A method for building automatic machine learning models for SQL in data service mode

Abstract

The present invention relates to the field of machine learning technology, and in particular to a method for constructing an automatic machine learning model for SQL under a data service mode, the method comprising preprocessing acquired customer information and feature mining to generate an AI form; determining the type of information in the AI ​​form according to the customer information; selecting the corresponding information type according to the location type when the customer browses the product, and determining the product recommendation level according to the number of search events for the product in the selected information type and the maximum time value of the search event; inputting data of a process of recommending products to customers into a model as training data for training to obtain an automatic machine learning model; determining the eligibility of the product recommendation process according to the product purchase rate of the customer on the product recommendation page. The present invention selects information in the AI ​​form to recommend products to customers according to the network type used by the customer when browsing the product, thereby improving the accuracy of process control for constructing the automatic machine learning model.

Description

Technical Field

[0001] The present invention relates to the field of machine learning technology, and in particular to a method for constructing an automatic machine learning model oriented to SQL under a data service mode. Background Art

[0002] Machine learning is now almost everywhere. Every company has accumulated huge amounts of data and hopes to extract as much information as possible from it to improve its business and customer satisfaction. E-shopping has become the mainstream way for people to consume, so recommending accurate products to customers has become an important topic in machine learning.

[0003] Chinese patent application publication number: CN116433313A discloses a method and device for recommending products based on federated learning. The method includes: reading the window width and the second time feature related to the time series in the second sample from a predetermined configuration, and aligning the first sample and the second sample according to the alignment feature to form a first set of data samples; determining the time window feature according to the first time feature and the window width; verifying the second time feature based on the time window feature to obtain the second set of data samples; performing multiple aggregation operations on the samples with the same first time feature and the same alignment feature in the second set of data samples to automatically construct multiple extended features; calculating the information value of the multiple extended features, retaining the extended features whose information value is greater than the preset value or discarding the extended features whose information value is less than the preset value; performing federated learning training on the training model at least based on the extended feature to obtain a recommendation model; and recommending products based on the recommendation model.

[0004] This shows that the existing technology has the problem of inaccurate classification of customer information, which leads to low accuracy of the customer product recommendation process and low performance of the constructed automatic machine learning model. Summary of the invention

[0005] To this end, the present invention provides an automatic machine learning model construction method for SQL under the data service mode, so as to overcome the problem in the prior art that inaccurate classification of customer information leads to low accuracy of the customer product recommendation process and thus low performance of the constructed automatic machine learning model.

[0006] To achieve the above object, the present invention provides a method for constructing an automatic machine learning model for SQL in a data service mode, comprising:

[0007] Preprocess the acquired customer information and mine the features of the preprocessed information to generate AI forms;

[0008] Determine the type of information in the AI ​​form based on customer information;

[0009] Selecting a corresponding information type according to the location type when the customer browses the product, and determining the product recommendation level according to the number of occurrences of the product search event in the selected information type and the maximum time value of the occurrence of the search event;

[0010] The data of the process of recommending products to customers is input into the model as training data to obtain an automatic machine learning model;

[0011] The eligibility of the product recommendation process is determined based on the product purchase rate of customers on the product recommendation page, and the construction process of the learning model is adjusted based on the determined eligibility, including adjusting the determination process of the product recommendation level, or recommending products of the same type with a lower price than the recommended product.

[0012] Furthermore, the process of determining the type of information in the AI ​​form according to the network type is to determine the type of information in the AI ​​form according to the network environment in which the customer information is located; wherein the network type includes a wireless network type and a mobile network type.

[0013] Further, in a wireless network type environment, the information type is determined to be a first type based on the IP address type being a company IP address, or the information type is determined to be a second type based on the IP address type being a home IP address.

[0014] Further, in a mobile network type environment, the information type is determined to be the first type according to the time type when the customer browses the goods being the working time type, or the information type is determined to be the second type according to the time type when the customer browses the goods being the rest time type.

[0015] Further, when recommending products to customers, a first information type is selected to recommend products to customers based on the customer's location type being a company type, or a second information type is selected to recommend products to customers based on the customer's location type being a home type.

[0016] Furthermore, when determining the product recommendation level, the product recommendation level is determined to be the second level based on the number of search events for the product in the selected information type being less than a preset number of occurrences, or the product recommendation level is determined to be a second level based on the number of search events for the product in the selected information type being greater than or equal to a preset number of occurrences.

