CN109214863B - Method for predicting urban house demand based on express delivery data - Google Patents
Method for predicting urban house demand based on express delivery data Download PDFInfo
- Publication number
- CN109214863B CN109214863B CN201810980223.7A CN201810980223A CN109214863B CN 109214863 B CN109214863 B CN 109214863B CN 201810980223 A CN201810980223 A CN 201810980223A CN 109214863 B CN109214863 B CN 109214863B
- Authority
- CN
- China
- Prior art keywords
- cell
- user
- express
- matrix
- time window
- 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
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000012384 transportation and delivery Methods 0.000 title claims description 61
- 230000008859 change Effects 0.000 claims abstract description 26
- 239000011159 matrix material Substances 0.000 claims description 86
- 230000006399 behavior Effects 0.000 claims description 51
- 239000013598 vector Substances 0.000 claims description 18
- 238000000354 decomposition reaction Methods 0.000 claims description 12
- 239000000126 substance Substances 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 8
- 238000012417 linear regression Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 238000011478 gradient descent method Methods 0.000 claims description 6
- 230000003993 interaction Effects 0.000 claims description 5
- 230000002093 peripheral effect Effects 0.000 claims description 4
- HSBKFSPNDWWPSL-CAHLUQPWSA-N 4-amino-5-fluoro-1-[(2r,5s)-5-(hydroxymethyl)-2,5-dihydrofuran-2-yl]pyrimidin-2-one Chemical compound C1=C(F)C(N)=NC(=O)N1[C@H]1C=C[C@@H](CO)O1 HSBKFSPNDWWPSL-CAHLUQPWSA-N 0.000 claims description 3
- 230000009471 action Effects 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 238000012821 model calculation Methods 0.000 claims description 3
- 239000004576 sand Substances 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000004931 aggregating effect Effects 0.000 claims 1
- 238000007418 data mining Methods 0.000 abstract description 3
- 238000007405 data analysis Methods 0.000 abstract description 2
- 230000000875 corresponding effect Effects 0.000 description 7
- 238000007726 management method Methods 0.000 description 4
- 238000013439 planning Methods 0.000 description 4
- 238000012935 Averaging Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/16—Real estate
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Business, Economics & Management (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Tourism & Hospitality (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Marketing (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Primary Health Care (AREA)
- Human Resources & Organizations (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Game Theory and Decision Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a method for predicting urban house demands based on express data, which relates to the field of urban multi-source data mining and analysis. According to the method, the possibility of the whole resident population flow of the community is estimated by analyzing the incidence relation between express behavior characteristics and the community where the user leaves the residence according to express behavior rules of the user expressed by express data, and further, the attraction of the community to the user and the community house demand change of the known event section are combined, a regression model is trained, and the community house demand of the next time section is estimated.
Description
Technical Field
The invention relates to the field of urban multi-source data mining and analysis, in particular to a method for predicting urban house requirements.
Background
With the development of economy and the improvement of the living standard of people, the house purchasing is the basic requirement of social life, the demand of houses in residential districts is predicted, the market management and regulation are facilitated, and the planning and construction of cities are facilitated. Meanwhile, with the development of the express service industry and the popularization of online shopping, more and more users choose to use express for conveying goods, and the express data generated by the method contains rich information (such as express receiving and sending time, express receiving and sending address, goods name, weight and the like), so that individual express behaviors and purchasing preference can be reflected. Meanwhile, the personal express delivery data in the same area are gathered together, and the behavior characteristics of people in the area (such as the whole express delivery frequency of community resident population, the inflow and outflow conditions of the community resident population and the like) can be reflected to a certain extent. One important aspect of reflecting the changes in the needs of a cell's premises is the mobility of the resident population in the cell. At present, a lot of work is carried out to analyze and estimate the movement situation of people in a certain range, the daily movement route or the travel mode of the people are generally analyzed and mined by using the call records or the travel records (such as the driving tracks, the riding records and the like) of the people in the area, however, the flow situation of resident people is a phenomenon which can be observed for a long time, the existing method is directed to the movement situation of the people in a short time, and therefore, the existing method is not suitable for analyzing the flow situation of the resident people in a cell by using the daily movement mode of the people. In addition, in the conventional methods, a method for analyzing and estimating the flow condition of the community resident population by using the express delivery data of the community residents is not provided. In addition to analyzing the population movement, factors that influence the choice of people to buy or sell the premises of the cell need to be considered in order to predict the needs of the premises of the cell. Some existing works consider a plurality of factors to estimate the room price, and patent CN107169847A proposes a system for dynamically adjusting the short rental room price, which combines the house source data, the landlord information, the resident demand information and the deal history information to extract features, and then estimates the room price through an artificial intelligence model. Patent CN103578057A proposes a real estate value estimation method based on an artificial neural network statistical model, which adopts a data mining and machine learning method in consideration of property attributes, environmental factors and the like of a cell, and selects a factor having the greatest influence on the real estate price. For people to choose whether to buy the housing of the community, the room price is only one of the factors, and the surrounding environment needs to be considered. In the existing methods, a method for predicting the house demand of a cell in combination of the flowing situation of resident population of the cell and the attraction of the cell to residents is not provided. Because the characteristics of the population can be reflected by the collection and analysis of a large amount of personal data, the flowing condition of resident population in a cell is mined through urban express delivery data, and the method has important significance in application scenes such as public safety, urban public security management and the like. In addition, the change of house requirements of the cell in the future is estimated by combining the population mobility situation with the attraction of the cell, and the method has the same significance in the application scenes of the business field, city planning, house price control and the like.
