CN108446351A - The hotel's screening technique and system based on user preference of OTA platforms - Google Patents
The hotel's screening technique and system based on user preference of OTA platforms Download PDFInfo
- Publication number
- CN108446351A CN108446351A CN201810191566.5A CN201810191566A CN108446351A CN 108446351 A CN108446351 A CN 108446351A CN 201810191566 A CN201810191566 A CN 201810191566A CN 108446351 A CN108446351 A CN 108446351A
- Authority
- CN
- China
- Prior art keywords
- hotel
- user
- user preference
- result
- prediction
- 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.)
- Granted
Links
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 the hotel's screening technique and system based on user preference of a kind of OTA platforms, hotel's screening technique includes：S1, the user information for obtaining each user；S2, user characteristic data is obtained according to the user information；S3, model training is carried out to the user characteristic data based on XGBOOST models, obtains user preference prediction model, and the first user preference prediction result is obtained according to the user preference prediction model；S4, hotel's the selection result that OTA platforms are obtained according to the first user preference prediction result.The present invention can obtain hotel's the selection result based on user preference, be saved greatly the computing resource of OTA platforms, while can the selection result quickly and efficiently be showed user, and the user experience is improved, achieve the purpose that personalized displaying.
Description
Technical field
The present invention relates to the technical field of information processing of OTA (Online Travel Agency, online travel agency) platform,
More particularly to the hotel's screening technique and system based on user preference of a kind of OTA platforms.
Background technology
With increase of the OTA softwares in people's routine use frequency, more and more hotels have been also begun to and OTA companies
It cooperates.With increasing for hotel's quantity, on the one hand represent OTA softwares can provide to the user it is more abundant and good
Service also implies that user will need to face interminable hotel's list, takes more time and gone really with energy from another point of view
Surely the hotel needed.
Practice have shown that it is minority just to have the user of clear hotel's target since entering OTA softwares, and more users can
It can not select to meet its demand from numerous hotels immediately.Even if can if auxiliary by using screening function
Screening item can be faced excessively to cause to increase laborious degree or be not apparent from target and do not know how asking for selection screening
Topic.Therefore, how to be effectively directed to different user and show its hotel needed, save the quality time for user, have become
One important striving direction of OTA platforms.
Under normal circumstances, if user, without search key, does not also carry out it when entering hotel's list page
His any screening, OTA platforms can provide hotel's list of default sort to the user, and still, such sequence does not simultaneously meet user
True preference.If it is intended to accomplishing to carry out each user completely personalized displaying, although technically completely may be used
Row, but also imply that huge calculation amount, it is not that can accomplish in one move.Therefore, OTA is relatively low again there is calculation amount is found
It disclosure satisfy that the demand of this scheme of user preference.
Invention content
The technical problem to be solved by the present invention is in order to overcome exist in the prior art OTA platforms exist can not according to
Family preference carry out the defect of hotel's sequence, and it is an object of the present invention to provide a kind of OTA platforms hotel's screening technique based on user preference
And system.
The present invention is to solve above-mentioned technical problem by following technical proposals：
The present invention provides a kind of hotel's screening technique based on user preference of OTA platforms, hotel's screening technique packet
It includes：
S1, the user information for obtaining each user；
Wherein, the user information includes user's history filter information, user's history browsing information and user's history order
At least one of information；
S2, user characteristic data is obtained according to the user information；
S3, model training is carried out to the user characteristic data based on XGBOOST models (a kind of machine learning algorithm), obtained
User preference prediction model is taken, and the first user preference prediction result is obtained according to the user preference prediction model；
S4, hotel's the selection result that OTA platforms are obtained according to the first user preference prediction result.
Preferably, further including before step S4：
S31, the urban information for obtaining each city；
Wherein, the urban information includes hotel's distributed intelligence and/or hotel's History Order information；
S32, mould is carried out to the first user preference prediction result and the urban information using XGBOOST models respectively
Type training obtains Urban Data correction model；
S33, obtained according to the Urban Data correction model each city each hotel correction result；
Wherein, hotel's parameter in each hotel of the correction result for characterizing each city and first user are inclined
Difference between the corresponding hotel's parameter of good prediction result.
Preferably, step S4 is specifically included：
The target cities for obtaining user's selection are used according to the corresponding correction result in the target cities and described first
Family preference prediction result obtains second user preference prediction result, and obtains OTA according to the second user preference prediction result
Hotel's the selection result of platform.
Preferably, step S3 further includes：
User preference prediction probability is obtained according to the user preference prediction model, and judges that the user preference prediction is general
Whether rate is more than given threshold, if so, retaining the first user preference prediction knot corresponding with the user preference prediction probability
Fruit；If it is not, then abandoning the first user preference prediction result corresponding with the user preference prediction probability.
Preferably, the user characteristic data includes the Hotel Star of user preference, hotel's price, hotel brand, hotel
Scoring, hotel's comment number, hotel's breakfast type, hotel's bed-type, fast-selling at least one of user and rate of exchange user；
Step S3 is specifically included：
Using XGBOOST models respectively to the Hotel Star, hotel brand, hotel's breakfast type, described
Hotel's bed-type, the fast-selling user and the rate of exchange user carry out model training, obtain corresponding Hotel Star prediction model, wine
Shop brand prediction model, hotel's breakfast type prediction model, hotel's bed-type prediction model, fast-selling user in predicting model and the rate of exchange are used
Family prediction model, the Hotel Star prediction result, hotel brand prediction result, hotel's breakfast type for then obtaining user preference are pre-
Survey result, hotel's bed-type prediction result, fast-selling user in predicting result and rate of exchange user in predicting result；
It is combined with quantile estimate model using XGBOOST models and model training is carried out to hotel's price, obtained
Then price expectation model in hotel's obtains hotel's price expectation result of user preference；
The hotel moved according to user information acquisition user's history using XGBOOST models is in the OTA platforms
The range of hotel's scoring obtains the minimum value of hotel's scoring of user preference；
The hotel moved according to user information acquisition user's history using XGBOOST models is in the OTA platforms
The range of number is commented in hotel, obtains the minimum value of hotel's comment number of user preference.
