed in the first group is equal to the group length; the distribution concentration of the first group corresponding to the group length is calculated to satisfy the expression: ; In the formula, Indicates the length of the packet The corresponding distribution concentration of the first group is Indicates the packet length, represents the reachable distance of the i-th enterprise to be processed in the first group, Represents the mean of the reachable distances of all enterprises to be processed in the first group; the distribution difference of the two groups corresponding to the group length is calculated to satisfy the expression: ; In the formula, Indicates the length of the packet The distribution difference between the two groups is Represents the mean of the reachable distances of all enterprises to be processed in the second group; Determine that the feasibility of grouping with different grouping lengths satisfies the expression: ; In the formula, Indicates the length of the packet The feasibility of grouping is positively correlated with the distribution concentration and distribution difference; the target grouping length is determined according to the size of the grouping feasibility; all the enterprises to be processed in the first group corresponding to the target grouping length are added as core enterprises to the result queue, and are deleted from the ordered queue; The model recommendation module is used to determine the interest level of enterprise customers in various types of enterprises, so as to determine the number of enterprises of each type to be recommended to the enterprise customers. The interest level is positively correlated with the collection status and browsing time.

2. The intelligent data analysis and recommendation system according to claim 1, characterized in that: The method for obtaining the reachable distance of the enterprise to be processed includes: Record the core enterprise newly added to the result queue as the target core enterprise; the core enterprise newly added to the result queue refers to the core enterprise added to the result queue when the result queue was iteratively updated last time; For any enterprise to be processed in the ordered queue, the distance between the operating status information of the enterprise to be processed and the operating status information of each target core enterprise is calculated. In response to the minimum value of the distance being less than the reachable distance of the enterprise to be processed, the minimum value of the distance is used as the reachable distance of the enterprise to be processed.

3. The intelligent data analysis and recommendation system according to claim 1, characterized in that: The distribution concentration of the first group corresponding to the packet length satisfies the expression: ； In the formula, Indicates the length of the packet The corresponding distribution concentration of the first group is Indicates the packet length, , They represent the maximum and minimum reachable distances of all enterprises to be processed in the first group, respectively. , They represent the lower and upper quartiles of the reachable distances of all processed enterprises in the first group, respectively.

4. The intelligent data analysis and recommendation system according to claim 1, characterized in that: The distribution difference of the two groups corresponding to the group length satisfies the expression: ； In the formula, Indicates the length of the packet The distribution difference between the two groups is Indicates the packet length, , They represent the maximum and minimum reachable distances of all enterprises to be processed in the first group, respectively. , They represent the maximum and minimum reachable distances of all enterprises to be processed in the second group, respectively. , , They represent the lower quartile, median and upper quartile of the reachable distances of all the enterprises to be processed in the first group, respectively. , , They represent the lower quartile, median and upper quartile of the reachable distances of all the processed enterprises in the second group, respectively.

5. The intelligent data analysis and recommendation system according to claim 1, characterized in that: The feasibility of the grouping of the grouping length satisfies the expression: ； In the formula, Indicates the length of the packet The feasibility of grouping Indicates the packet length, Indicates the length of the packet The corresponding distribution concentration of the first group is Indicates the length of the packet The distribution difference of the two corresponding groups, A represents the number of enterprises to be processed in the ordered queue, represents an exponential function with a natural constant as base, represents the distance between the operating status information of the jth enterprise to be processed in the first group and the operating status information of the first enterprise to be processed, represents the reachable distance of the first enterprise to be processed in the first group, represents the reachable distance of the jth enterprise to be processed in the first group, Represents the minimum function.

6. The intelligent data analysis and recommendation system according to claim 1, characterized in that: The method for obtaining the interest level of the enterprise customers in various types of enterprises includes: The central enterprise of each cluster is taken as the representative enterprise of each type; the representative enterprises of all types are taken as the first recommended enterprises and displayed on the recommendation page, so as to obtain the collection and browsing time of each first recommended enterprise by the enterprise customers; in response to the enterprise customer collecting the first recommended enterprise corresponding to the rth type, the interest degree of the enterprise customer in the enterprise of the rth type satisfies the expression: ; Otherwise, the interest level of the enterprise customer in the rth type of enterprise satisfies the expression: ;in, It indicates the browsing time of the first recommended enterprise corresponding to the rth type by the enterprise customers. It indicates the maximum browsing time of enterprise customers for the first recommended enterprise of all types. , represent the preset first parameter and second parameter respectively, and .

7. An intelligent data analysis recommendation engine, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the intelligent data analysis and recommendation system according to any one of claims 1 to 6 is implemented.

Images