A method and system for predicting e-commerce consumer preferences based on machine learning

Abstract

The present invention discloses a method and system for predicting e-commerce consumer preferences based on machine learning, and the method includes data collection, data reconstruction, data preprocessing, consumer preference inference and e-commerce consumer preference prediction. The present invention belongs to the field of e-commerce analysis technology, and specifically refers to a method and system for predicting e-commerce consumer preferences based on machine learning. The scheme connects text comment data and reconstructs heterogeneous data, thereby realizing simultaneous processing of numerical purchase record data and text comment data, and improving the data dimension of consumer preference prediction; adopts text filtering and word segmentation processing methods to preprocess text, and optimizes the data representation of text through word vector conversion, thereby improving data quality and providing good data support for subsequent modeling and prediction; adopts a multi-view latent Dirichlet allocation model to infer consumer preferences, thereby improving the interpretability of the model and thus improving the feasibility and effectiveness of analyzing consumer purchase motivations.

Description

Technical Field

[0001] The present invention belongs to the technical field of e-commerce analysis, and specifically refers to an e-commerce consumer preference prediction method and system based on machine learning. Background Art

[0002] E-commerce consumer preference prediction is of vital importance in the e-commerce industry. It uses machine learning technology to analyze and model the data generated by consumers on e-commerce platforms, predict consumer preferences, and make personalized recommendations, thereby improving user experience, increasing user stickiness, and thus improving the profitability of e-commerce platforms.

[0003] However, in the existing e-commerce consumer preference prediction process, there are technical problems that, under the e-commerce model, consumer preference prediction must consider both the objective data records of consumer purchasing behavior and the subjective comments and evaluations of consumers on products in order to more accurately reflect e-commerce consumer preferences; there are technical problems that the quality of comment text data applicable to Chinese is uneven, and the original data often has missing, text anomalies and symbol interference; there is a technical problem that the simultaneous processing of numerical purchase record data and text comment data has high requirements on the scalability and accuracy of the model, and the existing model has poor interpretability, which leads to limited method availability, and the dependency relationship between consumers' purchase record data and comments needs to be modeled and further predicted. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an e-commerce consumer preference prediction method and system based on machine learning. In view of the technical problem that in the existing e-commerce consumer preference prediction process, consumer preference prediction under the e-commerce model must consider both the objective data records of consumer purchasing behavior and the subjective comments and evaluations of consumers on the products in order to more accurately reflect the e-commerce consumer preferences, this solution creatively connects text comment data and reconstructs heterogeneous data, thereby realizing the simultaneous processing of numerical purchase record data and text comment data, and improving the data dimension of consumer preference prediction; in view of the technical problem that in the existing e-commerce consumer preference prediction process, the quality of comment text data applicable to Chinese is uneven, and the original data often has missing data, text anomalies and symbol interference. Technical issues: This solution uses text filtering and word segmentation methods to preprocess text, and optimizes the data representation of text through word vector conversion, which improves the data quality of preference prediction and provides good data support for subsequent modeling and prediction; in the existing e-commerce consumer preference prediction process, there is a simultaneous processing of numerical purchase record data and text comment data, which has high requirements on the scalability and accuracy of the model, and the existing model has poor interpretability, which leads to limited method availability. It is necessary to model the dependency relationship between consumer purchase record data and comments and further predict the technical problem. This solution creatively uses a multi-view latent Dirichlet allocation model to infer consumer preferences, improves the interpretability of the model, and thus improves the feasibility and effectiveness of analyzing consumers' specific purchasing motivations.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides an e-commerce consumer preference prediction method based on machine learning, which comprises the following steps:

[0006] Step S1: data collection;

[0007] Step S2: Data reconstruction, which is used to aggregate and reconstruct the original data, specifically, by reconstructing the original data of e-commerce consumer preference prediction to obtain preference prediction reconstructed data;

[0008] Step S3: data preprocessing, which is used to perform data optimization processing on the preference prediction reconstruction data, specifically, obtaining the preference prediction optimization data by performing text preprocessing on the preference prediction reconstruction data;

[0009] Step S4: consumer preference estimation;

[0010] Step S5: E-commerce consumer preference prediction.

