Online electronic commerce recommendation method based on big data

Abstract

The invention relates to the technical field of e-commerce recommendation, in particular to an online e-commerce recommendation method based on big data. The method comprises the following steps: judging according to user collaborative filtering recommended commodities generated by a user model and generated article collaborative filtering recommended commodities; according to the invention, the generation of user collaborative filtering recommended commodities is carried out according to the medium preference and high preference commodities in the user model, and then the generation of article collaborative filtering recommended commodities is carried out according to the medium preference and high preference commodities in the user model. The recommendation results of the generated user collaborative filtering recommended commodities and the generated article collaborative filtering recommended commodities are judged to be the same, and the recommendation results of the user and the article are compared, so that the accuracy of the recommendation system can be improved, and if the same commodities are recommended by the two methods, the recommendation efficiency is improved. Therefore, the credibility of the recommendation result can be enhanced, the user has a sense of identity to the recommendation result more possibly, and the recommendation accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of e-commerce recommendation, and in particular to an online e-commerce recommendation method based on big data. Background Art

[0002] In e-commerce, product recommendation is an important tool to improve customer experience and increase sales. Recommendation based on big data can provide more personalized product recommendations through users' past transaction records, browsing history, product reviews and social media information. Through this series of steps, it can better understand user needs and provide more personalized and accurate product recommendations, thereby improving user satisfaction and purchase conversion rate. When obtaining user behavior data, the recommendation system only relies on a single data source or a single recommendation strategy, and cannot fully understand the user's complex needs, which makes the recommendation results inaccurate. The risk, especially in user collaborative filtering, new users or a few products may have only a few ratings or interaction data, which will make it difficult to accurately predict user preferences, reduce the credibility of recommendation results, and thus greatly reduce the accuracy of recommendation results. In order to avoid relying on a single data source or a single recommendation strategy to reduce the accuracy of recommendations, we provide an online e-commerce recommendation method based on big data. Summary of the invention

[0003] The purpose of the present invention is to provide an online e-commerce recommendation method based on big data to solve the problems raised in the above background technology.

[0004] To achieve the above object, the present invention provides an online e-commerce recommendation method based on big data, comprising the following method steps:

[0005] S1. Obtain relevant online e-commerce data and extract user behavior data and product information, and then analyze preferred products based on the extracted user behavior data and extracted product information;

[0006] S2. Analyze the prediction of user preference trends based on user behavior data, and then establish a user model based on the predicted user preference trend data. The user model is used to collect preferred products for analysis of medium and high preferences;

[0007] S3. Generate user collaborative filtering recommended products and item collaborative filtering recommended products based on the medium preference and high preference products in the user model, and determine whether the recommendation results are the same through the generated user collaborative filtering recommended products and the generated item collaborative filtering recommended products.

[0008] As a further improvement of the technical solution, the S1 specifically comprises the following method steps:

[0009] S1, obtaining relevant online e-commerce data, and extracting user behavior data and product information from the obtained online relevant e-commerce data, and then analyzing preferred products based on the extracted user behavior data and the extracted product information, obtaining the user's stay time on the analyzed preferred product page based on the user's entry timestamp and exit timestamp on the analyzed preferred product page, and converting the stay time into a preference level by setting up a conversion rule;

[0010] User behavior data includes user browsing history, purchase history, search history, and clickstream data;

[0011] Product information includes product classification, price, brand, and specification information; user feedback data includes user ratings, review content, and return information.

[0012] As a further improvement of the technical solution, the specific conversion conditions in S1 include:

[0013] Case ①: When the dwell time is less than 1 minute, the analyzed preferred product is considered as low preference;

[0014] Case 2: When the dwell time is between 1 and 3 minutes, the analyzed preferred products are considered as medium preference;

[0015] Case ③: When the dwell time exceeds 3 minutes, the analyzed preferred products will be regarded as high preference.

