CN110956528B - Recommendation method and system for e-commerce platform - Google Patents
Recommendation method and system for e-commerce platform Download PDFInfo
- Publication number
- CN110956528B CN110956528B CN201910973429.1A CN201910973429A CN110956528B CN 110956528 B CN110956528 B CN 110956528B CN 201910973429 A CN201910973429 A CN 201910973429A CN 110956528 B CN110956528 B CN 110956528B
- Authority
- CN
- China
- Prior art keywords
- prediction
- algorithm
- training
- historical
- weight vectors
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Abstract
The invention discloses a recommendation method and a recommendation system for an e-commerce platform, wherein a first prediction algorithm and a second prediction algorithm are used as primary learners of a Stacking algorithm to predict a training set to obtain two prediction data sets; selecting a plurality of strong features, solving a plurality of groups of weight vectors corresponding to the strong features in a training set, respectively weighting two prediction data sets to obtain corresponding prediction data sets, and respectively training a secondary learner by using the data sets to obtain a plurality of corresponding training models; and predicting the data set to be predicted by using the training models, and taking the average value of the prediction results of the training models as the final prediction result. The invention makes up the defects of the previous model research, improves the prediction accuracy of the e-commerce platform on the purchase probability of the customer, has better model effect, and ensures that the model has wider application range on the e-commerce platform and is easier to generalize and popularize.
Description
Technical Field
The invention relates to the field of data mining, in particular to a recommendation method and a recommendation system for an e-commerce platform.
Background
Data mining uses correlation algorithms to extract correct, useful, unknown, comprehensive, and user-interesting knowledge from large, incomplete, noisy, fuzzy, random data, build models, models for decision support, methods, tools, and processes that provide predictive decision support. Data mining generally refers to the process of algorithmically searching a large amount of data for information hidden therein. With the advent of the internet era and the big outbreak of data, data mining technology is widely and urgently applied to various fields, such as industries of finance, telecommunication, insurance, medical treatment, catering and the like. A large amount of data are processed by using methods such as sorting, analyzing, summarizing, reasoning and the like, so that practical problems are guided and analyzed, relevant prediction results are obtained, and more favorable decisions are made.
Currently, as the living standard of consumers gradually rises, the consumption upgrading phenomenon appearing in the consumers is increasingly obvious, and a plurality of novel e-commerce modes, such as social e-commerce and the like, emerge. The e-commerce platform is full of vitality and competition tends to be fierce. Under the background, if the data can be better utilized, the data is analyzed and predicted by using a data mining technology, so that the e-commerce platform is energized, the platform is more accurate and personalized, more vitality is injected into the e-commerce platform, and the platform is better for consumers in goods and services. Meanwhile, the technology creates larger profit margin for the e-commerce platform.
Although the existing prediction algorithm has a relatively excellent effect in one aspect, in the actual application process, the application is based on a single prediction algorithm in many cases, and the advantages of all prediction algorithms cannot be well utilized to avoid the defects, so that the effective utilization of the fusion of the prediction algorithms is difficult to achieve.
Disclosure of Invention
The invention aims to provide a recommendation method and a recommendation system for an e-commerce platform, so that personalized recommendation prediction of the e-commerce platform is more accurate.
In order to realize the task, the invention adopts the following technical scheme:
a recommendation method for an e-commerce platform comprises the following steps:
the method comprises the steps that a first prediction algorithm and a second prediction algorithm are used as primary learners of a Stacking algorithm, firstly, a built training set is used as input to predict the two primary learners of the first prediction algorithm and the second prediction algorithm, and two prediction data sets are obtained;
selecting a plurality of strong features, solving a plurality of groups of weight vectors corresponding to the strong features in a training set, and weighting the two prediction data sets by using the plurality of groups of weight vectors to obtain a prediction data set corresponding to each group of weighted weight vectors;
respectively training a secondary learner by utilizing the corresponding prediction data set after each group of weight vectors are weighted to obtain a plurality of corresponding training models;
and predicting the data set to be predicted by utilizing the training models, and taking the average value of the prediction results of the training models as the final prediction result.
Furthermore, the first prediction algorithm adopts an XGboost algorithm, and the second prediction algorithm adopts a LightGBM algorithm.