[0011] Furthermore, in step S1, the data collection is used to collect the basic data set required for e-commerce consumer preference prediction, specifically, to collect data from the e-commerce platform to obtain the original data for e-commerce consumer preference prediction;

[0012] The e-commerce consumer preference prediction raw data specifically includes consumer basic information data, purchase record basic information data, consumer review data and product basic information data.

[0013] Furthermore, in step S2, the data reconstruction includes the following steps:

[0014] Step S21: Product category screening, which is used to screen the product categories for preference prediction, specifically, screening the e-commerce consumer preference prediction raw data belonging to the target category from the product basic information data in the e-commerce consumer preference prediction raw data, and removing the data that does not belong to the target category to obtain category filtering data;

[0015] Step S22: text comment data connection, used to connect text comments, specifically, based on the basic information data of consumers in the original data of e-commerce consumer preference prediction, data aggregation is performed on consumer comment data belonging to the same consumer, and the data aggregation specifically refers to connecting the text comments in the consumer comment data to obtain connected text comment data;

[0016] Step S23: heterogeneous data aggregation, specifically, based on the consumer basic information data in the e-commerce consumer preference prediction raw data, data aggregation is performed on the category filtering data and the connection text comment data belonging to the same consumer;

[0017] Step S24: data aggregation and reconstruction, specifically, performing data aggregation and reconstruction through the product category screening, the text review data connection and the heterogeneous data aggregation to obtain preference prediction and reconstruction data.

[0018] Furthermore, in step S3, the data preprocessing includes the following steps:

[0019] Step S31: text filtering, for eliminating non-Chinese characters and punctuation marks, specifically filtering the non-Chinese characters and punctuation marks in the preference prediction reconstruction data to obtain preference prediction filtered data;

[0020] Step S32: word segmentation and elimination of stop words, specifically, using a bidirectional maximum matching algorithm to perform word segmentation on the preference prediction filtering data, and performing a stop word elimination operation to obtain a preference prediction word segmentation result;

[0021] Step S33: word vector conversion, which is used to convert the preference prediction word segmentation result into a word vector, specifically, using the Doc2vec model to convert the preference prediction word segmentation result into a vector representation to obtain the preference prediction word vector;

[0022] Step S34: text preprocessing, through the text filtering, word segmentation processing and elimination of stop words and word vector conversion, text preprocessing is performed to obtain preference prediction optimization data.

[0023] Further, in step S4, the consumer preference inference is specifically inferring the consumer preference based on the preference prediction optimization data and the multi-view latent Dirichlet allocation model to obtain the consumer preference inference data, including the following steps:

[0024] Step S41: multi-view definition, specifically dividing the preference prediction optimization data into different data views, the data views specifically including comment topic view, product topic view and purchase record data topic view;

[0025] Step S42: define symmetric prior parameters, specifically setting the symmetric prior parameters of the Dirichlet distribution, wherein the symmetric prior parameters of the Dirichlet distribution specifically include the symmetric prior parameters of the comment , commodity symmetry prior parameters , symmetric prior parameters of review preference and the symmetric prior parameters of product preferences ;

[0026] Step S43: constructing a multi-view latent Dirichlet allocation model, including the following steps:

[0027] Step S431: Modeling the comment topic view, the calculation formula is:

[0028] ;

[0029] Where k is the review topic index, is the multinomial distribution of the k-th review topic, is the comment symmetry prior parameter, which is used to control the density of comment topics. is the comment symmetric prior parameter The Dirichlet distribution of is the sampling operation identifier;

[0030] Step S432: Modeling of commodity theme view, the calculation formula is:

[0031] ;

[0032] Where p is the product index, is the multinomial distribution of product topics, which specifically refers to the multinomial distribution of topics related to the pth product in the kth review topic, is a commodity symmetric prior parameter, which is used to control the density of commodity topics. is the commodity symmetry prior parameter The Dirichlet distribution of

[0033] Step S433: Consumer review preference modeling, the calculation formula is:

[0034] ;

[0035] Where d is the consumer index, is the dth consumer review preference vector, which is used to indicate the consumer’s preference for different review topics. is the symmetric prior parameter of review preference, is the comment preference symmetric prior parameter The Dirichlet distribution of

