CN109740072B - Hotel ordering method and system based on POI under OTA platform - Google Patents
Hotel ordering method and system based on POI under OTA platform Download PDFInfo
- Publication number
- CN109740072B CN109740072B CN201811638493.6A CN201811638493A CN109740072B CN 109740072 B CN109740072 B CN 109740072B CN 201811638493 A CN201811638493 A CN 201811638493A CN 109740072 B CN109740072 B CN 109740072B
- Authority
- CN
- China
- Prior art keywords
- hotel
- user
- hotels
- recommended
- ordering
- 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
Abstract
The invention discloses a hotel ordering method and system based on POI under an OTA platform, wherein the method comprises the following steps: acquiring a region where the POI is located; acquiring all optional hotels; calculating the distance between each selectable hotel and the POI, and setting the selectable hotels with the distance smaller than the distance threshold as first hotels; acquiring user dimension information of a user to be recommended; calculating preference degree of users to be recommended to each hotel under the OTA platform; acquiring hotel dimension information of each first hotel; calculating the probability of each first hotel to be ordered by the user to be recommended by using the prediction model; ordering the ordering probability of each first hotel for the user to be recommended; and outputting the ordered result. According to the online hotel ordering method and the online hotel ordering system, through a machine learning algorithm, based on information of users and hotels, the probability of the users for ordering the hotels is predicted in real time, so that intelligent ordering of online hotels is realized, personalized ordering can be realized for different users, sales of the hotels can be improved, and brand image of OTA is improved.
Description
Technical Field
The invention relates to the field of OTA (online travel agency), in particular to a hotel ordering method and system based on POI (point of interest) under an OTA platform.
Background
In the OTA industry, a hotel ordering method after POI information selected by a user is mainly a method for ordering from small to large based on the distance from a hotel to the POI. There is also a sorting method based on sales of hotels, which considers the popularity of hotels, and ranks the hotels with high sales in front and low sales in back to facilitate the user to make a selection, in practice, sorting is performed according to different OTA platforms and other kinds of sorting modes, such as sorting according to the price of hotels, sorting according to the user collection condition, and so on.
Disclosure of Invention
The technical problem to be solved by the invention is to overcome the defect that in the prior art, the OTA hotel ordering mode is single and does not meet the actual demands of users, and provide a POI-based hotel ordering method and system under an OTA platform.
The invention solves the technical problems by the following technical scheme:
the invention provides a hotel ordering method based on POI under an OTA platform, which comprises the following steps:
acquiring a region where a POI is located according to the POI input by a user to be recommended;
Acquiring all selectable hotels in the region;
calculating the distance between each selectable hotel and the POI, and setting the selectable hotels with the distance smaller than a distance threshold as first hotels;
acquiring user dimension information of the user to be recommended;
calculating preference degree of the user to be recommended to each hotel under the OTA platform;
acquiring hotel dimension information of each first hotel;
inputting the user dimension information of the user to be recommended, the hotel dimension information of each first hotel and the preference of the user to be recommended to each hotel into a prediction model, and calculating the probability of the user to be recommended to place an order to each first hotel by using the prediction model;
ordering the probability of each first hotel to be ordered by the user to be recommended;
and outputting the ordered result.
Preferably, the user dimension information includes basic information of the user, and the user dimension information further includes at least one of information of an order of a hotel ordered by the user in the past, information of a hotel browsed by the user, and information of a hotel collected by the user;
The hotel dimension information comprises basic information of the hotel, and further comprises at least one of price information of the hotel, good grade information of the hotel and information of the number of times the hotel is collected.
Preferably, the step of calculating the preference of the user to be recommended to each hotel under the OTA platform includes:
obtaining hidden variables of all users in the OTA platform and hidden variables of all hotels in the OTA platform by using a latent semantic model;
obtaining hidden variables of the users to be recommended according to the hidden variables of all the users;
and calculating the preference degree of the user to be recommended to each hotel by adopting the following formula:
wherein lfm is ij Representing the preference degree of the user i to be recommended to each hotel j, u ik A kth dimension value, h, representing a vector of hidden variables of the user i to be recommended jk And the kth dimension value of the vector of hidden variables representing each hotel j, and m represents the lengths of the user and the vector of hidden variables of the hotel.
Preferably, the step of obtaining the hidden variables of all users in the OTA platform and the hidden variables of all hotels in the OTA platform by using the latent semantic model includes:
Acquiring all users;
obtaining scoring data of all hotels;
and calculating hidden variables of all users and hidden variables of all hotels by using the latent semantic model by using the scoring data of all users and all hotels.
Preferably, the generating step of the prediction model is as follows:
acquiring historical order data, wherein each piece of the historical order data comprises corresponding historical users, a historical hotel and a result of whether to place an order or not;
processing each piece of historical order data to obtain historical feature vectors composed of the user dimension information corresponding to the historical user, the hotel dimension information corresponding to the historical hotels and the preference of the user to be recommended to each hotel;
marking each historical order according to the result of whether to place an order or not to obtain a label corresponding to each historical feature vector;
and constructing a classification model by using the historical feature vector and the corresponding label through an XGBOOST (an integrated algorithm) algorithm so as to obtain the prediction model.
The prediction model can be adjusted by an AUC (area under curve) index, and the specific steps of the adjustment are as follows: judging whether the prediction model reaches the AUC index, if not, adjusting parameters in the prediction model, and then obtaining the prediction model conforming to the AUC index according to the adjusted model.
Preferably, the ranking the probability of each first hotel to be ordered by the user to be recommended is:
ordering the probability of each first hotel to be recommended to the user to be recommended according to the high-to-low order to obtain a first sequence;
the hotel ordering method further comprises the following steps:
ranking the hotels in the first sequence according to the scores to obtain a second sequence;
and outputting the second sequence as a result of the output sequencing.
