CN110288125B - Commuting model establishing method based on mobile phone signaling data and application - Google Patents

Abstract

The invention provides a commuting model establishing method based on mobile phone signaling data, which is used for establishing a commuting model of a city according to mobile phone signaling data acquired by mobile phone base stations in the city, and is characterized by comprising the following steps: step S1, acquiring mobile phone signaling data and analyzing the mobile phone signaling data to obtain user commuting data; s2, calculating the distribution weight of the mobile phone base station and the urban space unit around the mobile phone base station according to the base station position information of the mobile phone base station, and S3, distributing user commuting data according to the distribution weight to obtain unit commuting data; s4, constructing a unit commuting model; s5, calculating to obtain a residual error of the unit commuting model according to the commuting amount calculated by the unit commuting model and the actual commuting amount; s6, clustering the residual errors in a spatial clustering mode and carrying out variable quantization processing to generate a residual error virtual variable; and S7, substituting the residual virtual variable into the unit commuting model to obtain a residual commuting model.

Description

A commuting model establishment method and application based on mobile phone signaling data

technical field

The invention belongs to the field of urban planning, and in particular relates to a method for establishing a commuting model based on mobile phone signaling data and an application thereof.

Background technique

The concept of the commuter model comes from urban traffic planning. In fact, in the traffic model system, the concept of the commuter model is equivalent to home-based work travel (home-based work, HBW), that is, starting from home to work. At present, the traditional gravity model (gravity model) is mainly used in the modeling of this kind of travel in the field of urban planning and urban transportation. Specifically, based on large-scale traffic survey data, a city-level commuting travel model is constructed, and with the help of this model, the impact of different residential land and job layouts on city-wide commuting behavior is simulated.

The current commuting model has three major flaws:

The first is the flaw in the data layer. Traditional data sources for commuting models are mainly large-scale census data, such as population census and traffic census. Compared with general questionnaire survey data, large-scale census data are characterized by large sample size and high coverage rate, but the disadvantage is that the survey cost is too high and consumes a lot of manpower and material resources, and it can only be conducted once every 10 years or so. Moreover, the population distribution and traffic conditions of cities will change dramatically within ten years. It is conceivable that the utility of census data will gradually decrease over time.

The second is the flaw in the method level. Taking Shanghai as an example, the current traffic model system in Shanghai is mainly based on traffic survey data and traditional gravity models (refer to the works of Chen Bizhuang, Lu Ximing, Dong Zhiguo, etc.). The shortcomings of the traffic survey data have shown that the shortcomings of the traditional gravity model are poor scalability, which directly leads to the poor prediction effect of the model on commuting behavior. In fact, in addition to urban planning and urban transportation, commuting models are also studied in other fields, typically the field of spatial econometrics. Compared with the traditional gravity model, the spatial regression model in the field of spatial measurement has better scalability, which means that more factors or variables can be considered in the model, but the defect of the spatial regression model is that the theory is relatively complicated. Not easy to master.

Finally, there are flaws at the application level. Insufficient data and methods directly lead to the poor fitting effect of the model, and indirectly lead to defects in the application of the model, that is, using a model with poor prediction effect to guide planning practice will produce deviations in results and unreasonable results.

Contents of the invention

In order to solve the above problems, a method for establishing a commuting model that optimizes the commuting model through mobile phone signaling big data is provided. The present invention adopts the following technical solutions:

The invention provides a method for establishing a commuting model based on mobile phone signaling data, which is used to establish the commuting model of the city according to the mobile phone signaling data collected by each mobile phone base station in the city, which is characterized in that it includes the following steps: Step S1 , obtain the mobile phone signaling data and analyze the user’s commuting data including the user’s departure base station and employment base station according to the mobile phone signaling data; step S2, calculate the predetermined number of mobile phone base stations and the surrounding areas of the mobile phone base station according to the base station location information of the mobile phone base station The distribution weight of the urban space unit, step S3, assign the user commuting data to the urban space unit according to the distribution weight to obtain the unit commuting data including the user's departure unit and employment unit; step S4, construct the unit commuting model, the unit The commuting model is of the form:

ln T _ij ＝κ _i +α _i ln N _j +β _i ln d _ij +ε _ij

In the formula, T _ij is the amount of commuting between urban spatial units, N _j is the number of jobs in the j-th employment unit, d _ij is the commuting cost between the i-th departure unit and the j-th employment unit , α _i , β _i are the population impact coefficient and employment impact coefficient of the i-th departure unit respectively, κ _i is the constant term of the i-th departure unit, ε _ij is the relationship between the i-th departure unit and the j-th Residuals between units of employment; step S5, calculate the residuals {R _n , X _n , Y _n } of the commuting model of the unit according to the commuting amount calculated by the commuting model of the unit and the actual commuting amount, where R _n Represents the absolute value of the residual corresponding to the nth urban spatial unit, X _n and Y _n represent the plane coordinates of the nth urban spatial unit; step S6, spatially aggregate the residual {R _n , X _n , Y _n } Clustering by class pattern and classifying according to the preset classification method to obtain 4 clustering types, and further variableize the 4 clustering types to generate a residual dummy variable; step S7, substitute the residual dummy variable into the unit commuting The model obtains the residual commuting model, and the form of the residual commuting model is as follows:

ln T _ij ＝κ _i +α _i ln N _j +β _i ln d _ij +∑ _k α _k D_SE _k +ε _ij

In the formula, D_SE _k is the residual of the clustering type corresponding to the kth class, the value of k is [0,1,2,3], the value of D_SE _k is [0,1], and α _k is the corresponding residual difference coefficient.

The commuting model optimization method based on mobile phone signaling data provided by the present invention can also have such technical features, wherein the mobile phone signaling data includes base station information and time information generated when mobile phone communication, and step S1 includes the following sub-steps: step S1-1, obtain the mobile phone signaling data; step S1-2, identify the user's departure base station and employment base station according to the base station information and time information; step S1-3, according to the mobile phone signaling data, departure base station and employment base station Generate user commute data.