Further, the data in the training set includes historical purchase amounts, a concentration degree of purchase categories and historical conversion rates of a plurality of customers.
Further, the selecting a plurality of strong features and solving a plurality of groups of weight vectors corresponding to the strong features in the training set includes:
and selecting the training centralized historical purchase quantity, the centralized degree of the purchase types and the historical conversion rate as strong features for normalization, and obtaining a weight vector of the historical purchase quantity, a weight vector of the centralized degree of the purchase types and a weight vector of the historical conversion rate.
A recommendation system for an e-commerce platform, comprising:
the prediction data set establishing module is used for taking a first prediction algorithm and a second prediction algorithm as primary learners of a Stacking algorithm, firstly predicting the two primary learners of the first prediction algorithm and the second prediction algorithm by taking an established training set as input to obtain two prediction data sets;
the prediction data set weighting module is used for selecting a plurality of strong features, solving a plurality of groups of weight vectors corresponding to the strong features in the training set, and weighting the two prediction data sets by utilizing the plurality of groups of weight vectors to obtain a prediction data set corresponding to each group of weight vectors after weighting;
the training model establishing module is used for respectively training the secondary learners by utilizing the corresponding prediction data sets after each group of weight vectors are weighted to obtain a plurality of corresponding training models;
and the prediction module is used for predicting the data set to be predicted by utilizing the training models and taking the average value of the prediction results of the training models as the final prediction result.
The invention has the following technical characteristics:
the invention optimizes and improves the stacking fusion algorithm, applies the data mining technology to the current active e-commerce platform, has higher social value, and ensures that the e-commerce platform has more individuation and vitality, thereby being beneficial to the purchase experience of users of the e-commerce platform and creating more profit space for the e-commerce platform. In the embodiment of the invention, the concept of combining the XGboost algorithm and the LightGBM algorithm makes up the defects of the previous model research, improves the accuracy of predicting the purchasing probability of customers by the e-commerce platform, has a better model effect, and ensures that the model has a wider application range on the e-commerce platform and is easier to generalize and popularize.
Drawings
Fig. 1 is a schematic diagram of a recommendation method for an e-commerce platform according to an embodiment of the present invention;
FIG. 2 is a comparison graph of accuracy of an XGboost algorithm model and a fusion model of the invention in different data scales;
FIG. 3 is a graph comparing accuracy of a LightGBM algorithm model and a fusion model of the invention for different data scales;
FIG. 4 is a graph of accuracy comparison for different models at different data scales.
Detailed Description
The invention optimizes and improves the stacking algorithm on the basis of the existing prediction algorithm, provides a fusion algorithm for avoiding the loss of strong feature relevance in the integration process, and applies the algorithm to the return visit prediction of the e-commerce platform.
The Stacking algorithm is largely divided into two layers, with the learner at layer 0 being referred to as the primary learner and the learner at layer 1 for combining being referred to as the secondary learner. A plurality of primary learners are trained by using original characteristic data as input, and then the outputs of the primary learners are combined to be used as characteristics for training a secondary learner.
In the Stacking algorithm integration, the characteristics of the original data set should be mined as much as possible, so as to avoid the loss of the correlation characteristics. In terms of customer return visit prediction of the e-commerce platform, three characteristics of historical purchase quantity, purchase category centralization degree and historical conversion rate are important for the model, and the correlation characteristics of the three characteristics are lost more or less in the process of carrying out stacking integration. Therefore, the method for compensating the lost information has a great effect on improving the accuracy of model prediction. The recommendation method of the E-commerce platform provided by the invention takes a first prediction algorithm and a second prediction algorithm as primary learners of a Stacking algorithm, and predicts the two primary learners of the first prediction algorithm and the second prediction algorithm by taking an established training set as input to obtain two prediction data sets; selecting a plurality of strong features, solving a plurality of groups of weight vectors corresponding to the strong features in the training set, and weighting the two prediction data sets by using the plurality of groups of weight vectors to obtain a prediction data set corresponding to each group of weight vectors after weighting; selecting a secondary learner, and respectively training the secondary learner by utilizing the corresponding prediction data set after each group of weight vectors are weighted to obtain a plurality of corresponding training models; and predicting the data set to be predicted by using a plurality of training models, wherein the average value of the prediction results of the plurality of training models is used as the final prediction result. In a specific embodiment, the method comprises the following steps:
Wherein, the training set can be obtained by the merchant through collecting the original sales data through the sales system and processing the data. The final training set may include the following fields: customer number, historical purchase amount, concentration of purchase categories, historical conversion rate, whether the customer purchases, etc.
