CN108446351B - Hotel screening method and system based on user preference of OTA platform - Google Patents
Hotel screening method and system based on user preference of OTA platform Download PDFInfo
- Publication number
- CN108446351B CN108446351B CN201810191566.5A CN201810191566A CN108446351B CN 108446351 B CN108446351 B CN 108446351B CN 201810191566 A CN201810191566 A CN 201810191566A CN 108446351 B CN108446351 B CN 108446351B
- Authority
- CN
- China
- Prior art keywords
- hotel
- user
- model
- prediction
- result
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/12—Hotels or restaurants
Abstract
The invention discloses a hotel screening method and a system based on user preference of an OTA platform, wherein the hotel screening method comprises the following steps: s1, acquiring user information of each user; s2, acquiring user characteristic data according to the user information; s3, performing model training on the user characteristic data based on the XGBOOST model to obtain a user preference prediction model, and obtaining a first user preference prediction result according to the user preference prediction model; and S4, obtaining a hotel screening result of the OTA platform according to the first user preference prediction result. According to the invention, the hotel screening result based on the user preference can be obtained, the computing resource of the OTA platform is greatly saved, meanwhile, the screening result can be rapidly and effectively displayed to the user, the user experience is improved, and the purpose of personalized display is achieved.
Description
Technical Field
The invention relates to the technical field of information processing of an OTA (on-line Travel Agency) platform, in particular to a hotel screening method and system based on user preference of the OTA platform.
Background
As OTA software increases in the frequency of people's daily use, more and more hotels have begun to collaborate with OTA companies. As the number of hotels increases, on one hand, OTA software is represented to provide users with richer and better services, and on the other hand, users are required to face a lengthy hotel list and spend more time and effort to determine the needed hotels.
Practice has shown that there are a few users who have a definite hotel goal from the moment the OTA software is entered, and more users may not be able to select the hotel that meets their needs at once. Even with the assistance of the use of screening functions, one may face the problem of too many screening terms resulting in increased effort or not having a clear target without knowing how to select the screening. Therefore, how to effectively show the hotels needed by different users and save precious time for the users becomes an important effort direction of the OTA platform.
Generally, when a user enters a hotel list page, if no search keyword is found and no other screening is carried out, the OTA platform provides a default ordered hotel list for the user, but the ordering does not accord with the real preference of the user. If it is desired to have a fully personalized presentation for each user, while technically fully feasible, it also means a huge amount of computation and is not always possible to kick on at a glance. Therefore, OTA needs to find such a solution that is computationally inexpensive and can satisfy user preferences.
Disclosure of Invention
The technical problem to be solved by the invention is to overcome the defect that the OTA platform in the prior art cannot sort hotels according to user preference, and the invention aims to provide a hotel screening method and system based on the user preference of the OTA platform.
The invention solves the technical problems through the following technical scheme:
the invention provides a hotel screening method based on user preference of an OTA platform, which comprises the following steps:
s1, acquiring user information of each user;
the user information comprises at least one of user history screening information, user history browsing information and user history order information;
s2, acquiring user characteristic data according to the user information;
s3, performing model training on the user characteristic data based on an XGBOOST model (a machine learning algorithm), obtaining a user preference prediction model, and obtaining a first user preference prediction result according to the user preference prediction model;
and S4, obtaining a hotel screening result of the OTA platform according to the first user preference prediction result.
Preferably, step S4 is preceded by:
s31, acquiring city information of each city;
the city information comprises hotel distribution information and/or hotel historical order information;
s32, performing model training on the first user preference prediction result and the city information by adopting an XGBOOST model to obtain a city data correction model;
s33, obtaining a correction result of each hotel of each city according to the city data correction model;
and the correction result is used for representing a difference value between the hotel parameter of each hotel in each city and the hotel parameter corresponding to the first user preference prediction result.
Preferably, step S4 specifically includes:
and acquiring a target city selected by a user, acquiring a second user preference prediction result according to the correction result corresponding to the target city and the first user preference prediction result, and acquiring a hotel screening result of the OTA platform according to the second user preference prediction result.
Preferably, step S3 further includes:
obtaining a user preference prediction probability according to the user preference prediction model, judging whether the user preference prediction probability is larger than a set threshold value, and if so, retaining a first user preference prediction result corresponding to the user preference prediction probability; and if not, giving up the first user preference prediction result corresponding to the user preference prediction probability.
Preferably, the user characteristic data comprises at least one of a hotel star rating, a hotel price, a hotel brand, a hotel score, a hotel comment number, a hotel breakfast type, a hotel bed type, a hot-sell user and a price-rate user of the user preference;
step S3 specifically includes:
respectively carrying out model training on the hotel star level, the hotel brand, the hotel breakfast type, the hotel bed type, the hot sales user and the price comparison user by adopting an XGB OST model to obtain a corresponding hotel star level prediction model, a hotel brand prediction model, a hotel breakfast type prediction model, a hotel bed type prediction model, a hot sales user prediction model and a price comparison user prediction model, and then obtaining a user-preferred hotel star level prediction result, a hotel brand prediction result, a hotel breakfast type prediction result, a hotel bed type prediction result, a hot sales user prediction result and a price comparison user prediction result;
performing model training on the hotel price by combining an XGBOOST model and a quantile regression model to obtain a hotel price prediction model, and then obtaining a hotel price prediction result preferred by a user;
acquiring a hotel grade range of a hotel which is historically checked in by a user on the OTA platform according to the user information by adopting an XGBOOST model, and acquiring a minimum value of the hotel grade preferred by the user;
and acquiring the hotel comment number range of the hotel which the user has historically checked in on the OTA platform according to the user information by adopting an XGBOOST model, and acquiring the minimum value of the hotel comment number preferred by the user.
The invention also provides a hotel screening system based on the user preference of the OTA platform, which comprises a user information acquisition module, a characteristic data acquisition module, a prediction model acquisition module, a prediction result acquisition module and a screening module;
the user information acquisition module is used for acquiring the user information of each user;
the user information comprises at least one of user history screening information, user history browsing information and user history order information;
the characteristic data acquisition module is used for acquiring user characteristic data according to the user information;
the prediction model acquisition module is used for carrying out model training on the user characteristic data based on an XGBOOST model, acquiring a user preference prediction model and calling the prediction result acquisition module;
the prediction result obtaining module is used for obtaining a first user preference prediction result according to the user preference prediction model;
and the screening module is used for acquiring a hotel screening result of the OTA platform according to the first user preference prediction result.