[0016] As a further improvement of the technical solution, the S2 specifically comprises the following method steps:

[0017] S2. Predict user preference trends based on user behavior data through big data analysis technology, and then establish a user model based on the predicted user preference trend data. The user model is used to collect preference products for analysis of medium and high preferences. At the same time, the user model automatically scores the preference products collected for analysis of medium and high preferences. When the user model collects preference products for analysis of medium and high preferences, the user model classifies the product types of the collected preference products.

[0018] As a further improvement of the technical solution, the S3 specifically comprises the following method steps:

[0019] S3. Generate user collaborative filtering recommended products based on medium preference and high preference products in the user model through user collaborative filtering algorithm, and then generate item collaborative filtering recommended products based on medium preference and high preference products in the user model through item collaborative filtering algorithm, and judge whether the recommendation results generated by user collaborative filtering recommended products and item collaborative filtering recommended products are the same.

[0020] As a further improvement of the technical solution, the specific judgment situation in S3 includes:

[0021] Case 1: When the generated user collaborative filtering recommended products are the same as the generated item collaborative filtering recommended products, the generated user collaborative filtering recommended products or the generated item collaborative filtering recommended products are recommended;

[0022] Case 2: When the generated user collaborative filtering recommended products are different from the generated item collaborative filtering recommended products, the generated user collaborative filtering recommended products and the generated item collaborative filtering recommended products are calculated by accuracy, and then the calculated user collaborative filtering index is compared with the calculated item collaborative filtering index. The specific comparison includes:

[0023] Case 1: When the calculated user collaborative filtering index is less than the calculated item collaborative filtering index, it indicates that data sparsity exists in the user collaborative filtering algorithm, and the user collaborative filtering algorithm is optimized. Then, the optimized user collaborative filtering algorithm is used to optimize the generation of user collaborative filtering recommended products based on the medium-preference and high-preference products in the user model, and the generated optimized user collaborative filtering recommended products are judged to have the same recommendation results as the generated item collaborative filtering recommended products.

[0024] Case 2: When the calculated user collaborative filtering index is greater than the calculated item collaborative filtering index, it indicates that data sparsity exists in the item collaborative filtering algorithm, and the user collaborative filtering algorithm is optimized. The optimized item collaborative filtering algorithm is then used to optimize the generation of item collaborative filtering recommended products based on the medium preference and high preference products in the user model. The generated user collaborative filtering recommended products and the generated optimized item collaborative filtering recommended products are judged to have the same recommendation results.

[0025] As a further improvement of the technical solution, the implementation steps of generating product recommendations through collaborative filtering of items in S3 are as follows:

[0026] Step 1: First, create a user and preferred product interaction matrix. The rows of the matrix represent users, the columns represent the preferred products to be analyzed, and the cell value is the rating status. 1 means rated and 0 means unrated. Then, get the user set from the user model. , specifically refers to the preferred products for analysis and analyze preferred products The collection of all users who have provided ratings, user collection , Refers to the first user in the user set. Refers to the first users, and sum the user set to obtain the user sum ;

[0027] Step 2: Obtain the analyzed preferred products from the user model and analyze preferred products Corresponding rating and rating , and measure the preference of two analyzed products by cosine similarity and analyze preferred products The angle between the rating vectors in the vector space and the cosine similarity , the specific algorithm formula is:

[0028] ;

[0029] Step 3: Get all users’ preferred products for analysis from the user model and analyze preferred products 's ratings and obtain the preferred products analyzed Rating collection and analyze preferred products Rating collection , and then the scoring set and score collection conduct Calculation of average score and Calculate the average score and obtain the calculated Average score and calculated Average score ;

[0030] Step 4: Measure the preferred products analyzed by Pearson correlation coefficient and analyze preferred products Linear correlation between rating vectors, summed over users ,score ,score , calculated Average score and calculated Average score Perform calculations and obtain linear results , the specific algorithm formula is:

[0031] ;

[0032] Step 5: Obtain cosine similarity To evaluate and analyze the preferred products and analyze preferred products The similarity between them is obtained by obtaining the linear result of the calculation Weighing and analyzing preferred products and analyze preferred products The linear relationship between the rating vectors. For each analyzed preference item, find the L most similar other analyzed preference items. This step usually involves sorting the similarity of each analyzed preference item, and then selecting the highest L values. The analyzed preference items corresponding to the highest L values ​​are the generated collaborative filtering recommended items.