Step 2, taking the first prediction algorithm and the second prediction algorithm as primary learners of the Stacking algorithm, and respectively predicting the training sets to obtain two groups of prediction data sets
The prediction data set will be the input to the next level learner after being weighted.
Where the predicted value represents a probability that the customer will purchase an item in a time period in the future, such as a month in the future. Here, the
The probabilities are predicted by models trained by the first prediction algorithm and the second prediction algorithm respectively.
In the embodiment of the present application, the first prediction algorithm and the second prediction algorithm may use a prediction algorithm existing in the prior art.
The XGboost algorithm and the LightGBM algorithm are widely applied at the present stage and provide convenience for life generation of people. Such as application of the XGBoost algorithm in home revenue and expenditure, resistivity estimation, application of the LightGBM algorithm in three-dimensional garment prototyping, and the like. The two algorithms have strong algorithm functions and good prediction effects, and are mainstream algorithms for industrial and data science competitions.
In the fusion algorithm, the XGboost algorithm and the LightGBM algorithm are respectively used as the first prediction algorithm and the second prediction algorithm, so that the fusion algorithm has ideal effect. Besides the two prediction algorithms, the first prediction algorithm and the second prediction algorithm can also adopt other prediction methods in the prior art.
Step 3, selecting the historical purchase quantity in the training set, the centralized degree of the purchase types and the historical conversion rate as strong features to be normalized to obtain the weight vector of the historical purchase quantity
Weight vector of concentration of purchase categories
And weight vector of historical conversion
There are three sets of weight vectors.
For example, the weight of the historical purchase amount for the ith customer is:
in the above equation, buy _ num i Representing the historical purchase amount of the ith customer,
representing the sum of the historical purchase amounts of all customers.
The concentration weight for the purchase category of the ith customer is expressed as:
in the above formula, conc _ deg i Represents a purchase category concentration degree value of the ith customer,
represents the sum of the purchase category concentration degree values of all customers.
The historical conversion weight for the ith customer is expressed as:
in the above formula, the conv _ rate i Representing the historical conversion rate of the ith customer,
representing the sum of the historical conversion rates of all customers.
W of all customers buyi 、w conci 、w convi Weight vector respectively constituting historical purchase amount
Weight vector of concentration of purchase categories
And weight vector of historical conversion
Step 4, using the three groups of weight vectors obtained in the step 2 to respectively carry out the two groups of prediction data sets obtained in the step 1
And giving rights to obtain three batches of data sets. In this step, the
Respectively weighting the prediction data sets to obtain two groups of predictions of which the data sets are weightedA data set.
And 5, training a secondary learner of the Stacking algorithm by using the three batches of data sets respectively, and saving three correspondingly generated training models as fusion models. Optionally, the secondary learning employs a LightGBM algorithm.
Step 6, processing the data set to be predicted according to the same method of the steps 2 to 4, and respectively inputting the processed data set into the three training models generated in the step 5 to obtain three prediction results pred buy ,pred conc ,pred conv (ii) a And averaging the three prediction results to obtain a final prediction value:
namely, step 6 uses the data set to be predicted as the training set in step 2, processes the data set according to steps 2 to 4, and finally obtains three data sets, and uses the three training models in step 5 to predict the three data sets, and the average value of the prediction results is used as the final prediction result.