Preferably, the hotel screening system further comprises a city information acquisition module, a correction model acquisition module and a correction result acquisition module;
the city information acquisition module is used for acquiring city information of each city;
the city information comprises hotel distribution information and/or hotel historical order information;
the correction model acquisition module is used for respectively carrying out model training on the user preference prediction result and the city information by adopting an XGBOOST model to acquire a city data correction model;
the correction result acquisition module acquires the correction result of each hotel of each city according to the city data correction model;
and the correction result is used for representing a difference value between the hotel parameter of each hotel in each city and the hotel parameter corresponding to the first user preference prediction result.
Preferably, the prediction result obtaining module is further configured to obtain a target city selected by a user, and obtain a second user preference prediction result according to the correction result corresponding to the target city and the first user preference prediction result;
and the screening module is also used for acquiring a hotel screening result of the OTA platform according to the second user preference prediction result.
Preferably, the hotel screening system further comprises a prediction probability obtaining module and a judging module;
the prediction probability obtaining module is used for obtaining the user preference prediction probability according to the user preference prediction model and calling the judging module;
the judging module is used for judging whether the user preference prediction probability is larger than a set threshold value, if so, a first user preference prediction result corresponding to the user preference prediction probability is reserved; and if not, giving up the first user preference prediction result corresponding to the user preference prediction probability.
Preferably, the user characteristic data comprises at least one of a hotel star rating, a hotel price, a hotel brand, a hotel score, a hotel comment number, a hotel breakfast type, a hotel bed type, whether the user is a hot-selling user and whether the user is a price-matching user preferred by the user;
the prediction model acquisition module is used for respectively carrying out model training on the hotel star level, the hotel brand, the hotel breakfast type, the hotel bed type, the hot sales user and the price comparison user by adopting an XGBOOST model to acquire a corresponding hotel star level prediction model, a hotel brand prediction model, a hotel breakfast type prediction model, a hotel bed type prediction model, a hot sales user prediction model and a price comparison user prediction model;
the prediction result acquisition module is used for respectively acquiring a hotel star-level prediction result, a hotel brand prediction result, a hotel breakfast type prediction result, a hotel bed type prediction result, a hot sale user prediction result and a price comparison user prediction result of user preference according to the hotel star-level prediction model, the hotel brand prediction model, the hotel breakfast type prediction model, the hot sale user prediction model and the price comparison user prediction model;
the prediction model acquisition module is also used for performing model training on the hotel price by combining an XGBOOST model and a quantile regression model to acquire a hotel price prediction model;
the prediction result obtaining module is also used for obtaining a hotel price prediction result preferred by the user according to the hotel price prediction model;
the prediction model acquisition module is also used for acquiring the hotel score range of the hotel which the user has historically checked in on the OTA platform according to the user information by adopting an XGBOOST model;
the prediction result acquisition module is further used for acquiring the minimum value of the hotel score preferred by the user according to the range of the hotel score;
the prediction model acquisition module is also used for acquiring the hotel comment number range of the hotel in the OTA platform, which is historically checked in by the user, by adopting an XGBOOST model according to the user information;
the prediction result acquisition module is also used for acquiring the minimum value of the hotel comment number preferred by the user according to the range of the hotel comment number.
The positive progress effects of the invention are as follows:
according to the method, user characteristic data are obtained through user information, model training is carried out on each user characteristic data, and a first user preference prediction result is obtained; meanwhile, the first user preference prediction result is combined with the city information of each city to obtain the correction result of each hotel of each city; after the user logs in the OTA platform, the result is predicted according to the user preference of the current user, and the correction result corresponding to the target city selected by the user is directly called, so that the hotel screening result based on the user preference is obtained, the computing resource of the OTA platform is greatly saved, meanwhile, the screening result can be rapidly and effectively displayed to the user, the user experience is improved, and the purpose of personalized display is achieved.
Drawings
Fig. 1 is a flowchart of a hotel screening method based on user preferences according to embodiment 1 of the present invention;
fig. 2 is a flowchart of a hotel screening method based on user preferences according to embodiment 2 of the present invention;
fig. 3 is a schematic block diagram of a hotel screening system based on user preferences according to embodiment 3 of the present invention;
fig. 4 is a module schematic diagram of a hotel screening system based on user preferences according to embodiment 4 of the present invention.
Detailed Description
The invention is further illustrated by the following examples, which are not intended to limit the scope of the invention.
Example 1
As shown in fig. 1, the hotel screening method based on user preferences for the OTA platform of the embodiment includes:
s101, acquiring user information of each user;
the user information comprises at least one of user history screening information, user history browsing information and user history order information;
s102, acquiring user characteristic data according to user information;
the user characteristic data comprises at least one of a hotel star rating, a hotel price, a hotel brand, a hotel score, a hotel comment number, a hotel breakfast type, a hotel bed type, whether the user is a hot sale user or not and whether the user is a price comparison user or not of user preference.
S103, performing model training on the user characteristic data by adopting an XGBOOST model to obtain a user preference prediction model, and obtaining a first user preference prediction result according to the user preference prediction model;
specifically, the user characteristic data and the user characteristic form are shown in the following table:
user characteristic data | Characteristic form |
|
2 stars with highest user occupancy frequency |
Hotel price | Price interval selected by user |
Hot-sell user | Whether the user is a hot-selling user |
Hotel brand | Hotel brand with user's intention |
Hotel scoring | Minimum value of user intention score |
Number of people commenting in hotel | Minimum number of people users comment on |
Breakfast type of hotel | Preference type of user breakfast |
Hotel bed type | Preference type of user's bed type |
User of price comparison | Whether the user is a price comparison user |
The method comprises the steps of selecting and training some users from hotel order data of an OTA platform in a past period of time in a sampling mode, extracting historical order information of the users, and obtaining a characteristic value corresponding to characteristic data of each user to serve as a target label of a training set. The method comprises the steps that a hotel star level, a hotel price, a hotel breakfast type and a hotel bed type can be obtained from order data of a user; when the user ordering area is a sales volume area located in the first three of the current cities, defining the user as a hot sales user; when the hotel subscribed by the user is a brand hotel, the user is defined to be interested in the brand hotel.