Disclosure of Invention
In order to overcome the defects of the prior art and to overcome the limitation that the existing crowd mobility and room price prediction method can not predict the house demand of a residential area according to the mobility of resident population of the residential area, the invention provides a method for predicting the house demand of an urban residential area based on express delivery data and the relevant multisource data of the residential area, the method mainly estimates the inflow and outflow conditions of resident population in a cell in a certain time range by analyzing information reflected by express delivery data, combines other data sources (cell information, peripheral interest point information and the like) to measure the attraction of the cell to people, and under the condition of knowing the related information of the previous time period, predicting the change of the demand of the house of the cell in the next time period, thereby correlating the multi-source data, the method has important significance in application scenes of public safety, urban public security management, business fields, urban planning, house price control and the like.
The technical scheme adopted by the invention for solving the technical problem comprises the following steps:
step 1: according to the delivery address and the receiving address of each piece of express data, classifying the user data belonging to the same cell according to the cell classification;
the method comprises the steps that firstly, longitude and latitude corresponding to each address in express data are obtained through an interface provided by an online map, clustering is carried out on each address according to the distance between the longitude and the latitude by using a DBSCAN clustering algorithm, all clustered clusters comprise other clusters besides a cell cluster, then the cluster corresponding to each cell is screened out according to keywords related to the cells and contained in the addresses in the clusters, and data containing the cell cluster after screening are obtained;
Step 2: aiming at the user data containing the cell cluster screened in the step 1, a user telephone field in the user data is used for uniquely identifying the user, historical express delivery data of the same user in the cell are aggregated and classified according to the user telephone information, and the personal express delivery behavior of each user is extractedThe characteristics of the individual express delivery behavior comprise: express delivery frequency of a user, average time interval of two express deliveries of the user, article type of express delivery of the user, article weight of express delivery of the user, and physical distance between two addresses of express delivery interaction of the user, and further a matrix R of express delivery behavior characteristics of each user in the same cell is constructedt mThe matrix Rt mEach row of (a) represents a user, and each column represents an express behavior feature;
and step 3: the matrix R obtained in step 2t mIf there is some missing part, the matrix R is decomposed by using the Regularized Singular Value Decomposition (RSVD) methodt mThe method comprises the following steps:
step 2.1: firstly, the express behavior characteristic matrixes of all cells in the same administrative area are merged into the same matrix
After that, will
The matrix is decomposed and expressed as formula (1):
u represents the number of users, F represents the number of express behavior characteristics, K represents the order of the decomposed matrix, and K is 40; d represents each administrative district (discrict), t represents a time window, and 
And represents the express delivery behavior feature matrix of all cells located in the administrative district d within the time window t, and the matrix is U x F dimension,
and
representation decomposition
The dimensionality of the two obtained matrixes is U multiplied by K and K multiplied by F respectively;
step 2.2: matrices derived from decomposition
And
calculating the vacancy value in the original matrix, and expressing the vacancy value as a formula (2):
wherein p isukTo represent
Line u of the matrix, qkfTo represent
Data of the f-th column of the matrix, buIs an offset, representing the characteristics of the cell itself,
from pukAnd q iskfPerforming inner product operation to obtain and express
The value of the position of the ith row and the fth column of the matrix is decomposed to obtain the matrix
And
obtained by learning through a gradient descent method, when the formula (3) reaches the minimum value,namely determine
And
the last three terms are penalty factors, the parameter lambda is 0.01, and the parameters are continuously updated by a gradient descent method
And
the intermediate value, finally obtaining the matrix which makes the formula (3) reach the minimum value, namely the decomposition matrix which is closest to the original matrix, and according to the original matrix
The position of the vacancy value will correspond to
And
performing inner product calculation on the rows and the columns to obtain a compensation value;
and 4, step 4: according to the characteristics supplemented in the step 3, the flow possibility of each user leaving the cell in three aspects of express delivery time, express delivery article type, express delivery article weight and interactive party distance is calculated, and the flow possibility of each user, namely the possibility of leaving the cell, is calculated, and the detailed steps are as follows:
Firstly, establishing a relation model between express behavior characteristics of a user and mobility of the user, and respectively defining time of express behavior generation, an article type of express delivery of the user, a relation between weight and distance of express articles of the user and the possibility that the user leaves a cell, wherein the relation between the time of express behavior and the possibility that the user leaves the cell is expressed as follows:
wherein the content of the first and second substances,
representing the estimated probability of leaving the cell of the nth user of the mth cell in the time window t in terms of the time of the express action, tsAnd teIndicating the start and end times of a time window, tfAnd tlExpressing the time of occurrence of the first express behavior and the last express behavior in a time window, wherein formula (4) expresses the possibility that a user who has an excessively fast express behavior recently leaves the cell within a period of time in the time window, and the size of the time window is determined according to the distribution of two express intervals of the user;
the relationship between the type of the item which is delivered by the user and the possibility that the user leaves the cell is expressed as follows:
wherein the content of the first and second substances,