Preferably, the hotel ordering method further comprises:
setting the selectable hotels with the distance greater than or equal to the distance threshold as second hotels, and sequencing the second hotels to obtain a third sequence;
ordering the third sequence after the second sequence to obtain a fourth sequence;
and outputting the fourth sequence as a result of the output sequencing.
Preferably, all the optional hotels that are not pictorial and/or indefinite are set to a fifth sequence, which is ordered after the fourth sequence to obtain a sixth sequence. And outputting the sixth sequence as a result of the output sequencing.
The invention also provides a hotel ordering system based on the POI under the OTA platform, which comprises the following steps:
The region acquisition module is used for acquiring the region where the POI is located according to the POI input by the user to be recommended;
the selectable hotel acquisition module is used for acquiring all selectable hotels in the region;
the first hotel setting module is used for calculating the distance between each selectable hotel and the POI, and setting the selectable hotels with the distance smaller than a distance threshold as first hotels;
the user dimension acquisition module is used for acquiring the user dimension information of the user to be recommended;
the preference degree calculating module is used for calculating preference degree of the user to be recommended to each hotel under the OTA platform;
the hotel dimension acquisition module is used for acquiring hotel dimension information of each first hotel;
the prediction module is used for inputting the user dimension information of the user to be recommended, the hotel dimension information of each first hotel and the preference degree of the user to be recommended for each hotel into a prediction model, and calculating the probability of the user to be recommended for each first hotel by using the prediction model;
the ordering module is used for ordering the probability of each first hotel to be ordered by the user to be recommended;
And the output module is used for outputting the ordered result.
Preferably, the user dimension information includes basic information of the user, and the user dimension information further includes at least one of information of an order of a hotel ordered by the user in the past, information of a hotel browsed by the user, and information of a hotel collected by the user;
the hotel dimension information comprises basic information of the hotel, and further comprises at least one of price information of the hotel, good grade information of the hotel and information of the number of times the hotel is collected.
Preferably, the preference calculating module includes:
the first hidden variable acquisition unit is used for acquiring hidden variables of all users in the OTA platform and hidden variables of all hotels in the OTA platform by using a latent semantic model;
the second hidden variable acquisition unit is used for obtaining hidden variables of the user to be recommended according to the hidden variables of all the users;
the preference degree calculating unit is used for calculating the preference degree of the user to be recommended to each hotel according to the following formula:
wherein lfm is ij Representing the preference degree of the user i to be recommended to each hotel j, u ik A kth dimension value, h, representing a vector of hidden variables of the user i to be recommended jk And the kth dimension value of the vector of hidden variables representing each hotel j, and m represents the lengths of the user and the vector of hidden variables of the hotel.
Preferably, the first hidden variable obtaining unit includes:
a first obtaining subunit, configured to obtain all the users;
the second acquisition subunit is used for acquiring the scoring data of all hotels;
and the hidden variable calculation operator unit is used for calculating hidden variables of all users and hidden variables of all hotels by using the score data of all users and all hotels by using the latent semantic model.
Preferably, the hotel ordering system further comprises a model generating unit, the model generating unit comprising:
the order acquisition unit is used for acquiring historical order data, wherein each piece of historical order data comprises a corresponding historical user, a corresponding historical hotel and a corresponding result of whether to place an order or not;
the data processing unit is used for processing each piece of historical order data to obtain historical feature vectors composed of the user dimension information corresponding to the historical user, the hotel dimension information corresponding to the historical hotels and the preference degree of the user to be recommended to each hotel;
The marking unit is used for marking each historical order according to the result of whether the order is placed or not so as to obtain a label corresponding to each historical feature vector;
and the model construction unit is used for constructing a classification model by using the history feature vector and the corresponding label through an XGBOOST algorithm so as to obtain the prediction model.
The prediction model can be adjusted through an AUC index, and the specific steps of the adjustment are as follows: judging whether the prediction model reaches the AUC index, if not, adjusting parameters in the prediction model, and then obtaining the prediction model conforming to the AUC index according to the adjusted model.
Preferably, the sorting module includes: a first sorting unit and a second sorting unit;
the first ordering unit is used for ordering the probability of each first hotel to be recommended by the user to be recommended according to the high-low probability so as to obtain a first sequence;
the second ordering unit is used for ordering hotels in the first sequence according to scores so as to obtain a second sequence;
and outputting the second sequence as a result of the output sequencing in the output module.
Preferably, the sorting module further comprises: a third sorting unit and a fourth sorting unit;
The third sorting unit is configured to set the selectable hotels with the distance greater than or equal to the distance threshold as second hotels, and sort the second hotels to obtain a third sequence;
the fourth sorting unit is used for sorting the third sequence after the second sequence to obtain a fourth sequence;
and outputting the fourth sequence as a result of the output sequencing in the output module.
Preferably, all the optional hotels that are not pictorial and/or indefinite are set to a fifth sequence, which is ordered after the fourth sequence to obtain a sixth sequence. And outputting the sixth sequence as a result of the output sequencing.
The invention has the positive progress effects that:
according to the hotel ordering method and system based on the POI under the OTA platform, the probability of the user for ordering the hotel can be judged in real time through the machine learning model according to the historical data of the user, the historical data of the hotel and the preference of the user for the hotel, so that the intelligent ordering of online hotels is realized, hotels conforming to the consumption habit of the user can be ordered rapidly, the retrieval strength of the user is reduced, the sales of the hotels is improved, and the brand image of the OTA is improved.
Drawings
Fig. 1 is a flowchart of a hotel ordering method based on POI under the OTA platform of embodiment 1.
Fig. 2 is a flowchart of step S15 in fig. 1.
Fig. 3 is a flowchart of the steps of obtaining a latent semantic model of a POI-based hotel ordering method under the OTA platform of embodiment 1.