And respectively establishing models aiming at the 9 user characteristic data, acquiring corresponding user preference prediction models, and screening hotels of the OTA platform according to user preference prediction results. Specifically, model training of 9 user feature data is divided into three different types:
1) the XGBOST model is adopted to respectively carry out model training on a hotel star, a hotel brand, a hotel breakfast type, a hotel bed type, a hot sales user and a price comparison user, obtain a corresponding hotel star prediction model, a hotel brand prediction model, a hotel breakfast type prediction model, a hotel bed type prediction model, a hot sales user prediction model and a price comparison user prediction model, and then obtain a user-preferred hotel star prediction result, a hotel brand prediction result, a hotel breakfast type prediction result, a hotel bed type prediction result, a hot sales user prediction result and a price comparison user prediction result.
Specifically, the XGBOOST model is an improved version of the conventional GBDT algorithm (a machine learning algorithm), and has many advantages of high speed, good effect, and the like. Unlike the GBDT algorithm, the XGB OST model modifies its objective function, which is specifically the following:
wherein, Obj(t)Is the function of the object of the function,for computing the running sum of residuals, where yiIs the true target value for the ith sample,for the predicted target value in the (t-1) th iteration, ft(xi) Represents the score, ω (f), of each leaf nodet) In order to regularize the parameters, constant is an arbitrary constant value for adjusting the model in order to ensure that the model has high robustness. The function may be approximated using Taylor expansion such that the final objective function depends only on the first and second derivatives of each data point on the error function, thereby making the objectiveThe function can be solved approximately.
Specifically, in the solving, the two-classification model adopts a logistic regression mode, the S function is taken as a target function, and the final result is a probability value between 0 and 1. There are only two possibilities, namely yes or no, for whether the user is a hot-selling user and whether the user is a price comparison user, if the probability value is >0.5, the prediction result is regarded as yes, otherwise, no; there are more than two possibilities for each of the hotel star, the hotel breakfast type and the hotel bed type, so a model is used to predict probability values for each possibility of the hotel star, the hotel breakfast type and the hotel bed type. The two stars with the highest probability of the star level of the hotel are selected as the prediction results, and the possibility with the highest probability of the breakfast type and the bed type of the hotel is selected as the only prediction result.
2) For the hotel price prediction model, too many hotels can be filtered due to too small specific price range, so that the specific price range has no substantial significance for hotel screening and a price interval of a user target needs to be obtained. If the XGB OST model classification method is adopted for prediction, the prediction result is often not flexible enough, and when the types of price intervals are increased, the training time of the XGB OST model is greatly increased, and the training accuracy is reduced. However, if a regression method is adopted to predict hotel prices preferred by users, only a certain prediction result can be obtained, and whether the result is reliable or not cannot be known, and no way is provided for extending the result to a corresponding price interval.
Therefore, in the embodiment, the XGBOOST model is combined with the quantile regression model, and a model for predicting upper and lower price limits at a certain confidence level is established by modifying the objective function of the XGBOOST model, so that a price interval corresponding to a high confidence level is obtained;
wherein, the formula of the upper limit function is: [ max (y-up,0) ] < Lambda + alpha > (y-up) < Lambda 2, the formula for the lower bound function is: [ max (down-y,0) ] < Lambda > + alpha (y-down) < Lambda > 2;
up is the upper limit predicted value, down is the lower limit predicted value, y is the actual user order price, and alpha and beta are adjustable coefficients respectively. [ max (y-up,0) ] < Lambda > and [ max (down-y,0) ] < Lambda > are used for limiting the actual price y of the user to meet the upper and lower limits as much as possible, otherwise, the max function is greater than 0; alpha (y-up) 2 and alpha (y-down) 2 are used for limiting the interval between the upper limit and the lower limit to be as small as possible, and the problem that the prediction range is too large and the actual effect is not achieved is avoided. And respectively calculating a first derivative and a second derivative of the objective function, and solving by adopting an algorithm to obtain a final prediction model, thereby obtaining a hotel price prediction result preferred by a user.
3) Hotel scoring and hotel review population
The method comprises the steps that an XGBOOST model is adopted to obtain a hotel score range of a hotel which a user historically lives in on an OTA platform according to user information, and a minimum value of the hotel score preferred by the user is obtained, namely when the score of the hotel is lower than the corresponding minimum value, the OTA platform does not sort the hotel; otherwise, the hotel is ranked and displayed.
The method comprises the steps that an XGBOOST model is adopted to obtain the hotel comment number range of a hotel which a user historically lives in on an OTA platform according to user information, and the minimum value of the hotel comment number preferred by the user is obtained, namely when the hotel comment number is lower than the corresponding minimum value, the OTA platform does not sequence the hotel; otherwise, the hotel is ranked and displayed.
And S104, obtaining a hotel screening result of the OTA platform according to the first user preference prediction result.
According to the embodiment, user characteristic data are obtained through user information, then an XGBOOST model is adopted to carry out model training on each user characteristic data, and a user preference prediction model is obtained to obtain a first user preference prediction result; after the user logs in the OTA platform, the hotel screening result preferred by the user is obtained according to the first user preference prediction result of the current user, the screening result can be rapidly and effectively displayed to the user, the user experience is improved, and the purpose of personalized display is achieved.
Example 2
As shown in fig. 2, the hotel screening method based on user preferences for the OTA platform of the embodiment is further improved on the basis of embodiment 1, specifically:
step S104 is preceded by:
s1031, obtaining a user preference prediction probability according to the user preference prediction model, judging whether the user preference prediction probability is larger than a set threshold value, and keeping a first user preference prediction result corresponding to the user preference prediction probability; if not, giving up the first user preference prediction result corresponding to the user preference prediction probability;
s1032, acquiring city information of each city;
the city information comprises hotel distribution information and/or hotel historical order information;
s1033, performing model training on the first user preference prediction result and the city information by adopting an XGBOOST model respectively to obtain a city data correction model;
s1034, obtaining a correction result of each hotel of each city according to the city data correction model;
and the correction result is used for representing the difference value between the hotel parameter of each hotel in each city and the hotel parameter corresponding to the first user preference prediction result.
Step S104 specifically includes:
s1041, obtaining a target city selected by a user, obtaining a second user preference prediction result according to a correction result corresponding to the target city and the first user preference prediction result, obtaining a hotel screening result of the OTA platform according to the second user preference prediction result, and displaying the hotel screening result.
In particular, because the distribution characteristics of hotels may differ under different cities. For example, the hotel price and the hotel star rating of a small city are lower than those of a large city, and the number of the hotel comment persons in a tourist city is generally larger than that in a non-tourist city. If a unified model is used to predict user preferences, some user characteristics may not be satisfied in some cities. If hotel screening is performed only by means of the prediction result obtained by the user preference prediction model, the number of hotels meeting requirements is extremely low or even none, so that user experience is influenced, and therefore correction of the user preference prediction result is required according to relevant data in each city.
The manner in which each hotel in each city is modified according to the user characteristic data is shown in the following table:
user characteristic data | Correction method |
Hotel star level | Star grade shift size (as-1 represents: turning one star grade down) |
Hotel price | Size of price movement |
Hot-sell user | User restriction of whether to release hot pin |
Hotel brand | Whether to release hotel brand restrictions |
Hotel scoring | Hotel scoring movement size |
Number of people commenting in hotel | Mobile number of hotel comment people |
Breakfast type of hotel | Whether to release breakfast restrictions |
Hotel bed type | Limitation of whether to release bed |
User of price comparison | Whether to release the price comparison limit |
As shown in the table above, the correction result of each hotel in each city is obtained by combining the first user preference prediction result, the hotel distribution information and the hotel history order information, when the target city is selected by the user, the second user preference prediction result is obtained according to the correction result corresponding to the target city and the first user preference prediction result, and the hotel screening result of the OTA platform is obtained according to the second user preference prediction result, so that the screening result can be rapidly and effectively displayed to the user, the user experience is improved, and the purpose of personalized display is achieved.
When a user opens a client of the OTA platform, the user firstly enters a hotel page and then enters a hotel search page; after the OTA platform acquires the city selected by the user, judging whether the OTA platform is provided with an intelligent screening button screening option or not, and if not, displaying default hotel sequencing; otherwise, the correction processing is carried out on the user preference prediction result and the correction result of the city on line, if the correction mode is numerical value movement, if the hotel star correction result is-1, and the hotel star of the user preference predicted by the user preference prediction model is 3-level or 4-level, the hotel star of the current city actually screened by the OTA platform is 2-level or 3-level; and if the correction mode is to release the limit, if the user predicted by the user preference prediction model is the hot sales user, the corresponding correction result is 1, namely the hot sales user limit is released, the OTA platform does not actually screen the hot sales hotels, and then the hotel sequencing is adjusted according to the user preference prediction result and the correction result of the city. When the user clicks the intelligent screening button screening item, a union set of screening conditions corresponding to all correction results is taken, hotels which do not accord with the screening conditions are filtered, and hotel sorting results which accord with the preference of the user are displayed to the user.
According to the method, user characteristic data are obtained through user information, then an XGBOOST model is adopted to carry out model training on each user characteristic data, and a user preference prediction model is obtained to obtain a user preference prediction result; meanwhile, combining the user preference prediction result with the city information of each city to obtain the correction result of each city; after the user logs in the OTA platform, the result is predicted according to the user preference of the current user, and the correction result corresponding to the target city selected by the user is directly called, so that the hotel screening result preferred by the user is obtained, the calculation workload of the OTA platform is reduced, meanwhile, the screening result can be rapidly and effectively displayed to the user, the user experience is improved, and the purpose of personalized display is achieved.
Example 3
As shown in fig. 4, the hotel screening system based on user preferences for the OTA platform of the embodiment includes a user information obtaining module 1, a feature data obtaining module 2, a prediction model obtaining module 3, a prediction result obtaining module 4, and a screening module 5.
Specifically, the user information obtaining module 1 is configured to obtain user information of each user;
the user information comprises at least one of user history screening information, user history browsing information and user history order information;
the characteristic data acquisition module 2 is used for acquiring user characteristic data according to the user information;
the user characteristic data comprises at least one of a hotel star rating, a hotel price, a hotel brand, a hotel score, a hotel comment number, a hotel breakfast type, a hotel bed type, whether the user is a hot sale user or not and whether the user is a price comparison user or not of user preference.
The prediction model acquisition module 3 is used for carrying out model training on user characteristic data based on the XGBOOST model, acquiring a user preference prediction model and calling the prediction result acquisition module 4;
the prediction result obtaining module 4 is used for obtaining a first user preference prediction result according to the user preference prediction model;
and the screening module 5 is used for acquiring a hotel screening result of the OTA platform according to the first user preference prediction result.
Specifically, the prediction model obtaining module 3 is configured to perform model training on a hotel star, a hotel brand, a hotel breakfast type, a hotel bed type, a hot sales user and a price comparison user respectively by using an XGBOOST model, and obtain a corresponding hotel star prediction model, hotel brand prediction model, hotel breakfast type prediction model, hotel bed type prediction model, hot sales user prediction model and price comparison user prediction model;
the prediction result acquisition module 4 is used for respectively acquiring a hotel star prediction result, a hotel brand prediction result, a hotel breakfast type prediction result, a hotel bed type prediction result, a hot sales user prediction result and a price comparison user prediction result of user preference according to a hotel star prediction model, a hotel brand prediction model, a hotel breakfast type prediction model, a hot sales user prediction result and a price comparison user prediction model;
the prediction model obtaining module 3 is also used for performing model training on the hotel price by combining the XGB OST model and the quantile regression model to obtain a hotel price prediction model;
the prediction result obtaining module 4 is further configured to obtain a hotel price prediction result preferred by the user according to the hotel price prediction model;
the prediction model acquisition module 3 is also used for acquiring the hotel score range of the hotel which the user has historically checked in on the OTA platform according to the user information by adopting the XGB OST model;
the prediction result acquisition module 4 is further configured to acquire a minimum hotel score value preferred by the user according to the hotel score range;
the prediction model acquisition module 3 is also used for acquiring the hotel comment number range of the hotel in the OTA platform for the user to stay in history by adopting the XGBOOST model according to the user information;
the prediction result obtaining module 4 is further configured to obtain a minimum value of the hotel comment population preferred by the user according to the range of the hotel comment population.
The user characteristic data and the user characteristic form are shown in the following table:
user characteristic data | Characteristic form |
|
2 stars with highest user occupancy frequency |
Hotel price | Price interval selected by user |
Hot-sell user | Whether the user is a hot-selling user |
Hotel brand | Whether the user is a hotel brand |
Hotel scoring | Minimum value of user intention score |
Number of people commenting in hotel | Minimum number of people users comment on |
Breakfast type of hotel | Preference type of user breakfast |
Hotel bed type | Preference type of user's bed type |
User of price comparison | Whether the user is a price comparison user |
The method comprises the steps of selecting and training some users from hotel order data of an OTA platform in a past period of time in a sampling mode, extracting historical order information of the users, and obtaining a characteristic value corresponding to characteristic data of each user to serve as a target label of a training set. Wherein, the hotel star level, the hotel price, the breakfast type and the bed type can be obtained from the order data of the user; when the user ordering area is a sales volume area located in the first three of the current cities, defining the user as a hot sales user; when the hotel subscribed by the user is a brand hotel, the user is defined as a brand hotel user.
And respectively establishing models aiming at the 9 user characteristic data, acquiring corresponding user preference prediction models, and screening hotels of the OTA platform according to user preference prediction results. Specifically, model training of 9 user feature data is divided into three different types:
1) the XGBOST model is adopted to respectively carry out model training on a hotel star, a hotel brand, a hotel breakfast type, a hotel bed type, a hot sales user and a price comparison user, obtain a corresponding hotel star prediction model, a hotel brand prediction model, a hotel breakfast type prediction model, a hotel bed type prediction model, a hot sales user prediction model and a price comparison user prediction model, and then obtain a user-preferred hotel star prediction result, a hotel brand prediction result, a hotel breakfast type prediction result, a hotel bed type prediction result, a hot sales user prediction result and a price comparison user prediction result.
Specifically, the XGBOOST model is an improved version of the conventional GBDT algorithm, and has many advantages of high speed, good effect, and the like. Unlike the GBDT algorithm, the XGB OST model modifies its objective function, which is specifically the following:
the function may be approximated using taylor expansion such that the final objective function depends only on the first and second derivatives of each data point on the error function, thereby allowing the objective function to be approximately solved.
Specifically, in the solving, the two-classification model adopts a logistic regression mode, the S function is taken as a target function, and the final result is a probability value between 0 and 1. There are only two possibilities, namely yes or no, for whether the user is a hot-selling user and whether the user is a price comparison user, if the probability value is >0.5, the prediction result is regarded as yes, otherwise, no; there are more than two possibilities for each of the hotel star, breakfast type and bed type, so a model is used to predict probability values for each possibility of the hotel star, breakfast type and bed type. The two stars with the highest probability of the star level of the hotel are selected as the prediction results, and the possibility with the highest probability is selected as the only prediction result for the breakfast type and the bed type.
2) For the hotel price prediction model, too many hotels can be filtered due to too small specific price range, so that the specific price range has no obvious guiding significance for hotel screening and needs to obtain the price interval of the user target. If the XGB OST model classification method is adopted for prediction, the prediction result is often not flexible enough, and when the types of price intervals are increased, the training time of the XGB OST model is greatly increased, and the training accuracy is reduced. However, if a regression method is adopted to predict hotel prices preferred by users, only a certain prediction result can be obtained, and whether the result is reliable or not cannot be known, and no way is provided for extending the result to a corresponding price interval.
Therefore, in the embodiment, the XGBOOST model is combined with the quantile regression model, and a model for predicting upper and lower price limits at a certain confidence level is established by modifying the objective function of the XGBOOST model, so that a price interval corresponding to a high confidence level is obtained;
wherein, the formula of the upper limit function is: [ max (y-up,0) ] < Lambda + alpha > (y-up) < Lambda 2, the formula for the lower bound function is: [ max (down-y,0) ] < Lambda > + alpha (y-down) < Lambda > 2;
up is the upper limit predicted value, down is the lower limit predicted value, y is the actual user order price, and alpha and beta are adjustable coefficients respectively. [ max (y-up,0) ] < Lambda > and [ max (down-y,0) ] < Lambda > are used for limiting the actual price y of the user to meet the upper and lower limits as much as possible, otherwise, the max function is greater than 0; alpha (y-up) 2 and alpha (y-down) 2 are used for limiting the interval between the upper limit and the lower limit to be as small as possible, and the problem that the prediction range is too large and the actual effect is not achieved is avoided. And respectively calculating a first derivative and a second derivative of the objective function, and solving by adopting an algorithm to obtain a final prediction model, thereby obtaining a hotel price prediction result preferred by a user.
3) Hotel scoring and hotel review population
The method comprises the steps that an XGBOOST model is adopted to obtain a hotel score range of a hotel which a user historically lives in on an OTA platform according to user information, and a minimum value of the hotel score preferred by the user is obtained, namely when the score of the hotel is lower than the corresponding minimum value, the OTA platform does not sort the hotel; otherwise, the hotel is ranked and displayed.
The method comprises the steps that an XGBOOST model is adopted to obtain the hotel comment number range of a hotel which a user historically lives in on an OTA platform according to user information, and the minimum value of the hotel comment number preferred by the user is obtained, namely when the hotel comment number is lower than the corresponding minimum value, the OTA platform does not sequence the hotel; otherwise, the hotel is ranked and displayed.
According to the embodiment, user characteristic data are obtained through user information, then an XGBOOST model is adopted to carry out model training on each user characteristic data, and a user preference prediction model is obtained to obtain a first user preference prediction result; after the user logs in the OTA platform, the hotel screening result preferred by the user is obtained according to the first user preference prediction result of the current user, the screening result can be rapidly and effectively displayed to the user, the user experience is improved, and the purpose of personalized display is achieved.
Example 4
As shown in fig. 4, the hotel screening system based on user preference of the OTA platform of the present embodiment is further improved on the basis of embodiment 3, specifically:
the hotel screening system further comprises a city information acquisition module 6, a correction model acquisition module 7, a correction result acquisition module 8, a prediction probability acquisition module 9 and a judgment module 10.