[0017] Furthermore, when making a secondary judgment on the product recommendation level, the product recommendation level is determined to be the first level based on the maximum time value of the product search event being greater than a preset value, or the product recommendation level is determined to be the third level based on the maximum time value of the product search event being less than or equal to a preset value.

[0018] Furthermore, when determining the eligibility of the product recommendation process, the product recommendation process is determined to be unqualified based on the fact that the product purchase rate of customers on the product recommendation page is less than a preset product purchase rate.

[0019] Furthermore, when it is determined that the product recommendation process is unqualified, the preset number of occurrences is adjusted by an adjustment coefficient based on the fact that the browsing time of the customer on the product recommendation page is less than the preset browsing time.

[0020] Furthermore, when it is determined that the product recommendation process is unqualified, a product of the same type with a lower price than the recommended product is replaced based on the customer's browsing time on the product recommendation page being greater than or equal to a preset browsing time.

[0021] Compared with the prior art, the beneficial effect of the present invention lies in that the present invention determines the type of customer information in the AI ​​form based on the type of network used by the customer when browsing the goods. When the customer is in a wireless network environment when browsing the goods, the customer's geographical location is determined to be in the company or at home through the wireless network IP address used by the customer, and then the information of the customer in the two geographical locations is classified. Through the above method, the accuracy of customer information analysis is improved, and accurate customer information is obtained.

[0022] Furthermore, the present invention determines the type of customer information in the AI ​​form based on the type of network used by the customer when browsing products. When the customer browses products in a mobile network environment, the present invention determines whether the customer is in the company or at home based on the time type when the customer browses products. When the customer browses products during a break, the present invention determines that the customer is at home; when the customer browses products during working hours, the present invention determines that the customer is in the company. The above method avoids the phenomenon of inaccurate classification of customer information due to inaccurate analysis of the customer's geographical location.

[0023] Furthermore, the present invention determines whether the customer is in the company or at home through analyzing the customer's location type and selects corresponding information types as data for recommending products. When the customer is in the company, the first information type is selected as data for recommending products, and when the customer is at home, the second information type is selected as data for recommending products. The above method not only reduces the reference data for recommending products and thus improves the efficiency and accuracy of product recommendations, but also takes into account the phenomenon that when the customer is at home, the recommended products may be inaccurate because the products are not purchased by the customer himself, thereby improving the accuracy and intelligence of product recommendations.

[0024] Furthermore, the present invention determines the recommendation level of the product based on the comparison result of the number of occurrences of the product search event and the preset number of occurrences. When the number of occurrences of the product search event is less than the preset number of occurrences, the product recommendation level is determined to be the second level. When the number of occurrences of the product search event is greater than or equal to the preset number of occurrences, a secondary judgment is performed. During the secondary judgment process, the maximum time value of the product search event is compared with the maximum time value of the product purchase event to determine whether the product is the product that has just been purchased and then determine the product recommendation level. The above method improves the accuracy of determining the product recommendation level. Taking into account that the customer has purchased the product with a search event occurrence number greater than the preset number of occurrences, the product recommendation level is determined to be the third level, thereby improving the accuracy of product recommendations.

[0025] Furthermore, the present invention determines the customer's satisfaction with the recommended products based on the customer's product purchase rate on the product recommendation page. When the customer's product purchase rate on the product recommendation page is less than the preset product purchase rate, the product recommendation process is determined to be unqualified. When the customer's product purchase rate on the product recommendation page is greater than or equal to the preset product purchase rate, the product recommendation process is determined to be qualified. The above method is used to intuitively determine the customer's product purchase rate for the recommended products and then determine the qualification of the product recommendation process, thereby improving the accuracy of the product recommendation process analysis and optimizing the automatic machine learning model.

[0026] Furthermore, the present invention determines an adjustment method for the product recommendation process based on a comparison result of a customer's browsing time on a product recommendation page and a preset browsing time when the product recommendation process is unqualified. When the browsing time is less than the preset browsing time, it means that the customer is not satisfied with the recommended products on the product recommendation page. The adjustment coefficient is used to increase the preset number of occurrences and thus improve the accuracy of the recommended product grade determination. When the browsing time is greater than or equal to the preset browsing time, it means that the customer did not purchase the recommended product due to financial problems, resulting in a low product purchase rate. The recommended price is determined to be lower than that of the same type of products recommended, thereby improving the accuracy of product recommendations and optimizing the automatic machine learning model. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a workflow diagram of a method for constructing an automatic machine learning model for SQL under a data service mode according to an embodiment of the present invention;

