Consumer preference behavior information recommendation method and system based on big data

Abstract

The invention provides a consumer preference behavior information recommendation method and system based on big data, and belongs to the technical field of information interaction, and the method comprises the steps: obtaining preference behavior data of consumers, and obtaining a preference behavior database; extracting user personal characteristics in the preference behavior data, and obtaining user language habits according to the preference behavior data; correcting the personal characteristics according to a standard database to obtain corrected characteristics, calculating characteristic values by using the corrected characteristics and the language habits, and matching a preference behavior database; and establishing a consumer preference behavior information recommendation model according to the preference behavior database, inputting the to-be-answered data into the consumer preference behavior information recommendation model, and outputting a recommendation result. By constructing the preference behavior database, answers can be recommended to the user more accurately, individually and pertinently, so that the user experience and the problem solving efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the field of information interaction technology, and in particular to a method and system for recommending consumer preference behavior information based on big data. Background Art

[0002] In the current digital age, consumer behavior data has become a key resource for companies to formulate marketing strategies, optimize product services, and improve user experience. Traditional consumer preference analysis methods often rely on highly subjective methods such as questionnaires and market research, which makes it difficult to fully and real-time capture consumers' true preferences. With the rapid development of big data technology, using big data to analyze consumer preference behavior has become a more efficient and accurate method.

[0003] Currently, there are many consumer preference analysis systems based on big data on the market, but most of them focus on analyzing single-dimensional data, ignoring the impact of multi-dimensional information such as user personal characteristics and language habits on preference behavior. In addition, these systems often lack standardized data correction mechanisms, resulting in deviations in analysis results and affecting recommendation effects.

[0004] In order to overcome the shortcomings of the prior art, the present invention proposes a consumer preference behavior information recommendation method based on big data. The method comprehensively analyzes multi-dimensional information such as consumers' preference behavior data, personal characteristics, language habits, etc., combines with a standardized database for feature correction, and establishes an accurate consumer preference behavior information recommendation model to achieve personalized and intelligent recommendation services. Summary of the invention

[0005] In response to the above problems, the present invention provides a method and system for recommending consumer preference behavior information based on big data, which overcomes the problems of insufficient accuracy and lack of personalization in the prior art in answering questions about oilfield environmental protection and safety standards, and enables users to have a more satisfactory experience through intelligent question-and-answer interaction.

[0006] In a first aspect, the present invention provides a method for recommending consumer preference behavior information based on big data, comprising:

[0007] Obtaining consumer preference behavior data and obtaining a preference behavior database;

[0008] Extracting user personal characteristics from the preference behavior data, and obtaining user language habits based on the preference behavior data;

[0009] Modify the personal features according to the standard database to obtain modified features, calculate the characteristic value using the modified features and the language habits and match the preferred behavior database;

[0010] A consumer preference behavior information recommendation model is established according to the preference behavior database, the data to be answered is input into the consumer preference behavior information recommendation model, and a recommendation result is output.

[0011] As a further improvement of the present invention, the step of obtaining the consumer's preference behavior data to obtain the preference behavior database includes preprocessing the interaction data to obtain the interaction database, wherein the preprocessing includes removing duplicate data, vectorization, data conversion, data integration and data reduction;

[0012] The extracting of the user's personal features from the preference behavior data and acquiring the user's language habits according to the preference behavior data includes extracting the user's personal features from the preprocessed preference behavior data and acquiring the user's language habits according to the preference behavior data.