The city information acquisition module 6 is used for acquiring city information of each city;
the city information comprises hotel distribution information and/or hotel historical order information;
the correction model obtaining module 7 is used for performing model training on the first user preference prediction result and the city information by adopting an XGBOOST model to obtain a city data correction model;
the correction result acquisition module 8 acquires the correction result of each hotel of each city according to the city data correction model;
and the correction result is used for representing the difference value between the hotel parameter of each hotel in each city and the hotel parameter corresponding to the first user preference prediction result.
The prediction probability obtaining module 9 is used for obtaining the user preference prediction probability according to the user preference prediction model and calling the judging module 10;
the judging module 10 is configured to judge whether the user preference prediction probability is greater than a set threshold, and if so, retain a first user preference prediction result corresponding to the user preference prediction probability; and if not, giving up the first user preference prediction result corresponding to the user preference prediction probability.
The prediction result obtaining module 4 is further configured to obtain a target city selected by the user, and obtain a second user preference prediction result according to a correction result corresponding to the target city and the first user preference prediction result;
the screening module 5 is further configured to obtain a hotel screening result of the OTA platform according to the second user preference prediction result.
And updating the second user preference prediction result every day, and calling the second user preference prediction result of the previous day on line.
In particular, because the distribution characteristics of hotels may differ under different cities. For example, the hotel price and the hotel star rating of a small city are lower than those of a large city, and the number of the hotel comment persons in a tourist city is generally larger than that in a non-tourist city. If a unified model is used to predict user preferences, some user characteristics may not be satisfied in some cities. If hotel screening is performed only by means of the prediction result obtained by the user preference prediction model, the number of hotels meeting requirements is extremely low or even none, so that user experience is influenced, and therefore correction of the user preference prediction result is required according to relevant data in each city.
The manner of correcting the hotels in each city according to the user characteristic data is shown in the following table:
user characteristic data | Correction method |
Hotel star level | Star grade shift size (as-1 represents: turning one star grade down) |
Hotel price | Size of price movement |
Hot-sell user | User restriction of whether to release hot pin |
Hotel brand | Whether to release hotel brand restrictions |
Hotel scoring | Hotel scoring movement size |
Number of people commenting in hotel | Mobile number of hotel comment people |
Breakfast type of hotel | Whether to release breakfast restrictions |
Hotel bed type | Limitation of whether to release bed |
Whether to compare prices | Whether to release the price comparison limit |
As shown in the table above, the difference between the actual hotel in each city and the preference prediction result of the user is obtained by combining the preference prediction result of the user, the hotel distribution information and the historical order information of the hotel, so that the screening result can be quickly and effectively displayed to the user, the user experience is improved, and the purpose of personalized display is achieved.
When a user opens a client of the OTA platform, the user firstly enters a hotel page and then enters a hotel search page; after the OTA platform acquires the city selected by the user, judging whether the OTA platform is provided with an intelligent screening button screening option or not, and if not, displaying default hotel sequencing; otherwise, the correction processing is carried out on the user preference prediction result and the correction result of the city on line, if the correction mode is numerical value movement, if the hotel star correction result is-1, and the hotel star of the user preference predicted by the user preference prediction model is 3-level or 4-level, the hotel star of the current city actually screened by the OTA platform is 2-level or 3-level; and if the correction mode is to release the limit, if the user predicted by the user preference prediction model is the hot sales user, the corresponding correction result is 1, namely the hot sales user limit is released, the OTA platform does not actually screen the hot sales hotels, and then the hotel sequencing is adjusted according to the user preference prediction result and the correction result of the city. When the user clicks the intelligent screening button screening item, a union set of screening conditions corresponding to all correction results is taken, hotels which do not accord with the screening conditions are filtered, and hotel sorting results which accord with the preference of the user are displayed to the user.
According to the method, user characteristic data are obtained through user information, then an XGBOOST model is adopted to carry out model training on each user characteristic data, and a user preference prediction model is obtained to obtain a user preference prediction result; meanwhile, combining the user preference prediction result with the city information of each city to obtain the correction result of each city; after the user logs in the OTA platform, the result is predicted according to the user preference of the current user, and the correction result corresponding to the target city selected by the user is directly called, so that the hotel screening result preferred by the user is obtained, the calculation workload of the OTA platform is reduced, meanwhile, the screening result can be rapidly and effectively displayed to the user, the user experience is improved, and the purpose of personalized display is achieved.
While specific embodiments of the invention have been described above, it will be appreciated by those skilled in the art that these are by way of example only, and that the scope of the invention is defined by the appended claims. Various changes and modifications to these embodiments may be made by those skilled in the art without departing from the spirit and scope of the invention, and these changes and modifications are within the scope of the invention.
Claims (6)
1. A hotel screening method based on user preference of an OTA platform is characterized by comprising the following steps:
s1, acquiring user information of each user;
the user information comprises at least one of user history screening information, user history browsing information and user history order information;
s2, acquiring user characteristic data according to the user information;
s3, performing model training on the user characteristic data based on the XGBOOST model to obtain a user preference prediction model, and obtaining a first user preference prediction result according to the user preference prediction model;
s4, obtaining a hotel screening result of the OTA platform according to the first user preference prediction result;
the user characteristic data comprises at least one of a hotel star rating, a hotel price, a hotel brand, a hotel score, a hotel comment number, a hotel breakfast type, a hotel bed type, a hot sales user and a price comparison user preferred by the user;
step S3 specifically includes:
respectively carrying out model training on the hotel star level, the hotel brand, the hotel breakfast type, the hotel bed type, the hot sales user and the price comparison user by adopting an XGB OST model to obtain a corresponding hotel star level prediction model, a hotel brand prediction model, a hotel breakfast type prediction model, a hotel bed type prediction model, a hot sales user prediction model and a price comparison user prediction model, and then obtaining a user-preferred hotel star level prediction result, a hotel brand prediction result, a hotel breakfast type prediction result, a hotel bed type prediction result, a hot sales user prediction result and a price comparison user prediction result;
performing model training on the hotel price by combining an XGBOOST model and a quantile regression model to obtain a hotel price prediction model, and then obtaining a hotel price prediction result preferred by a user;
acquiring a hotel grade range of a hotel which is historically checked in by a user on the OTA platform according to the user information by adopting an XGBOOST model, and acquiring a minimum value of the hotel grade preferred by the user;
obtaining the hotel comment number range of the hotel which the user has historically checked in on the OTA platform according to the user information by adopting an XGBOOST model, and obtaining the minimum value of the hotel comment number preferred by the user;
step S4 is preceded by:
s31, acquiring city information of each city;
the city information comprises hotel distribution information and/or hotel historical order information;
s32, performing model training on the first user preference prediction result and the city information by adopting an XGBOOST model to obtain a city data correction model;
s33, obtaining a correction result of each hotel of each city according to the city data correction model;
and the correction result is used for representing a difference value between the hotel parameter of each hotel in each city and the hotel parameter corresponding to the first user preference prediction result.
