CN109933716A - A kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference - Google Patents
A kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference Download PDFInfo
- Publication number
- CN109933716A CN109933716A CN201910034827.7A CN201910034827A CN109933716A CN 109933716 A CN109933716 A CN 109933716A CN 201910034827 A CN201910034827 A CN 201910034827A CN 109933716 A CN109933716 A CN 109933716A
- Authority
- CN
- China
- Prior art keywords
- hotel
- array
- data
- client
- index
- 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.)
- Pending
Links
- 230000009471 action Effects 0.000 title claims abstract description 13
- 239000011159 matrix material Substances 0.000 claims abstract description 46
- 239000013598 vector Substances 0.000 claims abstract description 30
- 238000000034 method Methods 0.000 claims abstract description 21
- 239000003337 fertilizer Substances 0.000 claims abstract description 12
- 230000008569 process Effects 0.000 claims abstract description 8
- 230000002087 whitening effect Effects 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 10
- 238000012935 Averaging Methods 0.000 claims description 6
- 241001269238 Data Species 0.000 claims description 6
- 238000003491 array Methods 0.000 claims description 6
- 230000001174 ascending effect Effects 0.000 claims description 4
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims 1
- 238000005259 measurement Methods 0.000 description 3
- 230000003542 behavioural effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000010903 husk Substances 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000001568 sexual effect Effects 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference, the present invention relates to hotel management technical fields；Determine the hotel's fertilizer index for being directed to client；According to fertilizer index developing algorithm input matrix；Data needed for obtaining fertilizer index；Preceding four rows data are pre-processed；Corresponding eigen vector is obtained with pca algorithm process to pretreated first four achievement data；Generalization bounds are chosen according to the size cases of characteristic value；Use quicksort as specific sort method；Standoff distance or pca scoring after determining sequence correspond to which hotel in initial data.It solves the problems, such as customer personalized recommendation；Increase the 5th index " standoff distance " and carry out aid decision under special circumstances, makes algorithm closed loop.
Description
Technical field
The present invention relates to hotel management technical fields, and in particular to a kind of Personalized wine based on customer action Habit Preference
Shop intelligent recommendation algorithm.
Background technique
Universal and e-commerce the fast development of computer networking technology, people come into information-intensive society and network warp
Ji epoch, e-commerce website provide more and more selections to people.For client, due to a sexual preference, educational level
The factors such as gentle economic condition are different, and different clients are also different to the hobby of commodity.On the line to be got up based on e-commerce development
Order the major way for becoming the predetermined hotel of people in hotel.And the hotel of numerous businessmans and magnanimity are faced, client is often difficult to look for
To the suitable hotel for meeting oneself interest preference, the hotel that businessman also makes it difficult for oneself to provide take off grain husk in terms of customer experience and
Out.Therefore, it is closed how according to customer action Habit Preference for lead referral in businessman hotels optionally numerous for existing customer
Suitable hotel becomes problem urgently to be resolved.
In existing recommended method, the introduction according to the essential information in hotel includes star, house type, price etc., and
It is common method that the information such as the overall ranking in hotel's picture and hotel, which push hotel to client, and this method can not be in conjunction with client's
Personal experience and preference meet the hotel of its interest to lead referral, personalized service of being also far from being, cannot effectively with visitor
Family preference is agreed with, can not be compared with approval of the raising client to the hotel recommended of limits, it would be highly desirable to improve.
Summary of the invention
In view of the defects and deficiencies of the prior art, the present invention intends to provide a kind of structure is simple, design rationally, make
With personalized hotel's intelligent recommendation algorithm easily based on customer action Habit Preference, solve the problems, such as that measurement index is single；
It solves the problems, such as customer personalized recommendation；Increase the 5th index " standoff distance " to assist determining under special circumstances
Plan makes algorithm closed loop.