The present invention also provides a kind of hotel's screening system based on user preference of OTA platforms, hotel's screening systems
Including User profile acquisition module, characteristic acquisition module, prediction model acquisition module, prediction result acquisition module and screening
Module；
The User profile acquisition module is used to obtain the user information of each user；
Wherein, the user information includes user's history filter information, user's history browsing information and user's history order
At least one of information；
The characteristic acquisition module is used to obtain user characteristic data according to the user information；
The prediction model acquisition module is used to carry out model instruction to the user characteristic data based on XGBOOST models
Practice, obtains user preference prediction model, and call the prediction result acquisition module；
The prediction result acquisition module is used to obtain the prediction of the first user preference according to the user preference prediction model
As a result；
Screen knot in the hotel that the screening module is used to obtain OTA platforms according to the first user preference prediction result
Fruit.
Preferably, hotel's screening system further includes urban information acquisition module, correction model acquisition module and amendment
As a result acquisition module；
The urban information acquisition module is used to obtain the urban information in each city；
Wherein, the urban information includes hotel's distributed intelligence and/or hotel's History Order information；
The correction model acquisition module is used for using XGBOOST models respectively to the user preference prediction result and institute
It states urban information and carries out model training, obtain Urban Data correction model；
The correction result acquisition module obtains each hotel in each city according to the Urban Data correction model
Correction result；
Wherein, hotel's parameter in each hotel of the correction result for characterizing each city and first user are inclined
Difference between the corresponding hotel's parameter of good prediction result.
Preferably, the prediction result acquisition module is additionally operable to obtain the target cities of user's selection, according to the target
The corresponding correction result in city and the first user preference prediction result obtain second user preference prediction result；
Screen knot in the hotel that the screening module is additionally operable to obtain OTA platforms according to the second user preference prediction result
Fruit.
Preferably, hotel's screening system further includes prediction probability acquisition module and judgment module；
The prediction probability acquisition module is used to obtain user preference prediction probability according to the user preference prediction model,
And call the judgment module；
The judgment module is for judging whether the user preference prediction probability is more than given threshold, if so, retaining
The first user preference prediction result corresponding with the user preference prediction probability；If it is not, then abandoning pre- with the user preference
Survey the corresponding first user preference prediction result of probability.
Preferably, the user characteristic data includes the Hotel Star of user preference, hotel's price, hotel brand, hotel
Scoring, hotel comment number, hotel's breakfast type, hotel's bed-type, whether be fast-selling user and whether be in rate of exchange user extremely
Few one kind；
The prediction model acquisition module is used for using XGBOOST models respectively to the Hotel Star, hotel's product
Board, hotel's breakfast type, hotel's bed-type, the fast-selling user and the rate of exchange user carry out model training, obtain
Corresponding Hotel Star prediction model, hotel brand prediction model, hotel's breakfast type prediction model, hotel's bed-type predict mould
Type, fast-selling user in predicting model and rate of exchange user in predicting model；
The prediction result acquisition module is used to predict mould according to the Hotel Star prediction model, the hotel brand
Type, hotel's breakfast type prediction model, hotel's bed-type prediction model, the fast-selling user in predicting model and the ratio
Valence user in predicting model obtains the Hotel Star prediction result, hotel brand prediction result, hotel's breakfast class of user preference respectively
Type prediction result, hotel's bed-type prediction result, fast-selling user in predicting result and rate of exchange user in predicting result；
The prediction model acquisition module is additionally operable to be combined to described with quantile estimate model using XGBOOST models
Hotel's price carries out model training, obtains hotel's price expectation model；
The prediction result acquisition module is additionally operable to obtain the hotel of user preference according to hotel's price expectation model
Price expectation result；
The prediction model acquisition module is additionally operable to obtain user's history according to the user information using XGBOOST models
The range that the hotel moved in scores in the hotel of the OTA platforms；
The hotel that the prediction result acquisition module is additionally operable to the range to score according to the hotel acquisition user preference is commented
The minimum value divided；
The prediction model acquisition module is additionally operable to obtain user's history according to the user information using XGBOOST models
The range of number is commented in the hotel of the OTA platforms in the hotel moved in；
The prediction result acquisition module is additionally operable to comment on the wine of the range acquisition user preference of number according to the hotel
Comment on the minimum value of number in shop.
The positive effect of the present invention is that：
The present invention obtains user characteristic data by user information, then carries out model training to each user characteristic data,
Obtain the first user preference prediction result；Meanwhile the urban information in each city is combined to obtain the first user preference prediction result
Take the correction result in each hotel in each city；After user logs in OTA platforms, predicted according to the user preference of active user
As a result, and directly invoke user selection the corresponding correction result in target cities, to obtain based on user preference hotel sieve
Choosing is as a result, be saved greatly the computing resource of OTA platforms, while can the selection result quickly and efficiently be showed use
Family, the user experience is improved, achievees the purpose that personalized displaying.
Description of the drawings
Fig. 1 is the flow chart of hotel's screening technique based on user preference of the embodiment of the present invention 1；
Fig. 2 is the flow chart of hotel's screening technique based on user preference of the embodiment of the present invention 2；
Fig. 3 is the module diagram of hotel's screening system based on user preference of the embodiment of the present invention 3；
Fig. 4 is the module diagram of hotel's screening system based on user preference of the embodiment of the present invention 4.
Specific implementation mode
It is further illustrated the present invention below by the mode of embodiment, but does not therefore limit the present invention to the reality
It applies among a range.
Embodiment 1
As shown in Figure 1, hotel's screening technique based on user preference of the OTA platforms of the present embodiment includes：
S101, the user information for obtaining each user；
Wherein, user information includes user's history filter information, user's history browsing information and user's history order information
At least one of；
S102, user characteristic data is obtained according to user information；
Wherein, user characteristic data includes the Hotel Star of user preference, hotel's price, hotel brand, hotel's scoring, wine
Number is commented on, hotel's breakfast type, hotel's bed-type, whether is fast-selling user and whether is at least one of rate of exchange user in shop.