indicating the estimated probability of leaving the cell of the nth user of the mth cell in terms of the type of the item to be delivered in the time window t, IntIndicates the type of item, ω, delivered by the nth user IA weight representing the item type reflecting the likelihood of the user leaving the cell, said item type using 17 classifications of e-commerce platform for goods;
the relationship between the weight and distance of the express item for the user and the possibility that the user leaves the cell is expressed as formula (6)
Wherein the content of the first and second substances,
represents the estimated probability of leaving the cell of the nth user of the mth cell in the time window t in terms of the express item weight and the transport distance, wntIndicating the weight, dis, of the express itemntIndicating the distance of express delivery;
the flow probability for each user is represented by a vector as:
then the flow probability construction matrix for all users in the same cell is:
averaging each column of the matrix to obtain the overall population mobility of the mth cell with respect to the three aspects, which is expressed by a vector:
the method comprises the following steps of establishing a matrix for the whole population mobility vectors of all m cells screened by express address data clustering in a city as follows:
the method is used for training a residential building demand prediction model;
and 5: acquiring cell information including cell house selling price, cell building year, cell house selling amount, cell administrative district and cell peripheral interest point information, calculating the diversity of cell POI by formula (7), and constructing a feature matrix of cell attraction by the features related to the cell 
Wherein CAt(Community Activity interaction) indicates that in the time window tThe attractiveness of all cells to the user,
the vector is a vector representing the attraction of the mth cell to users and consists of four characteristics representing the attraction of the cell, namely an administrative district where the cell is located, the construction year of the cell, the mean price of houses of the cell and the information entropy of POI types;
acquiring relevant information of each cell according to the cell name obtained in the step 1, wherein the relevant information comprises the number of sold houses and house source number of the cell in each month, the average price of the houses in each month, the building time of the cell and the administrative district where the cell is located; acquiring the number and the types of points of interest (POI) in a circular area of 500 meters around the cell by using an interface of an online map according to the longitude and latitude of the central Point of the cell range obtained in the step 1, extracting features capable of reflecting attraction of the cells to residents according to the related information, constructing a matrix for expressing the attraction features of each cell, mainly considering the administrative district where the cell is located, the year of building the cell, the average price of houses of the cell, the number and diversity of points of interest (POI) in a certain range around the cell, obtaining information such as the administrative district where the cell is located, the year of building, the average price of houses and the like from a network online house trading platform, obtaining the number and type of POI around the cell from an interface provided by an online map, and in addition, for the measurement of the diversity of POIs within the range of 500 meters around the cell, the information entropy of the POI category is calculated, as shown in formula (7):
Wherein p isiRepresenting the number of i-th POI around the cell;
constructing a matrix by using factors influencing the attraction of each cell to users:
CAt(Community Activity interaction) TableShowing the attraction of all cells to the user during the time window t,
indicating the attraction of the mth cell to the user;
step 6: the probability matrix of the cell resident population flowing obtained by calculation in the step 4 and the step 5
And a matrix CA of the attraction of a cell to a usertMerging and constructing demand characteristic matrix DFt(Demand Features), and
establishing a Linear Regression model (LR) according to a time window and demand characteristics of the change of house demand, taking a cell demand characteristic matrix as input, changing house demand of the next time window into output compared with the house demand of the time window, training the Linear Regression model, predicting the change of house demand of the next time window compared with the time window based on the cell demand characteristics of the time window through the trained Regression model, and predicting the change of house demand of the next time period compared with the previous time period through the trained Regression model on the premise of knowing the population flow condition of the cell of a certain time period, the attraction of the cell to users and the change of house demand of the cell;
The model calculation formula is as follows:
HDt＝DFt·W (8)
HDt(Housing Demand) represents the house Demand change of all cells in the city in a time window t compared with the house Demand change of a time window t-1, and is expressed by a vector
Known changes in house requirements
Calculated by equation (9):
wherein the content of the first and second substances,
representing the number of sources in the mth cell in time window t,
representing the house volume of the mth cell in the time window t,
indicating the number of sources in the mth cell in time window t-1,
representing the house volume of the mth cell in the time window t-1, and finally calculating the DF by the calculation formula (10)tWeight W corresponding to each feature:
after W is obtained, the DF for a new time window is input in reverset+1Calculating the change HD of the house demand in the new time window t +1 compared with the last time window t according to the formula (8)t+1。
The method has the advantages that the express delivery behavior rule of the user is expressed through express delivery data, the incidence relation between express delivery behavior characteristics and the residential area where the user leaves the residence is analyzed, the possibility that the whole resident population of the residential area flows is estimated, and then a regression model is trained according to the attraction of the residential area to the user and the change of the residential area house demand of the residential area in the known event section, and the residential area house demand of the next time section is estimated.
Drawings
Fig. 1 is a flowchart of a method for predicting urban house demand based on express delivery data in an embodiment of the present invention.
Fig. 2 is a distribution diagram of time intervals of express delivery receiving and sending behaviors of a user.
Detailed Description
The invention is further illustrated with reference to the following figures and examples.