The commuting model optimization method based on mobile phone signaling data provided by the present invention can also have such technical features, wherein, step S2 includes the following sub-steps: Step S2-1, obtaining the base station location information of the mobile phone base station and the unit of the urban space unit Position information; step S2-2, obtain the quantity of its surroundings for each mobile phone base station as the predetermined quantity and the nearest urban space unit as the surrounding space unit of each mobile phone base station; step S2-3, according to the base station position information and the unit position information Calculate respectively the adjacent distances of each mobile phone base station and each surrounding space unit corresponding to it; Step S2-4, successively determines that the maximum value in all adjacent distances corresponding to the mobile phone base station is used as the maximum adjacent distance for each mobile phone base station, and further Calculate and obtain the bandwidth distance corresponding to the mobile phone base station according to the maximum adjacent distance; step S2-5, calculate the distribution weight of the mobile phone base station and surrounding space units according to the bandwidth distance and the adjacent distance.

The commuting model optimization method based on mobile phone signaling data provided by the present invention can also have such technical features, wherein the preset classification method is to classify according to the first residual and the second residual, and the first residual is urban space The residual of the unit, and the second residual is the residual of the urban spatial unit surrounding the urban spatial unit corresponding to the first residual.

The commuting model optimization method based on mobile phone signaling data provided by the present invention may also have such a technical feature, wherein the urban space unit is a neighborhood committee unit.

The commuting model optimization method based on mobile phone signaling data provided by the present invention may also have such a technical feature, wherein the commuting cost is commuting time or commuting distance.

The commuting model optimization method based on mobile phone signaling data provided by the present invention may also have such technical features, wherein the predetermined number is 30.

The present invention also provides an application of the residual commuting model established according to the commuting model establishment method, which is characterized in that, through an urban commuting data analysis system with a residual commuting model based on the job post in the city preset by the analyst Quantity Analyze the mobile phone signaling data of the city to obtain the commuting data of the city's residents. The urban commuting data analysis system includes: a commuting model storage department, which stores the residual commuting model; The mobile phone signaling data collected by each mobile phone base station; the signaling data analysis and acquisition part is used to analyze the mobile phone signaling data to obtain user commuting data including the user's departure base station and employment base station; the distribution weight calculation part uses According to the base station location information of the mobile phone base station, the distribution weight of the mobile phone base station and a predetermined number of urban space units around the mobile phone base station is calculated; the commuting data distribution part distributes the commuting data of the user to the urban space unit according to the distribution weight, so as to obtain the starting place including the user The unit commuting data of the unit and the unit of employment; the commuting data calculation and acquisition department calculates the unit commuting data and the number of jobs through the residual commuting model to obtain the residents' commuting data.

The urban commuting data calculation system provided by the present invention may also have such technical features, wherein the residents' commuting data is the average commuting distance or average commuting time of the city.

Invention function and effect

According to the method for establishing a commuting model based on mobile phone signaling data of the present invention, the user commuting data corresponding to each mobile phone base station is calculated according to the mobile phone signaling data, and the distribution weights between the mobile phone base station and each urban space unit are calculated, which realizes the user The commuting data is evenly distributed to each urban space unit to obtain the unit commuting data corresponding to each urban space unit, which solves the problem that the urban traffic data used in urban planning cannot evenly cover each unit, thereby optimizing the data source used in model building; Also, since the signaling data of the mobile phone can be updated quickly, the real-time performance of the commuting model can be better optimized. At the same time, the present invention establishes a unit commuting model on the basis of the traditional gravity model, which solves the problem that the traditional gravity model cannot analyze the "direction" of commuting; further, the present invention also obtains the residual commuting model by calculating residual optimization , so that the goodness of fit of the model is further improved, thereby greatly improving the prediction effect of the model, and better guiding the urban land use layout.

Description of drawings

Fig. 1 is the flow chart of the commuting model establishment method in the embodiment of the present invention;

Fig. 2 is a schematic diagram of unit commuting data in an embodiment of the present invention;

Fig. 3 is the schematic diagram of employment attractiveness difference in the embodiment of the present invention;

Fig. 4 is a comparative schematic diagram of the goodness of fit of each model in the embodiment of the present invention;

Fig. 5 is the structural block diagram of city commuting data analysis system in the embodiment of the present invention;

Fig. 6 is a schematic diagram of the commuting distance of the whole city of Shanghai in the embodiment of the present invention; and

Fig. 7 is a flow chart of the communication data analysis process in the embodiment of the present invention.

Detailed ways

In recent years, urban commuting research based on mobile positioning big data has begun to emerge, and more accurate commuting models can be constructed through big data. However, the current research on big data mostly stays at the level of qualitative description of the status quo of urban commuting space, and there are few in-depth studies on urban models, which cannot guide planning practice well. In layman's terms, the existing big data research is to answer the question of "what", rather than the question of "why" and "how" in the future. The present invention aims to excavate the commuting rules of the city and construct the commuting model of the city through the big data of mobile phone signaling, so as to conduct quantitative analysis to deeply explore the factors behind the commuting volume and answer the question of "why"; secondly, when the commuting model is constructed , can be combined with the actual needs of planning to answer the question of "how" in the future.

In order to make the technical means, creative features, goals and effects of the present invention easy to understand, the method for establishing a commuting model based on mobile phone signaling data in the present invention will be described in detail below in conjunction with the embodiments and accompanying drawings.

<Example>

Fig. 1 is a flowchart of a method for establishing a commuting model in an embodiment of the present invention.