[0036] Step S434: Consumer product preference modeling is used to model the dependency relationship between the consumer's purchase record data and comments. The calculation formula is:

[0037] ;

[0038] In the formula, is a consumer product preference vector, which is used to indicate the degree of consumer product preference. The consumer product preference vector specifically refers to the product preference probability vector of the dth consumer for the kth review topic. is the symmetric prior parameter of product preference, is the symmetric prior parameter of product preference The Dirichlet distribution of

[0039] Step S435: iterative parameter update, specifically, iteratively updating the symmetric prior parameters of the Dirichlet distribution through the Bayesian inference method until the model converges;

[0040] Step S44: training the consumer preference inference model, specifically, training the model through the multi-view definition, the symmetric prior parameter definition and the multi-view latent Dirichlet allocation model construction to obtain the consumer preference inference model Model Cpre ;

[0041] Step S45: Using the consumer preference inference model Model Cpre Conduct consumer preference inference and obtain consumer preference inference data.

[0042] Furthermore, in step S5, the e-commerce consumer preference prediction is used to perform e-commerce consumer preference prediction, specifically to perform e-commerce consumer preference prediction based on consumer preference inference data, to obtain consumer preference distribution analysis data and product theme preference association distribution analysis data, and further obtain e-commerce consumer preference prediction data, the e-commerce consumer preference prediction data including consumer preference distribution, product theme preference association distribution and e-commerce consumer preference reference information.

[0043] The present invention provides an e-commerce consumer preference prediction system based on machine learning, comprising: a data acquisition module, a data reconstruction module, a data preprocessing module, a consumer preference inference module and an e-commerce consumer preference prediction module;

[0044] The data collection module obtains the original data of e-commerce consumer preference prediction through data collection, and sends the original data of e-commerce consumer preference prediction to the data reconstruction module;

[0045] The data reconstruction module reconstructs the original data of the e-commerce consumer preference prediction to obtain preference prediction reconstructed data, and sends the preference prediction reconstructed data to the data preprocessing module;

[0046] The data preprocessing module performs text preprocessing on the preference prediction reconstruction data to obtain preference prediction optimization data, and sends the preference prediction optimization data to the consumer preference inference module;

[0047] The consumer preference inference module infers consumer preferences through a multi-view latent Dirichlet allocation model to obtain consumer preference inference data, and sends the consumer preference inference data to the e-commerce consumer preference prediction module;

[0048] The e-commerce consumer preference prediction module is used for e-commerce consumer preference prediction, specifically to perform e-commerce consumer preference prediction based on the consumer preference inference data, obtain consumer preference distribution analysis data and product theme preference association distribution analysis data, and further obtain e-commerce consumer preference prediction data.

[0049] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0050] (1) In view of the technical problem that in the existing e-commerce consumer preference prediction process, consumer preference prediction must consider both the objective data records of consumer purchasing behavior and the subjective comments and evaluations of consumers on products in order to more accurately reflect the e-commerce consumer preferences, this solution creatively connects text comment data and reconstructs heterogeneous data, thereby realizing the simultaneous processing of numerical purchase record data and text comment data, thereby improving the data dimension of consumer preference prediction.

[0051] (2) In view of the technical problems that the quality of Chinese review text data is uneven and the original data often contains missing data, text anomalies and symbol interference in the existing e-commerce consumer preference prediction process, this solution uses text filtering and word segmentation methods to preprocess the text, and optimizes the data representation of the text through word vector conversion, thereby improving the data quality of preference prediction and providing good data support for subsequent modeling and prediction.

[0052] (3) In the existing e-commerce consumer preference prediction process, there is the simultaneous processing of numerical purchase record data and text review data, which places high requirements on the scalability and accuracy of the model. In addition, the existing model has poor interpretability, which leads to limited method availability. It is necessary to model the dependency relationship between consumers' purchase record data and comments and further predict them. In view of the technical problem, this scheme creatively adopts a multi-view latent Dirichlet allocation model to infer consumer preferences, improves the interpretability of the model, and thus improves the feasibility and effectiveness of analyzing consumers' specific purchasing motivations. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A flowchart of a method for predicting e-commerce consumer preferences based on machine learning provided by the present invention;

[0054] Figure 2 A schematic diagram of an e-commerce consumer preference prediction system based on machine learning provided by the present invention;

[0055] Figure 3 is a schematic flow chart of step S2;

[0056] Figure 4 is a schematic flow chart of step S3;

[0057] Figure 5 It is a schematic diagram of the process of step S4.

[0058] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0060] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0061] Example 1, see Figure 1 The present invention provides a method for predicting e-commerce consumer preferences based on machine learning, the method comprising the following steps:

[0062] Step S1: data collection;

[0063] Step S2: Data reconstruction, which is used to aggregate and reconstruct the original data, specifically, by reconstructing the original data of e-commerce consumer preference prediction to obtain preference prediction reconstructed data;

[0064] Step S3: data preprocessing, which is used to perform data optimization processing on the preference prediction reconstruction data, specifically, obtaining the preference prediction optimization data by performing text preprocessing on the preference prediction reconstruction data;

[0065] Step S4: consumer preference estimation;

[0066] Step S5: E-commerce consumer preference prediction.

[0067] Example 2, see Figure 1 , this embodiment is based on the above embodiment. In step S1, the data collection is used to collect the basic data set required for e-commerce consumer preference prediction, specifically, to collect data from the e-commerce platform to obtain the original data for e-commerce consumer preference prediction;

[0068] The e-commerce consumer preference prediction raw data specifically includes consumer basic information data, purchase record basic information data, consumer review data and product basic information data.

[0069] Example 3, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S2, the data reconstruction includes the following steps:

[0070] Step S21: Product category screening, which is used to screen the product categories for preference prediction, specifically, screening the e-commerce consumer preference prediction raw data belonging to the target category from the product basic information data in the e-commerce consumer preference prediction raw data, and removing the data that does not belong to the target category to obtain category filtering data;

[0071] The target categories specifically refer to clothing products, shoes products, and jewelry products;

[0072] The category filtering data specifically refers to the basic information data of the purchase records containing the target category commodities;

[0073] Step S22: text comment data connection, used to connect text comments, specifically, based on the basic information data of consumers in the original data of e-commerce consumer preference prediction, data aggregation is performed on consumer comment data belonging to the same consumer, and the data aggregation specifically refers to connecting the text comments in the consumer comment data to obtain connected text comment data;

[0074] Step S23: heterogeneous data aggregation, specifically, based on the consumer basic information data in the e-commerce consumer preference prediction raw data, data aggregation is performed on the category filtering data and the connection text comment data belonging to the same consumer;

[0075] Step S24: data aggregation and reconstruction, specifically, performing data aggregation and reconstruction through the product category screening, the text review data connection and the heterogeneous data aggregation to obtain preference prediction and reconstruction data;

[0076] By performing the above operations, in view of the technical problem that in the existing e-commerce consumer preference prediction process, consumer preference prediction must consider both the objective data records of consumer purchase behavior and the subjective comments and evaluations of consumers on commodities in the e-commerce model in order to more accurately reflect the e-commerce consumer preferences, this solution creatively connects text comment data and reconstructs heterogeneous data, thereby realizing the simultaneous processing of numerical purchase record data and text comment data, and improving the data dimension of consumer preference prediction. .

[0077] Example 4, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S3, the data preprocessing includes the following steps:

[0078] Step S31: text filtering, for eliminating non-Chinese characters and punctuation marks, specifically filtering the non-Chinese characters and punctuation marks in the preference prediction reconstruction data to obtain preference prediction filtered data;

[0079] Step S32: word segmentation and elimination of stop words, specifically, using a bidirectional maximum matching algorithm to perform word segmentation on the preference prediction filtering data, and performing a stop word elimination operation to obtain a preference prediction word segmentation result;

[0080] Step S33: word vector conversion, which is used to convert the preference prediction word segmentation result into a word vector, specifically, using the Doc2vec model to convert the preference prediction word segmentation result into a vector representation to obtain the preference prediction word vector;

[0081] Step S34: text preprocessing, through the text filtering, the word segmentation processing and the elimination of stop words and the word vector conversion, the text preprocessing is performed to obtain preference prediction optimization data;

[0082] By performing the above operations, in view of the technical problems that the quality of review text data applicable to Chinese is uneven and the original data often has missing, text anomalies and symbol interference in the existing e-commerce consumer preference prediction process, this solution adopts text filtering and word segmentation processing methods to perform text preprocessing, and optimizes the data representation of the text through word vector conversion, thereby improving the data quality of preference prediction and providing good data support for subsequent modeling and prediction.

[0083] Example 5, see Figure 1 and Figure 5 This embodiment is based on the above embodiment. In step S4, the consumer preference inference is specifically to infer the consumer preference based on the multi-view latent Dirichlet allocation model according to the preference prediction optimization data to obtain the consumer preference inference data, including the following steps:

[0084] Step S41: multi-view definition, specifically dividing the preference prediction optimization data into different data views, the data views specifically including comment topic view, product topic view and purchase record data topic view;

[0085] Step S42: define symmetric prior parameters, specifically setting the symmetric prior parameters of the Dirichlet distribution, wherein the symmetric prior parameters of the Dirichlet distribution specifically include the symmetric prior parameters of the comment , commodity symmetry prior parameters , symmetric prior parameters of review preference and the symmetric prior parameters of product preferences ;

[0086] Step S43: constructing a multi-view latent Dirichlet allocation model, including the following steps:

[0087] Step S431: Modeling the comment topic view, the calculation formula is:

[0088] ;

[0089] Where k is the review topic index, is the multinomial distribution of the k-th review topic, is the comment symmetry prior parameter, which is used to control the density of comment topics. is the comment symmetric prior parameter The Dirichlet distribution of is the sampling operation identifier;

[0090] Step S432: Modeling of commodity theme view, the calculation formula is:

[0091] ;

[0092] Where p is the product index, is the multinomial distribution of product topics, which specifically refers to the multinomial distribution of topics related to the pth product in the kth review topic, is a commodity symmetric prior parameter, which is used to control the density of commodity topics. is the commodity symmetry prior parameter The Dirichlet distribution of

[0093] Step S433: Consumer review preference modeling, the calculation formula is:

[0094] ;

[0095] Where d is the consumer index, is the dth consumer review preference vector, which is used to indicate the consumer’s preference for different review topics. is the symmetric prior parameter of review preference, is the comment preference symmetric prior parameter The Dirichlet distribution of

[0096] Step S434: Consumer product preference modeling is used to model the dependency relationship between the consumer's purchase record data and comments. The calculation formula is:

[0097] ;

[0098] In the formula, is the consumer product preference vector, which specifically refers to the product preference probability vector of the dth consumer for the kth review topic, is the symmetric prior parameter of product preference, is the symmetric prior parameter of product preference The Dirichlet distribution of

[0099] Step S435: iterative parameter update, specifically, iteratively updating the symmetric prior parameters of the Dirichlet distribution through the Bayesian inference method until the model converges;

[0100] Step S44: training the consumer preference inference model, specifically, training the model through the multi-view definition, the symmetric prior parameter definition and the multi-view latent Dirichlet allocation model construction to obtain the consumer preference inference model Model Cpre ;

[0101] Step S45: Using the consumer preference inference model Model Cpre Conduct consumer preference inference and obtain consumer preference inference data;

[0102] By performing the above operations, in view of the technical problem that in the existing e-commerce consumer preference prediction process, there is the simultaneous processing of numerical purchase record data and text comment data, which has high requirements on the scalability and accuracy of the model, and the existing model has poor interpretability, which leads to limited method availability, and the dependency relationship between consumers' purchase record data and comments needs to be modeled and further predicted, this scheme creatively adopts a multi-view latent Dirichlet allocation model to infer consumer preferences, improves the interpretability of the model, and thus improves the feasibility and effectiveness of analyzing consumers' specific purchasing motivations.

[0103] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S5, the e-commerce consumer preference prediction is used to perform e-commerce consumer preference prediction, specifically, based on consumer preference inference data, perform e-commerce consumer preference prediction, obtain consumer preference distribution analysis data and product theme preference association distribution analysis data, and further obtain e-commerce consumer preference prediction data, the e-commerce consumer preference prediction data includes consumer preference distribution, product theme preference association distribution and e-commerce consumer preference reference information.