As shown in fig. 1 and 2, the method for predicting the house demand of the urban community based on the express delivery data comprises the following specific processes:
step 1: according to the delivery address and the receiving address of each piece of express data, classifying the user data belonging to the same cell according to the cell classification;
the address writing mode of the same cell is multiple, firstly, the longitude and latitude corresponding to each address in express data are obtained through an interface provided by an online map, as the address contained in the express data is not only the type of the cell, the DBSCAN clustering algorithm is utilized to cluster each address according to the distance between the longitude and latitude, all clustered clusters also contain other clusters (such as enterprises, schools and the like) besides the cell cluster, then the cluster corresponding to each cell is screened out according to whether the address in the cluster contains keywords (such as keywords which can express the meaning of the cell) relevant to the cell, such as 'cell', 'community', 'home' and 'garden', etc.), and the data containing the cell cluster after screening is obtained;
Step 2: aiming at the user data containing the cell cluster screened in the step 1, a user telephone field in the user data is used for uniquely identifying the user, historical express delivery data of the same user in the cell is aggregated and classified according to user telephone information, and personal express delivery behavior characteristics of each user are extracted, wherein the method mainly comprises the following steps: the method comprises the steps of constructing a matrix of express behavior characteristics of each user in the same cell by the aid of express frequency (several express behaviors are shared in a time window), average time interval of two express behaviors of the user (the average time interval of two express behaviors and two adjacent express behaviors), article type of express delivery of the user, article weight of express delivery of the user and physical distance between addresses of two express interaction parties of the user, and further constructing a matrix of express behavior characteristics of each user in the same cell
Each row of the matrix represents a user, and each column represents an express behavior characteristic;
and step 3: because the data of the users in the original express delivery data are partially lost, the matrix of the behavior characteristics of the cell user express delivery obtained according to the original express delivery data in the step 2
There will also be some missing, so the matrix is sparse, and the matrix is decomposed by using the Regularized Singular Value Decomposition (RSVD) method 
The method for completing the vacancy value comprises the following steps:
step 2.1: firstly, the express behavior characteristic matrixes of all cells in the same administrative area are merged into the same matrix
Wherein d denotes each administrative district, which will be described later
The matrix is decomposed and expressed as formula (1):
u represents the number of users, F represents the number of express behavior characteristics, K represents the order of the decomposed matrix, and K is 40; d represents each administrative district (discrict), t represents a time window, and
and represents the express delivery behavior feature matrix of all cells located in the administrative district d within the time window t, and the matrix is U x F dimension,
and
representation decomposition
The dimensionality of the two obtained matrixes is U multiplied by K and K multiplied by F respectively; taking K of the invention as 40;
step 2.2: matrices derived from decomposition
And
calculating the vacancy value in the original matrix, and expressing the vacancy value as a formula (2):
wherein p isukTo represent
Line u of the matrix, qkfTo represent
Data of the f-th column of the matrix, buIs an offset quantity, represents the characteristics of the cell (the house selling price of the cell, the building year of the cell, the interest point information around the cell),
from pukAnd q iskfPerforming inner product operation to obtain and express
The value of the position of the ith row and the fth column of the matrix is originally a vacancy value, and because the similarity or difference exists among users of different cells, the characteristics of the cells are added as bias, so that the vacancy value is enabled to be Supplemented by the user characteristics in the cells with higher similarity, the characteristics of the cells themselves used in particular will be explained in detail in step 5, and the matrix obtained by decomposition
And
obtained by learning through a gradient descent method, when the formula (3) reaches the minimum value, the method determines that
And
the final three terms are penalty factors to prevent overfitting, the parameter lambda is 0.01, and the parameters are continuously updated by a gradient descent method
And
the intermediate value, finally obtaining the matrix which makes the formula (3) reach the minimum value, namely the decomposition matrix which is closest to the original matrix, and according to the position of the missing value of the original matrix, corresponding to the original matrix
And
performing inner product calculation on the rows and the columns to obtain a compensation value;
and 4, step 4: according to the characteristics supplemented in the step 3, calculating the flowing possibility of each user leaving the cell in three aspects of express delivery time, express delivery article type, express delivery article weight and interactive party distance, and the flowing possibility of each user, namely the possibility of leaving the cell;
firstly, establishing a relation model between express behavior characteristics of a user and mobility of the user, and respectively defining time of express behavior generation, an article type of express delivery of the user, a relation between weight and distance of express articles of the user and the possibility that the user leaves a cell, wherein the relation between the time of express behavior and the possibility that the user leaves the cell is expressed as follows:
Wherein the content of the first and second substances,
representing the estimated probability of leaving the cell of the nth user of the mth cell in the time window t in terms of the time of the express action, tsAnd teIndicating the start and end times of a time window, tfAnd tlThe time of the first express behavior and the time of the last express behavior in the time window are represented, the formula (4) represents the possibility that a user with the latest express behavior leaves the cell within a period of time in the time window, the size of the time window is determined according to the distribution of two express intervals of the user, and the time window is set to be two months;
the relationship between the type of the item which is delivered by the user and the possibility that the user leaves the cell is expressed as follows:
wherein the content of the first and second substances,
indicating the estimated probability of leaving the cell of the nth user of the mth cell in terms of the type of the item to be delivered in the time window t, IntIndicates the type of item, ω, delivered by the nth userIIndicating that the item type is reflecting the userWeight in probability of leaving the cell, said item type using 17 classifications of e-commerce platform for goods;