As shown in Figure 1, the commuting model establishment method based on mobile phone signaling data mainly includes the following steps:

Step S1, obtain the mobile phone signaling data and analyze the user's commuting data including the user's departure base station and employment base station according to the mobile phone signaling data. For specific steps, see step S1-1 to step S1-3.

Step S1-1, acquiring signaling data of the mobile phone. The mobile phone signaling data is the mobile phone signaling generated when each mobile phone base station in a city communicates with the mobile phone of each user (ie, the owner of the mobile phone). When the user's mobile phone switches on and off, sends and receives text messages, and receives and makes calls, it will "exchange information" with the base station, that is, the mobile phone is recorded by a specific surrounding base station (spatial location) at a specific time (the time when the behavior occurs), and the time-space information. If the user does not do anything, the location of the mobile phone will be "periodically updated", that is, the location will be updated periodically every 2 hours or so, that is to say, even if the user's mobile phone is not in use, it will be recorded every 2 hours one point.

In this embodiment, the mobile phone signaling data includes mobile phone information (such as mobile phone number), time information (that is, the time when communication occurs) and base station information (such as base station number) during communication.

Step S1-2, identifying the user's departure base station and employment base station according to the base station information and time information in the signaling data of the mobile phone.

In step S1-2 of this embodiment, the mobile phone signaling data for two consecutive weeks is obtained, and the mobile phone signaling data is counted according to the mobile phone information, so as to obtain the mobile phone signaling data of each user within two weeks. Further, for each user's mobile phone signaling data, analyze the user's life trajectory according to the time information in the mobile phone signaling data in order to identify the user's residence (ie, the base station of departure) and the place of work (ie, the base station of employment).

Specifically, if the point recorded by the user at night (8:00 p.m. ) is the user's place of residence. Similarly, if the recorded point of the user during the day (9:00 am to 6:00 pm) is relatively fixed (called "daytime high-frequency recording point"), then this point is likely to be the user's work place. Namely: the high-frequency recording point at night represents the place of residence, and the high-frequency recording point during the day represents the work place.

Step S1-3, generating commuting data of the user according to the signaling data of the mobile phone, the base station of departure and the base station of employment.

In this embodiment, the commuting data of the user includes the base station information of the departure place corresponding to the departure place, the corresponding departure time, the base station information of the employment place corresponding to the employment place, the corresponding employment time and mobile phone information.

In step S1 of this embodiment, using the mobile 2G mobile phone signaling data for two consecutive weeks in the first half of 2014 in Shanghai, a total of about 13.7 million users with relatively stable places of residence and employment were identified, accounting for 24 million users in Shanghai. About 60% of the resident population.

Step S2, according to the base station location information of the mobile phone base station, calculate the distribution weights of the mobile phone base station and a predetermined number of urban space units around the mobile phone base station. For specific steps, see steps S2-1 to step S2-5.

Step S2-1, obtaining base station location information of mobile phone base stations and unit location information of urban space units. Base station location information and unit location information are obtained from public urban planning information.

In this embodiment, the city planning information is the city information issued by the sixth national census or city information and geographic information from other sources, and the city planning information is specific to urban space units (such as street space units or neighborhood committee space units) The division is carried out, and the base station location information of the location of each mobile phone base station is recorded.

In step S2-2, for each mobile phone base station, obtain a predetermined number of surrounding urban space units and the nearest urban space units as the surrounding space units of each mobile phone base station.

In step S2-2 of this embodiment, the predetermined number is 30, that is, the nearest urban space units with 30 surrounding areas of each mobile phone base station are used as surrounding space units. Simultaneously, because the urban space unit adopted in the present embodiment is the neighborhood committee space unit, its spatial distribution is irregular, therefore adopt the position of the centroid of each neighborhood committee space unit to carry out corresponding calculation (similarly, the unit location information of the neighborhood committee space unit also based on the position of the centroid).

Step S2-3, calculating the adjacent distances between each mobile phone base station and each corresponding surrounding space unit according to the base station location information and the unit location information.

In this embodiment, the adjacent distance is the distance between the centroids of the mobile phone base station and its surrounding space units. Since each mobile phone base station corresponds to 30 surrounding space units, each mobile phone base station has 30 adjacent distances corresponding to each surrounding space unit. .

In step S2-4, for each mobile phone base station, determine the largest adjacent distance among all adjacent distances corresponding to the mobile phone base station as the maximum adjacent distance, and further calculate and obtain the bandwidth distance of the corresponding mobile phone base station according to the maximum adjacent distance.

In this embodiment, the bandwidth distance is the maximum range of base station data allocation, that is, when the straight-line distance between a city space unit and a certain base station exceeds the bandwidth distance, the base station will no longer allocate data to the city space unit.

In step S2-4 of the present embodiment, the maximum adjacent distance is that among the 30 urban space units (i.e. peripheral space units) corresponding to the mobile phone base station, the centroid of the urban space unit farthest from the mobile phone base station is adjacent to the mobile phone base station. distance. In the present embodiment, the data of the bandwidth distance is that the numerical value of the maximum adjacent distance is the same, and simultaneously, the maximum of the bandwidth distance of the present embodiment is no more than 4km, that is, when the value of the maximum adjacent distance is greater than 4km, the bandwidth distance of the mobile phone base station is as follows: 4km for calculation.

Step S2-5, calculating the allocation weights of the mobile phone base station and the surrounding space units according to the bandwidth distance and the adjacent distance.

In step S2-5 of this embodiment, the basic calculation method of the distribution weight is:

Also meet:

是编号为k的居委会空间单元获得的编号为i的手机基站的分配权重,d_(i)k是基站i与居委会单元k的形心的相邻距离，θ是带宽距离。公式(2)保证了公式(1)计算出的分配权重的和为1。In the formula,

is the allocation weight of cell phone base station number i obtained by the neighborhood committee space unit number k, d _(i)k is the adjacent distance between base station i and the centroid of neighborhood committee unit k, and θ is the bandwidth distance. Formula (2) ensures that the sum of the distribution weights calculated by formula (1) is 1.