Fig. 4 is a flowchart of the steps of generating a predictive model of a hotel ordering method based on POI under the OTA platform of embodiment 1.
Fig. 5 is a flowchart of step S18 in fig. 1.
Fig. 6 is a schematic block diagram of a hotel ordering system based on POI under the OTA platform of embodiment 2.
Fig. 7 is a schematic block diagram of the preference calculating module in fig. 6.
Fig. 8 is a schematic block diagram of the first hidden variable obtaining unit in fig. 7.
Fig. 9 is a schematic block diagram of a model generating unit of the hotel ordering system based on POI under the OTA platform of embodiment 2.
Fig. 10 is a block diagram of the sorting module of fig. 6.
Detailed Description
The invention is further illustrated by means of the following examples, which are not intended to limit the scope of the invention.
Example 1
The embodiment provides a hotel ordering method based on POI under an OTA platform, and the specific flow is shown in figure 1 and comprises the following steps:
step S11, acquiring a region where a POI is located according to the POI input by a user to be recommended;
In this embodiment, the region may be a city, may be a region, may be a street, and specifically, the range of the region may be selected according to the requirement of the user to be recommended.
Step S12, obtaining all optional hotels in the area;
the selectable hotels are hotels which can be reserved currently, and specifically comprise hotels in the traditional sense, and also comprise civilian hosts, guest stacks, young hotels and the like.
Step S13, calculating the distance between each selectable hotel and the POI, and setting the selectable hotels with the distance smaller than a distance threshold as first hotels;
the distance threshold is a value preset according to needs, for example, may be 1 km.
Step S14, acquiring user dimension information of the user to be recommended;
the user dimension information comprises basic information of the user, and further comprises at least one of information of an order of a hotel ordered by the user in the past, information of a hotel browsed by the user and information of a hotel collected by the user; the basic information of the user may include information of age, gender, academic history, hometown, etc. of the user.
If the user does not have the basic information, setting the basic information of the user to be null; if the user does not subscribe to the hotel, setting the information of the order of the hotel which is subscribed to by the user in the past to be empty; if the user does not browse the hotel, setting the information of the hotel browsed by the user to be empty; and if the user does not collect the hotel, setting the information of the hotel collected by the user to be empty.
Step S15, calculating preference of the user to be recommended to each hotel under the OTA platform;
step S16, acquiring hotel dimension information of each first hotel;
the hotel dimension information comprises basic information of the hotel, and further comprises at least one of price information of the hotel, good grade information of the hotel and collection times information of the hotel; the basic information of the hotel can comprise information of star level, room type and the like of the hotel.
If the hotel does not have the basic information, setting the basic information of the hotel to be empty; if the hotel does not have price information, setting the price information of the hotel to be empty; if the hotel does not have the good score information, setting the good score information of the hotel to be null; and if the hotel does not have the information of the number of times of collection, setting the information of the number of times of collection of the hotel to be empty.
Step S17, calculating the probability of each first hotel to be ordered by the user to be recommended by using the prediction model;
the specific implementation manner of step S17 is that the user dimension information of the user to be recommended, the hotel dimension information of each first hotel and the preference of the user to be recommended for each hotel are input into a prediction model, and then the probability of the user to be recommended for each first hotel is calculated by using the prediction model.
Step S18, ordering the probability of each first hotel to be ordered by the user to be recommended;
and step S19, outputting the ordered result.
In this embodiment, through the POI customized by the user to be recommended, the hotels near the POI are reasonably ordered, different consumption behavior habits of different users to be recommended are combined with the quality of the hotels, the hotels which are most in line with the user demands are ordered according to the ordering probability of the users, and the ordering result is output, so that the process of screening the hotels by the users is reduced, the operation cost of the users is reduced, and the sales of the hotels is improved.
As shown in fig. 2, in the present embodiment, step S15 includes the steps of:
Step S151, using a latent semantic model to obtain hidden variables of all users in the OTA platform and hidden variables of all hotels in the OTA platform;
step S152, obtaining hidden variables of the users to be recommended according to the hidden variables of all the users;
wherein lfm is ij Representing the preference degree of the user i to be recommended to each hotel j, u ik A kth dimension value, h, representing a vector of hidden variables of the user i to be recommended jk And the kth dimension value of the vector of hidden variables representing each hotel j, and m represents the lengths of the user and the vector of hidden variables of the hotel.
In this embodiment, as shown in fig. 3, a flowchart of the step of obtaining the lingo-semantic model in step S151 includes:
step S1511, obtaining all users;
step S1512, scoring data of all hotels is obtained;
and step S1513, calculating hidden variables of all users and hidden variables of all hotels by using the latent meaning model by using the scoring data of all users and all hotels.
Wherein the data used by the latent semantic model comprises both data of users who have scored and data of users who have not scored. It should be appreciated that if a user does not score a hotel, the user's score data for the hotel is null.
In this embodiment, hidden variables of all users and hidden variables of all hotels are calculated through a latent meaning model, and hidden variables of users to be recommended and hidden variables of all hotels are obtained according to hidden variables of all users and hidden variables of all hotels, so that preference of users to be recommended to each hotel is predicted, wherein the preference of users to be recommended can be set to be updated once per day, that is, hidden variables of all hotels and hidden variables of all users are calculated once per day through the latent meaning model, and when preference of users to be recommended to each hotel needs to be predicted, data corresponding to users to be recommended only need to be obtained from the calculated hidden variables of all users and hidden variables of all hotels, so that requirements of different users are met, and the calculation process is faster.
As shown in fig. 4, the generating step of the prediction model in step 17 in this embodiment includes:
step S171, acquiring historical order data, wherein each piece of the historical order data comprises corresponding historical users, historical hotels and whether an ordering result is obtained;
step S172, processing each piece of historical order data to obtain a historical feature vector;
the specific implementation manner of step S172 is to process each piece of the historical order data to obtain the user dimension information corresponding to the historical user, the hotel dimension information corresponding to the historical hotel, and a historical feature vector formed by the preference of the user to be recommended to each hotel.