the relationship between the weight and distance of the express item for the user and the possibility that the user leaves the cell is expressed as formula (6)
Wherein the content of the first and second substances,
represents the estimated probability of leaving the cell of the nth user of the mth cell in the time window t in terms of the express item weight and the transport distance, wntIndicating the weight, dis, of the express itemntIndicating the distance of express delivery;
the flow probability for each user is represented by a vector as:
then the flow probability construction matrix for all users in the same cell is:
averaging each column of the matrix to obtain the overall population mobility of the mth cell with respect to the three aspects, which is expressed by a vector:
the method comprises the following steps of establishing a matrix for the whole population mobility vectors of all m cells screened by express address data clustering in a city as follows:
the method is used for training a residential building demand prediction model;
and 5: obtaining the information of the residential area, including the selling price of the residential area, the year of the residential area and the residential area houseThe sale amount Of houses, administrative districts where the districts are located and interest Point information around the districts are mainly considered, the years Of building the districts, the average price Of houses Of the districts and the number and diversity Of interest Points (POI) in a circular area with the radius Of 500 meters around the districts are mainly considered, the diversity Of the POI Of the districts is calculated by a formula (7), and the feature matrix Of the attraction Of the districts is constructed by the features related to the districts 
Wherein CAt(Community attraction) indicates the attraction of all cells to the user during the time window t,
the vector is a vector representing the attraction of the mth cell to users and consists of four characteristics representing the attraction of the cell, namely an administrative district where the cell is located, the construction year of the cell, the mean price of houses of the cell and the information entropy of POI types;
acquiring relevant information of each cell according to the cell name obtained in the step 1, wherein the relevant information comprises the number of sold houses and house source number of the cell in each month, the average price of the houses in each month, the building time of the cell and the administrative district where the cell is located; according to the longitude and latitude of the central Point of the cell range obtained in the step 1, the number and the types of interest Points (POI) in a circular area of 500 meters around the cell are obtained by using an interface of an online map, the number and the types of buildings with certain functions, such as a shop, a subway station, a park, a restaurant and the like, are extracted according to the information, the characteristics capable of reflecting the attraction of the cell to residents are reflected, and a matrix representing the attraction characteristics of each cell is constructed, wherein the administrative area where the cell is located, the year of building the cell, the average price of houses of the cell and the number and the diversity of the interest Points (POI) in a certain range around the cell are mainly considered. It should be noted that information such as an administrative district, a construction year, a house average price and the like where the cell is located can be obtained from a network online house trading platform, the number and types of POIs around the cell can be obtained from an interface provided by an online map, and in addition, for the measurement of the diversity of POIs within a certain range around the cell, the information entropy of the POI category is mainly calculated, as shown in formula (7):
Wherein p isiThe number of the ith POI around the cell is represented, and the larger the value is, the stronger the diversity of the POI in a certain range around the cell is;
constructing a matrix by using factors influencing the attraction of each cell to users:
CAt(Community attraction) indicates the attraction of all cells to the user during the time window t,
indicating the attraction of the mth cell to the user;
step 6: the probability matrix of the cell resident population flowing obtained by calculation in the step 4 and the step 5
And a matrix CA of the attraction of a cell to a usertMerging and constructing demand characteristic matrix DFt(Demand Features), and
establishing a Linear Regression model (LR) according to a time window and demand characteristics of the change of the house demand, taking a cell demand characteristic matrix as input, comparing the house demand of the following time window with the house demand of the time window to change into output, training the Linear Regression model, predicting the house demand change of the following time window compared with the time window based on the cell demand characteristics of the time window, and further knowing the population flow condition of the cell and the cell pairing use condition of the cell in a certain time period through the trained Regression model On the premise that the attraction of the user and the housing demand of the community change, the change of the housing demand of the community in the next time period compared with the previous time period can be predicted.
The model calculation formula is as follows:
HDt＝DFt·W (8)
HDt(Housing Demand) represents the house Demand change of all cells in the city in a time window t compared with the house Demand change of a time window t-1, and is expressed by a vector
Known changes in house requirements
Calculated by equation (9):
wherein the content of the first and second substances,
representing the number of sources in the mth cell in time window t,
representing the house volume of the mth cell in the time window t,
indicating the number of sources in the mth cell in time window t-1,
representing the house volume of the mth cell in time window t-1,
the larger the value of (D) is, the more users leave the cell is, the more users move in the cell, the demand is reduced, and finally the DF is obtained through the calculation formula (10)tWeight W corresponding to each feature:
after W is obtained, the DF for a new time window is input in reverset+1Calculating the change HD of the house demand in the new time window t +1 compared with the last time window t according to the formula (8)t+1。
The invention provides a method for predicting house requirements of a city community based on express data and community-related multi-source data, which is characterized in that express behavior rules of a user expressed by express data are analyzed, incidence relations between express behavior characteristics and communities where the user leaves the residence are analyzed, possibility of flowing of the whole resident population of the community is estimated, attractiveness of the community to people is measured by combining other data sources (community information, peripheral interest point information and the like), a regression model is trained under the condition that relevant information of a previous time period is known, changes of house requirements of the community in the next time period compared with the previous time period are predicted, and therefore multi-source data are correlated, and the method has important significance in application scenes of public safety, city public security management, the business field, city planning, house price control and the like.