Through the formulas (1) and (2), the allocation weights of each mobile phone base station and its corresponding 30 surrounding space units can be obtained.

Step S3, assigning the user's commuting data to urban space units according to the distribution weights to obtain unit commuting data including the user's departure unit and employment unit.

In step S3 of this embodiment, the commuting data is a connection between two points, including both the place of residence and the place of employment, that is, two base stations of the place of residence and the place of employment. Therefore, when allocating the commuting data in step S3, it is necessary to allocate the residence base station data and the employment base station data at the same time. The specific formula algorithm is as follows:

and:

Formulas (3)(4)(5)(6) are based on formulas (1) and (2), and are described in turn below:

是出发地(可以理解为通勤者的居住地)单元o获得的基站i的权重，

是目的地(可以理解为通勤者的工作地)单元d获得的基站j的权重，T_od是根据权重计算得到的出发地居委会单元o至目的地居委会单元d的通勤量。For formula (3), λ _ij is the commuting amount from base station i to base station j, and base station i and base station j are the base stations near neighborhood committee o and neighborhood committee d respectively,

is the weight of the base station i obtained by the departure point (which can be understood as the commuter’s residence) unit o,

is the weight of the base station j obtained by the destination (can be understood as the commuter’s work place) unit d, and T _od is the commuting amount calculated from the starting neighborhood unit o to the destination neighborhood unit d based on the weight.

For formula (4) and formula (5), its meaning is the same as that of formula (1). The distance calculation of the distance between the base station and the surrounding neighborhood committee units is used to obtain the distribution weights of the departure base station and the destination base station to the surrounding neighborhood committee units, where , d _(i)o is the distance between base station i and neighborhood unit o, d _(j)d is the distance between base station j and neighborhood unit d, _θi and _θj are the weight allocation bandwidth of base station and base station respectively.

For Formula 6, its meaning is the same as Formula 2, and the sum of the calculated weights is guaranteed to be 1, so that the total amount of data when the base station distributes data to the neighborhood committee units remains unchanged.

According to the formula (3)(4)(5)(6), the user's commuting flow data is allocated as the commuting flow between the neighborhood committee and the neighborhood committee, that is, the user's commuting data corresponding to each mobile phone base station is evenly distributed to each surrounding area according to the distribution weight Spatial unit, so that each urban spatial unit obtains user commuting data distributed by surrounding base stations to form unit commuting data.

In the present embodiment, the unit commuting data includes information such as the resident population, employment positions, commuting time between the neighborhood committee and the neighborhood committee, and the specific form of the data is as shown in Figure 2. Among the figures, pcq_O represents the commuting departure number (i.e. the place of residence), pcq_D represents the commuter destination number (work place), num_home_O represents the total resident population of the departure place (hereinafter abbreviated as P _i ), num_work_D represents the total employment positions of the destination (hereinafter abbreviated as N _j ), num represents the place of residence and The amount of commuting between work places (hereinafter abbreviated as T _ij ), dist represents the commuting distance (hereinafter abbreviated as d _ij ), dura_car represents the car commuting time (by car), and dura_bus represents the bus commuting time (by bus , such as ground buses and subways). In the actual modeling process, only one of commuting distance, car commuting time, and bus commuting time can be selected. In this embodiment, commuting distance is used for modeling.

Step S4, building a unit commuting model, the form of the unit commuting model is as follows:

ln T _ij ＝κ _i +α _i ln N _j +β _i ln d _ij +ε _ij (7)

In the formula, T _ij is the amount of commuting between urban spatial units, i represents the i-th departure unit, j represents the j-th employment unit, N _j is the number of jobs in the j-th employment unit, d _ij is The commuting cost between the i-th departure unit and the j-th employment unit (including commuting time or commuting distance, this implementation uses commuting distance), α _i and β _i are the population of the i-th departure unit Influence coefficient and employment position influence coefficient, under normal circumstances, the coefficient is positive, κ _i is the constant term of the i-th departure unit, ε _ij is the residual between the i-th departure unit and the j-th employment unit. (The symbol ln means to take the "logarithm" of the corresponding variable, which is equivalent to a numerical transformation of the original variable. The commuting amount, resident population, employment position and commuting distance are all transformed in the formula.)

The traditional commuting model adopts the global commuting model (ie global model, global model), and the form of the global model is as follows:

ln T _ij ＝κ+αln P _i +βln N _j +γln d _ij +ε (8)

In the formula, the meaning of the corresponding variables is consistent with the unit commuting model, P _i is the population of the i-th departure unit, α and β are the impact coefficients of population and employment, respectively, and γ is the distance attenuation coefficient. The coefficient is negative, κ is the constant term, and ε is the residual.

The average goodness of fit of the global model (goodness of fit, which is the most important indicator for evaluating the model effect, takes a value between 0 and 1, the higher the value, the better the model effect) is 0.65, and the fitting effect is acceptable. However, in this global model, there is an important flaw, that is, it cannot take into account the difference in attractiveness to employees between two units, or the “direction” of commuting, as illustrated below with an example.

As shown in Figure 3, two units A and B are located on both sides of the same subway line, unit A is located in the suburbs, and unit B is located within the inner ring of the central city. Because it is located on both sides of the same subway line, the commuting time and distance from A to B and from B to A should be the same, then according to the global model, employment from unit A to unit B and employment from unit B to unit A The number of people should be about the same, but in fact, the number of people employed from unit A to unit B should be much higher than the number of people employed from unit B to unit A. Because unit B is located in the central city, and unit A is located in the suburbs, the attraction of B to A is greater than that of A to B. This is the actual situation in Shanghai. Many people in the suburbs work in the inner ring of the central city, but the reverse is true. Few people. The global model cannot reflect this difference in "attractiveness". To reflect this difference, a sub-unit model (that is, a unit commuting model) must be adopted.