S103, model training is carried out to user characteristic data using XGBOOST models, obtains user preference prediction model,
And the first user preference prediction result is obtained according to user preference prediction model；
Specifically, user characteristic data and user characteristics form are as shown in the table：
| User characteristic data | Characteristic formp |
| Hotel Star | User moves in highest 2 stars of frequency |
| Hotel's price | The price range of user's selection |
| Fast-selling user | Whether user is fast-selling user |
| Hotel brand | The hotel brand that user takes notice of |
| It scores in hotel | User takes notice of the minimum value of scoring |
| Comment on number in hotel | User takes notice of the minimum value of comment number |
| Hotel's breakfast type | The type of preferences of user's breakfast |
| Hotel's bed-type | The type of preferences of user's bed-type |
| Rate of exchange user | Whether user is rate of exchange user |
Wherein, some use are trained by selection of sampling from hotel's order data of the OTA platforms in the past period
Family, and these user's history order informations are extracted, obtain mesh of the corresponding characteristic value of each user characteristic data as training set
Mark label.Wherein, Hotel Star, hotel's price, hotel's breakfast type and hotel's bed-type can be obtained from the order data of user；
When it is positioned at the sales volume region of the front three of current city that user, which orders region, it is fast-selling user to define the user；Work as user
When the hotel of order is brand hotel, then defines the user and take notice of the brand hotel.
It is respectively established for above-mentioned 9 user characteristic datas, obtains corresponding user preference prediction model, further according to
User preference prediction result screens the hotel of OTA platforms.Specifically, the model training of 9 user characteristic datas is divided into
Three kinds of different types：
1) use XGBOOST models respectively to Hotel Star, hotel brand, hotel's breakfast type, hotel's bed-type, fast-selling use
Family and rate of exchange user carry out model training, obtain corresponding Hotel Star prediction model, hotel brand prediction model, hotel's breakfast
Type prediction model, hotel's bed-type prediction model, fast-selling user in predicting model and rate of exchange user in predicting model, then obtain user
The Hotel Star prediction result of preference, hotel brand prediction result, hotel's breakfast type prediction result, hotel's bed-type prediction knot
Fruit, fast-selling user in predicting result and rate of exchange user in predicting result.
Specifically, XGBOOST models are the modified versions to traditional GBDT algorithms (a kind of machine learning algorithm), have speed
The multiple advantages such as degree is fast and effect is good.It is different from GBDT algorithms, its object function of XGBOOST model modifications, the object function
Formula is specific as follows：
Wherein, Obj(t)It is object function,Company for calculating residual error sums it up,
In, yiFor the real goal value of i-th of sample,The predicted target values in iteration, f are taken turns for (t-1)t(xi) represent often
The score of a leaf node, ω (ft) it is regularization parameter, it is to adjust to ensure that model has higher robustness, constant
Save the arbitrary constant value of model.The function can carry out approximate processing using Taylor expansion so that final object function only relies on
In the first derivative and second dervative on error function of each data point, so that object function can be with approximate solution.
Wherein, giAnd hiRespectivelyFirst derivative and second dervative.
Specific in solution, wherein two disaggregated models use logistic regression mode, using S function as object function, finally
As a result the probability value between 0-1.For whether be fast-selling user and whether be rate of exchange user all only there are two types of may, i.e. "Yes"
Or "No", if probability value>0.5, then it is "Yes" to be considered as prediction result, otherwise is "No"；For Hotel Star, hotel's breakfast class
The hotels Xing He bed-type there is a possibility that be more than two kinds the case where, therefore using model to Hotel Star, hotel's breakfast type and
Each possibility of hotel's bed-type all carries out the prediction of probability value.Wherein, highest two stars of selection Hotel Star probability are made
For prediction result, for hotel's breakfast type and hotel's bed-type, then the highest possibility of select probability is as unique prediction knot
Fruit.
2) for hotel's price expectation model, since specific Price Range is too small, too many hotel can be filtered out, therefore have
The Price Range of body screens no essential meaning to hotel, needs the price range for obtaining ownership goal.If using XGBOOST
The method of category of model predicted, prediction result often underaction, and when price range type becomes more, XGBOOST moulds
The training time of type can be significantly increased, and training accuracy can also decrease.And it is inclined if carrying out user using the method returned
Good hotel's price expectation, then can only obtain a determining prediction result, and not can know that whether the result is reliable, also not have
There is method to expand to corresponding price range.
Therefore XGBOOST models are combined with quantile estimate model in the present embodiment, by XGBOOST models
Object function is modified, and the model of price bound of the prediction under confidence degree is established, to obtain high confidence level pair
The price range answered；
Wherein, the formula of upper limit function is：[max (y-up, 0)] ^ β+α (y-up) ^2, the formula of lower limit function are：[max
(down-y,0)]^β+α(y-down)^2；
Up is upper limit predicted value, and down is lower limit predicted value, and y is actual user's ordering price, and α and β are respectively adjustable
Coefficient.[max (y-up, 0)] ^ β and [max (down-y, 0)] ^ β need to meet as possible for limiting user's real price y
Between lower limit, otherwise max functions are more than 0；α (y-up) ^2 and α (y-down) ^2 is small as possible for limiting bound section, avoids
Estimation range does not have greatly very much the problem of actual effect.Single order by calculating separately object function is led leads with second order, brings algorithm into
Solve and obtain final prediction model, to obtain hotel's price expectation result of user preference.
3) number is commented in hotel's scoring and hotel
The hotel that user's history is moved in is obtained using XGBOOST models according to user information to score in the hotel of OTA platforms
Range, obtain the minimum value of hotel's scoring of user preference, when the scoring in the hotels Ji Dang is less than corresponding minimum value, then OTA
The hotels platform Bu Duigai are ranked up；Otherwise, which is ranked up and is shown.
The hotel that user's history is moved in is obtained using XGBOOST models according to user information to comment in the hotel of OTA platforms
The range of number obtains the minimum value of hotel's comment number of user preference, and the hotels Ji Dang comment on number and are less than corresponding minimum
When value, then the hotels OTA platforms Bu Duigai are ranked up；Otherwise, which is ranked up and is shown.
S104, hotel's the selection result that OTA platforms are obtained according to the first user preference prediction result.