[0104] Embodiment 7, see Figure 2 , this embodiment is based on the above embodiment, and the present invention provides an e-commerce consumer preference prediction system based on machine learning, including: a data acquisition module, a data reconstruction module, a data preprocessing module, a consumer preference inference module and an e-commerce consumer preference prediction module;

[0105] The data collection module obtains the original data of e-commerce consumer preference prediction through data collection, and sends the original data of e-commerce consumer preference prediction to the data reconstruction module;

[0106] The data reconstruction module reconstructs the original data of the e-commerce consumer preference prediction to obtain preference prediction reconstructed data, and sends the preference prediction reconstructed data to the data preprocessing module;

[0107] The data preprocessing module performs text preprocessing on the preference prediction reconstruction data to obtain preference prediction optimization data, and sends the preference prediction optimization data to the consumer preference inference module;

[0108] The consumer preference inference module infers consumer preferences through a multi-view latent Dirichlet allocation model to obtain consumer preference inference data, and sends the consumer preference inference data to the e-commerce consumer preference prediction module;

[0109] The e-commerce consumer preference prediction module is used for e-commerce consumer preference prediction, specifically to perform e-commerce consumer preference prediction based on the consumer preference inference data, obtain consumer preference distribution analysis data and product theme preference association distribution analysis data, and further obtain e-commerce consumer preference prediction data.

[0110] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0111] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0112] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A method for predicting e-commerce consumer preferences based on machine learning, characterized by: The method comprises the following steps: Step S1: Data collection, which is used to collect the basic data set required for e-commerce consumer preference prediction, specifically, to collect data from the e-commerce platform to obtain the original data for e-commerce consumer preference prediction; The e-commerce consumer preference prediction raw data specifically includes consumer basic information data, purchase record basic information data, consumer review data, and product basic information data; Step S2: Data reconstruction, which is used to aggregate and reconstruct the original data. Specifically, the original data of e-commerce consumer preference prediction is reconstructed to obtain preference prediction reconstructed data, including the following steps: Step S21: Product category screening, which is used to screen the product categories for preference prediction, specifically, screening the e-commerce consumer preference prediction raw data belonging to the target category from the product basic information data in the e-commerce consumer preference prediction raw data, and removing the data that does not belong to the target category to obtain category filtering data; Step S22: text comment data connection, used to connect text comments, specifically, based on the basic information data of consumers in the original data of e-commerce consumer preference prediction, data aggregation is performed on consumer comment data belonging to the same consumer, and the data aggregation specifically refers to connecting the text comments in the consumer comment data to obtain connected text comment data; Step S23: heterogeneous data aggregation, specifically, based on the consumer basic information data in the e-commerce consumer preference prediction raw data, data aggregation is performed on the category filtering data and the connection text comment data belonging to the same consumer; Step S24: data aggregation and reconstruction, specifically, performing data aggregation and reconstruction through the product category screening, the text review data connection and the heterogeneous data aggregation to obtain preference prediction and reconstruction data; Step S3: data preprocessing, which is used to perform data optimization processing on the preference prediction reconstruction data, specifically, obtaining the preference prediction optimization data by performing text preprocessing on the preference prediction reconstruction data, including the following steps: Step S31: text filtering, for eliminating non-Chinese characters and punctuation marks, specifically filtering the non-Chinese characters and punctuation marks in the preference prediction reconstruction data to obtain preference prediction filtered data; Step S32: word segmentation and elimination of stop words, specifically, using a bidirectional maximum matching algorithm to perform word segmentation on the preference prediction filtering data, and performing a stop word elimination operation to obtain a preference prediction word segmentation result; Step S33: word vector conversion, which is used to convert the preference prediction word segmentation result into a word vector, specifically, using the Doc2vec model to convert the preference prediction word segmentation result into a vector representation to obtain the preference prediction word vector; Step S34: text preprocessing, through the text filtering, the word segmentation processing and the elimination of stop words and the word vector conversion, text preprocessing is performed to obtain preference prediction optimization data; Step S4: inferring consumer preferences, specifically inferring consumer preferences based on the preference prediction optimization data and the multi-view latent Dirichlet allocation model to obtain consumer preference inference data, including the following steps: Step S41: multi-view definition, specifically dividing the preference prediction optimization data into different data views, the data views specifically including comment topic view, product topic view and purchase record data topic view; Step S42: define symmetric prior parameters, specifically setting the symmetric prior parameters of the Dirichlet distribution, wherein the symmetric prior parameters of the Dirichlet distribution specifically include the symmetric prior parameters of the comment , commodity symmetry prior parameters , symmetric prior parameters of review preference and the symmetric prior parameters of product preferences ; Step S43: constructing a multi-view latent Dirichlet allocation model; Step S44: training the consumer preference inference model, specifically, training the model through the multi-view definition, the symmetric prior parameter definition and the multi-view latent Dirichlet allocation model construction to obtain the consumer preference inference model Model Cpre ; Step S45: Using the consumer preference inference model Model Cpre Conduct consumer preference inference and obtain consumer preference inference data; Step S5: e-commerce consumer preference prediction is used to perform e-commerce consumer preference prediction, specifically, based on consumer preference inference data, perform e-commerce consumer preference prediction, obtain consumer preference distribution analysis data and product theme preference association distribution analysis data, and further obtain e-commerce consumer preference prediction data, the e-commerce consumer preference prediction data includes consumer preference distribution, product theme preference association distribution and e-commerce consumer preference reference information.