The sub-unit model is equivalent to "split" the global gravity model, and a total of 4991 sub-unit models (corresponding to the number of residential committee space units) are constructed (hereinafter referred to as the basic model). The form of the basic model is similar to the global model, the only difference is that since the data is split into 4991 subsets, each subset builds a model separately, and for each subset, its resident population is a constant (constant That is, a constant number, which is different from variables, which will change between different units), so the model lacks a variable of the population of the departure unit, that is, P _i . The form of the basic model is shown in the above-mentioned unit commuting model. Although the number of models in the basic model is large, the computational load is not large. Compared with the global model, there is almost no difference in computing time.

The average goodness-of-fit of the basic model is 0.76, which is much better than the 0.65 of the global model. The reason why sub-unit modeling was not carried out in the past is that the data (traffic survey data) is not enough to cover every urban spatial unit, and the mobile phone signaling data can perfectly solve this problem. In the following, the present invention will further optimize the basic model through the residual to improve the goodness of fit of the model, thereby improving the prediction accuracy and better guiding the planning practice.

Step S5, calculate the residuals {R _n , X _n , Y _n } of the unit commuting model according to the commuting volume calculated by the unit commuting model and the actual commuting volume, where R _n represents the residual of the nth urban space unit The absolute value of the difference, X _n and Y _n represent the plane coordinates of the nth urban spatial unit.

In this embodiment, the residual is the difference between the actual value and the predicted value. Taking an example between any two places, A is the starting point and B is the destination. First, there is an actual commuting amount between the two places. Denoted as T ₁ , secondly, for A, there is a coefficient of the sub-unit model, and the coefficient is brought into the model to calculate the commuting amount predicted by the model, which is abbreviated as T ₂ , and the residual is (T ₁ -T ₂ ), denoted as R for short, for the same starting point A, the residuals to different destinations B (B ₁ , B ₂ , B ₃ ... B _n ) are different, denoted as R ₁ , R ₂ , R ₃ respectively ...R _n , and because the spatial position of each destination is different, each residual has a spatial position attribute, and the value of the residual and the spatial position attribute are recorded as {{R ₁ , X ₁ , Y ₁ }, {R ₂ , X ₂ , Y ₂ }...{R _n , X _n , Y _n }}. Where R represents the absolute value of the residual, and X and Y represent the plane coordinates (ie latitude and longitude) of the residual position.

Step S6, clustering the residuals {R _n , X _n , Y _n } in the spatial clustering mode and classifying them according to the preset classification method to obtain 4 clustering types, and further variableize the 4 clustering types Processing generates residual dummy variables.

In this embodiment, the above residuals {R _n , X _n , Y _n } are clustered on the ArcGIS software platform to obtain the spatial clustering mode of the residuals. Specifically, use the built-in local spatial autocorrelation calculation tool in ArcGIS, and use python to write a loop algorithm to calculate all urban spatial units in turn. Thus, the residuals are finally divided into four typical types. The connotation of these types is the "special connection" between each unit, that is, in addition to employment positions and commuting time, there are many factors that will affect the amount of commuting between the two places. In this embodiment, clustering is performed based on the residuals of residents starting from one unit to work in a certain area, and the residuals of each unit around the area, and further four types of clustering are obtained based on the levels of the above two residuals The results are high-high clustering (HH cluster), low-low clustering (LL cluster), high-low clustering (HL cluster) and low-high clustering (LH cluster). For example, if a spatial unit is located along a subway line, then most of the employment of residents in this spatial unit is likely to be located along the subway line, and the residual error along the subway line may be very high, which means "high clustering" along the subway line.

Further, these 4 types are used as 4 dummy variables, that is, variable processing is performed to obtain residual dummy variables. In this embodiment, "dummy variable" can simply be understood as abstracting the form of a variable into two values of 1 and 0. See Table 1 for details:

Table: Residual dummy variable setting description table

Take "high-high clustering (HH)" as an example, high-high clustering corresponds to the dummy variable D_SE_0, and this value is 1, and the other three values, namely D_SE_1, D_SE_2, and D_SE_3 are all 0, and so on, each Each aggregation type corresponds to one of the dummy variables, only one is 1, and the other three are all 0. For units without significant aggregation characteristics, all four dummy variables are 0.

Step S7, substituting the residual dummy variable into the unit commuting model to obtain the residual commuting model, the form of the residual commuting model is as follows:

ln T _ij ＝κ _i +α _i ln N _j +β _j ln d _ij +∑ _k α _k D_SE _k +ε _ij (9)

In the formula, D_SE _k is the residual of the clustering type corresponding to the kth class, the value of k is [0,1,2,3], the value of D_SE _k is [0,1], and α _k is the corresponding residual difference coefficient, and the meanings of other parameters are similar to those in the unit commuting model.

As shown in Figure 4, the average goodness of fit of the residual commuting model reaches 0.92, far exceeding the 0.76 of the basic model and the 0.65 of the global model (that is, the traditional gravity model in the figure). The improvement of the goodness of fit means that the prediction effect There has been a substantial improvement.

Table 1: Comparison of traditional gravity model, basic model and residual model

In Table 1, the global model has 3 variables, the basic model has 2 variables, and the residual model has 3 variables. The three variables of the global model are the resident population in the place of residence, the number of jobs in the place of employment, and the commuting time between the place of residence and the place of employment. The two variables in the base model are the number of jobs at the place of employment and the commute time from the place of residence to the place of employment. The two conventional variables of the residual model are the number of jobs and commuting time, which are consistent with the basic model. On this basis, a new residual variable is added. A residual variable actually contains four dummy variables, and the four dummy variables are obtained through the residual The spatial clustering analysis of the difference obtains that the 4 spatial types of the clustered residuals are used as 4 dummy variables. In this embodiment, the 4 dummy variables are collectively referred to as a residual dummy variable, and the residual dummy variable After adding the basic model, the basic model is further optimized.