Step S173, marking each historical order according to the result of whether to place an order, so as to obtain a label corresponding to each historical feature vector;
in this embodiment, the hotel label for the order is 1, and the hotel label for browsing without the order is 0.
And step 174, constructing a classification model by using the history feature vector and the corresponding label through an XGBOOST algorithm so as to obtain the prediction model.
In this embodiment, in the process of obtaining the prediction model, a certain evaluation index may be used to verify the effect of the prediction model, in this embodiment, an AUC index is used to verify, where the higher the AUC index is, the better the prediction model is, and in practice, other indexes may be used to verify the prediction model, such as accuracy, recall, etc., and in particular, which index needs to be selected according to the actual situation.
And if the prediction model does not reach the evaluation index, adjusting parameters of the prediction model according to the evaluation index and retraining the prediction model until the prediction model meets the evaluation index.
In this embodiment, the integrated learning method in machine learning is adopted, so that the generalization capability of the model can be improved. And after the POI information is selected by the user, carrying out distance screening and filtering according to the distance from the hotel to the POI. And excavating influence factors (user dimension information, hotel dimension information and preference degree of users to be recommended for each first hotel) of users on hotel ordering, and constructing a machine learning model. The machine learning model is used for predicting the user ordering probability of the hotels, and the hotels are ordered according to the predicted ordering probability, so that the hotels are automatically ordered according to different users, and the requirements of different users are met.
In this embodiment, a certain evaluation index is set to train a prediction model, so as to optimize the prediction model, so that the ranking predicted by the prediction model is closer to the ranking required by the user to be recommended.
As shown in fig. 5, step S18 in this embodiment specifically includes:
step S181, ordering the probability of each first hotel to be recommended to the user to be recommended according to the high-to-low order to obtain a first sequence;
step S182, ranking hotels in the first sequence according to scores to obtain a second sequence;
In this embodiment, the specific implementation method for sorting hotels in the first sequence according to the scores includes setting a point score threshold, and placing hotels in the first sequence which do not meet the point score threshold behind hotels meeting the point score threshold to obtain a second sequence.
Step S183, setting the selectable hotels with the distance greater than or equal to the distance threshold as second hotels, and sorting the second hotels to obtain a third sequence;
step S184, sorting the third sequence after the second sequence to obtain a fourth sequence;
step S185, setting all the optional hotels without pictures and/or indeterminate as a fifth sequence;
wherein after the indeterminate hotels of all selectable hotels are ordered to the photo-free hotels, the photo-free hotels of all selectable hotels of the indeterminate are ordered last.
Step S186, sorting the fifth sequence after the fourth sequence to obtain a sixth sequence;
and outputting the sixth sequence as a result of the output sequencing.
And the sixth sequence can also comprise hotels which are filtered out and do not meet the distance requirement, the filtered hotels are ranked from small to large according to the distance POI distance, and the filtered hotels are supplemented behind the hotels with low scores, so that the user is given enough opportunities to select the hotels.
In this embodiment, by using the latent meaning model and the prediction model, the ranking of hotels meeting the demands of users to be recommended can be automatically predicted, and the results meeting the demands of users can be predicted online in real time according to different conditions of different users, so that the uniqueness of hotel prediction is overcome, the complexity of manual hotel selection of users is reduced while hotels are reasonably ranked, and the hotel ranking speed is faster and the user experience is better.
According to the method for machine learning, hotels with a certain distance range are ranked according to POI information selected by a user to be recommended, probability that the user to be recommended will place an order on the hotels is predicted, quality of the hotels is corrected, and more selection space is provided for the user for hotels outside the distance range. Different hotel ordering schemes are purposefully carried out on different users, so that inconvenience of users to be recommended when selecting hotels is reduced, user experience is improved, hotel order sales are promoted, and brand image is optimized.
Example 2
As shown in fig. 6, this embodiment provides a hotel ordering system based on POI under an OTA platform, including: the system comprises a region acquisition module 11, an optional hotel acquisition module 12, a first hotel setting module 13, a user dimension acquisition module 14, a preference calculation module 15, a hotel dimension acquisition module 16, a prediction module 17, a ranking module 18 and an output module 19.
The region acquisition module 11 is configured to acquire a region where the POI is located according to the POI input by the user to be recommended.
In this embodiment, the region may be a city, may be a region, may be a street, and specifically, the range of the region may be selected according to the requirement of the user to be recommended.
The alternative hotel acquisition module 12 is used to acquire all alternative hotels within the area.
The alternative hotels may include both those in the traditional sense, as well as those in the civilian, the hotel, the young hotel, etc.
The first hotel setting module 13 is configured to calculate a distance between each of the alternative hotels and the POI, and set the alternative hotels with the distance smaller than a distance threshold as first hotels.
And the user dimension acquisition module 14 is used for acquiring the user dimension information of the user to be recommended.
The user dimension information comprises basic information of the user, and further comprises at least one of information of an order of a hotel ordered by the user in the past, information of a hotel browsed by the user and information of a hotel collected by the user; the basic information of the user comprises information of the age, sex, academic calendar, hometown and the like of the user.
If the user does not have the basic information, setting the basic information of the user to be null; if the user does not subscribe to the hotel, setting the information of the order of the hotel which is subscribed to by the user in the past to be empty; if the user does not browse the hotel, setting the information of the hotel browsed by the user to be empty; and if the user does not collect the hotel, setting the information of the hotel collected by the user to be empty. And the preference degree calculating module 15 is used for calculating the preference degree of the user to be recommended to each hotel under the OTA platform.
And the hotel dimension acquisition module 16 is configured to acquire hotel dimension information of each first hotel.