[0013] Furthermore, the extracting of user personal characteristics from the preference behavior data includes:

[0014] Build a bridge between the original time series and multidimensional space analysis, reconstruct the phase space, and make the one-dimensional time series obtain the reconstructed time series in the multidimensional space:

[0015] W i =(w i ,w i+τ ,…,w (m+1)τ ) T

[0016] Where τ is the time delay, m is the embedding dimension, i is the preference behavior data, and W i is the time series of preference behavior data i, w i is the phase point of preference behavior data i;

[0017] Calculate the distance between phase points:

[0018] D=∑ i≠j H(r-‖w i -w j ‖)

[0019] Where H(·) is the Heaviside function, r is the given distance, and w j is the phase point of preference behavior data j;

[0020] Calculate the ratio of preference behavior data less than r to the total number of phase point pairs:

[0021]

[0022] Where N is the number of preference behavior data, Q N (r) is the logarithmic proportion of distances less than r;

[0023] Compute the distance between two points:

[0024]

[0025] Calculate the ratio of the number of phase point pairs that are associated to all pairs:

[0026]

[0027] In the formula, Q m (r) is the correlation dimension whose distance is less than r, and the information gain weight of the correlation dimension is calculated as:

[0028] Output the correlation dimension as user personal features.

[0029] Furthermore, acquiring the user's language habits according to the preference behavior data includes:

[0030] Calculate the support of transaction items in the user preference behavior database:

[0031]

[0032] Where D is the number of interaction terms, R e is the number of transaction items containing the interaction term e, u e is the support of the interaction term e, v e is the weight of the interaction term e, ψ e Adjust the support parameter for the interaction term e;

[0033] Perform support statistics on the preference items in the user preference behavior database, delete the preference items with support below the threshold from the interaction data, and obtain a prefix with a length of 1;

[0034] Find the projection database corresponding to the prefix. If the interaction data is empty, return recursively.

[0035] Compute interaction support counts:

[0036]

[0037] Where Q is the total number of projected database transactions, θ e is the support count of the interaction term e;

[0038] If the support counts of all interaction items are lower than the transaction item count threshold, return recursively;

[0039] Merge the interaction items and prefixes that satisfy the support count to obtain a combined prefix;

[0040] Add 1 to the prefix length, and the combined prefix is ​​the prefix of the merged item. Recursively return to determine whether the support count of the interaction item is lower than the transaction item count threshold;

[0041] Repeat the recursive mining of prefixes whose lengths meet the support requirements until no new prefixes can be mined, and output the combined prefixes as user language habits.

[0042] Furthermore, the step of modifying the personal characteristics according to the standard database to obtain modified characteristics includes:

[0043] The personal characteristic data are input into the discrete evaluation model, giving the maximum likelihood estimate:

[0044]

[0045] In the formula, G is the maximum likelihood estimate, a is the number of personal characteristics, (·) T is the transpose of the matrix, z κ is the κth personal feature matrix;

[0046] The discrete assessment model is adjusted according to the maximum likelihood estimation, the personal characteristics are fitted, and the degree of deviation of the personal characteristics is calculated:

[0047]

[0048] In the formula, q is the number of personal characteristics s, g(s) is the value of personal characteristics, ρ s is the degree of deviation of personal characteristic s, c s is the cluster center value of personal feature s;

[0049] Calculate the correlation coefficient between personal characteristics and reference characteristics in the standard database:

[0050]

[0051] Where x0(s) is the reference feature in the standard database of the sth feature, μ i (s) is the correlation coefficient of the sth personal characteristic of the i-th preference behavior data, x i (s) is the sth personal feature of the ith interaction data, and min is x0(s) and x i (s), λ is the resolution coefficient, σ is the adjustment constant, and max is the value of x0(s) and x i (s) the maximum difference;

[0052] When the correlation coefficient is greater than a given value, the reference feature is used to replace the personal feature, and the replaced feature is output as the modified feature.

[0053] Furthermore, the calculating of the characteristic value by using the correction feature and the language habit and matching the preference behavior database comprises:

[0054] Calculate the property value:

[0055]

[0056] In the formula, A e is the feature set of the e-th preference behavior database data, B is the modified feature set, ω1 is the weight of the acoustic feature, C is the language habit set of the instruction text, ω2 is the weight of the language habit, |A e | is the modified feature set A e The value of |ω1B+ω2C| is the value of the preference behavior data, ρ is the interaction matching factor, ν is the characteristic change constant, and the e-th characteristic value is Y e ;

[0057] When the characteristic value is less than a given value, the interaction data does not match the preference behavior database data, and the preference behavior database data is replaced until the corrected characteristic value is greater than or equal to the given value, and the e-th preference behavior database data is associated with the preference behavior data;

[0058] When the personal characteristic value is greater than or equal to a given value, the e-th preference behavior database data is associated with the preference behavior data;

[0059] The preference behavior data is replaced to match the preference behavior database data until all instruction texts are traversed.