2. The method for predicting e-commerce consumer preferences based on machine learning according to claim 1, characterized in that: In step S43, the multi-view latent Dirichlet allocation model is constructed, including the following steps: Step S431: Modeling the comment topic view, the calculation formula is: ； Where k is the review topic index, is the multinomial distribution of the k-th review topic, is the comment symmetry prior parameter, which is used to control the density of comment topics. is the comment symmetric prior parameter The Dirichlet distribution of is the sampling operation identifier; Step S432: Modeling of commodity theme view, the calculation formula is: ； Where p is the product index, is the multinomial distribution of product topics, which specifically refers to the multinomial distribution of topics related to the pth product in the kth review topic, is a commodity symmetric prior parameter, which is used to control the density of commodity topics. is the commodity symmetry prior parameter The Dirichlet distribution of Step S433: Consumer review preference modeling, the calculation formula is: ； Where d is the consumer index, is the dth consumer review preference vector, which is used to indicate the consumer’s preference for different review topics. is the symmetric prior parameter of review preference, is the comment preference symmetric prior parameter The Dirichlet distribution of Step S434: Consumer product preference modeling is used to model the dependency relationship between the consumer's purchase record data and comments. The calculation formula is: ； In the formula, is the consumer product preference vector, which specifically refers to the product preference probability vector of the dth consumer for the kth review topic, is the symmetric prior parameter of product preference, is the symmetric prior parameter of product preference The Dirichlet distribution of Step S435: iterative parameter update, specifically, iteratively updating the symmetric prior parameters of the Dirichlet distribution through the Bayesian inference method until the model converges.

3. An e-commerce consumer preference prediction system based on machine learning, used to implement an e-commerce consumer preference prediction method based on machine learning as described in any one of claims 1-2, characterized in that: It includes data collection module, data reconstruction module, data preprocessing module, consumer preference inference module and e-commerce consumer preference prediction module.

4. The e-commerce consumer preference prediction system based on machine learning according to claim 3, characterized in that: The data collection module obtains the original data of e-commerce consumer preference prediction through data collection, and sends the original data of e-commerce consumer preference prediction to the data reconstruction module; The data reconstruction module reconstructs the original data of the e-commerce consumer preference prediction to obtain preference prediction reconstructed data, and sends the preference prediction reconstructed data to the data preprocessing module; The data preprocessing module performs text preprocessing on the preference prediction reconstruction data to obtain preference prediction optimization data, and sends the preference prediction optimization data to the consumer preference inference module; The consumer preference inference module infers consumer preferences through a multi-view latent Dirichlet allocation model to obtain consumer preference inference data, and sends the consumer preference inference data to the e-commerce consumer preference prediction module; The e-commerce consumer preference prediction module is used for e-commerce consumer preference prediction, specifically to perform e-commerce consumer preference prediction based on the consumer preference inference data, obtain consumer preference distribution analysis data and product theme preference association distribution analysis data, and further obtain e-commerce consumer preference prediction data.

Images