The hotel dimension information comprises basic information of the hotel, and further comprises at least one of price information of the hotel, good grade information of the hotel and collection times information of the hotel; the basic information of the hotel comprises information such as star level, room type and the like of the hotel.
If the hotel does not have the basic information, setting the basic information of the hotel to be empty; if the hotel does not have price information, setting the price information of the hotel to be empty; if the hotel does not have the good score information, setting the good score information of the hotel to be null; and if the hotel does not have the information of the number of times of collection, setting the information of the number of times of collection of the hotel to be empty. The prediction module 17 is configured to input the user dimension information of the user to be recommended, the hotel dimension information of each first hotel, and the preference of the user to be recommended for each hotel into a prediction model, and calculate the probability of each first hotel being placed by the user to be recommended using the prediction model.
The ranking module 18 is configured to rank the probability that the user to be recommended places each of the first hotels.
The output module 19 outputs the ordered result, which in this embodiment is the sixth sequence.
As shown in fig. 7, in this embodiment, the preference calculating module 15 specifically includes: a first hidden variable acquisition unit 151, a second hidden variable acquisition unit 152, and a preference calculation unit 153.
The first hidden variable obtaining unit 151 is configured to obtain hidden variables of all users in the OTA platform and hidden variables of all hotels in the OTA platform using a latent semantic model.
The second hidden variable obtaining unit 152 is configured to obtain hidden variables of the user to be recommended according to hidden variables of all the users.
The preference degree calculating unit 153 is configured to calculate and obtain the preference degree of the user to be recommended for each hotel by using a first formula, where the first formula is:
wherein lfm is ij Representing the preference degree of the user i to be recommended to each hotel j, u ik A kth dimension value, h, representing a vector of hidden variables of the user i to be recommended jk And the kth dimension value of the vector of hidden variables representing each hotel j, and m represents the lengths of the user and the vector of hidden variables of the hotel.
As shown in fig. 8, in the present embodiment, the first hidden variable acquisition unit 151 specifically includes: a first acquisition subunit 1511, a second acquisition subunit 1512, and a hidden-variable computation subunit 1513.
The first acquisition subunit 1511 is configured to acquire all the users.
The second obtaining subunit 1512 is configured to obtain scoring data of all hotels.
The hidden variable computation subunit 1513 is configured to calculate hidden variables of the all users and hidden variables of the all hotels by using the score data of the all users and the all hotels by using the latent semantic model.
Wherein the data used by the latent semantic model comprises both data of users who have scored and data of users who have not scored. It should be appreciated that if a user does not score a hotel, the user's score data for the hotel is null.
In this embodiment, hidden variables of all users and hidden variables of all hotels are calculated through a latent meaning model, and hidden variables of users to be recommended and hidden variables of all hotels are obtained according to hidden variables of all users and hidden variables of all hotels, so that preference of users to be recommended to each hotel is predicted, wherein the preference of users to be recommended can be set to be updated once per day, that is, hidden variables of all hotels and hidden variables of all users are calculated once per day through the latent meaning model, and when preference of users to be recommended to each hotel needs to be predicted, data corresponding to users to be recommended only need to be obtained from the calculated hidden variables of all users and hidden variables of all hotels, so that requirements of different users are met, and the calculation process is faster.
As shown in fig. 9, in this embodiment, the hotel ordering system further includes a model generating unit 177, where the model generating unit specifically includes: an order acquisition unit 171, a data processing unit 172, a labeling unit 173, and a model construction unit 174.
The order acquisition unit 171 is configured to acquire historical order data, where each piece of the historical order data includes a corresponding historical user, a corresponding historical hotel, and a result of whether to place an order.
The data processing unit 172 is configured to process each piece of the historical order data to obtain a historical feature vector composed of the user dimension information corresponding to the historical user, the hotel dimension information corresponding to the historical hotel, and the preference of the user to be recommended to each hotel.
The marking unit 173 is configured to mark each of the historical orders according to the result of whether to place an order, so as to obtain a label corresponding to each of the historical feature vectors.
In this embodiment, the hotel label for the order is 1, and the hotel label for browsing without the order is 0.
The model construction unit 174 is configured to perform construction of a classification model by using XGBOOST algorithm by using the historical feature vector and the corresponding label, so as to obtain the prediction model.
In this embodiment, in the process of obtaining the prediction model, a certain evaluation index may be used to verify the effect of the prediction model, in this embodiment, an AUC index is used to verify, where the higher the AUC index is, the better the prediction model is, and in practice, other indexes may be used to verify the prediction model, such as accuracy, recall, etc., and in particular, which index needs to be selected according to the actual situation.
And if the prediction model does not reach the evaluation index, adjusting parameters of the prediction model according to the evaluation index and retraining the prediction model until the prediction model meets the evaluation index.
In this embodiment, the prediction model is trained by setting a certain evaluation index, so as to optimize the prediction model, so that the ranking predicted by the prediction model is closer to the ranking required by the user to be recommended.
As shown in fig. 10, in this embodiment, the sorting module 18 specifically includes: a first ordering unit 181, a second ordering unit 182, a third ordering unit 183, a fourth ordering unit 184, a fifth ordering unit 185, a sixth ordering unit 186.
The first ranking unit 181 is configured to rank the probability of each first hotel to be ordered by the user to be recommended according to the from high to low, so as to obtain a first sequence.
The second ranking unit 182 is configured to rank the hotels in the first sequence according to the scores, so as to obtain a second sequence.
In this embodiment, the specific implementation method for sorting hotels in the first sequence according to the scores includes setting a point score threshold, and placing hotels in the first sequence which do not meet the point score threshold behind hotels meeting the point score threshold to obtain a second sequence.
The third ranking unit 183 is configured to set the optional hotel with the distance greater than or equal to the distance threshold as a second hotel, and rank the second hotel to obtain a third sequence.