[0028] Figure 2 A workflow diagram for determining a product recommendation level in a SQL-oriented automatic machine learning model construction method under a data service mode in an embodiment of the present invention;

[0029] Figure 3 A workflow diagram for determining the eligibility of a product recommendation process in a SQL-oriented automatic machine learning model construction method under a data service mode of an embodiment of the present invention;

[0030] Figure 4 A workflow diagram for determining an adjustment method for a product recommendation process in a SQL-oriented automatic machine learning model construction method under a data service mode in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0032] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0033] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0034] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0035] See also Figure 1-Figure 4 As shown, Figure 1 This is a workflow diagram of a method for constructing an automatic machine learning model for SQL under a data service mode according to an embodiment of the present invention; Figure 2 A workflow diagram for determining a product recommendation level in a SQL-oriented automatic machine learning model construction method under a data service mode in an embodiment of the present invention; Figure 3 A workflow diagram for determining the eligibility of a product recommendation process in a SQL-oriented automatic machine learning model construction method under a data service mode of an embodiment of the present invention; Figure 4 A workflow diagram for determining an adjustment method for a product recommendation process in a SQL-oriented automatic machine learning model construction method under a data service mode in an embodiment of the present invention.

[0036] The method for constructing an automatic machine learning model for SQL in a data service mode according to an embodiment of the present invention includes:

[0037] Step S1, preprocessing the acquired customer information, and mining the features of the preprocessed information to generate an AI form;

[0038] Step S2, determining the type of information in the AI ​​form based on the customer information;

[0039] Step S3, selecting a corresponding information type according to the location type when the customer browses the product, and determining the product recommendation level according to the number of occurrences of the product search event in the selected information type and the maximum time value of the occurrence of the search event;

[0040] Step S4, inputting the data of the process of recommending products to customers as training data into the model for training to obtain an automatic machine learning model;

[0041] Step S5, determining the eligibility of the product recommendation process based on the product purchase rate of the customer on the product recommendation page, and adjusting the construction process of the learning model based on the determined eligibility, including adjusting the determination process of the product recommendation level, or recommending products of the same type with a lower price than the recommended product.

[0042] In the embodiment of the present invention, the preprocessing of the acquired customer information includes but is not limited to "filling in missing information and deleting duplicate information". In the embodiment of the present invention, the filling in missing information includes, when the wireless network IP address is missing in the customer information, determining to fill in the wireless network IP address of the customer company based on the time information of the missing wireless network IP address information being a working day, and determining to fill in the wireless network IP address of the home based on the time information of the missing wireless network IP address information being not a working day. The deleting duplicate information includes deleting redundant customer information and retaining only one piece of customer information when exactly the same customer information exists.

[0043] Specifically, when determining the type of information in the AI ​​form, the method for determining the type of information is determined based on the type of network the customer uses when shopping;

[0044] When the network type during shopping by the customer is a wireless network type, the information type is determined by a first determination method;

[0045] When the network type used by the customer for shopping is a mobile network type, the information type is determined using a second determination method.

[0046] Specifically, when the information type is determined by the first determination method, the information type is determined according to the IP address type of the wireless network;

[0047] When the IP address type of the wireless network is the first type, determining that the information in the network environment is the first type;

[0048] When the IP address type of the wireless network is the second type, it is determined that the information in the network environment is of the second type.

[0049] In the embodiment of the present invention, the first type of IP address is a company network IP address, and the second type of IP address is a home network IP address. The method for determining the type of IP address in the embodiment of the present invention is preferably to observe the range of the IP address and the subnet mask. In home networks and company networks, different IP address segments and subnet masks are usually used. For example, a home network may use a private IP address segment (such as 192.168.xx or 10.xxx), while a company network may use a dedicated IP address segment (such as 172.16.xx to 172.31.xx).

[0050] Specifically, the present invention determines the type of customer information in the AI ​​form based on the type of network used by the customer when browsing products. When the customer is in a wireless network environment when browsing products, the customer's geographical location is determined to be at work or at home through the wireless network IP address used by the customer, and then the information of the customer in the two geographical locations is classified. Through the above method, the accuracy of customer information analysis is improved, and accurate customer information is obtained.