Claims (1)
1. A method for predicting urban house demand based on express delivery data is characterized by comprising the following steps:
step 1: according to the delivery address and the receiving address of each piece of express data, classifying the user data belonging to the same cell according to the cell classification;
the method comprises the steps that firstly, longitude and latitude corresponding to each address in express data are obtained through an interface provided by an online map, clustering is carried out on each address according to the distance between the longitude and the latitude by using a DBSCAN clustering algorithm, all clustered clusters comprise other clusters besides a cell cluster, then the cluster corresponding to each cell is screened out according to keywords related to the cells and contained in the addresses in the clusters, and data containing the cell cluster after screening are obtained;
step 2: needleFor the user data containing the cell cluster screened in the step 1, uniquely identifying the user by using a 'user telephone' field in the user data, aggregating and classifying historical express delivery data of the same user in the cell according to user telephone information, and extracting the personal express delivery behavior characteristics of each user, wherein the personal express delivery behavior characteristics comprise: express delivery frequency of a user, average time interval of two express deliveries of the user, article type of express delivery of the user, article weight of express delivery of the user, and physical distance between addresses of two parties of express delivery interaction of the user, and further a matrix of express delivery behavior characteristics of each user in the same cell is constructed 
Matrix array
Each row of (a) represents a user, and each column represents an express behavior feature;
and step 3: the matrix obtained in step 2
If there is some missing, the matrix is decomposed by using regularized singular value
The method comprises the following steps:
step 2.1: firstly, the express behavior characteristic matrixes of all cells in the same administrative area are merged into the same matrix
After that, will
The matrix is decomposed and expressed as formula (1):
wherein, U represents the number of users, F represents the number of express behavior characteristics, K represents the order number of the decomposed matrix, and K is 40; d represents each administrative district, and t represents a time window, then
And represents the express delivery behavior feature matrix of all cells located in the administrative district d within the time window t, and the matrix is U x F dimension,
and
representation decomposition
The dimensionality of the two obtained matrixes is U multiplied by K and K multiplied by F respectively;
step 2.2: matrices derived from decomposition
And
calculating the vacancy value in the original matrix, and expressing the vacancy value as a formula (2):
wherein p isukTo represent
Line u of the matrix, qkfTo represent
Matrix arrayData of the f-th column, buIs an offset, representing the characteristics of the cell itself,
from pukAnd q iskfPerforming inner product operation to obtain and express 
The value of the position of the ith row and the fth column of the matrix is decomposed to obtain the matrix
And
obtained by learning through a gradient descent method, when the formula (3) reaches the minimum value, the method determines that
And
the last three terms are penalty factors, the parameter lambda is 0.01, and the parameters are continuously updated by a gradient descent method
And
the intermediate value, finally obtaining the matrix which makes the formula (3) reach the minimum value, namely the decomposition matrix which is closest to the original matrix, and according to the original matrix
The position of the vacancy value will correspond to
And
performing inner product calculation on the rows and the columns to obtain a compensation value;
and 4, step 4: according to the characteristics supplemented in the step 3, the flow possibility of each user leaving the cell in three aspects of express delivery time, express delivery article type, express delivery article weight and interactive party distance is calculated, and the flow possibility of each user, namely the possibility of leaving the cell, is calculated, and the detailed steps are as follows:
establishing a relation model between express behavior characteristics of a user and mobility of the user, and respectively defining time of express behavior generation, an article type of express delivery of the user, a relation between weight and distance of express articles of the user and the possibility of the user leaving a cell, wherein the relation between the time of express behavior and the possibility of the user leaving the cell is expressed as follows:
Wherein the content of the first and second substances,
representing the estimated probability of leaving the cell of the nth user of the mth cell in the time window t in terms of the time of the express action, tsAnd teIndicating the start and end times of a time window, tfAnd tlExpressing the time of occurrence of the first express behavior and the last express behavior in a time window, wherein formula (4) expresses the possibility that a user who has an excessively fast express behavior recently leaves the cell within a period of time in the time window, and the size of the time window is determined according to the distribution of two express intervals of the user;
the relationship between the type of the item which is delivered by the user and the possibility that the user leaves the cell is expressed as follows:
wherein the content of the first and second substances,
indicating the estimated probability of leaving the cell of the nth user of the mth cell in terms of the type of the item to be delivered in the time window t, IntIndicates the type of item, ω, delivered by the nth userIA weight representing the item type reflecting the likelihood of the user leaving the cell, said item type using 17 classifications of e-commerce platform for goods;
the relationship between the weight and distance of the express item for the user and the possibility that the user leaves the cell is expressed as formula (6)
Wherein the content of the first and second substances,
represents the estimated probability of leaving the cell of the nth user of the mth cell in the time window t in terms of the express item weight and the transport distance, w ntIndicating the weight, dis, of the express itemntIndicating the distance of express delivery;