To achieve the above object, the technical solution adopted by the present invention is that: steps are as follows for its calculating:
1, the hotel's fertilizer index for being directed to client is determined:
1.1, determine that four preferences for measuring hotels and client agree with the index of degree, that is, client moves in the time；Client enters
Firmly number；Client moves in unit price；Customer experience is divided equally；
1.2, the 5th index is determined, that is, standoff distance, in all very similar situation of aforementioned four index, it will make
Which determined to make algorithm closed loop for customer recommendation hotel with " standoff distance "；
2, according to fertilizer index developing algorithm input matrix:
In the description of step 1, there are five indexs in each hotel, construct a 5*1 column to each alternative hotel at this
Vector, the 5*1 vector be followed successively by from top to bottom client moves in the time, client moves in number, client move in unit price, customer experience it is equal
Divide, standoff distance；If there is n similar optional hotels, just there is n 5*1 vector here, this n 5*1 vector is from left to right spelled
It picks up and, just obtained a 5*n matrix, arrived this, the building of algorithm input matrix finishes；The input matrix is from the point of view of row, and first
Row indicates the specific number of days of the upper anniversary each hotel occupancy in optional n hotel of the client；Second row indicates on the client
The specific number of anniversary each hotel occupancy in optional n hotel；The third line indicated the client upper anniversary in optional n
The average unit price of each hotel occupancy in a hotel；And so on；Meanwhile each column of the input matrix all indicate that the customer selects
5 indexs when one hotel；
3, data needed for obtaining fertilizer index:
To alternative n similar hotels, each hotel described in step 1 and the customer are taken out from database
Five indexs for agreeing with degree of preference construct the matrix of a 5*n by the data of taking-up according to the description in step 2, as
The input matrix of algorithm；
4, preceding four rows data are pre-processed:
After step 1- step 3, the matrix for obtaining a 5*n is inputted as algorithm, but because only that preceding four
Client of the client that row data indicate in a certain hotel moves in the time, client moves in number, client moves in unit price, customer experience
Pca processing is respectively done as input (standoff distance is the index of an aid decision, does not make pca processing), it is therefore desirable to right
The algorithm input matrix of 5*n is handled, and is extracted preceding four row and is constituted a 4*n matrix；For this 4*n matrix, because it is every
The meaning of each index represented by a line is different, and the order of magnitude also differs larger, therefore before doing and handling in next step, to data into
Row whitening pretreatment；
5, corresponding characteristic value is obtained with pca algorithm process to by the pretreated first four achievement data of step 4
With feature vector；
6, Generalization bounds are chosen according to the size cases of characteristic value: if four characteristic values obtained are all (little close to 0
Thought in 0.1 close to 0), then according to the ranking results of the 5th index standoff distance, the lesser wine of preferential recommendation standoff distance
Shop；If four characteristic values obtained not necessarily all close to 0 (being thought if no more than 0.1 close to 0), the situation is according to first four index
The ranking results of pca pivot scoring, the biggish hotel of recommendation pca pivot scoring (and client's personal behavior preference compatible degree is more
It is high)；
7, use quicksort as specific sort method:
Generalization bounds are selected according to the size cases of characteristic value, if the Generalization bounds selected herein are according to the 5th finger
Mark standoff distance size recommend when, then the 5th index standoff distance in all optional hotels is ranked up, obtain it is specific which
A hotel is closer apart from client, to determine to which hotel of the lead referral；If the Generalization bounds selected are bases herein
Pca scoring is recommended, then algorithm is ranked up to the pca scoring in all optional hotels in next step, obtains which specific hotel
Pca scoring it is higher, and then determine recommend which hotel；
8, standoff distance or pca scoring after determining sequence correspond to which hotel in initial data: due to simple
Quicksort be that perhaps pca scoring from small big actually which standoff distance or pca scoring is discharged to standoff distance
Which, corresponding to hotel, family in former data, do not showed in above-mentioned sequence；So needing in addition to generate a hotel Ge Yu
The equal generic sequence array of quantity, represents the subscript of raw data matrix, its correspondence is followed in quicksort in step 7
Data it is mobile, finally export two arrays, first array is to carry out the 5th index in the optional hotel after ascending sort
" standoff distance " or pca score data, second array are the corresponding original subscripts of the data after sequence；
9, the output of algorithm and Generalization bounds selection: pass through above-mentioned steps, the output of algorithm is described in step 8
Two arrays, the pca scoring in first each hotel of array representation or standoff distance；Second array representation, first number
The serial number in hotel corresponding to the pca scoring of group or standoff distance；In practical application, according to the Generalization bounds of selection, if
It is then preferentially to select the serial number pair of first element in second array according to the ranking results of the 5th index " standoff distance "
The hotel answered is as recommendation target；If according to the ranking results that first four index p ca pivot scores, then preferential recommendation second
The corresponding hotel of the serial number of the last one element in array.
Further, the client in the step 1 moved in the time as the hotel client Ru Zhumou in a upper anniversary
Number of days (unit is day)；It was within a upper anniversary that client, which moves in number, which moves in the cumulative frequency in certain hotel, and (unit is
Number)；Client moved in unit price as within a upper anniversary, which moves in the average unit price (unit is member) in certain hotel；Client
Experience was divided within a upper anniversary, and the client is to the average score for moving in hotel；Standoff distance is in place for existing customer institute
The distance (unit is km) in the hotel Zhi Yu.
Further, the whitening pretreatment in the step 4, it is desirable to make the input of learning algorithm by whitening process
With following property, that is, correlation is lower between feature, all feature variances having the same；Therefore, x is enabledj (i)For 4*n square
Definite Index element in battle array, i indicate that the i-th column, j indicate that jth row, n indicate every a total of n data of row, concrete operation step
It is as follows:
4.1, zero averaging processing is carried out to data, enabling data mean value is 0；To every data line, SchillingX is used againj (i)-μjInstead of xj (i), realize zero averaging, wherein μjFor the mean value of all sample datas of jth row；
4.2, data are normalized, enabling data variance is 1；To every data line, SchillingIt uses againTo replaceWherein σjFor the standard deviation of all sample datas of jth row.
Further, the pca pivot scoring in the step 6, acquisition methods are as follows: the 4* that will be found out in step 5
4 feature vectors are denoted as
Wherein from left to right
The 4*1 vector A of first row1For pivot；The matrix that first four achievement data in step 4 Jing Guo whitening pretreatment is constituted is denoted asIt enablesWhereinFor 1*4 vector, B is 4*n square
Battle array, then C=[scores1 scores2…scoresn] it is 1*n vector, scoresiFor the pca pivot scoring in i-th of hotel.
Further, steps are as follows for the realization of quicksort in the step 7:
7.1, assume there is an array A [n] to be sorted, n to indicate the element number of array A [n]；If array A's [n]
Length n then terminates this step no more than 1 (can obtain the length of an array in matlab with " length " interface)；If number
The length of group A [n] is greater than 1, then performs the following operations: on the basis of first element of array A [n], being denoted as a, the position of a
It is denoted as k, since second element of array A [n], successively array A [n] is traversed, if the element ratio base element a traversed
Greatly, then it directly skips the element and continues to traverse next element until all elements traversal terminates；If the element ratio traversed
Base element a is small, then enables k+1 as new k value, element corresponding to the element greater than base element a and new k value is handed over
It changes, then proceedes to traverse next element until terminating；After to array A [n] traversal, by first element of array A [n]
A exchanges one wheel sequence of completion with new k value corresponding element；So far, the element on the left side a is respectively less than element a, the right a in array A [n]
Element be all larger than element a；
7.2, the array that all elements on the left side a in array A [n] are constituted is regarded as a new array, a in array A [n]
The element on the right equally regards another new array as, continues to repeat to walk the operation for playing a.
After adopting the above structure, the invention has the following beneficial effects: it is of the present invention a kind of based on customer action Habit Preference
Personalized hotel's intelligent recommendation algorithm, solve the problems, such as that measurement index is single；Solve asking for customer personalized recommendation
Topic；Increase the 5th index " standoff distance " and carry out aid decision under special circumstances, make algorithm closed loop, the present invention has structure letter
The advantages that list, setting is reasonable, low manufacture cost.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is the flow chart of specific embodiment.
Fig. 2 is the policy map that specific embodiment party is recommended according to " standoff distance ".
Fig. 3 is the policy map that specific embodiment is recommended according to pca marking.
Specific embodiment
The present invention will be further described below with reference to the drawings.
Referring to as shown in Figure 1-Figure 3, present embodiment the technical solution adopted is that: steps are as follows for its calculating:
1, the hotel's fertilizer index for being directed to client is determined:
1.1, determine that four preferences for measuring hotels and client agree with the index of degree, that is, client moves in the time --- on
The number of days (unit is day) in the hotel client Ru Zhumou in one anniversary；Client moves in number --- and it, should within a upper anniversary
Client moves in the cumulative frequency in certain hotel, and unit is number；Client moves in unit price --- and within a upper anniversary, which is moved in
The average unit price in certain hotel, unit are member；Customer experience is divided equally --- and within a upper anniversary, which puts down to hotel is moved in
Score；
1.2, the 5th index is determined, that is, at a distance from hotel, unit is for standoff distance --- existing customer position
Km, in all very similar situation of aforementioned four index, it will determine where be directed to the customer recommendation using " standoff distance "
A hotel makes algorithm closed loop；
2, according to fertilizer index developing algorithm input matrix:
In the description of step 1, there are five indexs in each hotel, construct a 5*1 column to each alternative hotel at this
Vector, the 5*1 vector be followed successively by from top to bottom client moves in the time, client moves in number, client move in unit price, customer experience it is equal
Divide, standoff distance；If there is n similar optional hotels, n takes 8 in the present embodiment, indicates that 8 hotels participate in recommending, here
There is n 5*1 vector, this n 5*1 vector is from left to right stitched together, a 5*n matrix has just been obtained, has arrived this, algorithm is defeated
Enter matrix building to finish；For the input matrix from the point of view of row, the first row indicated that the client upper anniversary was each in optional n hotel
The specific number of days of hotel occupancy；Second row indicate the client upper anniversary in optional n hotel each hotel occupancy it is specific
Number；The third line indicates the average unit price of the upper anniversary each hotel occupancy in optional n hotel of the client；And so on；
Meanwhile each column of the input matrix all indicate 5 indexs when the customer selects a hotel；
3, data needed for obtaining fertilizer index:
To alternative 8 similar hotels, each hotel described in step 1 and the customer are taken out from database
Five indexs for agreeing with degree of preference construct the matrix of a 5*n by the data of taking-up according to the description in step 2, as
The input matrix of algorithm；
4, preceding four rows data are pre-processed:
After step 1- step 3, the matrix for obtaining a 5*n is inputted as algorithm, but because only that preceding four
Client of the client that row data indicate in a certain hotel moves in the time, client moves in number, client moves in unit price, customer experience
Pca processing is respectively done as input (standoff distance is the index of an aid decision, does not make pca processing), it is therefore desirable to right
The algorithm input matrix of 5*n is handled, and is extracted preceding four row and is constituted a 4*n matrix；For this 4*n matrix, because it is every
The meaning of each index represented by a line is different, and the order of magnitude also differs larger, therefore before doing and handling in next step, to data into
Row whitening pretreatment；Wish to make by whitening process the input of learning algorithm that there is following property, that is, correlation between feature
It is lower, all feature variances having the same；Therefore, x is enabledj (i)For the definite Index element in 4*n matrix, i indicates the i-th column, j
Indicate that jth row, n indicate every a total of n data of row, specific steps are as follows:
4.1, zero averaging processing is carried out to data, enabling data mean value is 0；To every data line, SchillingX is used againj (i)-μjInstead of xj (i), realize zero averaging, wherein μjFor the mean value of all sample datas of jth row；
4.2, data are normalized, enabling data variance is 1；To every data line, SchillingIt uses againTo replaceWherein σjFor the standard deviation of all sample datas of jth row.
5, corresponding characteristic value is obtained with pca algorithm process to by the pretreated first four achievement data of step 4
With feature vector；
6, Generalization bounds are chosen according to the size cases of characteristic value: Generalization bounds one: as the spy of four selected indexs
Value indicative is all close to 0, i.e., when being all not more than 0.1, indicate four, each hotel finger target value of optional recommendation very close to,
Which hotel of preferential recommendation is determined according to the 5th index " standoff distance " at this time；As shown in attached drawing two, " data " expression is obtained
Certain customer's taken moves in hotel's primary data, " latent " indicates the characteristic value of preceding four rows data, at this time preceding four rows data
Characteristic value carries out hotel's recommendation according to the 5th index " standoff distance " all close to 0, therefore at this time, can by standoff distance sequence
Know, the 5th hotel's standoff distance is nearest, at this time the 5th hotel of preferential recommendation.
Generalization bounds two: if the case where characteristic value of above-mentioned four selected indexs is not both less than 0.1, is commented using pca
Which divide preferential recommendation hotel determined.As shown in Figure 3, the characteristic value for the preceding four rows data that " latent " is indicated not necessarily all connects
It is bordering on 0, is recommended at this time using pca scoring, by pca marking and queuing it is found that the 4th hotel pca scoring highest, indicates and care for
Objective behavioural habits preference compatible degree highest is directed to the 4th hotel of customer's preferential recommendation at this time；
The method for obtaining the scoring of pca pivot is as follows: the 4*4 feature vector that will be found out in step 5 is denoted as
Wherein from left to right
The 4*1 vector A of first row1For pivot；The matrix that first four achievement data in step 4 Jing Guo whitening pretreatment is constituted is denoted asIt enablesWhereinFor 1*4 vector, B is 4*n square
Battle array, then C=[scores1 scores2…scoresn] it is 1*n vector, scoresiFor the pca pivot scoring in i-th of hotel；With
For matlab, the 4*4 eigenvectors matrix obtained by pca algorithm process, each column are an ingredient coefficients, and from
Left-to-right descending sort, the i.e. first row from left to right of the 4*4 matrix are pivot direction, the first row 4*1 from left to right of the expression pivot
Vector transposition is 1*4 vector, then with by pretreated " data_df " matrix multiple, a 1*n pivot scoring can be obtained
Vector；
7, use quicksort as specific sort method:
Generalization bounds are selected according to the size cases of characteristic value, if the Generalization bounds selected herein are according to the 5th finger
Mark standoff distance size recommend when, then the 5th index standoff distance in all optional hotels is ranked up, obtain it is specific which
A hotel is closer apart from client, to determine to which hotel of the lead referral；If the Generalization bounds selected are bases herein
Pca scoring is recommended, then algorithm is ranked up to the pca scoring in all optional hotels in next step, obtains which specific hotel
Pca scoring it is higher, and then determine recommend which hotel；Steps are as follows for the realization of quicksort:
7.1, assume there is an array A [n] to be sorted, n to indicate the element number of array A [n]；If array A's [n]
Length n then terminates this step no more than 1 (can obtain the length of an array in matlab with " length " interface)；If number
The length of group A [n] is greater than 1, then performs the following operations: on the basis of first element of array A [n], being denoted as a, the position of a
It is denoted as k, since second element of array A [n], successively array A [n] is traversed, if the element ratio base element a traversed
Greatly, then it directly skips the element and continues to traverse next element until all elements traversal terminates；If the element ratio traversed
Base element a is small, then enables k+1 as new k value, element corresponding to the element greater than base element a and new k value is handed over
It changes, then proceedes to traverse next element until terminating；After to array A [n] traversal, by first element of array A [n]
A exchanges one wheel sequence of completion with new k value corresponding element；So far, the element on the left side a is respectively less than element a, the right a in array A [n]
Element be all larger than element a；
7.2, the array that all elements on the left side a in array A [n] are constituted is regarded as a new array, a in array A [n]
The element on the right equally regards another new array as, continues to repeat to walk the operation for playing a；
8, standoff distance or pca scoring after determining sequence correspond to which hotel in initial data: due to simple
Quicksort be that perhaps pca scoring from small big actually which standoff distance or pca scoring is discharged to standoff distance
Which, corresponding to hotel, family in former data, do not showed in above-mentioned sequence；So needing in addition to generate a hotel Ge Yu
The equal generic sequence array of quantity, represents the subscript of raw data matrix, its correspondence is followed in quicksort in step 7
Data it is mobile, finally export two arrays, first array is to carry out the 5th index in the optional hotel after ascending sort
" standoff distance " or pca score data, second array are the corresponding original subscripts of the data after sequence；
9, the output of algorithm and Generalization bounds selection: pass through above-mentioned steps, the output of algorithm is described in step 8
Two arrays, the pca scoring in first each hotel of array representation or standoff distance (pca scoring or waiting time,
Such as figure three " scores ")；Hotel corresponding to the pca scoring of second array representation, first array or standoff distance
Serial number；In practical application, according to the Generalization bounds of selection, if according to the ranking results of the 5th index " standoff distance ",
Then preferentially select the corresponding hotel of serial number of first element in second array as recommendation target；If being referred to according to first four
Mark pca pivot scoring ranking results, then in second array of preferential recommendation the last one element the corresponding hotel of serial number.No
By being recommended according to " standoff distance ", is recommended according further to pca scoring, require to be ranked up data.This hair
It is bright using special to target under initial data described in quicksort method described in step 7 in " summary of the invention " and step 8
Processing, has obtained simulation result shown in figure two and figure three.In the simulation result, " scores " indicates the data of ascending sort,
Subscript of " number " expression " scores " in raw data matrix, to identify " scores " correspond to it is actual which
Hotel.Therefore the subscript in " number " can be used directly and recommend hotel, and in this scenario, however, it is determined that Generalization bounds be according to
According to the 5th index " standoff distance ", then " scores " scores smaller hotel more by preferential recommendation, i.e., more leans in " number "
Hotel corresponding to preceding subscript is more by preferential recommendation；If which hotel of preferential recommendation determined using pca scoring,
" scores " scores bigger hotel more by preferential recommendation, i.e., the hotel corresponding to the subscript in " number " more rearward the excellent
First recommend
After adopting the above structure, present embodiment has the beneficial effect that
1, multi objective measures the degree of agreeing with for recommending hotel and customer action Habit Preference, solves that measurement index is single to ask
Topic；
2, client moves in the time to reasonable selection, client moves in number, client moves in unit price, customer experience divides equally four fingers
Mark selects the behavioural habits preference in hotel from four dimensions analysis client, solves the problems, such as customer personalized recommendation；
3, increase the 5th index " standoff distance " and carry out aid decision under special circumstances, make algorithm closed loop.
The above is only used to illustrate the technical scheme of the present invention and not to limit it, and those of ordinary skill in the art are to this hair
The other modifications or equivalent replacement that bright technical solution is made, as long as it does not depart from the spirit and scope of the technical scheme of the present invention,
It is intended to be within the scope of the claims of the invention.
Claims (5)
1. a kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference, it is characterised in that: its calculating step
It is rapid as follows:
(1), the hotel's fertilizer index for being directed to client is determined:
(1.1), determine that four preferences for measuring hotels and client agree with the index of degree, that is, client moves in the time；Client moves in
Number；Client moves in unit price；Customer experience is divided equally；
(1.2), the 5th index is determined, that is, standoff distance, in all very similar situation of aforementioned four index, it will use
" standoff distance " determines to make algorithm closed loop (2) for which hotel of the customer recommendation
(2), according to fertilizer index developing algorithm input matrix:
In the description of step (1), each hotel there are five index, at this to each alternative hotel construct a 5*1 arrange to
Amount, the 5*1 vector be followed successively by from top to bottom client moves in the time, client moves in number, client move in unit price, customer experience it is equal
Divide, standoff distance；If there is n similar optional hotels, just there is n 5*1 vector here, this n 5*1 vector is from left to right spelled
It picks up and, just obtained a 5*n matrix, arrived this, the building of algorithm input matrix finishes；The input matrix is from the point of view of row, and first
Row indicates the specific number of days of the upper anniversary each hotel occupancy in optional n hotel of the client；Second row indicates on the client
The specific number of anniversary each hotel occupancy in optional n hotel；The third line indicated the client upper anniversary in optional n
The average unit price of each hotel occupancy in a hotel；And so on；Meanwhile each column of the input matrix all indicate that the customer selects
5 indexs when one hotel；
(3), data needed for obtaining fertilizer index:
To alternative n similar hotels, each hotel described in step (1) is taken out from database and the customer is inclined
Good five indexs for agreeing with degree construct the matrix of a 5*n by the data of taking-up according to the description in step 2, as calculation
The input matrix of method；
(4), preceding four rows data are pre-processed:
After step (1)-step (3), the matrix for obtaining a 5*n is inputted as algorithm, but because only that preceding four
Client of the client that row data indicate in a certain hotel moves in the time, client moves in number, client moves in unit price, customer experience
Respectively pca processing is done as input, it is therefore desirable to handle the algorithm input matrix of 5*n, extract preceding four row and constitute
One 4*n matrix；For this 4*n matrix because the meaning of each index represented by its every a line is different, the order of magnitude also differ compared with
Greatly, therefore before doing processing in next step, whitening pretreatment is carried out to data；
(5), corresponding characteristic value is obtained with pca algorithm process to by step (4) pretreated first four achievement data
With feature vector；
(6), Generalization bounds are chosen according to the size cases of characteristic value: if four characteristic values obtained all close to 0, according to the
The ranking results of five index standoff distances, the lesser hotel of preferential recommendation standoff distance；If four characteristic values obtained are not
All close to 0, the ranking results which then scores according to first four index p ca pivot recommend the biggish wine of pca pivot scoring
Shop；
(7), use quicksort as specific sort method:
Generalization bounds are selected according to the size cases of characteristic value, if the Generalization bounds selected herein are according to the 5th index phase
When gauge is recommended from size, then the 5th index standoff distance in all optional hotels is ranked up, which specific wine obtained
Shop is closer apart from client, to determine to which hotel of the lead referral；If the Generalization bounds selected are commented according to pca herein
Divide and recommend, then algorithm is ranked up to the pca scoring in all optional hotels in next step, obtains the pca in which specific hotel
It scores higher, and then determines which hotel recommended；
(8), standoff distance or pca scoring after determining sequence correspond to which hotel in initial data: due to simple
Only standoff distance, perhaps pca scores big actually which standoff distance or pca scoring pair be discharged to from small to quicksort
Which hotel, family, does not show in above-mentioned sequence in Ying Yuyuan data；So needing in addition to generate the hotel a Ge Yu number
Equal generic sequence array is measured, the subscript of raw data matrix is represented, follows its correspondence in the quicksort in step (7)
Data it is mobile, finally export two arrays, first array is to carry out the 5th index in the optional hotel after ascending sort
" standoff distance " or pca score data, second array are the corresponding original subscripts of the data after sequence；
(9), the output of algorithm and Generalization bounds selection: pass through above-mentioned steps, the output of algorithm is described in step (8)
Two arrays, the pca scoring in first each hotel of array representation or standoff distance；Second array representation, first array
Pca scoring or standoff distance corresponding to hotel serial number；In practical application, according to the Generalization bounds of selection, if
According to the ranking results of the 5th index " standoff distance ", then preferentially the serial number of first element is corresponding in second array of selection
Hotel as recommend target；If according to the ranking results that first four index p ca pivot scores, then second number of preferential recommendation
The corresponding hotel of the serial number of the last one element in group.
2. a kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference according to claim 1,
Be characterized in that: the client in the step (1) moves in the number of days that the time was the hotel client Ru Zhumou in a upper anniversary；Visitor
It was within a upper anniversary that number is moved at family, which moves in the cumulative frequency in certain hotel；Client moves in unit price as at upper one
In anniversary, the average unit price customer experience which moves in certain hotel was divided within a upper anniversary, and the client is to moving in wine
The average score in shop；Standoff distance is existing customer position at a distance from hotel.
3. a kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference according to claim 1,
It is characterized in that: the whitening pretreatment in the step (4), it is desirable to by whitening process the input of learning algorithm be had such as
Lower property, that is, correlation is lower between feature, all feature variances having the same；Therefore, x is enabledj (i)For in 4*n matrix
Definite Index element, i indicate that the i-th column, j indicate that jth row, n indicate every a total of n data of row, specific steps are as follows:
(4.1), zero averaging processing is carried out to data, enabling data mean value is 0；To every data line, Schilling
X is used againj (i)-μjInstead of xj (i), realize zero averaging, wherein μjFor the mean value of all sample datas of jth row；
(4.2), data are normalized, enabling data variance is 1；To every data line, Schilling
It uses againTo replaceWherein σjFor the standard deviation of all sample datas of jth row.
4. a kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference according to claim 1,
Be characterized in that: the pca pivot scoring in the step (6), acquisition methods are as follows: the 4*4 feature that will be found out in step (5)
Vector is denoted as
Wherein from left to right first
The 4*1 vector A of column1For pivot；The matrix that first four achievement data in step 4 Jing Guo whitening pretreatment is constituted is denoted asIt enablesWhereinFor 1*4 vector, B is 4*n square
Battle array, then C=[scores1 scores2 … scoresn] it is 1*n vector, scoresiFor the pca pivot scoring in i-th of hotel.
5. a kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference according to claim 1,
Be characterized in that: steps are as follows for the realization of quicksort in the step (7):
(7.1), assume there is an array A [n] to be sorted, n to indicate the element number of array A [n]；If the length of array A [n]
It spends n and is not more than 1, then terminate this step；If the length of array A [n] is greater than 1, perform the following operations: with the of array A [n]
On the basis of one element, it is denoted as a, the position of a is denoted as k, since second element of array A [n], successively traverses array A
[n] directly skips the element and continues to traverse next element until all if the element ratio base element a traversed is big
Element traversal terminates；If the element ratio base element a traversed is small, enable k+1 as new k value, this is greater than benchmark member
The element of plain a is exchanged with element corresponding to new k value, then proceedes to traverse next element until terminating；To array A [n] time
After going through, first element a of array A [n] is exchanged to one wheel sequence of completion with new k value corresponding element；So far, array A [n]
The element on the middle left side a is respectively less than element a, and the element on the right of a is all larger than element a；
(7.2), the array that all elements on the left side a in array A [n] are constituted is regarded as a new array, a is right in array A [n]
The element on side equally regards another new array as, continues to repeat to walk the operation for playing a.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910034827.7A CN109933716A (en) | 2019-01-15 | 2019-01-15 | A kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910034827.7A CN109933716A (en) | 2019-01-15 | 2019-01-15 | A kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109933716A true CN109933716A (en) | 2019-06-25 |
Family
ID=66985024
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910034827.7A Pending CN109933716A (en) | 2019-01-15 | 2019-01-15 | A kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109933716A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112330391A (en) * | 2020-10-26 | 2021-02-05 | 武汉鼎森世纪科技有限公司 | Product recommendation method based on clients and employees |
| CN113032690A (en) * | 2019-12-09 | 2021-06-25 | 上海擎感智能科技有限公司 | Hotel recommendation method, recommendation system, readable storage medium and electronic equipment |
| CN115829421A (en) * | 2023-02-20 | 2023-03-21 | 杭银消费金融股份有限公司 | Intelligent analysis method and system for business requirements |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100293011A1 (en) * | 2009-05-18 | 2010-11-18 | Xavier Lebreton | Method and system of booking management |
| CN102855333A (en) * | 2012-09-27 | 2013-01-02 | 南京大学 | Service selection system based on group recommendation and selection method thereof |
| US20150248707A1 (en) * | 2014-03-03 | 2015-09-03 | Adara, Inc. | Hotel Recommendation Engine Based on Customer Data from Multiple Online Sources |
| CN105095477A (en) * | 2015-08-12 | 2015-11-25 | 华南理工大学 | Recommendation algorithm based on multi-index grading |
| CN106920151A (en) * | 2017-03-06 | 2017-07-04 | 携程旅游网络技术（上海）有限公司 | The recommendation method and system in the alternative pond in hotel |
| 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 |
-
2019
- 2019-01-15 CN CN201910034827.7A patent/CN109933716A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100293011A1 (en) * | 2009-05-18 | 2010-11-18 | Xavier Lebreton | Method and system of booking management |
| CN102855333A (en) * | 2012-09-27 | 2013-01-02 | 南京大学 | Service selection system based on group recommendation and selection method thereof |
| US20150248707A1 (en) * | 2014-03-03 | 2015-09-03 | Adara, Inc. | Hotel Recommendation Engine Based on Customer Data from Multiple Online Sources |
| CN105095477A (en) * | 2015-08-12 | 2015-11-25 | 华南理工大学 | Recommendation algorithm based on multi-index grading |
| CN106920151A (en) * | 2017-03-06 | 2017-07-04 | 携程旅游网络技术（上海）有限公司 | The recommendation method and system in the alternative pond in hotel |
| 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 |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113032690A (en) * | 2019-12-09 | 2021-06-25 | 上海擎感智能科技有限公司 | Hotel recommendation method, recommendation system, readable storage medium and electronic equipment |
| CN112330391A (en) * | 2020-10-26 | 2021-02-05 | 武汉鼎森世纪科技有限公司 | Product recommendation method based on clients and employees |
| CN112330391B (en) * | 2020-10-26 | 2022-07-08 | 武汉鼎森世纪科技有限公司 | Product recommendation method based on clients and employees |
| CN115829421A (en) * | 2023-02-20 | 2023-03-21 | 杭银消费金融股份有限公司 | Intelligent analysis method and system for business requirements |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103678518B (en) | Method and device for adjusting recommendation lists | |
| CN103325061B (en) | A kind of community discovery method and system | |
| CN111553754A (en) | Updating method and device of behavior prediction system | |
| CN109933716A (en) | A kind of personalized hotel's intelligent recommendation algorithm based on customer action Habit Preference | |
| US20150205801A1 (en) | Systems and methods for wine ranking | |
| CN109299372A (en) | A kind of wisdom curricula-variable recommended method based on collaborative filtering | |
| CN108573041B (en) | Probability matrix decomposition recommendation method based on weighted trust relationship | |
| CN107220365A (en) | Accurate commending system and method based on collaborative filtering and correlation rule parallel processing | |
| CN106168980A (en) | Multimedia resource recommends sort method and device | |
| CN110020121A (en) | Software crowdsourcing item recommendation method and system based on transfer learning | |
| WO2012162238A2 (en) | Method and system for creating events and matching users via blended profiles | |
| Wang et al. | Interactive social recommendation | |
| CN109508428A (en) | The point of interest recommended method excavated based on the true popularity of point of interest and implicit trust | |
| CN105354260B (en) | The Mobile solution of a kind of mosaic society's network and item characteristic recommends method | |
| CN103034963B (en) | A kind of service selection system and system of selection based on correlation | |
| CN105138508A (en) | Preference diffusion based context recommendation system | |
| Wu et al. | Continuous fuzzy kano model and fuzzy AHP model for aesthetic product design: case study of an electric scooter | |
| CN108829847A (en) | Commodity search method and system based on multi-modal shopping preferences | |
| CN107025311A (en) | A kind of Bayes's personalized recommendation method and device based on k nearest neighbor | |
| CN110263245A (en) | The method and apparatus for pushing object to user based on intensified learning model | |
| CN106777069A (en) | Personalized recommendation method and commending system based on matrix factorisation | |
| US20180197149A1 (en) | Automated methods and systems to schedule activities | |
| CN110008411A (en) | It is a kind of to be registered the deep learning point of interest recommended method of sparse matrix based on user | |
| CN109657048A (en) | One kind being applied to answerer's recommended method in open source Ask-Answer Community | |
| CN106204153A (en) | A kind of two-staged prediction Top N proposed algorithm based on attribute proportion similarity |
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 |