[0051] Specifically, when determining the information type by the second determination method, the information type is determined according to the time type when the customer browses the product;

[0052] When the time type is the first time type, determining that the information under the time type is the first category;

[0053] When the time type is the second time type, it is determined that the information under the time type is of the second category.

[0054] In the embodiment of the present invention, the first time type is a working time type, and the second time type is a rest time type. The method for determining the time type in the embodiment of the present invention is preferably to determine the time type according to company policies and statutory holidays. For example, when the customer's company has a two-day weekend system and the working hours are from 9 am to 6 pm, the customer uses the mobile network to purchase goods at 7 pm on Friday afternoon, and the information is determined to be the second type of information; the customer uses the mobile network to purchase goods at 3 pm on Friday afternoon, and the information type is determined to be the first type of information.

[0055] The AI ​​form in the embodiment of the present invention includes two main classification information: "customer" and "interaction". The "customer" includes but is not limited to "customer ID information, network type information used by the customer, and product name information selected by the customer", and the "interaction" includes but is not limited to "preset occurrence times, purchase event occurrence times, search event occurrence times, search event occurrence times, purchase event occurrence times, and product recommendation levels".

[0056] Specifically, the present invention determines the type of customer information in the AI ​​form based on the network type used by the customer when browsing products. When the customer browses products in a mobile network environment, the customer's geographical location is determined to be in the company or at home based on the time type when the customer browses products. When the customer browses products during break time, the customer's geographical location is determined to be at home; when the customer browses products during working hours, the customer's geographical location is determined to be in the company. The above method avoids the phenomenon of inaccurate classification of customer information due to inaccurate analysis of the customer's geographical location.

[0057] Specifically, when recommending products to a customer, the type of information selected is determined based on the customer's location type;

[0058] When the location type of the customer is the first location type, selecting the first information type to recommend products to the customer;

[0059] When the location type of the customer is the second location type, the second information type is selected to recommend products to the customer.

[0060] In the embodiment of the present invention, the method for determining the first location type and the second location type is the same as the method for determining the information type. The first location type is a company type, and the second location type is a family type.

[0061] Specifically, the present invention determines whether the customer is in the company or at home through analyzing the customer's location type and selects corresponding information types as data for recommending products. When the customer is in the company, the first information type is selected as data for recommending products, and when the customer is at home, the second information type is selected as data for recommending products. The above method not only reduces the reference data for recommending products and thus improves the efficiency and accuracy of product recommendations, but also takes into account the phenomenon that when the customer is at home, the recommended products may be inaccurate because the products are not purchased by the customer himself, thereby improving the accuracy and intelligence of product recommendations.

[0062] Specifically, when determining the product recommendation level, the product recommendation level is determined based on the comparison result of the number of occurrences N of the search event of the product in the selected information type and the preset number of occurrences N0;

[0063] When N<N0, the commodity recommendation level is determined to be the second level;

[0064] When N≥N0, determine to make a secondary judgment on the product recommendation level;

[0065] Among them, the value of the preset occurrence number N0 is the average value of the search times corresponding to the goods purchased by the customer, and its value is set to 5, but the above value is not limited to this, and technical personnel in this field can also adjust the value according to actual needs.

[0066] Specifically, when determining to make a secondary determination on the product recommendation level, the product recommendation level is determined according to the comparison result of the maximum time value M of the occurrence of the search event of the product and the preset value M0;

[0067] When M>M0, the commodity recommendation level is determined to be the first level;

[0068] When M≤M0, the product recommendation level is determined to be the third level;

[0069] Among them, the value of the preset value M0 is the maximum time value of the purchase event corresponding to the commodity, but the above value is not limited to this, and technical personnel in this field can also adjust the value according to actual needs.

[0070] The maximum time value for the occurrence of the search event in the embodiment of the present invention is determined by the following method. When the time value for the search event for the product badminton includes 20210305, 20210315 and 20210405, the maximum time value is 20210405. The maximum time value for the occurrence of the product purchase event is determined by the same method.

[0071] In the embodiment of the present invention, after determining the commodity recommendation level, commodities are recommended to the customer in the order of the first level, the second level and the third level.