[0060] Further, when the characteristic value is less than 0.413, the preference behavior data does not match the preference behavior database data, and the preference behavior database data is replaced until the corrected characteristic value is greater than or equal to 0.413, and the e-th preference behavior database data is associated with the preference behavior data;

[0061] When the personal characteristic value is greater than or equal to 0.413, the e-th preference behavior database data is associated with the preference behavior data;

[0062] The preference behavior data is replaced to match the preference behavior database data until all instruction texts are traversed.

[0063] Furthermore, the establishment of the intelligent question-answering interaction model according to the preference behavior database includes a given clustering objective function:

[0064]

[0065] Where E is an n×p 0, 1 matrix, S is the cluster center data set, For the The center point of the class, For the data, Center point With data The distance between the categorical variables, n is the number of data, and p is the number of center points;

[0066] Calculate the attribute weight of the attribute value:

[0067]

[0068] In the formula, For the The local attributes of the data, For the The lth angle measurement of the data, For local attributes Angle measurement The single attribute weight of For the data object f l In local properties The equivalence class of For the Related attributes, For related attributes Angle measurement The multi-attribute weights of For attribute value The weight of For attribute value and attribute values The number of co-occurrences of is the number of relevant attributes, and the coefficient of deviation is δ;

[0069] Randomly find the cluster center initially, average the weight set and use the weight of the attribute value to calculate the distance between the data object and the cluster:

[0070]

[0071] In the formula, For the The y-th attribute value of the data, for The y-th attribute value of the data,

[0072] For attribute value and attribute values The distance between clusters, For attribute value The weight of For attribute value The weight of

[0073] According to cluster analysis, the risk degree and deviation degree are divided into different groups, the risk level of different groups is calculated, and the intelligent question-answering interactive model is output. The expression is:

[0074]

[0075] In the formula, For the The risk level of behavioral data For data The risk level weight, For the The deviation weight of each data, For the The risk level of the data For the The degree of deviation of the data.

[0076] In a second aspect, a consumer preference behavior information recommendation system based on big data, the system includes a preference behavior data acquisition unit, a personal feature extraction unit, a language habit acquisition unit, a correction feature acquisition unit, a preference behavior database matching unit, a consumer preference behavior information recommendation model construction unit, and a recommendation result output unit;

[0077] Wherein, the preference behavior data acquisition unit is used to acquire consumer preference behavior information recommendation data based on big data to obtain a preference behavior database;

[0078] The personal feature extraction unit is used to extract the user's personal features from the preference behavior data;

[0079] The language habit acquisition unit is used to acquire the user's language habits according to the preference behavior data;

[0080] The modified feature acquisition unit is used to modify the personal features according to the standard database to obtain the modified features;

[0081] The preference behavior database matching unit is used to calculate the characteristic value and match the preference behavior database using the correction feature and the language habit;

[0082] The consumer preference behavior information recommendation model building unit is used to build a consumer preference behavior information recommendation model according to the preference behavior database;

[0083] The recommendation result output unit is used to input the data to be answered into the consumer preference behavior information recommendation model and output the recommendation result.

[0084] As a further improvement of the present invention, it also includes a preprocessing unit for preprocessing the consumer preference behavior data acquired by the acquisition unit.

[0085] The present invention provides a method and system for recommending consumer preference behavior information based on big data, which optimizes preference behavior by extracting user personal characteristics from preference behavior data and obtaining user language habits based on the preference behavior data, and correcting the personal characteristics based on a standard database, thereby making user retrieval more targeted and accurate, saving resources and labor costs, and being able to provide more personalized retrieval services for different users, thereby improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 It is a flowchart of a method for recommending consumer preference behavior information based on big data according to the first embodiment of the present invention. DETAILED DESCRIPTION

[0087] The following is combined with specific embodiments and appendix Figure 1 The invention is described in detail so that those skilled in the art can more fully understand the purpose, features and effects of the invention.