The fourth sorting unit 184 is configured to sort the third sequence to the second sequence to obtain a fourth sequence.
The fifth ranking unit 185 is used to set all of the alternative hotels that are photo-free and/or indeterminate to a fifth sequence.
Wherein after ordering all the unsubscribable alternative hotels to the photo-free hotels, the photo-free and unfixed alternative hotels are ranked last.
The sixth sorting unit 186 is configured to sort the fifth sequence to the fourth sequence to obtain a sixth sequence.
And outputting the sixth sequence as a result of the output sequencing.
In this embodiment, through obtaining the POI that the user input, through using lingering meaning model and prediction model, can automatic prediction satisfy the ordering of the hotel of waiting to recommend user demand to can be according to the different circumstances of different users, the result that accords with the user demand is predicted in real time on line, thereby overcome the singleness of hotel prediction, when rationally ordering the hotel, reduce the loaded down with trivial details nature of user's manual selection hotel, make hotel ordering speed faster, user experience better.
While specific embodiments of the invention have been described above, it will be appreciated by those skilled in the art that this is by way of example only, and 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 principles and spirit of the invention, but such changes and modifications fall within the scope of the invention.
Claims (12)
1. The hotel ordering method based on POI under the OTA platform is characterized by comprising the following steps:
Acquiring a region where a POI is located according to the POI input by a user to be recommended;
acquiring all selectable hotels in the region;
calculating the distance between each selectable hotel and the POI, and setting the selectable hotels with the distance smaller than a distance threshold as first hotels;
acquiring user dimension information of the user to be recommended;
calculating preference degree of the user to be recommended to each hotel under the OTA platform; the preference degree is related to scoring data of all hotels in the OTA platform;
acquiring hotel dimension information of each first hotel;
inputting the user dimension information of the user to be recommended, the hotel dimension information of each first hotel and the preference of the user to be recommended to each hotel into a prediction model, and calculating the probability of the user to be recommended to place an order to each first hotel by using the prediction model;
ordering the probability of each first hotel to be ordered by the user to be recommended;
outputting the ordered result;
the ranking of the probability of each first hotel to be ordered by the user to be recommended is as follows:
ordering the probability of each first hotel to be recommended to the user to be recommended according to the high-to-low order to obtain a first sequence;
The hotel ordering method further comprises the following steps:
ranking the hotels in the first sequence according to the scores to obtain a second sequence;
the hotel ordering method further comprises the following steps:
setting the selectable hotels with the distance greater than or equal to the distance threshold as second hotels, and sequencing the second hotels to obtain a third sequence;
ordering the third sequence after the second sequence to obtain a fourth sequence;
and outputting the fourth sequence as a result of the output sequencing.
2. The method for POI-based hotel ordering under an OTA platform of claim 1,
the user dimension information comprises basic information of the user, and at least one of information of an order of a hotel ordered by the user in the past, information of a hotel browsed by the user and information of a hotel collected by the user;
the hotel dimension information comprises basic information of the hotel, and further comprises at least one of price information of the hotel, good grade information of the hotel and information of the number of times the hotel is collected.
3. The method of claim 1, wherein the step of calculating the preference of the user to be recommended for each hotel under the OTA platform comprises:
Obtaining hidden variables of all users in the OTA platform and hidden variables of all hotels in the OTA platform by using a latent semantic model;
obtaining hidden variables of the users to be recommended according to the hidden variables of all the users;
and calculating the preference degree of the user to be recommended to each hotel by adopting the following formula:
wherein lfm is ij Representing the user i to be recommended to each user iSaid preference degree of hotel j, u ik A kth dimension value, h, representing a vector of hidden variables of the user i to be recommended jk And the k dimension value of the vector representing the hidden variable of each hotel j, and m represents the lengths of the user and the vector of the hidden variable of each hotel.
4. The method for POI-based hotel ordering under an OTA platform of claim 3, wherein the step of using a latent semantic model to obtain hidden variables for all users in the OTA platform and hidden variables for all hotels in the OTA platform is:
acquiring all users;
obtaining scoring data of all hotels;
and calculating hidden variables of all users and hidden variables of all hotels by using the latent semantic model by using the scoring data of all users and all hotels.
5. The hotel ordering method based on POIs under the OTA platform according to claim 1 wherein the generating step of the prediction model is:
acquiring historical order data, wherein each piece of the historical order data comprises corresponding historical users, a historical hotel and a result of whether to place an order or not;
processing each piece of historical order data to obtain historical feature vectors composed of the user dimension information corresponding to the historical user, the hotel dimension information corresponding to the historical hotels and the preference of the user to be recommended to each hotel;
marking each historical order according to the result of whether to place an order or not to obtain a label corresponding to each historical feature vector;
and constructing a classification model by using the history feature vector and the corresponding label through an XGBOOST algorithm so as to obtain the prediction model.
6. The method of claim 1, wherein after obtaining the second sequence, the outputting the ordered result is outputting the second sequence.
7. A hotel ordering system based on POIs under an OTA platform, comprising:
The region acquisition module is used for acquiring the region where the POI is located according to the POI input by the user to be recommended;
the selectable hotel acquisition module is used for acquiring all selectable hotels in the region;
the first hotel setting module is used for calculating the distance between each selectable hotel and the POI, and setting the selectable hotels with the distance smaller than a distance threshold as first hotels;
the user dimension acquisition module is used for acquiring the user dimension information of the user to be recommended;
the preference degree calculating module is used for calculating preference degree of the user to be recommended to each hotel under the OTA platform; the preference degree is related to scoring data of all hotels in the OTA platform;
the hotel dimension acquisition module is used for acquiring hotel dimension information of each first hotel;
the prediction module is used for inputting the user dimension information of the user to be recommended, the hotel dimension information of each first hotel and the preference degree of the user to be recommended for each hotel into a prediction model, and calculating the probability of the user to be recommended for each first hotel by using the prediction model;
the ordering module is used for ordering the probability of each first hotel to be ordered by the user to be recommended;
The output module is used for outputting the ordered results;
the sorting module comprises a first sorting unit and a second sorting unit;
the first ordering unit is used for ordering the probability of each first hotel to be recommended by the user to be recommended according to the high-low probability so as to obtain a first sequence;
the second ordering unit is used for ordering hotels in the first sequence according to scores so as to obtain a second sequence;
the ranking module further includes: a third sorting unit and a fourth sorting unit;
the third sorting unit is configured to set the selectable hotels with the distance greater than or equal to the distance threshold as second hotels, and sort the second hotels to obtain a third sequence;
the fourth sorting unit is used for sorting the third sequence after the second sequence to obtain a fourth sequence;
and outputting the fourth sequence as a result of the output sequencing in the output module.
8. The OTA platform POI-based hotel ordering system of claim 7,
the user dimension information comprises basic information of the user, and at least one of information of an order of a hotel ordered by the user in the past, information of a hotel browsed by the user and information of a hotel collected by the user;
The hotel dimension information comprises basic information of the hotel, and further comprises at least one of price information of the hotel, good grade information of the hotel and information of the number of times the hotel is collected.
9. The OTA platform POI-based hotel ordering system of claim 7 wherein the preference calculation module comprises:
the first hidden variable acquisition unit is used for acquiring hidden variables of all users in the OTA platform and hidden variables of all hotels in the OTA platform by using a latent semantic model;
the second hidden variable acquisition unit is used for obtaining hidden variables of the user to be recommended according to the hidden variables of all the users;
the preference degree calculating unit is used for calculating the preference degree of the user to be recommended to each hotel according to the following formula:
wherein lfm is ij Representing the preference degree of the user i to be recommended to each hotel j, u ik A kth dimension value, h, representing a vector of hidden variables of the user i to be recommended jk And the kth dimension value of the vector of hidden variables representing each hotel j, and m represents the length of the vector of hidden variables of the user and each hotel.
10. The OTA platform POI-based hotel ordering system of claim 9, wherein the first hidden variable acquisition unit comprises:
A first obtaining subunit, configured to obtain all the users;
the second acquisition subunit is used for acquiring the scoring data of all hotels;
and the hidden variable calculation operator unit is used for calculating hidden variables of all users and hidden variables of all hotels by using the score data of all users and all hotels by using the latent semantic model.
11. The OTA platform POI-based hotel ordering system of claim 7, further comprising a model generation unit comprising:
the order acquisition unit is used for acquiring historical order data, wherein each piece of historical order data comprises a corresponding historical user, a corresponding historical hotel and a corresponding result of whether to place an order or not;
the data processing unit is used for processing each piece of historical order data to obtain historical feature vectors composed of the user dimension information corresponding to the historical user, the hotel dimension information corresponding to the historical hotels and the preference degree of the user to be recommended to each hotel;
the marking unit is used for marking each historical order according to the result of whether the order is placed or not so as to obtain a label corresponding to each historical feature vector;
And the model construction unit is used for constructing a classification model by using the history feature vector and the corresponding label through an XGBOOST algorithm so as to obtain the prediction model.
12. The POI-based hotel ordering system under an OTA platform of claim 7 wherein outputting the ordered result in the output module is outputting the second sequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811638493.6A CN109740072B (en) | 2018-12-29 | 2018-12-29 | Hotel ordering method and system based on POI under OTA platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811638493.6A CN109740072B (en) | 2018-12-29 | 2018-12-29 | Hotel ordering method and system based on POI under OTA platform |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109740072A CN109740072A (en) | 2019-05-10 |
| CN109740072B true CN109740072B (en) | 2023-07-04 |
Family
ID=66362360
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811638493.6A Active CN109740072B (en) | 2018-12-29 | 2018-12-29 | Hotel ordering method and system based on POI under OTA platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109740072B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110619558B (en) * | 2019-09-18 | 2021-02-05 | 北京三快在线科技有限公司 | Search sorting method, system and equipment for online lodging products |
| CN110633370B (en) * | 2019-09-19 | 2023-07-04 | 携程计算机技术（上海）有限公司 | OTA hotel label generation method, system, electronic device and 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 |
| CN113763112A (en) * | 2021-02-25 | 2021-12-07 | 北京沃东天骏信息技术有限公司 | Information pushing method and device |
| CN113590937B (en) * | 2021-07-05 | 2022-04-19 | 深圳市天下房仓科技有限公司 | Hotel searching and information management method and device, electronic equipment and storage medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5948040A (en) * | 1994-06-24 | 1999-09-07 | Delorme Publishing Co. | Travel reservation information and planning system |
| JP2004061503A (en) * | 2002-07-18 | 2004-02-26 | Alpine Electronics Inc | Point-of-interest (poi) information display method and navigation system |
| CN102867031A (en) * | 2012-08-27 | 2013-01-09 | 百度在线网络技术（北京）有限公司 | Method and system for optimizing point of interest (POI) searching results, mobile terminal and server |
| EP2581703A1 (en) * | 2011-10-12 | 2013-04-17 | Mapquest, Inc. | Systems and methods for ranking points of interest |
| CN103150309A (en) * | 2011-12-07 | 2013-06-12 | 清华大学 | Method and system for searching POI (Point of Interest) points of awareness map in space direction |
| CN104504064A (en) * | 2014-12-19 | 2015-04-08 | 百度在线网络技术（北京）有限公司 | Information recommendation method and device |
| CN105589899A (en) * | 2014-11-18 | 2016-05-18 | 北京四维图新科技股份有限公司 | Display method and display device of point of interest in electronic map |
| CN105653736A (en) * | 2016-03-01 | 2016-06-08 | 北京师范大学 | Interest point group recommendation method based on geographical locations |
| US9857177B1 (en) * | 2012-06-20 | 2018-01-02 | Amazon Technologies, Inc. | Personalized points of interest for mapping applications |
| CN108717640A (en) * | 2018-01-24 | 2018-10-30 | 北京穷游天下科技发展有限公司 | The data processing method and electronic equipment of travel information |
Family Cites Families (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7826965B2 (en) * | 2005-06-16 | 2010-11-02 | Yahoo! Inc. | Systems and methods for determining a relevance rank for a point of interest |
| US7831381B2 (en) * | 2005-08-04 | 2010-11-09 | Microsoft Corporation | Data engine for ranking popularity of landmarks in a geographical area |
| US20100305842A1 (en) * | 2009-05-27 | 2010-12-02 | Alpine Electronics, Inc. | METHOD AND APPARATUS TO FILTER AND DISPLAY ONLY POIs CLOSEST TO A ROUTE |
| US8669884B2 (en) * | 2011-02-02 | 2014-03-11 | Mapquest, Inc. | Systems and methods for generating electronic map displays with points of-interest information |
| US20120221363A1 (en) * | 2011-02-25 | 2012-08-30 | Hipmunk, Inc. | System and method for displaying hotel information |
| US20130262479A1 (en) * | 2011-10-08 | 2013-10-03 | Alohar Mobile Inc. | Points of interest (poi) ranking based on mobile user related data |
| US20150073941A1 (en) * | 2011-12-13 | 2015-03-12 | Emma Burrows | Hotel finder interface |
| US9080885B2 (en) * | 2012-06-05 | 2015-07-14 | Apple Inc. | Determining to display designations of points of interest within a map view |
| 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 |
| CN107507042A (en) * | 2017-09-15 | 2017-12-22 | 携程计算机技术（上海）有限公司 | Marketing method and system based on user's portrait |
-
2018
- 2018-12-29 CN CN201811638493.6A patent/CN109740072B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5948040A (en) * | 1994-06-24 | 1999-09-07 | Delorme Publishing Co. | Travel reservation information and planning system |
| JP2004061503A (en) * | 2002-07-18 | 2004-02-26 | Alpine Electronics Inc | Point-of-interest (poi) information display method and navigation system |
| EP2581703A1 (en) * | 2011-10-12 | 2013-04-17 | Mapquest, Inc. | Systems and methods for ranking points of interest |
| CN103150309A (en) * | 2011-12-07 | 2013-06-12 | 清华大学 | Method and system for searching POI (Point of Interest) points of awareness map in space direction |
| US9857177B1 (en) * | 2012-06-20 | 2018-01-02 | Amazon Technologies, Inc. | Personalized points of interest for mapping applications |
| CN102867031A (en) * | 2012-08-27 | 2013-01-09 | 百度在线网络技术（北京）有限公司 | Method and system for optimizing point of interest (POI) searching results, mobile terminal and server |
| CN105589899A (en) * | 2014-11-18 | 2016-05-18 | 北京四维图新科技股份有限公司 | Display method and display device of point of interest in electronic map |
| CN104504064A (en) * | 2014-12-19 | 2015-04-08 | 百度在线网络技术（北京）有限公司 | Information recommendation method and device |
| CN105653736A (en) * | 2016-03-01 | 2016-06-08 | 北京师范大学 | Interest point group recommendation method based on geographical locations |
| CN108717640A (en) * | 2018-01-24 | 2018-10-30 | 北京穷游天下科技发展有限公司 | The data processing method and electronic equipment of travel information |
Non-Patent Citations (2)
| Title |
|---|
| Web application for recommending personalised mobile tourist routes;D. Gavalas;《 IET Software》;第6卷(第4期);313 – 322 * |
| 基于位置社交网络的数据挖掘;连德富;《中国博士学位论文全文数据库 信息科技》;I138-31 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109740072A (en) | 2019-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109740072B (en) | Hotel ordering method and system based on POI under OTA platform | |
| CN109271574A (en) | A kind of hot word recommended method and device | |
| CN107609960A (en) | Rationale for the recommendation generation method and device | |
| WO2015188699A1 (en) | Item recommendation method and device | |
| CN109189904A (en) | Individuation search method and system | |
| CN107507016A (en) | A kind of information push method and system | |
| US11774264B2 (en) | Method and system for providing information to a user relating to a point-of-interest | |
| CN106708821A (en) | User personalized shopping behavior-based commodity recommendation method | |
| CN109165350A (en) | A kind of information recommendation method and system based on deep knowledge perception | |
| Gomathi et al. | Restaurant recommendation system for user preference and services based on rating and amenities | |
| CN108446351B (en) | Hotel screening method and system based on user preference of OTA platform | |
| CN104991966A (en) | Ordering method and system of media content | |
| CN109509039A (en) | Method for building up and system, the Method of Commodity Recommendation and system of price expectation model | |
| WO2019218654A1 (en) | Product ordering method | |
| CN109241451B (en) | Content combination recommendation method and device and readable storage medium | |
| CN107230381A (en) | Recommend method, server and client in a kind of parking stall | |
| CN105138508A (en) | Preference diffusion based context recommendation system | |
| CN108256970A (en) | A kind of method that Products Show is carried out based on shopping need | |
| CN111444405A (en) | User interaction method and device for searching, mobile terminal and storage medium | |
| CN109885776A (en) | Recommended models can be explained in open source community PR reviewer | |
| CN108694211B (en) | Application distribution method and device | |
| CN108197241A (en) | A kind of method for searching path based on user preference, system, storage medium and processor | |
| CN110209916B (en) | Method and device for recommending point of interest images | |
| CN110377841B (en) | Similarity calculation method and system applied to collaborative filtering method | |
| CN112699311A (en) | Information pushing method, storage medium and electronic equipment |
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 |