[0072] Specifically, the present invention determines the recommendation level of a product based on a comparison result of the number of occurrences of the product search event and a preset number of occurrences. When the number of occurrences of the product search event is less than the preset number of occurrences, the product recommendation level is determined to be the second level. When the number of occurrences of the product search event is greater than or equal to the preset number of occurrences, a secondary judgment is performed. During the secondary judgment, the maximum time value of the product search event is compared with the maximum time value of the product purchase event to determine whether the product is a product that has just been purchased, and then determine the product recommendation level. The above method improves the accuracy of determining the product recommendation level. Taking into account that the customer has purchased a product with a search event number greater than the preset number of occurrences, the product recommendation level is determined to be the third level, thereby improving the accuracy of product recommendations.

[0073] Specifically, when determining the eligibility of the product recommendation process, the eligibility of the product recommendation process is determined based on the comparison result between the product purchase rate W on the product recommendation page and the preset product purchase rate W0;

[0074] When W<W0, it is determined that the product recommendation process is unqualified;

[0075] When W≥W0, it is determined that the product recommendation process is qualified;

[0076] The value of the preset product purchase rate W0 is the historical average of the product purchase rates when customers browse products on the recommended product page, but the above value is not limited thereto, and those skilled in the art may also adjust the value according to actual needs.

[0077] The value of the product purchase rate in the embodiment of the present invention is the ratio of the number of products purchased by customers on the product recommendation page to the total number of products on the product recommendation page. In one embodiment of the present invention, the number of products on the product recommendation page is set to 10, and the preset product purchase rate is set to 0.15.

[0078] Specifically, the present invention determines the customer's satisfaction with the recommended products based on the customer's product purchase rate on the product recommendation page. When the customer's product purchase rate on the product recommendation page is less than the preset product purchase rate, the product recommendation process is determined to be unqualified. When the customer's product purchase rate on the product recommendation page is greater than or equal to the preset product purchase rate, the product recommendation process is determined to be qualified. The above method is used to intuitively determine the customer's product purchase rate for the recommended products and then determine the qualification of the product recommendation process, thereby improving the accuracy of the product recommendation process analysis and optimizing the automatic machine learning model.

[0079] Specifically, when it is determined that the recommended product is unqualified, the adjustment method is determined according to the comparison result of the browsing time T of the customer on the product recommendation page and the preset browsing time T0;

[0080] When T<T0, it is determined to adjust the product recommendation process in a first adjustment manner;

[0081] When T≥T0, determine to adjust the product recommendation process in a second adjustment manner;

[0082] Among them, the value of the preset browsing time T0 is the historical average time that customers browse the recommended product page when purchasing products on the recommended product page, and its value is set to 32, in seconds, but the above value is not limited to this, and technical personnel in this field can also adjust the value according to actual needs.

[0083] Specifically, when it is determined to adjust the product recommendation process in the first adjustment manner, it is determined to adjust the preset number of occurrences by the adjustment coefficient K;

[0084] When it is determined that the commodity recommendation process is adjusted in the second adjustment manner, it is determined that a commodity of the same category with a lower price than the recommended commodity is recommended.

[0085] Specifically, the adjustment coefficient K1 is calculated and set by the following formula:

[0086]

[0087] The adjusted preset number of occurrences is set to N0'=K×N0.

[0088] Specifically, the present invention determines an adjustment method for the product recommendation process based on a comparison result of the customer's browsing time on the product recommendation page and the preset browsing time when the product recommendation process is unqualified. When the browsing time is less than the preset browsing time, it means that the customer is not satisfied with the recommended products on the product recommendation page. The adjustment coefficient is used to increase the preset number of occurrences and thus improve the accuracy of the recommended product level judgment. When the browsing time is greater than or equal to the preset browsing time, it means that the customer did not purchase the recommended product due to financial problems, resulting in a low product purchase rate. The recommended price is determined to be lower than that of the same type of products recommended, thereby improving the accuracy of product recommendations and optimizing the automatic machine learning model.

[0089] Example 1

[0090]

[0091]

[0092] In Example 1, the number of search event occurrences for badminton in the AI ​​table is less than the preset number of occurrences, so the product recommendation level of badminton is the second level, the number of search event occurrences for volleyball is greater than the preset number of occurrences and the search event occurrence time value is greater than the purchase event occurrence time value, so the product recommendation level of volleyball is the first level; the number of search event occurrences for basketball in the AI ​​table is greater than the preset number of occurrences and the search event occurrence time value is less than the purchase event occurrence time value, so the product recommendation level of badminton is the third level, and the number of search event occurrences for sweaters is less than the preset number of occurrences, so the product recommendation level of sweaters is the second level.