[0033] As a further improvement of the technical solution, the implementation steps of the user collaborative filtering recommended product index generated by accuracy calculation in the second case of S3 are as follows:

[0034] Step 1: The recommended user collaborative filtering recommended products that are actually adopted by the user are regarded as true positives TP, the recommended user collaborative filtering recommended products that are not actually adopted by the user are regarded as false positives FP, the unrecommended user collaborative filtering recommended products that are not actually adopted by the user are regarded as true negatives TN, and the unrecommended user collaborative filtering recommended products that should have been adopted by the user are regarded as false negatives FN;

[0035] Step 2: Calculate the generated user collaborative filtering recommended product index based on the true positive class TP, false positive class FP, true negative class TN, and false negative class FN, and obtain the calculated user collaborative filtering index , the specific algorithm formula is:

[0036] .

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] This online e-commerce recommendation method based on big data generates user collaborative filtering recommended products according to medium-preference and high-preference products in the user model, and then generates item collaborative filtering recommended products according to medium-preference and high-preference products in the user model, and judges whether the recommendation results generated by user collaborative filtering recommended products and generated item collaborative filtering recommended products are the same. By comparing the recommendation results in two dimensions, namely, user and item, the accuracy of the recommendation system can be improved. When both methods recommend the same product, the credibility of the recommendation result can be enhanced, and the user is more likely to identify with the recommendation result, thereby improving the accuracy of the recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flowchart of the overall steps of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] Example 1

[0042] The present invention provides an online e-commerce recommendation method based on big data. Figure 1 , comprising the following method steps:

[0043] S1. Obtain relevant online e-commerce data and extract user behavior data and product information, and then analyze preferred products based on the extracted user behavior data and extracted product information;

[0044] The S1 specifically includes the following steps:

[0045] S1. Obtain relevant online e-commerce data, extract user behavior data and product information from the obtained online relevant e-commerce data, and then analyze the preferred products based on the extracted user behavior data and extracted product information. According to the entry timestamp and exit timestamp of the user on the analyzed preferred product page, obtain the user's stay time on the analyzed preferred product page, and convert the stay time into the preference level by setting up a conversion rule. By analyzing the user behavior data, the user's preferences, interests and purchasing habits can be understood, so as to provide personalized product recommendations for the user, which can improve the user experience, increase the user's stickiness to the platform, and promote the growth of transaction volume;

[0046] The specific conversion conditions in S1 include:

[0047] Case ①: When the dwell time is less than 1 minute, the analyzed preferred product is considered as low preference;

[0048] Case 2: When the dwell time is between 1 and 3 minutes, the analyzed preferred products are considered as medium preference;

[0049] Case ③: When the dwell time exceeds 3 minutes, the analyzed preferred products are considered as high preference;

[0050] The online e-commerce data obtained include user behavior data, product information and user feedback data;

[0051] User behavior data includes user browsing history, purchase history, search history, and clickstream data;

[0052] Product information includes product classification, price, brand, and specification information; user feedback data includes user ratings, review content, and return information.

[0053] S2. Analyze the prediction of user preference trends based on user behavior data, and then establish a user model based on the predicted user preference trend data. The user model is used to collect preferred products for analysis of medium and high preferences;

[0054] The S2 specifically includes the following steps:

[0055] S2. Use big data analysis technology to predict user preference trends based on user behavior data. By predicting user preference trends, more personalized services and recommendations can be provided to users, thereby improving user satisfaction and loyalty. Then, a user model is established based on the predicted user preference trend data. Based on the established user model, a personalized recommendation function can be implemented to recommend products that best meet the user's needs based on the user's preference trend, thereby improving user experience and user loyalty. The user model is used to collect the preferred products of medium and high preference analyses. At the same time, the user model automatically scores the preferred products of medium and high preference analyses. When the user model collects the preferred products of medium and high preference analyses, the user model classifies the collected analyzed preferred products into product types.