For a model, poor scalability means fewer variables can be considered. For example, the basic assumption of the traditional gravity model is that the amount of commuting between two places is directly proportional to the resident population at the departure place and the number of jobs at the destination, and inversely proportional to the travel time or distance between the two places, that is, the two places The higher the total population, the greater the number of jobs, the shorter the travel time or the closer the distance, the greater the amount of commuting between the two regions, which is in line with the general rule. The "resident population", "number of jobs" and "traffic time" are the three variables here, and the traditional gravity model can only consider these three variables. However, there are still many factors that affect commuting volume, such as the scale effect of large employment centers, historical and cultural reasons, etc. Traditional gravity models cannot take these factors into account.

Above, the method for establishing a commuting model based on mobile phone signaling data in the present invention has been described. In addition, the residual commuting model established by this method can be applied in urban planning, employment planning and other planning systems that need to calculate the commuting cost of residents. The following describes a method based on the residual commuting model and the preset number of jobs in conjunction with the accompanying drawings. An urban commuting data analysis system for analyzing commuting data of residents in the city.

Fig. 5 is a structural block diagram of an urban commuting data analysis system in an embodiment of the present invention.

As shown in Figure 5, the urban commuting data analysis system 100 includes a commuting model storage unit 11, a signaling data storage unit 12, a signaling data analysis acquisition unit 13, a distribution weight calculation unit 14, a commuting data distribution unit 15, and a commuting data calculation and acquisition unit. part 16, input display part 17, communication part 18 and control part 19.

Wherein, the communication unit 17 is used for communicating and interacting with each component of the urban commuting data analysis system 100 , and the control unit 18 is used for controlling the work of each component of the urban commuting data analysis system 100 .

The commuting model storage unit 11 stores the residual commuting model established by the method for establishing the commuting model based on the signaling data of the mobile phone.

The signaling data storage unit 12 stores mobile phone signaling data obtained in advance from various mobile phone base stations in the city.

The signaling data analysis and acquisition unit 13 is used to analyze the signaling data of the mobile phone to acquire the commuting data of the user including the base station of the user's departure base station and the base station of the employment site.

The allocation weight calculation unit 14 is used for calculating the allocation weights of the mobile phone base station and a predetermined number of urban space units around the mobile phone base station according to the base station location information of the mobile phone base station.

The commuting data allocation unit 15 allocates the user's commuting data to urban space units according to the allocation weights to obtain urban commuting data including the user's departure unit and employment unit.

In this embodiment, the processing methods of the signaling data analysis and acquisition unit 13 , the distribution weight calculation unit 14 and the commuting data distribution unit 15 are respectively consistent with the processing methods of step S1 , step S2 and step S3 of the above-mentioned commuting model establishment method.

The commuting data calculation and acquisition unit 16 calculates the urban commuting data and the number of jobs through the residual commuting model stored in the commuting model storage unit 11 , so as to acquire the residents' commuting data.

In this embodiment, the number of jobs is the number of jobs input by the analyst through the input display unit 17 according to the specific analysis situation. Resident commuting data is the average commuting distance of the city, and the average commuting distance corresponds to the commuting distance in the unit commuting data used when establishing the residual commuting model. In other embodiments, the residents' commuting data can also be the average commuting time, which is divided into the average bus commuting time and the average car commuting time. The average car commuting time corresponds to the car commuting time used when establishing the residual commuting model.

Theoretically, if a certain number of jobs are added in any unit, the commuting volume of all units will change. If the distribution in some units can shorten the average commuting time of the city to the greatest extent, then the performance of these units is the best . Therefore, the goal of the calculation is to increase the number of jobs in certain units, and the average commuting distance (or time) of the city decreases the most.

Taking Shanghai as an example, assuming that 10,000 jobs need to be added (that is, the number of jobs is 10,000), the commuting data calculation and acquisition department 16 respectively calculates the changes in the city's average commuting distance for 4,991 neighborhood committee space units after the jobs are added. Figure 6 shows the change distribution of the neighborhood committee space unit with shorter commuting distance in the figure, the better the effect of increasing employment.

In Figure 6, Jiangwan-Wujiaochang, Jinqiao, Zhangjiang-Chuansha, and Xinzhuang-Qibao all belong to the sub-center level of the city. For these areas, the overall planning requirements are to speed up industrial transformation and spatial adjustment, appropriately increase jobs, and promote Integration of industry and city. In addition, Luodian-Gucun, Jinqiao, South Railway Station-Caohejing, Caolu, Gongkang-Daning, Gaoqing Road-Yuqiao, etc. all belong to the regional center level. To meet the needs of development, realize the balanced distribution of public services and employment positions, and mainly serve the surrounding areas. Other areas are also basically the current main employment centers, including Minhang Economic Development Zone, Zizhu High-tech Zone, Waigaoqiao and other areas. In addition, increasing jobs in Pujiang and Zhoupu areas can also improve the commuting situation in the city.

At the same time, it can also be seen that the core area of the inner ring, including Lujiazui, Nanjing East Road, and Nanjing West Road, currently has a very concentrated number of jobs and is the highest-level employment center in the city. However, the calculation results of the model show that there is no need to further increase employment in these areas, because too many jobs will attract residents from further areas to come to work, resulting in a further increase in the city's average commuting distance.

Furthermore, according to the calculation results of the model, policy recommendations for the layout of jobs in Shanghai can be put forward, as shown in Table 2. Among them, the Zhangjiang-Chuanshadi and Jinqiao areas, because they belong to the sub-center of the city, have relatively comprehensive functions, and the calculation results show that there is a large demand for job creation, so they are the areas that Shanghai should focus on in the future. Secondly, there is also a large demand for additional jobs in the areas of South Railway Station-Caohejing and Luodian-Gucun. The average commuting distance of residents in these areas is relatively long. Therefore, the increase in jobs will have a positive effect on the commuting conditions of residents in the surrounding areas. Great improvement.

Table 2: Suggestions on Optimizing the Layout of Employment Posts in Shanghai

Therefore, through the calculation of the residual commuting model, the unit with the most shortened average distance in the city is the unit with the best effect. In the traditional gravity model (that is, the global model) environment, it is impossible to calculate which unit will shorten the city's average commuting distance the most after adding jobs.

The input and display part 17 is used to display the employment number input screen when the system is started so that analysts can input the number of jobs, and when the commuting data calculation and acquisition part 16 calculates the residents' commuting data, it displays the result display screen and displays the residents' commuting data on the screen. Data for analysts to view.

In this embodiment, the employment number input screen can be an input box, and the result display screen can display a data distribution diagram as shown in FIG. 6 .

Fig. 7 is a flow chart of the communication data analysis process in the embodiment of the present invention.

As shown in FIG. 7, when the analyst inputs the number of jobs and confirms the calculation in the number of jobs input screen displayed on the input display unit 17, the following steps begin:

Step T1, the signaling data analysis and acquisition unit 13 analyzes the mobile phone signaling data to obtain user commuting data, and then enters step T2;

Step T2, the distribution weight calculation unit 14 calculates the distribution weights of the mobile phone base station and a predetermined number of urban space units around the mobile phone base station according to the base station location information of the mobile phone base station, and then enters step T3;

In step T3, the commuting data allocation unit 15 allocates the user commuting data obtained in step T1 to urban space units according to the distribution weight obtained in step T2 to obtain unit commuting data, and then enters step T4;

Step T4, the commuting data calculation and acquisition unit 16 calculates the unit commuting data and the number of jobs input by the analyst through the residual commuting model, so as to obtain the resident commuting data, and then enters step T5;

In step T5, the input display unit 17 displays the result display screen and displays the resident commuting data calculated in step T4 on the screen for analysts to view, and then enters the end state.

In the communication data analysis process of this embodiment, the signaling data analysis and acquisition unit 13 , the distribution weight calculation unit 14 and the commuting data distribution unit 15 sequentially perform calculation processing to obtain unit commuting data. In other embodiments, the signaling data analysis and acquisition unit 13, the distribution weight calculation unit 14, and the commuting data distribution unit 15 can also complete the calculation of the unit commuting data in advance, so that after the analyst inputs the number of jobs, the commuting data can be directly calculated. The calculation acquisition unit 16 performs calculation of residents' commuting data to speed up the calculation.

The above-mentioned urban commuting data analysis system 100 is only an application level of the model, and the set goal is relatively simple (and shorten the commuting distance), and the simulation method is also relatively simple (and increase jobs). In actual planning applications, the scenarios faced will be more complex, and targeted simulations for specific situations are required, so that the simulation results can better guide planning practice.

Function and effect of embodiment

According to the method for establishing a commuting model based on mobile phone signaling data provided in this embodiment, since the user commuting data corresponding to each mobile phone base station is calculated according to the mobile phone signaling data, and the distribution weights between the mobile phone base station and each urban space unit are calculated, it is realized. Evenly distribute user commuting data to each urban space unit to obtain unit commuting data corresponding to each urban space unit, which solves the problem that the urban traffic data used in urban planning cannot evenly cover each unit, thereby optimizing the data used in model building source; also because the mobile phone signaling data can be updated quickly, it can better optimize the real-time performance of the commuting model. At the same time, the present invention establishes a unit commuting model on the basis of the traditional gravity model, which solves the problem that the traditional gravity model cannot analyze the "direction" of commuting; further, the present invention also obtains the residual commuting model by calculating residual optimization , so that the goodness of fit of the model is further improved, and many factors that originally affected commuting but could not be accurately obtained are incorporated into the model through residual variables, so that the applicability of the model is stronger, and the prediction effect of the model is greatly improved. A good guide to the land use layout of the city.

In an embodiment, since the adjacent distances of each mobile phone base station and its corresponding surrounding space units are calculated according to the base station location information and the unit location information, and the maximum adjacent distance of each mobile phone base station is calculated as the bandwidth distance of the mobile phone base station, further According to the bandwidth distance and adjacent distance calculation, the distribution weight of mobile phone base stations and surrounding space units can be calculated, so the bandwidth distance can be dynamically calculated according to the number of units within the range of each base station, so as to effectively avoid the coverage of the base station in the central city The problem of too many units and too little coverage in the suburbs can be solved, and the mobile phone base station data can be allocated to each urban space unit more accurately.

The above-mentioned embodiments are only used to illustrate the specific implementation manners of the present invention, and the present invention is not limited to the description scope of the above-mentioned embodiments.

Claims (7)

1. A commuting model establishing method based on mobile phone signaling data is used for establishing a commuting model of a city according to mobile phone signaling data collected by mobile phone base stations in the city, and is characterized by comprising the following steps:

step S1, acquiring the mobile phone signaling data and analyzing the mobile phone signaling data to obtain user commuting data including a departure place base station and a employment place base station of a user;

step S2, calculating the distribution weight of the mobile phone base station and a preset number of city space units around the mobile phone base station according to the base station position information of the mobile phone base station, wherein the city space units are street space units or living committee space units;

step S3, distributing the user commuting data to the urban space unit according to the distribution weight so as to obtain unit commuting data comprising a departure place unit and a employment place unit of the user, wherein the user commuting data are residence place base station data and employment place base station data;

s4, constructing a unit commute model, wherein the unit commute model is in the following form:

ln T _ij ＝κ _i +α _i ln N _j +β _i ln d _ij +ε _ij (1)

in the formula, T _ij For commuting amounts between city space units, N _j The number of employment posts of the jth employment location unit, d _ij For the commute cost, α, between the ith said origin unit and the jth said employment unit _i 、β _i The population number influence coefficient and employment position influence coefficient, kappa, of the ith departure place unit _i Is a constant term of the ith said origin unit, epsilon _ij The residual error between the ith departure place unit and the jth employment place unit is obtained;

step S5, calculating and obtaining a residual error { Rn, xn, yn } of the unit commuting model according to the commuting amount calculated by the unit commuting model and the actual commuting amount, wherein Rn represents an absolute value of the residual error corresponding to the nth city space unit, and Xn and Yn represent plane coordinates of the nth city space unit;

s6, clustering the residual errors { Rn, xn, yn } in a spatial clustering mode, classifying the residual errors according to a preset classification mode to obtain 4 clustering types, and further performing variable quantization processing on the 4 clustering types to generate residual error virtual variables;

and S7, substituting the residual virtual variable into the unit commuting model to obtain a residual commuting model, wherein the residual commuting model is in the following form:

ln T _ij ＝κ _i +α _i ln N _j +β _i ln d _ij +∑ _k α _k D_SE _k +ε _ij (2)

in the formula, D _ SE _k Is the residual virtual variable corresponding to the cluster type of the kth class, k being taken to be [0,1,2,3]，D_SE _k Is taken as value of [0,1]，α _k Are the corresponding residual coefficients;

wherein, the step S2 comprises the following substeps:

s2-1, acquiring base station position information of the mobile phone base station and unit position information of the urban space unit, wherein the user commuting data are residential area base station data and employment area base station data;

s2-2, acquiring the city space units with the peripheral quantity being the preset quantity and the nearest distance for each mobile phone base station as the peripheral space units of each mobile phone base station;

s2-3, respectively calculating the adjacent distance between each mobile phone base station and each peripheral space unit corresponding to the mobile phone base station according to the base station position information and the unit position information;

s2-4, sequentially judging the largest value of all adjacent distances corresponding to each mobile phone base station as the maximum adjacent distance for each mobile phone base station, and further calculating and acquiring the bandwidth distance corresponding to the mobile phone base station according to the maximum adjacent distance;

s2-5, calculating the distribution weight of the mobile phone base station and the peripheral space unit according to the bandwidth distance and the adjacent distance, wherein the basic calculation method of the distribution weight comprises the following steps:

simultaneously, the following requirements are met:

in the formula (I), the compound is shown in the specification,

is the assigned weight of the cell phone base station with number i obtained by the committee space unit with number k, d _(i)k Is the neighboring distance of the centroid of the base station i and the living committee unit k, theta is the bandwidth distance, the formula (4) ensures that the sum of the assigned weights calculated by the formula (3) is 1,

when the user commute data is distributed in the step S3, a specific formula algorithm is as follows:

and:

in the formula of lambda _ij Is the commute amount from base station i to base station j, base station i and base station j are base stations in the vicinity of residence commission o and residence commission d, respectively,

is the origin, i.e., the weight of base station i obtained by the commuter's residence unit o,

is the weight, T, of the base station j obtained by the destination, i.e. the commuter's work location unit d _od The amount of the commute from the departure place committee unit o to the destination committee unit d calculated from the weight, d _(i)o Is the distance between base station i and the residence committee unit o, d _(j)d Is the distance, θ, between the base station j and the residence committee unit d _i And theta _j The base station and the weight distribution bandwidth of the base station are respectively, and the formula (8) ensures that the sum of the calculated weights is 1.

2. The commute model building method based on handset signaling data according to claim 1, wherein:

wherein, the mobile phone signaling data comprises base station information and time information generated during mobile phone communication,

the step S1 includes the following substeps:

s1-1, acquiring the mobile phone signaling data;

s1-2, identifying a departure place base station and a employment place base station of the user according to the base station information and the time information;

and S1-3, generating the user commuting data according to the mobile phone signaling data, the departure place base station and the employment foundation station.

3. The commute model building method based on handset signaling data according to claim 1, wherein:

wherein the preset classification mode is to classify according to the first residual error and the second residual error,

the first residual is the residual of the city space unit, and the second residual is the residual of the city space unit around the city space unit corresponding to the first residual.

4. The commute model building method based on handset signaling data according to claim 1, wherein:

wherein the commute cost is a commute time or a commute distance.

5. The commute model building method based on handset signaling data according to claim 1, wherein:

wherein the predetermined number is 30.

6. The application of the residual commute model established by the commute model establishing method according to any one of claims 1 to 5 in a city commute data analysis system, for analyzing the mobile phone signaling data of a city based on the number of employment sites in the city preset by an analyst to obtain the resident commute data of the city, includes:

a commute model storage unit for storing the residual commute model;

a signaling data storage unit for storing the mobile phone signaling data acquired from each mobile phone base station in the city in advance;

the signaling data analysis and acquisition part is used for analyzing the mobile phone signaling data so as to acquire user commuting data comprising a departure place base station and a employment place base station of a user;

the distribution weight calculation part is used for calculating the distribution weights of the mobile phone base station and a preset number of city space units around the mobile phone base station according to the base station position information of the mobile phone base station;

a commute data allocation unit that allocates the user commute data to the city space unit according to the allocation weight to obtain unit commute data including a departure place unit and a employment place unit of the user;

a commute data calculation acquisition unit, through the residual commute model is right the unit commute data and the employment post quantity is calculated thereby to acquire the resident commute data.

7. The use of the residual commute model established by the commute model establishing method according to claim 6 in an urban commute data analysis system, wherein:

wherein the resident commute data is an average commute distance or an average commute time of the city.

Images