[0093] Example 2

[0094]

[0095]

[0096] In Example 2, the number of search event occurrences of potato chips in the AI ​​table is greater than the preset number of occurrences and the search event occurrence time value is greater than the purchase event occurrence time value, so the product recommendation level of potato chips is the first level, and the number of search event occurrences of jelly is less than the preset number of occurrences, so the product recommendation level of jelly is the second level; the number of search event occurrences of jerseys in the AI ​​table is greater than the preset number of occurrences and the search event occurrence time value is less than the purchase event occurrence time value, so the product recommendation level of jerseys is the third level, and the number of search event occurrences of sweatpants is less than the preset number of occurrences, so the product recommendation level of sweatpants is the second level.

[0097] Example 3

[0098]

[0099]

[0100] In Example 3, the number of search event occurrences of hats in the AI ​​table is less than the preset number of occurrences, so the product recommendation level of hats is the second level, the number of search event occurrences of shirts is greater than the preset number of occurrences and the search event occurrence time value is greater than the purchase event occurrence time value, so the product recommendation level of shirts is the first level; the number of search event occurrences of sports shoes in the AI ​​table is less than the preset number of occurrences, so the product recommendation level of sports shoes is the second level, the number of search event occurrences of swimsuits is less than the preset number of occurrences, so the product recommendation level of swimsuits is the second level.

[0101] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0102] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for constructing an automatic machine learning model for SQL in a data service mode, characterized in that: include: Preprocess the acquired customer information and mine the features of the preprocessed information to generate AI forms; Determine the type of information in the AI ​​form based on customer information; Selecting a corresponding information type according to the location type when the customer browses the product, and determining the product recommendation level according to the number of occurrences of the product search event in the selected information type and the maximum time value of the occurrence of the search event; The data of the process of recommending products to customers is input into the model as training data to obtain an automatic machine learning model; Determining the eligibility of the product recommendation process according to the product purchase rate of the customer on the product recommendation page, and adjusting the construction process of the learning model according to the determined eligibility, including adjusting the determination process of the product recommendation level, or recommending products of the same category with a lower price than the recommended product; The process of determining the type of information in the AI ​​form according to the network type is to determine the type of information in the AI ​​form according to the network environment where the customer information is located; wherein the network type includes a wireless network type and a mobile network type; In a wireless network type environment, the information type is determined to be a first type based on the IP address type being a company IP address, or the information type is determined to be a second type based on the IP address type being a home IP address; In a mobile network type environment, the information type is determined to be the first type based on the time type when the customer browses the product being the working time type, or the information type is determined to be the second type based on the time type when the customer browses the product being the rest time type.

2. The method for constructing an automatic machine learning model for SQL under the data service mode according to claim 1, characterized in that: When recommending products to customers, the first information type is selected based on the customer's location type being company type to recommend products to customers, or the second information type is selected based on the customer's location type being home type to recommend products to customers.

3. The method for constructing an automatic machine learning model for SQL under the data service mode according to claim 2, characterized in that: When determining the product recommendation level, the product recommendation level is determined to be the second level based on the number of search events for the product in the selected information type being less than the preset number of occurrences, or the product recommendation level is determined to be a second level based on the number of search events for the product in the selected information type being greater than or equal to the preset number of occurrences.

4. The method for constructing an automatic machine learning model for SQL under the data service mode according to claim 3, characterized in that: When making a secondary judgment on the product recommendation level, the product recommendation level is determined to be the first level based on the maximum time value of the product search event being greater than a preset value, or the product recommendation level is determined to be the third level based on the maximum time value of the product search event being less than or equal to a preset value.

5. The method for constructing an automatic machine learning model for SQL under the data service mode according to claim 4, characterized in that: When determining the eligibility of the product recommendation process, the product recommendation process is determined to be unqualified based on the fact that the product purchase rate of customers on the product recommendation page is less than a preset product purchase rate.

6. The method for constructing an automatic machine learning model for SQL under the data service mode according to claim 5, characterized in that: When it is determined that the product recommendation process is unqualified, the preset number of occurrences is adjusted by an adjustment coefficient based on the fact that the browsing time of the customer on the product recommendation page is less than the preset browsing time.

7. The method for constructing an automatic machine learning model for SQL under the data service mode according to claim 6, characterized in that: When it is determined that the product recommendation process is unqualified, a product of the same type with a lower price than the recommended product is replaced based on the customer's browsing time on the product recommendation page being greater than or equal to a preset browsing time.

Images