[0056] Steps to analyze user preference trends through big data technology:

[0057] Step 1: First, collect user behavior data, and pre-process the user behavior data to remove duplicates, correct errors, and handle missing values. Then, standardize the pre-processed user behavior data to meet the needs of the analysis model.

[0058] Step 2: Extract features based on the standardized data. The extracted features include the purchase frequency of the analyzed preferred products, the most frequently visited pages, and the characteristics of seasonal trends. The pre-processed user behavior data is then stored in the central database.

[0059] Step 3. Use statistical methods and data visualization tools to find patterns and trends in the preprocessed user behavior data, select a machine learning model (which is an existing technology) to train the preprocessed user behavior data, and train the machine learning model on the training set. Then use the validation set to adjust the parameters of the machine learning model to prevent overfitting. Use the machine learning model to predict the preprocessed user behavior data. The output of the machine learning model is the predicted user preference trend data.

[0060] S3, generating user collaborative filtering recommended products and item collaborative filtering recommended products based on the medium preference and high preference products in the user model, and determining whether the recommendation results are the same through the generated user collaborative filtering recommended products and the generated item collaborative filtering recommended products;

[0061] The S3 specifically includes the following steps:

[0062] S3, generating user collaborative filtering recommended products based on the medium preference and high preference products in the user model through the user collaborative filtering algorithm, and then generating item collaborative filtering recommended products based on the medium preference and high preference products in the user model through the item collaborative filtering algorithm, and determining whether the recommendation results of the generated user collaborative filtering recommended products and the generated item collaborative filtering recommended products are the same;

[0063] The implementation steps for generating product recommendations through user collaborative filtering are as follows:

[0064] Step ①: First, create a user and preferred product interaction matrix. The rows of the matrix represent users, the columns represent the preferred products to be analyzed, and the cell value is the rating status. 1 means rated and 0 means unrated. Then, get the user from the user model. Preference for analysis Ratings and users Preference for analysis , get the user's rating Rating Collection and with users Rating Collection , Refers to the user Preferred products for the first analysis Ratings, Refers to the user Preferred products for the Nth analysis Ratings, Refers to the user Preferred products for the first analysis Ratings, Refers to the user Preferred products for the Mth analysis Ratings;

[0065] Step ②According to the user Rating Collections and Users The scoring set is used to score the index set, and the preferred products for the same analysis are analyzed together. 's ratings and obtain the rating index collection , Refers to the user With users Common preference for the first analysis Ratings, Refers to the user With users Common preference items for the second analysis Ratings, Refers to the user With users Common preference items for the nth analysis Ratings, based on user rating index collection Sum the scores in to get the score sum ;

[0066] Step 3: Index the collection based on user ratings Ratings in user With users Calculate the average score and obtain the user Average score and users Average score , and then sum the scores ,user Average score ,user Average score Calculate user similarity and obtain the calculated user similarity , the specific algorithm formula is:

[0067] ;

[0068] in, Refers to the preferred product of user U in the i-th analysis Ratings, Refers to the preferred product of user V in the i-th analysis Ratings;

[0069] Step ④: After calculating the similarity between all users, find the most similar K users for each user. The set of these users is called "neighbors". Use the priority queue to select the K with the highest similarity, use the determined nearest neighbors to generate recommendations for the target user, and obtain the generated user collaborative filtering recommended products.

[0070] The implementation steps of generating product recommendations through collaborative filtering of items in S3 are as follows:

[0071] Step 1: First, create a user and preferred product interaction matrix. The rows of the matrix represent users, the columns represent the preferred products to be analyzed, and the cell value is the rating status. 1 means rated and 0 means unrated. Then, get the user set from the user model. , specifically refers to the preferred products for analysis and analyze preferred products The collection of all users who have provided ratings, user collection , Refers to the first user in the user set. Refers to the first users, and sum the user set to obtain the user sum ;

[0072] Step 2: Obtain the analyzed preferred products from the user model and analyze preferred products Corresponding rating and rating , and measure the preference of two analyzed products by cosine similarity and analyze preferred products The angle between the rating vectors in the vector space and the cosine similarity , the specific algorithm formula is:

[0073] ;

[0074] The formula is used to calculate cosine similarity. Cosine similarity can be used to build recommendations. According to the user's historical preference data, the similarity between the user and other products is calculated, so as to recommend products with similar preference trends to the user, thereby improving the accuracy of recommendations and user satisfaction.

[0075] Step 3: Get all users’ preferred products for analysis from the user model and analyze preferred products 's ratings and obtain the preferred products analyzed Rating collection and analyze preferred products Rating collection , and then the scoring set and score collection conduct Calculation of average score and Calculate the average score and obtain the calculated Average score and calculated Average score ;

[0076] Step 4: Measure the preferred products analyzed by Pearson correlation coefficient and analyze preferred products Linear correlation between rating vectors, summed over users ,score ,score , calculated Average score and calculated Average score Perform calculations and obtain linear results , the specific algorithm formula is:

[0077] ;

[0078] Among them, this formula is used to calculate linear results. Compared with other correlation coefficients, the Pearson correlation coefficient is more robust to outliers, that is, when evaluating the relationship between the rating vectors of preferred products, it is not easily affected by extreme values ​​and can better reflect the overall trend of the data;

[0079] Step 5: Obtain cosine similarity To evaluate and analyze the preferred products and analyze preferred products The similarity between them is obtained by obtaining the linear result of the calculation Weighing and analyzing preferred products and analyze preferred products The linear relationship between the rating vectors. For each analyzed preference item, find the L most similar other analyzed preference items. This step usually involves sorting the similarity of each analyzed preference item, and then selecting the highest L values. The analyzed preference items corresponding to the highest L values ​​are the generated collaborative filtering recommended items.

[0080] Specific judgment situations include:

[0081] Case 1: When the generated user collaborative filtering recommended products are the same as the generated item collaborative filtering recommended products, the generated user collaborative filtering recommended products or the generated item collaborative filtering recommended products are recommended;

[0082] Case 2: When the generated user collaborative filtering recommended products are different from the generated item collaborative filtering recommended products, the generated user collaborative filtering recommended products and the generated item collaborative filtering recommended products are calculated by accuracy, and then compared with the calculated user collaborative filtering index and the calculated item collaborative filtering index (the calculation process is the same as the calculation process of the calculated user collaborative filtering index, and the generated user collaborative filtering recommended products are replaced with the generated item collaborative filtering recommended products). The specific comparison situations include:

[0083] The implementation steps of the user collaborative filtering recommended product index generated by accuracy calculation in the second case of S3 are as follows:

[0084] Step 1: The recommended user collaborative filtering recommended products that are actually adopted by the user are regarded as true positives TP, the recommended user collaborative filtering recommended products that are not actually adopted by the user are regarded as false positives FP, the unrecommended user collaborative filtering recommended products that are not actually adopted by the user are regarded as true negatives TN, and the unrecommended user collaborative filtering recommended products that should have been adopted by the user are regarded as false negatives FN;

[0085] Step 2: Calculate the generated user collaborative filtering recommended product index based on the true positive class TP, false positive class FP, true negative class TN, and false negative class FN, and obtain the calculated user collaborative filtering index , the specific algorithm formula is:

[0086] ;

[0087] Among them, this formula is used to calculate the generated user collaborative filtering recommendation product index. By calculating the user collaborative filtering recommendation product index, more personalized recommendations can be achieved, thereby improving user satisfaction. By analyzing the user's historical behavior data, the system can identify the user's preferences and recommend products that match their interests;

[0088] Case 1: When the calculated user collaborative filtering index is less than the calculated item collaborative filtering index, it indicates that data sparsity exists in the user collaborative filtering algorithm, and the user collaborative filtering algorithm is optimized. Then, the optimized user collaborative filtering algorithm is used to optimize the generation of user collaborative filtering recommended products based on the medium-preference and high-preference products in the user model, and the generated optimized user collaborative filtering recommended products are judged to have the same recommendation results as the generated item collaborative filtering recommended products.