The present embodiment obtains user characteristic data by user information, then using XGBOOST models to each user characteristics
Data carry out model training, obtain user preference prediction model and obtain the first user preference prediction result；It is put down when user logs in OTA
After platform, according to the first user preference prediction result of active user obtain user preference hotel's the selection result, can quickly and
The selection result is effectively showed into user, the user experience is improved, achievees the purpose that personalized displaying.
Embodiment 2
As shown in Fig. 2, hotel's screening technique based on user preference of the OTA platforms of the present embodiment is in embodiment 1
On the basis of be further improved, specifically：
Further include before step S104：
S1031, user preference prediction probability is obtained according to user preference prediction model, and judges user preference prediction probability
Whether it is more than given threshold, then retains the first user preference prediction result corresponding with user preference prediction probability；If it is not, then putting
Abandon the first user preference prediction result corresponding with user preference prediction probability；
S1032, the urban information for obtaining each city；
Wherein, urban information includes hotel's distributed intelligence and/or hotel's History Order information；
S1033, model instruction is carried out to the first user preference prediction result and urban information using XGBOOST models respectively
Practice, obtains Urban Data correction model；
S1034, obtained according to Urban Data correction model each city each hotel correction result；
Wherein, correction result is used to characterize hotel's parameter in each hotel in each city and the prediction of the first user preference is tied
Difference between the corresponding hotel's parameter of fruit.
Step S104 is specifically included：
S1041, the target cities for obtaining user's selection, according to the corresponding correction result in target cities and the first user preference
Prediction result obtains second user preference prediction result, and the hotel of OTA platforms is obtained according to second user preference prediction result
Then the selection result shows hotel's the selection result.
Specifically, because under different cities, the distribution characteristics in hotel may be different.Such as hotel's price of small city and
Hotel Star is less than metropolitan hotel's price and Hotel Star, and hotel's comment number of tourist city is generally higher than non-tourism city
Comment on number in the hotel in city.If predicted user preference using unified model, some user characteristics may be certain
Satisfaction is unable to get under city.If only the prediction result obtained by user preference prediction model carries out hotel's screening, then can
Cause satisfactory hotel's number extremely low or even without to influence user experience, it is therefore desirable to according under each city
Related data carries out the amendment of user preference prediction result.
The mode being modified to each hotel in each city according to user characteristic data is as shown in the table：
| User characteristic data | Correcting mode |
| Hotel Star | Star moves size (as -1 indicates：Turn a star down) |
| Hotel's price | Price moves size |
| Fast-selling user | Whether fast-selling user limitation is decontroled |
| Hotel brand | Whether hotel brand limitation is decontroled |
| It scores in hotel | The mobile size of hotel's scoring |
| Comment on number in hotel | It comments on number and moves size in hotel |
| Hotel's breakfast type | Whether breakfast limitation is decontroled |
| Hotel's bed-type | Whether bed-type limitation is decontroled |
| Rate of exchange user | Whether rate of exchange limitation is decontroled |
As shown above, it in conjunction with the first user preference prediction result, hotel's distributed intelligence and hotel's History Order information, obtains
The correction result for taking each hotel in each city, when the target cities of user's selection, then according to the corresponding amendment in target cities
As a result second user preference prediction result is obtained with the first user preference prediction result, and according to second user preference prediction result
Hotel's the selection result of OTA platforms is obtained, the selection result quickly and efficiently can be showed user, improve user by realization
Experience achievees the purpose that personalized displaying.
After user opens the client of OTA platforms, hotel's page can be entered first, subsequently into hotel's searched page；
After OTA platforms get the city of user's selection, judge whether OTA platforms are equipped with " intelligent screening " button and screen item, if
No, then the sequence of acquiescence hotel is shown；Otherwise place is modified to the correction result of user preference prediction result and city online
Reason, if correcting mode is the movement of numerical value, if Hotel Star correction result is -1, and the use of user preference prediction model prediction
The Hotel Star of family preference is 3 grades or 4 grades, then the Hotel Star of the practical current city filtered out of OTA platforms is 2 grades or 3 grades；
If correcting mode is to decontrol to limit, if the user of user preference prediction model prediction is fast-selling user, corresponding amendment
As a result it is 1, that is, decontrols fast-selling user's limitation, then OTA platforms does not actually screen fast-selling hotel, then predict to tie according to user preference
The correction result in fruit and city adjusts hotel's sequence.When the user clicks when " intelligent screening " button screening item, then all amendments are taken
As a result the union of corresponding screening conditions filters out the hotel for not meeting screening conditions, and to meet user inclined to user's displaying
Good hotel's ranking results.
The present invention obtains user characteristic data by user information, then using XGBOOST models to each user characteristics number
According to model training is carried out, obtains user preference prediction model and obtain user preference prediction result；Meanwhile user preference being predicted to tie
Fruit obtains the correction result in each city in conjunction with the urban information in each city；After user logs in OTA platforms, used according to current
The user preference prediction result at family, and the corresponding correction result in target cities of user's selection is directly invoked, to obtain user
Hotel's the selection result of preference reduces the amount of calculation of OTA platforms, while can be quickly and efficiently by the selection result exhibition
Show that, to user, the user experience is improved, achievees the purpose that personalized displaying.
Embodiment 3
As shown in figure 4, hotel's screening system based on user preference of the OTA platforms of the present embodiment includes that user information obtains
Modulus block 1, characteristic acquisition module 2, prediction model acquisition module 3, prediction result acquisition module 4 and screening module 5.
Specifically, User profile acquisition module 1 is used to obtain the user information of each user；
Wherein, user information includes user's history filter information, user's history browsing information and user's history order information
At least one of；
Characteristic acquisition module 2 is used to obtain user characteristic data according to user information；
Wherein, user characteristic data includes the Hotel Star of user preference, hotel's price, hotel brand, hotel's scoring, wine
Number is commented on, hotel's breakfast type, hotel's bed-type, whether is fast-selling user and whether is at least one of rate of exchange user in shop.
Prediction model acquisition module 3 is used to carry out model training to user characteristic data based on XGBOOST models, obtains and uses
Family preference prediction model, and call prediction result acquisition module 4；
Prediction result acquisition module 4 is used to obtain the first user preference prediction result according to user preference prediction model；
Screening module 5 is used to obtain hotel's the selection result of OTA platforms according to the first user preference prediction result.