the flow probability for each user is represented by a vector as:
then the flow probability construction matrix for all users in the same cell is:
for matrixThe average operation of each column of (a) is to obtain the overall population mobility of the cell of the mth cell with respect to the three aspects, and the overall population mobility of the cell is expressed by a vector:
the method comprises the following steps of establishing a matrix for the whole population mobility vectors of all m cells screened by express address data clustering in a city as follows:
the method is used for training a residential building demand prediction model;
and 5: acquiring cell information including cell house selling price, cell building year, cell house selling amount, cell administrative district and cell peripheral interest point POI information, wherein the diversity of the cell POI is calculated by a formula (7), and the feature matrix of the cell attraction is constructed by the features related to the cell
Wherein CAtIndicating the attractiveness of all cells to the user during the time window t,
the vector is a vector representing the attraction of the mth cell to users and consists of four characteristics representing the attraction of the cell, namely an administrative district where the cell is located, the construction year of the cell, the mean price of houses of the cell and the information entropy of POI types;
Acquiring relevant information of each cell according to the cell name obtained in the step 1, wherein the relevant information comprises the number of sold houses and house source number of the cell in each month, the average price of the houses in each month, the building time of the cell and the administrative district where the cell is located; acquiring the number and types of interest points in a circular area of 500 meters around the cell according to the longitude and latitude of the central point of the cell range obtained in the step 1, extracting features capable of reflecting the attraction of the cell to residents according to related information, and constructing a matrix representing the attraction features of each cell, wherein the administrative area where the cell is located, the year of building the cell, the average price of houses of the cell and the number and diversity of the interest points POI in a certain range around the cell are mainly considered, the administrative area where the cell is located, the building year and the average price of houses are all obtained from a network online house trading platform, the number and types of the POI around the cell are obtained from an interface provided by an online map, and in addition, for the measurement of the diversity of the POI in the range of 500 meters around the cell, the entropy of the information of the POI types is calculated, and is shown in a formula (7):
wherein p isiRepresenting the number of i-th POI around the cell;
constructing a matrix by using factors influencing the attraction of each cell to users:
CAtIndicating the attractiveness of all cells to the user during the time window t,
indicating the attraction of the mth cell to the user;
step 6: the probability matrix of the cell resident population flowing obtained by calculation in the step 4 and the step 5
And a matrix CA of the attraction of a cell to a usertMerging and constructing demand characteristic matrix DFtAnd is and
establishing a linear regression model LR according to the time window and the demand characteristics of the house demand change, taking a cell demand characteristic matrix as input, changing the house demand of the following time window into output compared with the house demand of the time window, and training linearityThe method comprises the steps of predicting the house demand change of a next time window compared with a time window through a regression model obtained through training based on the cell demand characteristics of the time window, and predicting the house demand change of the next time period compared with the previous time period through the trained regression model on the premise that the population mobility condition of the cell, the attraction of the cell to users and the house demand change of the cell are known in a certain time period;
the model calculation formula is as follows:
HDt＝DFt·W (8)
HDtrepresenting the house requirement change of all cells in the city in a time window t compared with the house requirement change of a time window t-1, and expressed by a vector 
Known changes in house requirements
Calculated by equation (9):
wherein the content of the first and second substances,
representing the number of sources in the mth cell in time window t,
representing the house volume of the mth cell in the time window t,
indicating the number of sources in the mth cell in time window t-1,
representing the house volume of the mth cell in the time window t-1, and finally calculating the DF by the calculation formula (10)tWeight W corresponding to each feature:
after W is obtained, the DF for a new time window is input in reverset+1Calculating the change HD of the house demand in the new time window t +1 compared with the last time window t according to the formula (8)t+1。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810980223.7A CN109214863B (en) | 2018-08-27 | 2018-08-27 | Method for predicting urban house demand based on express delivery data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810980223.7A CN109214863B (en) | 2018-08-27 | 2018-08-27 | Method for predicting urban house demand based on express delivery data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109214863A CN109214863A (en) | 2019-01-15 |
| CN109214863B true CN109214863B (en) | 2021-08-03 |
Family
ID=64989752
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810980223.7A Active CN109214863B (en) | 2018-08-27 | 2018-08-27 | Method for predicting urban house demand based on express delivery data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109214863B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111523005B (en) * | 2019-02-02 | 2023-08-11 | 北京嘀嘀无限科技发展有限公司 | Network contract user analysis method and device and electronic equipment |
| CN111784018A (en) * | 2019-04-03 | 2020-10-16 | 北京嘀嘀无限科技发展有限公司 | Resource scheduling method and device, electronic equipment and storage medium |
| CN110717710A (en) * | 2019-09-24 | 2020-01-21 | 广东聚宝屋传媒有限公司 | Demand information analysis method and device based on express package |
| CN111369058B (en) * | 2020-03-05 | 2022-06-14 | 中国民用航空飞行学院 | Forest fire fighting helicopter demand prediction method and system |