[0088] Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as commonly understood by those skilled in the art to which the present invention belongs. When the definition of a term in the present invention conflicts with the meaning commonly understood by those skilled in the art to which the present invention belongs, the definition described in the present invention shall prevail.

[0089] The present invention improves the existing technology by providing a method and system for recommending consumer preference behavior information based on big data, combines artificial intelligence technology, realizes personalized response to user searches, significantly improves the intelligence level of preference behavior recommendations, and solves the shortcomings of the existing technology in personalized services and intelligent recommendations.

[0090] Embodiment 1

[0091] As a specific embodiment of the present invention, this embodiment provides a method for recommending consumer preference behavior information based on big data, referring to Figure 1 , the specific steps are as follows:

[0092] Obtaining consumer preference behavior data and obtaining a preference behavior database;

[0093] Extracting user personal characteristics from the preference behavior data, and obtaining user language habits based on the preference behavior data;

[0094] Modify the personal features according to the standard database to obtain modified features, calculate the characteristic value using the modified features and the language habits and match the preferred behavior database;

[0095] A consumer preference behavior information recommendation model is established according to the preference behavior database, the data to be answered is input into the consumer preference behavior information recommendation model, and a recommendation result is output.

[0096] Embodiment 2

[0097] As a specific embodiment of the present invention, this embodiment provides a method for recommending consumer preference behavior information based on big data, referring to Figure 1 , the specific steps are as follows:

[0098] As a specific embodiment of the present invention, this embodiment provides a method for recommending consumer preference behavior information based on big data, referring to Figure 1 , the specific steps are as follows:

[0099] In this embodiment, the obtaining of the consumer's preference behavior data to obtain the preference behavior database includes preprocessing the interaction data to obtain the interaction database, wherein the preprocessing includes removing duplicate data, vectorization, data conversion, data integration and data reduction;

[0100] The extracting of the user's personal features from the preference behavior data and acquiring the user's language habits according to the preference behavior data includes extracting the user's personal features from the preprocessed preference behavior data and acquiring the user's language habits according to the preference behavior data.

[0101] In this embodiment, further, extracting the user's personal characteristics from the preference behavior data includes:

[0102] Build a bridge between the original time series and multidimensional space analysis, reconstruct the phase space, and make the one-dimensional time series obtain the reconstructed time series in the multidimensional space:

[0103] W i =(w i ,w i+τ ,…,w (m+1)τ ) T

[0104] Where τ is the time delay, m is the embedding dimension, i is the preference behavior data, and W i is the time series of preference behavior data i, w i is the phase point of preference behavior data i;

[0105] Calculate the distance between phase points:

[0106] D=∑ i≠j H(r-‖w i -w j ‖)

[0107] Where H(·) is the Heaviside function, r is the given distance, and w jis the phase point of preference behavior data j;

[0108] Calculate the ratio of preference behavior data less than r to the total number of phase point pairs:

[0109]

[0110] Where N is the number of preference behavior data, Q N (r) is the logarithmic proportion of distances less than r;

[0111] Compute the distance between two points:

[0112]

[0113] Calculate the ratio of the number of phase point pairs that are associated to all pairs:

[0114]

[0115] In the formula, Q m (r) is the correlation dimension whose distance is less than r, and the information gain weight of the correlation dimension is calculated as:

[0116] Output the correlation dimension as user personal features.

[0117] In this embodiment, acquiring the user's language habits according to the preference behavior data includes:

[0118] Calculate the support of transaction items in the user preference behavior database:

[0119]

[0120] Where D is the number of interaction terms, R e is the number of transaction items containing the interaction term e, u e is the support of the interaction term e, v e is the weight of the interaction term e, ψ e Adjust the support parameter for the interaction term e;

[0121] Perform support statistics on the preference items in the user preference behavior database, delete the preference items with support below the threshold from the interaction data, and obtain a prefix with a length of 1;

[0122] Find the projection database corresponding to the prefix. If the interaction data is empty, return recursively.