Specifically, prediction model acquisition module 3 is used for using XGBOOST models respectively to Hotel Star, hotel brand, wine
Shop breakfast type, hotel's bed-type, fast-selling user and rate of exchange user carry out model training, obtain corresponding Hotel Star prediction mould
Type, hotel brand prediction model, hotel's breakfast type prediction model, hotel's bed-type prediction model, fast-selling user in predicting model and
Rate of exchange user in predicting model；
Prediction result acquisition module 4 is used for according to Hotel Star prediction model, hotel brand prediction model, hotel's breakfast class
It is inclined that type prediction model, hotel's bed-type prediction model, fast-selling user in predicting model and rate of exchange user in predicting model obtain user respectively
Good Hotel Star prediction result, hotel brand prediction result, hotel's breakfast type prediction result, hotel's bed-type prediction result,
Fast-selling user in predicting result and rate of exchange user in predicting result；
Prediction model acquisition module 3 is additionally operable to be combined to hotel's valence with quantile estimate model using XGBOOST models
Lattice carry out model training, obtain hotel's price expectation model；
Prediction result acquisition module 4 is additionally operable to obtain hotel's price expectation of user preference according to hotel's price expectation model
As a result；
Prediction model acquisition module 3 is additionally operable to obtain what user's history was moved according to user information using XGBOOST models
The range that hotel scores in the hotel of OTA platforms；
Prediction result acquisition module 4 is additionally operable to obtain the minimum of hotel's scoring of user preference according to the range that hotel scores
Value；
Prediction model acquisition module 3 is additionally operable to obtain what user's history was moved according to user information using XGBOOST models
The range of number is commented in the hotel of OTA platforms in hotel；
Comment on people in the hotel that prediction result acquisition module 4 is additionally operable to the range acquisition user preference for commenting on number according to hotel
Several minimum values.
Wherein, user characteristic data and user characteristics form are as shown in the table：
| User characteristic data | Characteristic formp |
| Hotel Star | User moves in highest 2 stars of frequency |
| Hotel's price | The price range of user's selection |
| Fast-selling user | Whether user is fast-selling user |
| Hotel brand | Whether user is hotel brand |
| It scores in hotel | User takes notice of the minimum value of scoring |
| Comment on number in hotel | User takes notice of the minimum value of comment number |
| Hotel's breakfast type | The type of preferences of user's breakfast |
| Hotel's bed-type | The type of preferences of user's bed-type |
| Rate of exchange user | Whether user is rate of exchange user |
Wherein, some use are trained by selection of sampling from hotel's order data of the OTA platforms in the past period
Family, and these user's history order informations are extracted, obtain mesh of the corresponding characteristic value of each user characteristic data as training set
Mark label.Wherein, Hotel Star, hotel's price, breakfast type and bed-type can be obtained from the order data of user；When user orders
When purchasing region as positioned at the sales volume region of the front three of current city, it is fast-selling user to define the user；When the wine that user orders
When shop is brand hotel, then it is brand hotel user to define the user.
It is respectively established for above-mentioned 9 user characteristic datas, obtains corresponding user preference prediction model, further according to
User preference prediction result screens the hotel of OTA platforms.Specifically, the model training of 9 user characteristic datas is divided into
Three kinds of different types：
1) use XGBOOST models respectively to Hotel Star, hotel brand, hotel's breakfast type, hotel's bed-type, fast-selling use
Family and rate of exchange user carry out model training, obtain corresponding Hotel Star prediction model, hotel brand prediction model, hotel's breakfast
Type prediction model, hotel's bed-type prediction model, fast-selling user in predicting model and rate of exchange user in predicting model, then obtain user
The Hotel Star prediction result of preference, hotel brand prediction result, hotel's breakfast type prediction result, hotel's bed-type prediction knot
Fruit, fast-selling user in predicting result and rate of exchange user in predicting result.
Specifically, XGBOOST models are the modified versions to traditional GBDT algorithms, and it is multinomial to have that speed is fast and effect is good etc.
Advantage.Different from GBDT algorithms, XGBOOST model modifications its object function, the objective function Equation are specific as follows：
The function can carry out approximate processing using Taylor expansion so that final object function only depends on each data point
The first derivative and second dervative on error function so that object function can be with approximate solution.
Specific in solution, wherein two disaggregated models use logistic regression mode, using S function as object function, finally
As a result the probability value between 0-1.For whether be fast-selling user and whether be rate of exchange user all only there are two types of may, i.e. "Yes"
Or "No", if probability value>0.5, then it is "Yes" to be considered as prediction result, otherwise is "No"；For Hotel Star, breakfast type and
Bed-type there is a possibility that the case where being more than two kinds, therefore can to each of Hotel Star, breakfast type and bed-type using model
Energy property all carries out the prediction of probability value.Wherein, select highest two stars of Hotel Star probability as prediction result, for morning
Type of eating and bed-type, then the highest possibility of select probability is as unique prediction result.
2) for hotel's price expectation model, since specific Price Range is too small, too many hotel can be filtered out, therefore have
The Price Range of body screens no apparent directive significance to hotel, needs the price range for obtaining ownership goal.If using
The method of XGBOOST categories of model predicted, prediction result often underaction, and when price range type becomes more,
The training time of XGBOOST models can be significantly increased, and training accuracy can also decrease.And if using return method into
Hotel's price expectation of row user preference can only then obtain a determining prediction result, and whether not can know that the result
Reliably, it also has no idea to expand to corresponding price range.
Therefore XGBOOST models are combined with quantile estimate model in the present embodiment, by XGBOOST models
Object function is modified, and the model of price bound of the prediction under confidence degree is established, to obtain high confidence level pair
The price range answered；
Wherein, the formula of upper limit function is：[max (y-up, 0)] ^ β+α (y-up) ^2, the formula of lower limit function are：[max
(down-y,0)]^β+α(y-down)^2；
Up is upper limit predicted value, and down is lower limit predicted value, and y is actual user's ordering price, and α and β are respectively adjustable
Coefficient.[max (y-up, 0)] ^ β and [max (down-y, 0)] ^ β need to meet as possible for limiting user's real price y
Between lower limit, otherwise max functions are more than 0；α (y-up) ^2 and α (y-down) ^2 is small as possible for limiting bound section, avoids
Estimation range does not have greatly very much the problem of actual effect.Single order by calculating separately object function is led leads with second order, brings algorithm into
Solve and obtain final prediction model, to obtain hotel's price expectation result of user preference.
3) number is commented in hotel's scoring and hotel
The hotel that user's history is moved in is obtained using XGBOOST models according to user information to score in the hotel of OTA platforms
Range, obtain the minimum value of hotel's scoring of user preference, when the scoring in the hotels Ji Dang is less than corresponding minimum value, then OTA
The hotels platform Bu Duigai are ranked up；Otherwise, which is ranked up and is shown.
The hotel that user's history is moved in is obtained using XGBOOST models according to user information to comment in the hotel of OTA platforms
The range of number obtains the minimum value of hotel's comment number of user preference, and the hotels Ji Dang comment on number and are less than corresponding minimum
When value, then the hotels OTA platforms Bu Duigai are ranked up；Otherwise, which is ranked up and is shown.
The present embodiment obtains user characteristic data by user information, then using XGBOOST models to each user characteristics
Data carry out model training, obtain user preference prediction model and obtain the first user preference prediction result；It is put down when user logs in OTA
After platform, according to the first user preference prediction result of active user obtain user preference hotel's the selection result, can quickly and
The selection result is effectively showed into user, the user experience is improved, achievees the purpose that personalized displaying.
Embodiment 4
As shown in figure 4, hotel's screening system based on user preference of the OTA platforms of the present embodiment is in embodiment 3
On the basis of be further improved, specifically：
Hotel's screening system further includes urban information acquisition module 6, correction model acquisition module 7, correction result acquisition mould
Block 8, prediction probability acquisition module 9 and judgment module 10.
Urban information acquisition module 6 is used to obtain the urban information in each city；
Wherein, urban information includes hotel's distributed intelligence and/or hotel's History Order information；
Correction model acquisition module 7 is used for using XGBOOST models respectively to the first user preference prediction result and city
Information carries out model training, obtains Urban Data correction model；
Correction result acquisition module 8 obtains the amendment knot in each hotel in each city according to Urban Data correction model
Fruit；
Wherein, correction result is used to characterize hotel's parameter in each hotel in each city and the prediction of the first user preference is tied
Difference between the corresponding hotel's parameter of fruit.
Prediction probability acquisition module 9 is used to obtain user preference prediction probability according to user preference prediction model, and calls
Judgment module 10；
Judgment module 10 is for judging whether user preference prediction probability is more than given threshold, if so, reservation and user
The corresponding first user preference prediction result of preference prediction probability；If it is not, then abandoning corresponding with user preference prediction probability
One user preference prediction result.
Prediction result acquisition module 4 is additionally operable to obtain the target cities of user's selection, according to the corresponding amendment in target cities
As a result second user preference prediction result is obtained with the first user preference prediction result；
Screening module 5 is additionally operable to obtain hotel's the selection result of OTA platforms according to second user preference prediction result.
Wherein, second user preference prediction result updates daily, and the second user preference prediction knot of the previous day is called on line
Fruit.
Specifically, because under different cities, the distribution characteristics in hotel may be different.Such as hotel's price of small city and
Hotel Star is less than metropolitan hotel's price and Hotel Star, and hotel's comment number of tourist city is generally higher than non-tourism city
Comment on number in the hotel in city.If predicted user preference using unified model, some user characteristics may be certain
Satisfaction is unable to get under city.If only the prediction result obtained by user preference prediction model carries out hotel's screening, then can
Cause satisfactory hotel's number extremely low or even without to influence user experience, it is therefore desirable to according under each city
Related data carries out the amendment of user preference prediction result.
The mode being modified to the hotel in each city according to user characteristic data is as shown in the table：
| User characteristic data | Correcting mode |
| Hotel Star | Star moves size (as -1 indicates：Turn a star down) |
| Hotel's price | Price moves size |
| Fast-selling user | Whether fast-selling user limitation is decontroled |
| Hotel brand | Whether hotel brand limitation is decontroled |
| It scores in hotel | The mobile size of hotel's scoring |
| Comment on number in hotel | It comments on number and moves size in hotel |
| Hotel's breakfast type | Whether breakfast limitation is decontroled |
| Hotel's bed-type | Whether bed-type limitation is decontroled |
| Whether the rate of exchange | Whether rate of exchange limitation is decontroled |
As shown above, it in conjunction with user preference prediction result, hotel's distributed intelligence and hotel's History Order information, obtains real
The difference in hotel and user preference prediction result under each city in border, realization can quickly and efficiently show the selection result
To user, the user experience is improved, achievees the purpose that personalized displaying.
After user opens the client of OTA platforms, hotel's page can be entered first, subsequently into hotel's searched page；
After OTA platforms get the city of user's selection, judge whether OTA platforms are equipped with " intelligent screening " button and screen item, if
No, then the sequence of acquiescence hotel is shown；Otherwise place is modified to the correction result of user preference prediction result and city online
Reason, if correcting mode is the movement of numerical value, if Hotel Star correction result is -1, and the use of user preference prediction model prediction
The Hotel Star of family preference is 3 grades or 4 grades, then the Hotel Star of the practical current city filtered out of OTA platforms is 2 grades or 3 grades；
If correcting mode is to decontrol to limit, if the user of user preference prediction model prediction is fast-selling user, corresponding amendment
As a result it is 1, that is, decontrols fast-selling user's limitation, then OTA platforms does not actually screen fast-selling hotel, then predict to tie according to user preference
The correction result in fruit and city adjusts hotel's sequence.When the user clicks when " intelligent screening " button screening item, then all amendments are taken
As a result the union of corresponding screening conditions filters out the hotel for not meeting screening conditions, and to meet user inclined to user's displaying
Good hotel's ranking results.
The present invention obtains user characteristic data by user information, then using XGBOOST models to each user characteristics number
According to model training is carried out, obtains user preference prediction model and obtain user preference prediction result；Meanwhile user preference being predicted to tie
Fruit obtains the correction result in each city in conjunction with the urban information in each city；After user logs in OTA platforms, used according to current
The user preference prediction result at family, and the corresponding correction result in target cities of user's selection is directly invoked, to obtain user
Hotel's the selection result of preference reduces the amount of calculation of OTA platforms, while can be quickly and efficiently by the selection result exhibition
Show that, to user, the user experience is improved, achievees the purpose that personalized displaying.
Although specific embodiments of the present invention have been described above, it will be appreciated by those of skill in the art that these
It is merely illustrative of, protection scope of the present invention is defined by the appended claims.Those skilled in the art is not carrying on the back
Under the premise of from the principle and substance of the present invention, various changes or modifications can be made to these embodiments, but these are changed
Protection scope of the present invention is each fallen with modification.
Claims (10)
1. a kind of hotel's screening technique based on user preference of OTA platforms, which is characterized in that hotel's screening technique packet
It includes：
S1, the user information for obtaining each user；
Wherein, the user information includes user's history filter information, user's history browsing information and user's history order information
At least one of；
S2, user characteristic data is obtained according to the user information；
S3, model training is carried out to the user characteristic data based on XGBOOST models, obtains user preference prediction model, and
The first user preference prediction result is obtained according to the user preference prediction model；
S4, hotel's the selection result that OTA platforms are obtained according to the first user preference prediction result.
2. hotel's screening technique based on user preference of OTA platforms as described in claim 1, which is characterized in that step S4
Further include before：
S31, the urban information for obtaining each city；
Wherein, the urban information includes hotel's distributed intelligence and/or hotel's History Order information；
S32, model instruction is carried out to the first user preference prediction result and the urban information using XGBOOST models respectively
Practice, obtains Urban Data correction model；
S33, obtained according to the Urban Data correction model each city each hotel correction result；
Wherein, hotel's parameter in each hotel of the correction result for characterizing each city and first user preference are pre-
Survey the difference between the corresponding hotel's parameter of result.
3. hotel's screening technique based on user preference of OTA platforms as claimed in claim 2, which is characterized in that step S4
It specifically includes：
The target cities of user's selection are obtained, it is inclined according to the corresponding correction result in the target cities and first user
Good prediction result obtains second user preference prediction result, and obtains OTA platforms according to the second user preference prediction result
Hotel's the selection result.
4. hotel's screening technique based on user preference of OTA platforms as described in claim 1, which is characterized in that step S3
Further include：
User preference prediction probability is obtained according to the user preference prediction model, and judges that the user preference prediction probability is
It is no to be more than given threshold, if so, retaining the first user preference prediction result corresponding with the user preference prediction probability；If
It is no, then abandon the first user preference prediction result corresponding with the user preference prediction probability.
5. hotel's screening technique based on user preference of OTA platforms as described in claim 1, which is characterized in that the use
Family characteristic includes the Hotel Star of user preference, hotel's price, hotel brand, hotel's scoring, hotel's comment number, hotel
Breakfast type, hotel's bed-type, fast-selling at least one of user and rate of exchange user；
Step S3 is specifically included：
Using XGBOOST models respectively to the Hotel Star, the hotel brand, hotel's breakfast type, the hotel
Bed-type, the fast-selling user and the rate of exchange user carry out model training, obtain corresponding Hotel Star prediction model, hotel's product
Board prediction model, hotel's breakfast type prediction model, hotel's bed-type prediction model, fast-selling user in predicting model and rate of exchange user are pre-
Model is surveyed, the Hotel Star prediction result, hotel brand prediction result, hotel's breakfast type prediction knot of user preference are then obtained
Fruit, hotel's bed-type prediction result, fast-selling user in predicting result and rate of exchange user in predicting result；
It is combined with quantile estimate model using XGBOOST models and model training is carried out to hotel's price, obtain hotel
Then price expectation model obtains hotel's price expectation result of user preference；
The hotel moved according to user information acquisition user's history using XGBOOST models is in the hotel of the OTA platforms
The range of scoring obtains the minimum value of hotel's scoring of user preference；
The hotel moved according to user information acquisition user's history using XGBOOST models is in the hotel of the OTA platforms
The range of comment number obtains the minimum value of hotel's comment number of user preference.
6. a kind of hotel's screening system based on user preference of OTA platforms, which is characterized in that hotel's screening system includes
User profile acquisition module, characteristic acquisition module, prediction model acquisition module, prediction result acquisition module and screening mould
Block；
The User profile acquisition module is used to obtain the user information of each user；
Wherein, the user information includes user's history filter information, user's history browsing information and user's history order information
At least one of；
The characteristic acquisition module is used to obtain user characteristic data according to the user information；
The prediction model acquisition module is used to carry out model training to the user characteristic data based on XGBOOST models, obtains
User preference prediction model is taken, and calls the prediction result acquisition module；
The prediction result acquisition module is used to obtain the first user preference prediction result according to the user preference prediction model；
The screening module is used to obtain hotel's the selection result of OTA platforms according to the first user preference prediction result.
7. hotel's screening system based on user preference of OTA platforms as claimed in claim 6, which is characterized in that the wine
Shop screening system further includes urban information acquisition module, correction model acquisition module and correction result acquisition module；
The urban information acquisition module is used to obtain the urban information in each city；
Wherein, the urban information includes hotel's distributed intelligence and/or hotel's History Order information；
The correction model acquisition module is used for using XGBOOST models respectively to the user preference prediction result and the city
City's information carries out model training, obtains Urban Data correction model；
The correction result acquisition module obtains the amendment in each hotel in each city according to the Urban Data correction model
As a result；
Wherein, hotel's parameter in each hotel of the correction result for characterizing each city and first user preference are pre-
Survey the difference between the corresponding hotel's parameter of result.
8. hotel's screening system based on user preference of OTA platforms as claimed in claim 7, which is characterized in that described pre-
It surveys result acquisition module to be additionally operable to obtain the target cities of user's selection, according to the corresponding correction result in the target cities
Second user preference prediction result is obtained with the first user preference prediction result；
The screening module is additionally operable to obtain hotel's the selection result of OTA platforms according to the second user preference prediction result.
9. hotel's screening system based on user preference of OTA platforms as claimed in claim 6, which is characterized in that the wine
Shop screening system further includes prediction probability acquisition module and judgment module；
The prediction probability acquisition module is used to obtain user preference prediction probability according to the user preference prediction model, and adjusts
With the judgment module；
The judgment module is for judging whether the user preference prediction probability is more than given threshold, if so, reservation and institute
State the corresponding first user preference prediction result of user preference prediction probability；If it is not, then abandoning general with user preference prediction
The corresponding first user preference prediction result of rate.
10. hotel's screening system based on user preference of OTA platforms as claimed in claim 6, which is characterized in that the use
Family characteristic includes the Hotel Star of user preference, hotel's price, hotel brand, hotel's scoring, hotel's comment number, hotel
Whether whether breakfast type, hotel's bed-type are fast-selling user and are at least one of rate of exchange user；
The prediction model acquisition module is used for using XGBOOST models respectively to the Hotel Star, the hotel brand, institute
It states hotel's breakfast type, hotel's bed-type, the fast-selling user and the rate of exchange user and carries out model training, obtain corresponding
Hotel Star prediction model, hotel brand prediction model, hotel's breakfast type prediction model, hotel's bed-type prediction model, fast sale
User in predicting model and rate of exchange user in predicting model；
The prediction result acquisition module is used for according to the Hotel Star prediction model, the hotel brand prediction model, institute
Hotel's breakfast type prediction model, hotel's bed-type prediction model, the fast-selling user in predicting model and the rate of exchange are stated to use
The Hotel Star prediction result, hotel brand prediction result, hotel's breakfast type that family prediction model obtains user preference respectively are pre-
Survey result, hotel's bed-type prediction result, fast-selling user in predicting result and rate of exchange user in predicting result；
The prediction model acquisition module is additionally operable to be combined to the hotel with quantile estimate model using XGBOOST models
Price carries out model training, obtains hotel's price expectation model；
The prediction result acquisition module is additionally operable to obtain hotel's price of user preference according to hotel's price expectation model
Prediction result；
The prediction model acquisition module is additionally operable to be moved according to user information acquisition user's history using XGBOOST models
Hotel the OTA platforms hotel score range；
The prediction result acquisition module is additionally operable to hotel's scoring of the range acquisition user preference to score according to the hotel
Minimum value；
The prediction model acquisition module is additionally operable to be moved according to user information acquisition user's history using XGBOOST models
Hotel the OTA platforms hotel comment on number range；
The hotel that the prediction result acquisition module is additionally operable to the range acquisition user preference for commenting on number according to the hotel is commented
By the minimum value of number.
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 true CN108446351A (en) | 2018-08-24 |
| CN108446351B 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) |
Cited By (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 |
| CN109801112A (en) * | 2019-02-02 | 2019-05-24 | 京东数字科技控股有限公司 | The method and apparatus for calculating user's score value |
| 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 |
| 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 |
| 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 |
| 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 |
Citations (11)
| 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 |
| 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 |
| 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 |
| 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 |
| 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 |
| CN107688662A (en) * | 2017-09-08 | 2018-02-13 | 携程计算机技术（上海）有限公司 | The recommendation method and system in OTA hotels |
-
2018
- 2018-03-08 CN CN201810191566.5A patent/CN108446351B/en active Active
Patent Citations (11)
| 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 |
| 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 |
| 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 |
| 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 |
| 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 |
| 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 |
| 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 |
| CN107688662A (en) * | 2017-09-08 | 2018-02-13 | 携程计算机技术（上海）有限公司 | The recommendation method and system in OTA hotels |
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)》 * |
| 王双: "国际酒店比价平台辅助系统的设计与实现", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
Cited By (7)
| 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 |
| CN109801112A (en) * | 2019-02-02 | 2019-05-24 | 京东数字科技控股有限公司 | The method and apparatus for calculating user's score value |
| 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 |
| 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 |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108446351B (en) | 2022-03-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108446351A (en) | The hotel's screening technique and system based on user preference of OTA platforms | |
| CN102279851B (en) | Intelligent navigation method, device and system | |
| CN106934498A (en) | The recommendation method and system of hotel's house type in OTA websites | |
| US20230043174A1 (en) | Method for pushing anchor information, computer device, and storage medium | |
| CN103218719B (en) | A kind of e-commerce website air navigation aid and system | |
| CN107424043A (en) | A kind of Products Show method and device, electronic equipment | |
| CN107315841A (en) | A kind of information search method, apparatus and system | |
| CN102841946B (en) | Commodity data retrieval ordering and Method of Commodity Recommendation and system | |
| CN105760400B (en) | A kind of PUSH message sort method and device based on search behavior | |
| CN108172301A (en) | A kind of mosquito matchmaker's epidemic Forecasting Methodology and system based on gradient boosted tree | |
| CN103970850B (en) | Site information recommends method and system | |
| CN107291841A (en) | A kind of method and system based on position and the social target of user's portrait intelligent Matching | |
| CN105718184A (en) | Data processing method and apparatus | |
| CN107274242A (en) | A kind of Method of Commodity Recommendation based on association analysis algorithm | |
| CN107194774A (en) | Personalized hotel's commending system and method in OTA websites | |
| CN106709767A (en) | Personalized recommendation method and system of OTA (online travel website) hotels | |
| CN103295145A (en) | Mobile phone advertising method based on user consumption feature vector | |
| CN109509039A (en) | Method for building up and system, the Method of Commodity Recommendation and system of price expectation model | |
| CN104239465A (en) | Scene information-based searching method and device | |
| CN106294661A (en) | A kind of extended search method and device | |
| CN109740072B (en) | Hotel ordering method and system based on POI under OTA platform | |
| CN109598534A (en) | Information determines method and device, electronic equipment and storage medium | |
| CN107391582A (en) | The information recommendation method of user preference similarity is calculated based on context ontology tree | |
| CN115203311B (en) | Industry data analysis mining method and system based on data brain | |
| CN103902549B (en) | The method and apparatus for searching for data sorting, the method and apparatus of data search |
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 |