2. The hotel screening method based on user preferences for OTA platform according to claim 1, wherein step S4 specifically comprises:
and acquiring a target city selected by a user, acquiring a second user preference prediction result according to the correction result corresponding to the target city and the first user preference prediction result, and acquiring a hotel screening result of the OTA platform according to the second user preference prediction result.
3. The method for hotel screening based on user preferences for OTA platform of claim 1 wherein step S3 further comprises:
obtaining a user preference prediction probability according to the user preference prediction model, judging whether the user preference prediction probability is larger than a set threshold value, and if so, retaining a first user preference prediction result corresponding to the user preference prediction probability; and if not, giving up the first user preference prediction result corresponding to the user preference prediction probability.
4. A hotel screening system based on user preference of an OTA platform is characterized by comprising a user information acquisition module, a characteristic data acquisition module, a prediction model acquisition module, a prediction result acquisition module and a screening module;
the user information acquisition module is used for acquiring the user information of each user;
the user information comprises at least one of user history screening information, user history browsing information and user history order information;
the characteristic data acquisition module is used for acquiring user characteristic data according to the user information;
the prediction model acquisition module is used for carrying out model training on the user characteristic data based on an XGBOOST model, acquiring a user preference prediction model and calling the prediction result acquisition module;
the prediction result obtaining module is used for obtaining a first user preference prediction result according to the user preference prediction model;
the screening module is used for acquiring a hotel screening result of the OTA platform according to the first user preference prediction result;
the user characteristic data comprises at least one of a hotel star rating, a hotel price, a hotel brand, a hotel score, a hotel comment number, a hotel breakfast type, a hotel bed type, whether the user is a hot sales user and whether the user is a price comparison user preferred by the user;
the prediction model acquisition module is used for respectively carrying out model training on the hotel star level, the hotel brand, the hotel breakfast type, the hotel bed type, the hot sales user and the price comparison user by adopting an XGBOOST model to acquire a corresponding hotel star level prediction model, a hotel brand prediction model, a hotel breakfast type prediction model, a hotel bed type prediction model, a hot sales user prediction model and a price comparison user prediction model;
the prediction result acquisition module is used for respectively acquiring a hotel star-level prediction result, a hotel brand prediction result, a hotel breakfast type prediction result, a hotel bed type prediction result, a hot sale user prediction result and a price comparison user prediction result of user preference according to the hotel star-level prediction model, the hotel brand prediction model, the hotel breakfast type prediction model, the hot sale user prediction model and the price comparison user prediction model;
the prediction model acquisition module is also used for performing model training on the hotel price by combining an XGBOOST model and a quantile regression model to acquire a hotel price prediction model;
the prediction result obtaining module is also used for obtaining a hotel price prediction result preferred by the user according to the hotel price prediction model;
the prediction model acquisition module is also used for acquiring the hotel score range of the hotel which the user has historically checked in on the OTA platform according to the user information by adopting an XGBOOST model;
the prediction result acquisition module is further used for acquiring the minimum value of the hotel score preferred by the user according to the range of the hotel score;
the prediction model acquisition module is also used for acquiring the hotel comment number range of the hotel in the OTA platform, which is historically checked in by the user, by adopting an XGBOOST model according to the user information;
the prediction result acquisition module is also used for acquiring the minimum value of the hotel comment number preferred by the user according to the range of the hotel comment number;
the hotel screening system also comprises a city information acquisition module, a correction model acquisition module and a correction result acquisition module;
the city information acquisition module is used for acquiring city information of each city;
the city information comprises hotel distribution information and/or hotel historical order information;
the correction model acquisition module is used for respectively carrying out model training on the user preference prediction result and the city information by adopting an XGBOOST model to acquire a city data correction model;
the correction result acquisition module acquires the correction result of each hotel of each city according to the city data correction model;
and the correction result is used for representing a difference value between the hotel parameter of each hotel in each city and the hotel parameter corresponding to the first user preference prediction result.
5. The hotel screening system of claim 4, wherein the prediction result obtaining module is further configured to obtain a target city selected by the user, and obtain a second user preference prediction result according to the correction result corresponding to the target city and the first user preference prediction result;
and the screening module is also used for acquiring a hotel screening result of the OTA platform according to the second user preference prediction result.
6. The hotel screening system of the OTA platform based on user preferences of claim 4, further comprising a prediction probability acquisition module and a decision module;
the prediction probability obtaining module is used for obtaining the user preference prediction probability according to the user preference prediction model and calling the judging module;
the judging module is used for judging whether the user preference prediction probability is larger than a set threshold value, if so, a first user preference prediction result corresponding to the user preference prediction probability is reserved; and if not, giving up the first user preference prediction result corresponding to the user preference prediction probability.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810191566.5A CN108446351B (en) | 2018-03-08 | 2018-03-08 | Hotel screening method and system based on user preference of OTA platform |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810191566.5A CN108446351B (en) | 2018-03-08 | 2018-03-08 | Hotel screening method and system based on user preference of OTA platform |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108446351A CN108446351A (en) | 2018-08-24 |
CN108446351B true CN108446351B (en) | 2022-03-22 |
Family
ID=63193893
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810191566.5A Active CN108446351B (en) | 2018-03-08 | 2018-03-08 | Hotel screening method and system based on user preference of OTA platform |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108446351B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109509039A (en) * | 2018-12-29 | 2019-03-22 | 携程计算机技术(上海)有限公司 | Method for building up and system, the Method of Commodity Recommendation and system of price expectation model |
CN109801112B (en) * | 2019-02-02 | 2021-03-05 | 京东数字科技控股有限公司 | Method and device for calculating user score |
CN111831892A (en) * | 2019-04-19 | 2020-10-27 | 拉扎斯网络科技(上海)有限公司 | Information recommendation method, information recommendation device, server and storage medium |
CN111127109A (en) * | 2019-12-27 | 2020-05-08 | 携程计算机技术(上海)有限公司 | Prediction method of different city heat values, model training method and system |
CN111445280A (en) * | 2020-03-10 | 2020-07-24 | 携程计算机技术(上海)有限公司 | Model generation method, restaurant ranking method, system, device and medium |
CN111429204A (en) * | 2020-03-10 | 2020-07-17 | 携程计算机技术(上海)有限公司 | Hotel recommendation method, system, electronic equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104346762A (en) * | 2013-07-31 | 2015-02-11 | 同程网络科技股份有限公司 | Large OTA (open travel alliance) and hotel supplier butt joint communication method |
CN105005579A (en) * | 2015-05-28 | 2015-10-28 | 携程计算机技术(上海)有限公司 | Personalized sorting method and system of hotel room types in OTA (Online Travel Agent) website |
CN105812830A (en) * | 2016-03-11 | 2016-07-27 | 传成文化传媒(上海)有限公司 | Recommendation method and system of hotel service content |
CN106909659A (en) * | 2017-02-27 | 2017-06-30 | 携程旅游网络技术(上海)有限公司 | Hotel's sort method based on traffic convenience degree in OTA websites |
CN106934498A (en) * | 2017-03-14 | 2017-07-07 | 携程旅游网络技术(上海)有限公司 | The recommendation method and system of hotel's house type in OTA websites |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090163183A1 (en) * | 2007-10-04 | 2009-06-25 | O'donoghue Hugh | Recommendation generation systems, apparatus and methods |
US20140222518A1 (en) * | 2013-02-07 | 2014-08-07 | TravelClick, Inc. | Methods and systems for setting optimal hotel property prices |
CN106203746A (en) * | 2016-09-30 | 2016-12-07 | 携程计算机技术(上海)有限公司 | Hotel group divides and the method for requirement forecasting |
CN106709767A (en) * | 2017-03-07 | 2017-05-24 | 携程计算机技术(上海)有限公司 | Personalized recommendation method and system of OTA (online travel website) hotels |
CN107194774A (en) * | 2017-05-22 | 2017-09-22 | 携程旅游网络技术(上海)有限公司 | Personalized hotel's commending system and method in OTA websites |
CN107688662B (en) * | 2017-09-08 | 2020-10-30 | 携程计算机技术(上海)有限公司 | OTA hotel recommendation method and system |
-
2018
- 2018-03-08 CN CN201810191566.5A patent/CN108446351B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104346762A (en) * | 2013-07-31 | 2015-02-11 | 同程网络科技股份有限公司 | Large OTA (open travel alliance) and hotel supplier butt joint communication method |
CN105005579A (en) * | 2015-05-28 | 2015-10-28 | 携程计算机技术(上海)有限公司 | Personalized sorting method and system of hotel room types in OTA (Online Travel Agent) website |
CN105812830A (en) * | 2016-03-11 | 2016-07-27 | 传成文化传媒(上海)有限公司 | Recommendation method and system of hotel service content |
CN106909659A (en) * | 2017-02-27 | 2017-06-30 | 携程旅游网络技术(上海)有限公司 | Hotel's sort method based on traffic convenience degree in OTA websites |
CN106934498A (en) * | 2017-03-14 | 2017-07-07 | 携程旅游网络技术(上海)有限公司 | The recommendation method and system of hotel's house type in OTA websites |
Non-Patent Citations (2)
Title |
---|
S. A. Akinwumi ; T.V Omotosho ; M R. Usikalu ; M. O. Adewusi ; O. O..Atmospheric gases attenuation in West Africa.《2016 IEEE Radio and Antenna Days of the Indian Ocean (RADIO)》.2016,第1-2页. * |
国际酒店比价平台辅助系统的设计与实现;王双;《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》;20150715;第I138-488页 * |
Also Published As
Publication number | Publication date |
---|---|
CN108446351A (en) | 2018-08-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108446351B (en) | Hotel screening method and system based on user preference of OTA platform | |
CN103544663B (en) | The recommendation method of network open class, system and mobile terminal | |
CN105677831B (en) | Method and device for determining recommended merchants | |
CN109118336B (en) | Information recommendation method and device, computer equipment and storage medium | |
CN104462611B (en) | Modeling method, sort method and model building device, the collator of information sorting model | |
CN104992347B (en) | A kind of method and device of video matching advertisement | |
CN107507016A (en) | A kind of information push method and system | |
CN107341176B (en) | Sample weight setting method and device and electronic equipment | |
CN106709767A (en) | Personalized recommendation method and system of OTA (online travel website) hotels | |
CN114723528A (en) | Commodity personalized recommendation method and system based on knowledge graph | |
CN109740072B (en) | Hotel ordering method and system based on POI under OTA platform | |
CN109509039A (en) | Method for building up and system, the Method of Commodity Recommendation and system of price expectation model | |
CN106897892A (en) | Advertisement placement method and device | |
CN106445954B (en) | Business object display method and device | |
CN112149352B (en) | Prediction method for marketing activity clicking by combining GBDT automatic characteristic engineering | |
CN108804577B (en) | Method for estimating interest degree of information tag | |
CN105138508A (en) | Preference diffusion based context recommendation system | |
CN111722766A (en) | Multimedia resource display method and device | |
CN109559152A (en) | A kind of network marketing method, system and computer storage medium | |
CN107767153A (en) | A kind of data processing method and device | |
CN110020152B (en) | Application recommendation method and device | |
CN113407854A (en) | Application recommendation method, device and equipment and computer readable storage medium | |
CN110110143B (en) | Video classification method and device | |
CN111523035A (en) | Recommendation method, device, server and medium for APP browsing content | |
CN109299389B (en) | Tourist attraction recommendation method and device |
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 |