[0123] Compute interaction support counts:

[0124]

[0125] Where Q is the total number of projected database transactions, g eis the support count of the interaction term e;

[0126] If the support counts of all interaction items are lower than the transaction item count threshold, return recursively;

[0127] Merge the interaction items and prefixes that satisfy the support count to obtain a combined prefix;

[0128] Add 1 to the prefix length, and the combined prefix is ​​the prefix of the merged item. Recursively return to determine whether the support count of the interaction item is lower than the transaction item count threshold;

[0129] Repeat the recursive mining of prefixes whose lengths meet the support requirements until no new prefixes can be mined, and output the combined prefixes as user language habits.

[0130] In this embodiment, the step of modifying the personal features according to the standard database to obtain modified features includes:

[0131] The personal characteristic data are input into the discrete evaluation model, giving the maximum likelihood estimate:

[0132]

[0133] In the formula, G is the maximum likelihood estimate, a is the number of personal characteristics, (·) T is the transpose of the matrix, z κ is the κth personal feature matrix;

[0134] The discrete assessment model is adjusted according to the maximum likelihood estimation, the personal characteristics are fitted, and the degree of deviation of the personal characteristics is calculated:

[0135]

[0136] In the formula, q is the number of personal characteristics s, g(s) is the value of personal characteristics, ρ s is the degree of deviation of personal characteristic s, c s is the cluster center value of personal feature s;

[0137] Calculate the correlation coefficient between personal characteristics and reference characteristics in the standard database:

[0138]

[0139] Where x0(s) is the reference feature in the standard database of the sth feature, μ i (s) is the correlation coefficient of the sth personal characteristic of the i-th preference behavior data, x i (s) is the sth personal feature of the ith interaction data, and min is x0(s) and x i (s), λ is the resolution coefficient, σ is the adjustment constant, and max is the value of x0(s) and xi (s) the maximum difference;

[0140] When the correlation coefficient is greater than a given value, the reference feature is used to replace the personal feature, and the replaced feature is output as the modified feature.

[0141] In this embodiment, the calculating of the characteristic value by using the correction feature and the language habit and matching the preference behavior database includes:

[0142] Calculate the property value:

[0143]

[0144] In the formula, A e is the feature set of the e-th preference behavior database data, B is the modified feature set, ω1 is the weight of the acoustic feature, C is the language habit set of the instruction text, ω2 is the weight of the language habit, |A e | is the modified feature set A e The value of |ω1B+ω2C| is the value of the preference behavior data, ρ is the interaction matching factor, ν is the characteristic change constant, and the e-th characteristic value is Y e ;

[0145] When the characteristic value is less than a given value, the interaction data does not match the preference behavior database data, and the preference behavior database data is replaced until the corrected characteristic value is greater than or equal to the given value, and the e-th preference behavior database data is associated with the preference behavior data;

[0146] When the personal characteristic value is greater than or equal to a given value, the e-th preference behavior database data is associated with the preference behavior data;

[0147] Replace the preference behavior data to match the preference behavior database data until all instruction texts are traversed. In this embodiment, when the characteristic value is less than 0.413, the preference behavior data does not match the preference behavior database data, and the preference behavior database data is replaced until the corrected characteristic value is greater than or equal to 0.413, and the e-th preference behavior database data is associated with the preference behavior data;

[0148] When the personal characteristic value is greater than or equal to 0.413, the e-th preference behavior database data is associated with the preference behavior data;

[0149] The preference behavior data is replaced to match the preference behavior database data until all instruction texts are traversed.

[0150] In this embodiment, the establishment of an intelligent question-answering interaction model based on the preference behavior database includes a given clustering objective function:

[0151]

[0152] Where E is an n×p 0, 1 matrix, S is the cluster center data set, For the The center point of the class, For the data, Center point With data The distance between the categorical variables, n is the number of data, and p is the number of center points;

[0153] Calculate the attribute weight of the attribute value:

[0154]

[0155] In the formula, For the The local attributes of the data, For the The lth angle measurement of the data, For local attributes Angle measurement The single attribute weight of For the data object f l In local properties The equivalence class of For the Related attributes, For related attributes Angle measurement The multi-attribute weights of For attribute value The weight of For attribute value and attribute values The number of co-occurrences of is the number of relevant attributes, and the coefficient of deviation is δ;

[0156] Randomly find the cluster center initially, average the weight set and use the weight of the attribute value to calculate the distance between the data object and the cluster:

[0157]

[0158] In the formula, For the The y-th attribute value of the data, for The y-th attribute value of the data,

[0159] For attribute value and attribute values The distance between clusters, For attribute value The weight of For attribute value The weight of

[0160] According to cluster analysis, the risk degree and deviation degree are divided into different groups, the risk level of different groups is calculated, and the intelligent question-answering interactive model is output. The expression is:

[0161]

[0162] In the formula, For the The risk level of behavioral data For data The risk level weight, For the The deviation weight of each data, For the The risk level of the data For the The degree of deviation of the data.

[0163] The consumer preference behavior information recommendation method based on big data of this embodiment overcomes the problems of insufficient accuracy and lack of personalization in the prior art in recommending consumer preference behavior information, and enables users to obtain a more satisfactory experience through retrieval of user preferences.

[0164] Embodiment 3

[0165] As a specific embodiment of the present invention, this embodiment provides a consumer preference behavior information recommendation system based on big data, including a preference behavior data acquisition unit, a personal feature extraction unit, a language habit acquisition unit, a correction feature acquisition unit, a preference behavior database matching unit, a consumer preference behavior information recommendation model construction unit, and a recommendation result output unit; wherein the preference behavior data acquisition unit is used to acquire consumer preference behavior information recommendation data based on big data to obtain a preference behavior database;

[0166] The personal feature extraction unit is used to extract the user's personal features from the preference behavior data;

[0167] The language habit acquisition unit is used to acquire the user's language habits according to the preference behavior data;

[0168] The modified feature acquisition unit is used to modify the personal features according to the standard database to obtain the modified features;

[0169] The preference behavior database matching unit is used to calculate the characteristic value and match the preference behavior database using the correction feature and the language habit;

[0170] The consumer preference behavior information recommendation model building unit is used to build a consumer preference behavior information recommendation model according to the preference behavior database;

[0171] The recommendation result output unit is used to input the data to be answered into the consumer preference behavior information recommendation model and output the recommendation result. Furthermore, it also includes a preprocessing unit, which is used to preprocess the interactive data of the intelligent question and answer of oilfield environmental protection and safety standards acquired by the acquisition unit, including deduplication, vectorization, data conversion, data integration and data reduction.

[0172] The method and system for recommending consumer preference behavior information based on big data of the present invention can also be modified and applied to other scenarios or fields without creative work.

[0173] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A method for recommending consumer preference behavior information based on big data, characterized in that: The method comprises: Obtaining consumer preference behavior data and obtaining a preference behavior database; Extracting user personal characteristics from the preference behavior data, and obtaining user language habits based on the preference behavior data; Modify the personal features according to the standard database to obtain modified features, calculate the characteristic value using the modified features and the language habits and match the preferred behavior database; A consumer preference behavior information recommendation model is established according to the preference behavior database, the data to be answered is input into the consumer preference behavior information recommendation model, and a recommendation result is output.

2. The method for recommending consumer preference behavior information based on big data according to claim 1, characterized in that: The acquiring of the consumer's preference behavior data to obtain the preference behavior database includes preprocessing the interaction data to obtain the interaction database, wherein the preprocessing includes removing duplicate data, vectorization, data conversion, data integration and data reduction; The extracting of the user's personal features from the preference behavior data and acquiring the user's language habits according to the preference behavior data includes extracting the user's personal features from the preprocessed preference behavior data and acquiring the user's language habits according to the preference behavior data.

3. The method for recommending consumer preference behavior information based on big data according to claim 1, characterized in that: The extracting of user personal characteristics from the preference behavior data includes: Build a bridge between the original time series and multidimensional space analysis, reconstruct the phase space, and make the one-dimensional time series obtain the reconstructed time series in the multidimensional space: IN i =(in i ,In i+τ ,…,In (m+1)τ ) T Where τ is the time delay, m is the embedding dimension, i is the preference behavior data, and W i is the time series of preference behavior data i, w i is the phase point of preference behavior data i; Calculate the distance between phase points: D=∑ i≠j H(r-‖w i -w j ‖) Where H(·) is the Heaviside function, r is the given distance, and w j is the phase point of preference behavior data j; Calculate the ratio of preference behavior data less than r to the total number of phase point pairs: Where N is the number of preference behavior data, Q N (r) is the logarithmic proportion of distances less than r; Compute the distance between two points: Calculate the ratio of the number of phase point pairs that are associated to all pairs: In the formula, Q m (r) is the correlation dimension whose distance is less than r, and the information gain weight of the correlation dimension is calculated as: Output the correlation dimension as user personal features.

4. The method for recommending consumer preference behavior information based on big data according to claim 1, characterized in that: The acquiring of the user's language habits according to the preference behavior data comprises: Calculate the support of transaction items in the user preference behavior database: Where D is the number of interaction terms, R e is the number of transaction items containing the interaction term e, u e is the support of the interaction term e, v e is the weight of the interaction term e, ψ e Adjust the support parameter for the interaction term e; Perform support statistics on the preference items in the user preference behavior database, delete the preference items with support below the threshold from the interaction data, and obtain a prefix with a length of 1; Find the projection database corresponding to the prefix. If the interaction data is empty, return recursively. Compute interaction support counts: Where Q is the total number of projected database transactions, θ e is the support count of the interaction term e; If the support counts of all interaction items are lower than the transaction item count threshold, return recursively; Merge the interaction items and prefixes that satisfy the support count to obtain a combined prefix; Add 1 to the prefix length, and the combined prefix is ​​the prefix of the merged item. Recursively return to determine whether the support count of the interaction item is lower than the transaction item count threshold; Repeat the recursive mining of prefixes whose lengths meet the support requirements until no new prefixes can be mined, and output the combined prefixes as user language habits.

5. The method for recommending consumer preference behavior information based on big data according to claim 1, characterized in that: The step of modifying the personal characteristics according to the standard database to obtain modified characteristics comprises: The personal characteristic data are input into the discrete evaluation model, giving the maximum likelihood estimate: In the formula, G is the maximum likelihood estimate, a is the number of personal characteristics, (·) T is the transpose of the matrix, z κ is the κth personal feature matrix; The discrete assessment model is adjusted according to the maximum likelihood estimation, the personal characteristics are fitted, and the degree of deviation of the personal characteristics is calculated: In the formula, q is the number of personal characteristics s, g(s) is the value of personal characteristics, ρ s is the degree of deviation of personal characteristic s, c s is the cluster center value of personal feature s; Calculate the correlation coefficient between personal characteristics and reference characteristics in the standard database: Where x0(s) is the reference feature in the standard database of the sth feature, μ i (s) is the correlation coefficient of the sth personal characteristic of the i-th preference behavior data, x i (s) is the sth personal feature of the ith interaction data, and min is x0(s) and x i (s), λ is the resolution coefficient, σ is the adjustment constant, and max is the value of x0(s) and x i (s) the maximum difference; When the correlation coefficient is greater than a given value, the reference feature is used to replace the personal feature, and the replaced feature is output as the modified feature.

6. The method for recommending consumer preference behavior information based on big data according to claim 5, characterized in that: The calculating of the characteristic value by using the correction feature and the language habit and matching the preference behavior database comprises: Calculate the property value: In the formula, A e is the feature set of the e-th preference behavior database data, B is the modified feature set, ω1 is the weight of the acoustic feature, C is the language habit set of the instruction text, ω2 is the weight of the language habit, |A e | is the modified feature set A e The value of |ω1B+ω2C| is the value of the preference behavior data, ρ is the interaction matching factor, ν is the characteristic change constant, and the e-th characteristic value is Y e ; When the characteristic value is less than a given value, the interaction data does not match the preference behavior database data, and the preference behavior database data is replaced until the corrected characteristic value is greater than or equal to the given value, and the e-th preference behavior database data is associated with the preference behavior data; When the personal characteristic value is greater than or equal to a given value, the e-th preference behavior database data is associated with the preference behavior data; The preference behavior data is replaced to match the preference behavior database data until all instruction texts are traversed.

7. The method for recommending consumer preference behavior information based on big data according to claim 6, characterized in that: When the characteristic value is less than 0.413, the preference behavior data does not match the preference behavior database data, and the preference behavior database data is replaced until the corrected characteristic value is greater than or equal to 0.413, and the e-th preference behavior database data is associated with the preference behavior data; When the personal characteristic value is greater than or equal to 0.413, the e-th preference behavior database data is associated with the preference behavior data; The preference behavior data is replaced to match the preference behavior database data until all instruction texts are traversed.

8. The method for recommending consumer preference behavior information based on big data according to claim 1, characterized in that: The establishment of an intelligent question-answering interaction model based on the preference behavior database includes a given clustering objective function: Where E is an n×p 0, 1 matrix, S is the cluster center data set, For the The center point of the class, For the data, Center point With data The distance between the categorical variables, n is the number of data, and p is the number of center points; Calculate the attribute weight of the attribute value: In the formula, For the The local attributes of the data, For the The lth angle measurement of the data, For local attributes Angle measurement The single attribute weight of For the data object f l In local properties The equivalence class of For the related attributes, For related attributes Angle measurement The multi-attribute weights of For attribute value The weight of For attribute value and attribute values The number of co-occurrences of is the number of relevant attributes, and the coefficient of deviation is δ; Randomly find the cluster center initially, average the weight set and use the weight of the attribute value to calculate the distance between the data object and the cluster: In the formula, For the The y-th attribute value of the data, for The y-th attribute value of the data, For attribute value and attribute values The distance between clusters, For attribute value The weight of For attribute value The weight of According to cluster analysis, the risk degree and deviation degree are divided into different groups, the risk level of different groups is calculated, and the intelligent question-answering interactive model is output. The expression is: In the formula, For the The risk level of behavioral data For data The risk level weight, For the The deviation weight of each data, For the The risk level of the data For the The degree of deviation of the data.

9. A consumer preference behavior information recommendation system based on big data, characterized in that: The system includes a preference behavior data acquisition unit, a personal feature extraction unit, a language habit acquisition unit, a correction feature acquisition unit, a preference behavior database matching unit, a consumer preference behavior information recommendation model construction unit, and a recommendation result output unit; Wherein, the preference behavior data acquisition unit is used to acquire consumer preference behavior information recommendation data based on big data to obtain a preference behavior database; The personal feature extraction unit is used to extract the user's personal features from the preference behavior data; The language habit acquisition unit is used to acquire the user's language habits according to the preference behavior data; The modified feature acquisition unit is used to modify the personal features according to the standard database to obtain the modified features; The preference behavior database matching unit is used to calculate the characteristic value and match the preference behavior database using the correction feature and the language habit; The consumer preference behavior information recommendation model building unit is used to build a consumer preference behavior information recommendation model according to the preference behavior database; The recommendation result output unit is used to input the data to be answered into the consumer preference behavior information recommendation model and output the recommendation result.

10. The consumer preference behavior information recommendation system based on big data according to claim 9, characterized in that: It also includes a preprocessing unit for preprocessing the consumer preference behavior data acquired by the acquisition unit.

Images