On the basis of the technical scheme, the invention further provides a recommendation system of the e-commerce platform, which comprises the following components:
the prediction data set establishing module is used for taking a first prediction algorithm and a second prediction algorithm as primary learners of a Stacking algorithm, firstly predicting the two primary learners of the first prediction algorithm and the second prediction algorithm by taking an established training set as input to obtain two prediction data sets;
the prediction data set weighting module is used for selecting a plurality of strong features, solving a plurality of groups of weight vectors corresponding to the strong features in the training set, and weighting the two prediction data sets by utilizing the plurality of groups of weight vectors to obtain a prediction data set corresponding to each group of weight vectors after weighting;
the training model establishing module is used for respectively training the secondary learners by utilizing the corresponding prediction data sets after weighting each group of weight vectors to obtain a plurality of corresponding training models;
and the prediction module is used for predicting the data set to be predicted by utilizing the training models and taking the average value of the prediction results of the training models as the final prediction result.
The description of each module in the system refers to the corresponding steps in the foregoing method embodiments, and is not repeated herein.
Model evaluation criteria
For the customer's prediction of a repurchase of a certain type of goods, the predicted error is of greater concern. Therefore, the invention adopts the auc as a standard for evaluating the model to have excellent effect.
Wherein, the auc formula is:
wherein the content of the first and second substances,
m indicates that there are M positive samples in the test sample and N indicates that there are N negative samples in the test sample.
auc represents the accuracy of model prediction. The larger the error, the higher the accuracy and the better the model effect.
Experiments and simulations
The data of the scheme is derived from JDATA platform open data. After data cleaning and preprocessing, 352568 records useful for the experiment are extracted, feature engineering is carried out on the preprocessed data, and 32 main features such as user id, age, gender, user grade, user historical browsing attention and other behaviors, user behavior and commodity attribute combination features, time features of user behaviors, goodness rate, historical purchase quantity, purchase category centralization degree and historical conversion rate are obtained.
After data are preprocessed and subjected to feature engineering, model training is performed on input data, wherein data except the is _ buy attribute in the table 2 are feature attributes, and the is _ buy attribute is a training label attribute. The model input data is shown in table 1.
Data entry for the model of Table 1
Suer _ lv _ cd: the fields in the raw data represent the user classes.
Table 2 is an example of the output of the XGBoost algorithm as a prediction model. Wherein, user _ id represents a target user, and purch _ proba represents the probability of repeated purchase of the target user.
Table 2 output example
Results and analysis of the experiments
Taking 1% to 100% of all data in the experiment, respectively training different classifiers to obtain different models, and comparing error indexes of the xgboost model and the Fusion model of the invention and the prediction accuracy of the lightGBM model and the Fusion model of the invention, as shown in fig. 2 and 3 (Fusion model represents the Fusion model of the invention, xgb model represents the xgboost model, lgb model represents the lightGBM model in fig. 2 to 4)
In order to further prove the stability, reliability and rigor of the experimental result of the scheme, four batches of different data in the experiment are randomly taken, 1% to 100% of each batch of data is respectively taken, the three models are trained in sequence to obtain the prediction accuracy of the three different models, and each batch of data respectively corresponds to one comparison graph to obtain the comparison graph shown in figure 4. The abscissa in each graph represents the data volume proportion under the data batch, and the ordinate represents the corresponding prediction accuracy. In the figure, the solid line, the dotted line and the dotted line respectively represent the prediction accuracy corresponding to the xgboost model, the lightGBM model and the fusion model of the present invention under different data scales. The change of the prediction accuracy of different models under different data scales is reflected visually in the graph.
Comparing and analyzing fig. 2 and fig. 3, comparing the prediction accuracy of the fusion model of the present invention with the prediction accuracy of the xgboost model and the lightGBM model, it can be known that the prediction accuracy of the fused model for predicting the probability of the customer purchasing a certain kind of commodity again is obviously higher than the xgboost model and the lightGBM model, and the model effect of the fusion model is better than the two models. Analyzing the experiment chart 4, under the same data scale, the model effect of the fusion model is superior to that of the lightGBM model, and the model effect of the lightGBM model is superior to that of the xgboost model. Moreover, for a certain model, as the data amount increases, the accuracy of predicting the probability of repurchasing a certain commodity by a model prediction customer is improved, the effect of the model is optimized, and when the data amount is in a certain range, the accuracy of predicting the probability of repurchasing a certain commodity by a model prediction customer is gradually improved. And analysis shows that the model effect of the fusion model is better than that of the single model from the overall view.