[0089] Case 2: When the calculated user collaborative filtering index is greater than the calculated item collaborative filtering index, it indicates that data sparsity exists in the item collaborative filtering algorithm, and the user collaborative filtering algorithm is optimized. The optimized item collaborative filtering algorithm is then used to optimize the generation of item collaborative filtering recommended products based on the medium preference and high preference products in the user model. The generated user collaborative filtering recommended products and the generated optimized item collaborative filtering recommended products are judged to have the same recommendation results.

[0090] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An online e-commerce recommendation method based on big data, characterized by: The method comprises the following steps: S1. Obtain online e-commerce data, extract user behavior data and product information, and then analyze preferred products based on the extracted user behavior data and product information; S2. Analyze the prediction of user preference trends based on user behavior data, and then establish a user model based on the predicted user preference trend data. The user model is used to collect preferred products for analysis of medium and high preferences; S3. Generate user collaborative filtering recommended products and item collaborative filtering recommended products based on the medium preference and high preference products in the user model, and determine whether the recommended results are the same by using the generated user collaborative filtering recommended products and the generated item collaborative filtering recommended products.

2. The online e-commerce recommendation method based on big data according to claim 1, characterized in that: The S1 specifically includes the following steps: S1, obtaining online e-commerce data, and extracting user behavior data and product information from the obtained online related e-commerce data, and then analyzing preferred products based on the extracted user behavior data and the extracted product information, obtaining the user's stay time on the analyzed preferred product page based on the user's entry timestamp and exit timestamp on the analyzed preferred product page, and converting the stay time into a preference level by setting up a conversion rule; User behavior data includes user browsing history, purchase history, search history, and clickstream data; Product information includes product classification, price, brand, and specification information; user feedback data includes user ratings, review content, and return information.

3. The online e-commerce recommendation method based on big data according to claim 2, characterized in that: The specific conversion conditions in S1 include: Case ①: When the dwell time is less than 1 minute, the analyzed preferred product is considered as low preference; Case 2: When the dwell time is between 1 and 3 minutes, the analyzed preferred products are considered as medium preference; Case ③: When the dwell time exceeds 3 minutes, the analyzed preferred products will be regarded as high preference.

4. The online e-commerce recommendation method based on big data according to claim 2, characterized in that: The S2 specifically includes the following steps: Predict user preference trends based on user behavior data through big data analysis technology; Then, a user model is established based on the predicted user preference trend data. The user model is used to collect preferred products for analysis of medium and high preferences. At the same time, the user model automatically scores the preferred items collected from the medium and high preference analyses; When the user model collects analyzed preferred commodities with medium preference and high preference, the user model classifies the collected analyzed preferred commodities into commodity types.

5. The online e-commerce recommendation method based on big data according to claim 4, characterized in that: The S3 specifically includes the following steps: Generate user collaborative filtering recommended products based on medium-preference and high-preference products in the user model through user collaborative filtering algorithm; Then, the collaborative filtering algorithm is used to generate recommended products based on the medium-preference and high-preference products in the user model. And the generated user collaborative filtering recommended products and the generated item collaborative filtering recommended products are used to determine if the recommendation results are the same.

6. The online e-commerce recommendation method based on big data according to claim 5, characterized in that: The specific judgment conditions in S3 include: Case 1: When the generated user collaborative filtering recommended products are the same as the generated item collaborative filtering recommended products, the generated user collaborative filtering recommended products or the generated item collaborative filtering recommended products are recommended; Case 2: When the generated user collaborative filtering recommended products are different from the generated item collaborative filtering recommended products, the generated user collaborative filtering recommended products and the generated item collaborative filtering recommended products are calculated by accuracy, and then the calculated user collaborative filtering index is compared with the calculated item collaborative filtering index. The specific comparison includes: Case 1: When the calculated user collaborative filtering index is less than the calculated item collaborative filtering index, it indicates that data sparsity exists in the user collaborative filtering algorithm. The user collaborative filtering algorithm is optimized, and then the optimized user collaborative filtering algorithm is used to optimize the generation of user collaborative filtering recommended products based on the medium-preference and high-preference products in the user model. The generated optimized user collaborative filtering recommended products and the generated item collaborative filtering recommended products are judged to have the same recommendation results. Case 2: When the calculated user collaborative filtering index is greater than the calculated item collaborative filtering index, it indicates that data sparsity exists in the item collaborative filtering algorithm, and the user collaborative filtering algorithm is optimized. The optimized item collaborative filtering algorithm is then used to optimize the generation of item collaborative filtering recommended products based on the medium preference and high preference products in the user model. The generated user collaborative filtering recommended products and the generated optimized item collaborative filtering recommended products are judged to have the same recommendation results.

7. The online e-commerce recommendation method based on big data according to claim 5, characterized in that: The implementation steps of generating product recommendations through collaborative filtering of items in S3 are as follows: Step 1: First, create a user and preferred product interaction matrix. The rows of the matrix represent users, the columns represent the preferred products to be analyzed, and the cell value is the rating status. 1 means rated and 0 means unrated. Then, get the user set from the user model. , specifically refers to the preferred products for analysis and analyze preferred products The collection of all users who have provided ratings, user collection , Refers to the first user in the user set. Refers to the first users, and sum the user set to obtain the user sum ; Step 2: Obtain the analyzed preferred products from the user model and analyze preferred products Corresponding rating and rating , and measure the preference of two analyzed products by cosine similarity and analyze preferred products The angle between the rating vectors in the vector space and the cosine similarity , the specific algorithm formula is: ； Step 3: Get all users’ preferred products for analysis from the user model and analyze preferred products Ratings and get the preferred products analyzed Rating collection and analyze preferred products Rating collection , and then the scoring set and score collection conduct Calculation of average score and Calculate the average score and obtain the calculated Average score and calculated Average score ; Step 4: Measure the preferred products analyzed by Pearson correlation coefficient and analyze preferred products Linear correlation between rating vectors, summed over users ,score ,score , calculated Average score and calculated Average score Perform calculations and obtain linear results , the specific algorithm formula is: ； Step 5: Obtain cosine similarity To evaluate and analyze the preferred products and analyze preferred products The similarity between them is obtained by obtaining the linear result of the calculation Weighing and analyzing preferred products and analyze preferred products The linear relationship between the rating vectors. For each analyzed preference product, find the L most similar other analyzed preference products. This step involves sorting the similarity of each analyzed preference product, and then selecting the highest L values. The analyzed preference products corresponding to the highest L values ​​are the generated collaborative filtering recommended products.

8. The online e-commerce recommendation method based on big data according to claim 6, characterized in that: The implementation steps of the user collaborative filtering recommended product index generated by accuracy calculation in the second case of S3 are as follows: Step 1: The recommended user collaborative filtering recommended products that are actually adopted by the user are regarded as true positives TP, the recommended user collaborative filtering recommended products that are not actually adopted by the user are regarded as false positives FP, the unrecommended user collaborative filtering recommended products that are not actually adopted by the user are regarded as true negatives TN, and the unrecommended user collaborative filtering recommended products that should have been adopted by the user are regarded as false negatives FN; Step 2: Calculate the generated user collaborative filtering recommended product index based on the true positive class TP, false positive class FP, true negative class TN, and false negative class FN, and obtain the calculated user collaborative filtering index , the specific algorithm formula is: 。

Images