| CN111651727B (en) * | 2020-05-29 | 2023-09-26 | 京东城市(北京)数字科技有限公司 | Method and device for calculating overall population transfer quantity among areas |
| CN117313981B (en) * | 2023-07-11 | 2024-05-14 | 厦门身份宝网络科技有限公司 | Dynamic urban village population analysis method, device, equipment and readable storage medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104616113A (en) * | 2015-02-13 | 2015-05-13 | 安徽科成信息科技有限公司 | Logistic management method and system based on big data |
| CN105224994A (en) * | 2014-07-03 | 2016-01-06 | 富士通株式会社 | The apparatus and method of the prediction residence time, the equipment determining delivery route and method |
| US9298196B2 (en) * | 2010-11-19 | 2016-03-29 | Google Inc. | Energy efficiency promoting schedule learning algorithms for intelligent thermostat |
| CN106710083A (en) * | 2016-12-02 | 2017-05-24 | 刘志超 | Intelligent express method based on data analysis and terminal system |
| CN107220911A (en) * | 2017-05-10 | 2017-09-29 | 深圳市易图资讯股份有限公司 | A kind of real estate market and information of real estate management system |
| CN107730404A (en) * | 2017-09-25 | 2018-02-23 | 璧垫尝 | A kind of service facility interoperability methods and commodity selling method based on platform |
| CN107766402A (en) * | 2017-06-27 | 2018-03-06 | 深圳市云房网络科技有限公司 | A kind of building dictionary cloud source of houses big data platform |
| CN107993179A (en) * | 2018-01-04 | 2018-05-04 | 南京市公安局栖霞分局 | A kind of police service platform population house data examination register method |
| CN108053096A (en) * | 2017-11-22 | 2018-05-18 | 链家网（北京）科技有限公司 | Items in commerce real estate address approach and device |
| CN108090643A (en) * | 2016-11-21 | 2018-05-29 | 莫元劲 | Express logistics dispatching method and device |
| CN109766956A (en) * | 2018-07-19 | 2019-05-17 | 西北工业大学 | Method for detecting abnormality based on express delivery big data |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2015201064A1 (en) * | 2014-03-03 | 2015-09-17 | Sanders Noonan Pty Ltd | Frameworks and methodologies configured to determine probabilistic desire for goods and/or services |
-
2018
- 2018-08-27 CN CN201810980223.7A patent/CN109214863B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9298196B2 (en) * | 2010-11-19 | 2016-03-29 | Google Inc. | Energy efficiency promoting schedule learning algorithms for intelligent thermostat |
| CN105224994A (en) * | 2014-07-03 | 2016-01-06 | 富士通株式会社 | The apparatus and method of the prediction residence time, the equipment determining delivery route and method |
| CN104616113A (en) * | 2015-02-13 | 2015-05-13 | 安徽科成信息科技有限公司 | Logistic management method and system based on big data |
| CN108090643A (en) * | 2016-11-21 | 2018-05-29 | 莫元劲 | Express logistics dispatching method and device |
| CN106710083A (en) * | 2016-12-02 | 2017-05-24 | 刘志超 | Intelligent express method based on data analysis and terminal system |
| CN107220911A (en) * | 2017-05-10 | 2017-09-29 | 深圳市易图资讯股份有限公司 | A kind of real estate market and information of real estate management system |
| CN107766402A (en) * | 2017-06-27 | 2018-03-06 | 深圳市云房网络科技有限公司 | A kind of building dictionary cloud source of houses big data platform |
| CN107730404A (en) * | 2017-09-25 | 2018-02-23 | 璧垫尝 | A kind of service facility interoperability methods and commodity selling method based on platform |
| CN108053096A (en) * | 2017-11-22 | 2018-05-18 | 链家网（北京）科技有限公司 | Items in commerce real estate address approach and device |
| CN107993179A (en) * | 2018-01-04 | 2018-05-04 | 南京市公安局栖霞分局 | A kind of police service platform population house data examination register method |
| CN109766956A (en) * | 2018-07-19 | 2019-05-17 | 西北工业大学 | Method for detecting abnormality based on express delivery big data |
Non-Patent Citations (3)
| Title |
|---|
| Mobile Crowd Sensing and Computing;Bin Guo;《 ACM Computing Surveys 》;20151231;全文 * |
| 社会感知计算：概念、问题及其研究进展;於志文;《计算机学报》;20120131;全文 * |
| 移动群智感知多任务参与者优选方法研究;刘琰;《计算机学报》;20170831;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109214863A (en) | 2019-01-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109214863B (en) | Method for predicting urban house demand based on express delivery data | |
| Sun et al. | Forecasting tourist arrivals with machine learning and internet search index | |
| Ma et al. | Analyzing driving factors of land values in urban scale based on big data and non-linear machine learning techniques | |
| Chen et al. | Understanding ridesplitting behavior of on-demand ride services: An ensemble learning approach | |
| He et al. | Customer preference and station network in the London bike-share system | |
| Straka et al. | Predicting popularity of electric vehicle charging infrastructure in urban context | |
| Gao et al. | Moment-based rental prediction for bicycle-sharing transportation systems using a hybrid genetic algorithm and machine learning | |
| CN105493109B (en) | Inferred using the air quality of multiple data sources | |
| Lv et al. | Mining user similarity based on routine activities | |
| Mimis et al. | Property valuation with artificial neural network: The case of Athens | |
| CN111582559B (en) | Arrival time estimation method and device | |
| CN111178611B (en) | Method for predicting daily electric quantity | |
| Lorenz et al. | Interpretable machine learning for real estate market analysis | |
| US20240086726A1 (en) | Systems and methods for big data analytics | |
| Kostic et al. | Deep survival modelling for shared mobility | |
| CN115422441A (en) | Continuous interest point recommendation method based on social space-time information and user preference | |
| Allahviranloo et al. | A comparison of time-use behavior in metropolitan areas using pattern recognition techniques | |
| Maleki et al. | Developing a supervised learning-based simulation method as a decision support tool for rebalancing problems in bike-sharing systems | |
| CN110309953B (en) | Urban security monitoring layout system and method adopting target mobility distribution prediction | |
| Wang et al. | Computer supported data-driven decisions for service personalization: a variable-scale clustering method | |
| CN115545758A (en) | Method and system for self-adaptive incremental site selection of urban service facilities | |
| Moosavi | Urban data streams and machine learning: a case of swiss real estate market | |
| Kim et al. | A comparative study of aggregate and disaggregate gravity models using Seoul metropolitan subway trip data | |
| Zhou et al. | Refined taxi demand prediction with ST-Vec | |
| Gong et al. | Automated prediction of shopping behaviours using taxi trajectory data and social media reviews |
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 |