The invention optimizes and improves the stacking fusion algorithm, applies the data mining technology to the current active e-commerce platform, has higher social value, and ensures that the e-commerce platform has more individuation and vitality, thereby being beneficial to the purchase experience of users of the e-commerce platform and creating more profit space for the e-commerce platform. In the embodiment of the invention, the XGboost and LightGBM algorithm are fused to make up the defects of the previous model research, the prediction accuracy of the e-commerce platform on the purchase probability of customers is improved, the model effect is better, the application range of the model on the e-commerce platform is wider, and the model is easier to generalize and popularize; within a certain range, the effect of the model is continuously optimized as the scale of the training data increases.
Claims (2)
1. A recommendation method for an e-commerce platform is applied to return visit prediction of the e-commerce platform and comprises the following steps:
the method comprises the steps that a first prediction algorithm and a second prediction algorithm are used as primary learners of a Stacking algorithm, firstly, a built training set is used as input to predict the two primary learners of the first prediction algorithm and the second prediction algorithm, and two prediction data sets are obtained; the first prediction algorithm adopts an XGboost algorithm, and the second prediction algorithm adopts a LightGBM algorithm; the data in the training set comprises historical purchase quantity, centralized degree of purchase types and historical conversion rate of a plurality of customers;
selecting a plurality of strong features, solving a plurality of groups of weight vectors corresponding to the strong features in a training set, and weighting the two prediction data sets by using the plurality of groups of weight vectors to obtain a prediction data set corresponding to each group of weighted weight vectors;
respectively training a secondary learner by utilizing the corresponding prediction data set after weighting each group of weight vectors to obtain a plurality of corresponding training models;
predicting a data set to be predicted by utilizing a plurality of training models, and taking the average value of the prediction results of the training models as a final prediction result; the prediction result is the probability that the customer purchases a certain commodity again;
the selecting a plurality of strong features and solving a plurality of groups of weight vectors corresponding to the strong features in the training set comprises the following steps:
and selecting the training concentrated historical purchase quantity, the concentrated degree of the purchase types and the historical conversion rate as strong features for normalization to obtain a weight vector of the historical purchase quantity, a weight vector of the concentrated degree of the purchase types and a weight vector of the historical conversion rate.
2. A recommendation system for an e-commerce platform, comprising:
the prediction data set establishing module is used for taking a first prediction algorithm and a second prediction algorithm as primary learners of a Stacking algorithm, firstly predicting the two primary learners of the first prediction algorithm and the second prediction algorithm by taking an established training set as input to obtain two prediction data sets; the first prediction algorithm adopts an XGboost algorithm, and the second prediction algorithm adopts a LightGBM algorithm; the data in the training set comprises historical purchase amount, centralized degree of purchase types and historical conversion rate of a plurality of customers;
the prediction data set weighting module is used for selecting a plurality of strong features, solving a plurality of groups of weight vectors corresponding to the strong features in the training set, and weighting the two prediction data sets by utilizing the plurality of groups of weight vectors to obtain a prediction data set corresponding to each group of weight vectors after weighting;
the training model establishing module is used for respectively training the secondary learners by utilizing the corresponding prediction data sets after each group of weight vectors are weighted to obtain a plurality of corresponding training models;
the prediction module is used for predicting the data set to be predicted by utilizing the training models and taking the average value of the prediction results of the training models as the final prediction result; the prediction result is the probability that a customer purchases a certain commodity again;
the selecting a plurality of strong features and solving a plurality of groups of weight vectors corresponding to the strong features in the training set comprises the following steps:
and selecting the training concentrated historical purchase quantity, the concentrated degree of the purchase types and the historical conversion rate as strong features for normalization to obtain a weight vector of the historical purchase quantity, a weight vector of the concentrated degree of the purchase types and a weight vector of the historical conversion rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910973429.1A CN110956528B (en) | 2019-10-14 | 2019-10-14 | Recommendation method and system for e-commerce platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910973429.1A CN110956528B (en) | 2019-10-14 | 2019-10-14 | Recommendation method and system for e-commerce platform |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110956528A CN110956528A (en) | 2020-04-03 |
| CN110956528B true CN110956528B (en) | 2022-11-04 |
Family
ID=69975674
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910973429.1A Active CN110956528B (en) | 2019-10-14 | 2019-10-14 | Recommendation method and system for e-commerce platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110956528B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112327666B (en) * | 2020-10-22 | 2023-02-07 | 智慧航海（青岛）科技有限公司 | Method for determining target function weight matrix of power cruise system control model |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107590741A (en) * | 2017-09-19 | 2018-01-16 | 广东工业大学 | A kind of method and system of predicted pictures popularity |
| CN108090510A (en) * | 2017-12-15 | 2018-05-29 | 北京大学 | A kind of integrated learning approach and device based on interval optimization |
| CN109272365A (en) * | 2018-10-15 | 2019-01-25 | 广东工业大学 | A kind of flow of restaurant guests prediction technique, device and equipment based on data mining |
| CN109741114A (en) * | 2019-01-10 | 2019-05-10 | 博拉网络股份有限公司 | A kind of user under big data financial scenario buys prediction technique |
-
2019
- 2019-10-14 CN CN201910973429.1A patent/CN110956528B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107590741A (en) * | 2017-09-19 | 2018-01-16 | 广东工业大学 | A kind of method and system of predicted pictures popularity |
| CN108090510A (en) * | 2017-12-15 | 2018-05-29 | 北京大学 | A kind of integrated learning approach and device based on interval optimization |
| CN109272365A (en) * | 2018-10-15 | 2019-01-25 | 广东工业大学 | A kind of flow of restaurant guests prediction technique, device and equipment based on data mining |
| CN109741114A (en) * | 2019-01-10 | 2019-05-10 | 博拉网络股份有限公司 | A kind of user under big data financial scenario buys prediction technique |
Non-Patent Citations (1)
| Title |
|---|
| 基于多样性加权相似度的集成局部加权偏最小二乘软测量建模;潘贝 等;《信息与控制》;20190430;第48卷(第02期);第217-222页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110956528A (en) | 2020-04-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108648049B (en) | Sequence recommendation method based on user behavior difference modeling | |
| CN111222332B (en) | Commodity recommendation method combining attention network and user emotion | |
| CN111797321B (en) | Personalized knowledge recommendation method and system for different scenes | |
| CN110717098B (en) | Meta-path-based context-aware user modeling method and sequence recommendation method | |
| CN110781409B (en) | Article recommendation method based on collaborative filtering | |
| CN111460130A (en) | Information recommendation method, device, equipment and readable storage medium | |
| CN111242729A (en) | Serialization recommendation method based on long-term and short-term interests | |
| CN113256367B (en) | Commodity recommendation method, system, equipment and medium for user behavior history data | |
| CN115917535A (en) | Recommendation model training method, recommendation device and computer readable medium | |
| CN112395487B (en) | Information recommendation method and device, computer readable storage medium and electronic equipment | |
| CN112085525A (en) | User network purchasing behavior prediction research method based on hybrid model | |
| CN110222838B (en) | Document sorting method and device, electronic equipment and storage medium | |
| Chen et al. | An effective matching algorithm with adaptive tie-breaking strategy for online food delivery problem | |
| CN110727872A (en) | Method and device for mining ambiguous selection behavior based on implicit feedback | |
| Choudhary et al. | SARWAS: Deep ensemble learning techniques for sentiment based recommendation system | |
| CN114861050A (en) | Feature fusion recommendation method and system based on neural network | |
| CN112819024A (en) | Model processing method, user data processing method and device and computer equipment | |
| Hassan et al. | Performance analysis of neural networks-based multi-criteria recommender systems | |
| CN110956528B (en) | Recommendation method and system for e-commerce platform | |
| CN117131261A (en) | E-commerce content recommendation system based on AI and big data analysis | |
| CN115829683A (en) | Power integration commodity recommendation method and system based on inverse reward learning optimization | |
| CN108133407B (en) | E-commerce recommendation technology and system based on soft set decision rule analysis | |
| CN115935067A (en) | Article recommendation method integrating semantics and structural view for socialized recommendation | |
| CN112464106B (en) | Object recommendation method and device | |
| Erkamim et al. | Decision Support System Using a Combination of COPRAS and Rank Reciprocal Approaches to Select Accounting Software |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |