Method for destination selection of non-commuting activities of urban residents based on data fusion

Abstract

The application discloses a city resident non-commuting activity destination selection method based on data fusion, comprising the following steps: based on mobile signaling data with activity destination labels, resident travel survey data, city POI data and city 500m*500m grid data, extracting activity attribute features and built environment features required for city resident non-commuting activity destination selection probability prediction; using multiple mathematical models to model city resident non-commuting activity destination selection probability prediction, and extracting an optimal model; putting the trained city resident non-commuting activity destination selection probability model into a non-commuting activity destination selection algorithm considering activity time constraints, and realizing non-commuting activity destination selection; the method provided by the application can not only make up for the limitation of time scale and coverage range in previous research, but also can more effectively control the future travel distribution structure of city residents, so as to provide more reliable city planning decision basis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of urban resident trip destination selection behavior analysis, in particular to a method for urban resident non-commuting activity destination selection based on data fusion. BACKGROUND

[0002] With the rapid development of urbanization, the concept and method of urban traffic planning are also developing and extending, and urban traffic demand prediction is an important part of urban traffic planning. Traffic trip demand model is an important carrier of traffic trip demand prediction, and two types of models are applied to urban traffic demand prediction, namely, trip-based traffic model and activity-based traffic model. In the trip-based model, resident trip demand is mainly aggregated in the OD matrix based on traffic zones, and the trip behavior of the group is mainly studied. In the activity-based model, the focus of traffic trip is transferred from the group to the individual, and it is believed that the trip of residents is derived from the activity demand scattered in time and space, that is, traffic is a derived demand.

[0003] In recent years, in order to make the trip demand prediction more detailed, activity-based models have been used. This type of model improves the traditional four-stage modeling method. The destination selection modeling of resident trip activity is a key link in the activity-based model, which is essentially to predict the distribution of urban residents' trip, which is an indispensable preliminary step in urban traffic planning and an important guarantee for the scientificity and reliability of future traffic planning.

[0004] In the past research, although there are many studies on the selection of commuting activity locations, only a few studies on the selection of non-commuting activity locations, and the construction of resident trip destination selection model still needs to be improved due to the limitation of data time scale and coverage. The present application fuses the characteristics of multi-source data, uses the advanced machine learning algorithm, and models the destination selection of non-commuting activities of urban residents. It can not only effectively make up for the limitation of time scale and coverage in the past research, but also obtain a more scientific and accurate destination selection model of non-commuting activities of urban residents, which can more effectively control the future trip distribution structure of urban residents and provide more reliable basis for urban planning decision. SUMMARY

[0005] This part aims to summarize some aspects of the embodiments of the present application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this part and the abstract and title of the specification of the present application to avoid obscuring the purpose of this part, the abstract and the title, and such simplifications or omissions cannot be used to limit the scope of the present application.

[0006] In view of the above problems, the present application is proposed.

[0007] In a first aspect of the embodiments of the present application, a method for selecting a non-commuting activity destination of urban residents based on data fusion is provided, comprising: extracting activity attribute features and built environment features required for predicting a non-commuting activity destination selection probability of urban residents based on mobile signaling data with an activity destination label, resident travel survey data, urban POI data, and urban 500m*500m grid data; modeling the non-commuting activity destination selection probability prediction of urban residents using multiple mathematical models, and extracting an optimal model; and incorporating the trained non-commuting activity destination selection probability model of urban residents into a non-commuting activity destination selection algorithm considering activity time constraints, to realize selection of a non-commuting activity destination.

[0008] As a preferred scheme of the method for selecting a non-commuting activity destination of urban residents based on data fusion, the modeling of the non-commuting activity destination selection probability prediction of urban residents comprises,

[0009] Extracting activity attribute features and built environment features required for predicting a non-commuting activity destination selection probability based on resident travel survey data, urban POI data, and 500m*500m grid data, modeling the non-commuting activity destination selection probability prediction of urban residents using multiple mathematical models, and extracting an optimal model;

[0010] The multiple mathematical models include a support vector machine model, a decision tree model, and an XGBoost model.

[0011] As a preferred scheme of the method for selecting a non-commuting activity destination of urban residents based on data fusion, the extraction of the optimal model comprises,

[0012] Normalizing the activity attribute features and built environment features, calculating the maximum value and minimum value of the required features for modeling the non-commuting activity destination selection probability, and converting the original data of the required features to the range of [0, 1] to obtain normalized original feature data;

[0013] The calculation of the normalization processing comprises,

[0014]

[0015] wherein, denotes the normalized original feature data, x i denotes the original feature data, x max denotes the maximum value of the original data set, x min denotes the minimum value of the original data set;

[0016] The normalized features are taken as a feature set X for model training, and the grid IDs of non-commuting activities in the resident travel survey data are taken as a sample label set Y, and the feature set X and the sample label set Y are constructed into a training set and a test set according to a ratio of 8:2;

[0017] The training set and the test set are input into a support vector machine model, a decision tree model and an XGBoost model for training, and the performance of the support vector machine model, the decision tree model and the XGBoost model is tested through cross-validation.

[0018] As a preferred scheme of the city resident non-commuting activity destination selection method based on data fusion provided in the application, wherein the extraction of the optimal model further includes,

[0019] The performance of the support vector machine model, the decision tree model and the XGBoost model is described by using an overall accuracy index, and the model with the highest overall accuracy index is selected to obtain the optimal model;

[0020] The calculation of the overall accuracy index includes,

[0021]

[0022]

[0023] Wherein, score represents whether the true value is correctly predicted, label Predic represents the predicted value of the label, label True represents the true value of the label, accuracy represents the accuracy, represents the number of test set samples, D test represents the test set.

[0024] As a preferred scheme of the city resident non-commuting activity destination selection method based on data fusion provided in the application, wherein the process of the non-commuting activity destination selection algorithm considering activity time constraints includes,

[0025] A1: Based on the resident travel survey data, according to commuting activities, the personal activity chain is divided into a sub-activity chain with the starting point and the ending point being commuting activities, and the non-commuting activity destination of the sub-activity chain is the destination to be selected;

[0026] A2: Combined with the resident travel survey data and 500m*500m grid data, a travel time-travel distance distribution curve is drawn with 5 minutes as an interval;

[0027] A3: combine the mobile phone signaling data with activity purpose label and 500m*500m grid data, classify the grid according to the classification of urban resident non-commuting activity purpose, and calculate the average access frequency of each grid;

[0028] A4: according to the classification result of the grid, combine the sub-activity chain and the distribution curve of travel time-travel distance to generate a non-commuting activity destination candidate set;

[0029] A5: extract individual activity characteristics and built environment characteristics, input into the optimal model of the urban resident non-commuting activity destination selection probability prediction model using multiple mathematical models, calculate the selected probability of part of the non-commuting activity destination in the candidate set, use the average access frequency obtained from the mobile phone signaling data with activity purpose label instead of the grid not covered by the resident travel survey data, and recalculate the comprehensive selected probability;

[0030] A6: select the non-commuting activity destination candidate set with the maximum comprehensive selected probability in the non-commuting activity destination candidate set as the final selection;

[0031] A7: for a sub-activity chain, repeat steps A4-A6 until all non-commuting activity destinations in the sub-activity chain are assigned, combine the travel time-travel distance distribution curve to verify whether the distance from the last non-commuting activity to the end point in the sub-activity chain is within the travel distance range under the travel time, if it is within the travel distance range, then enter the non-commuting activity destination assignment of the next sub-activity chain, otherwise, remove the last non-commuting activity destination in the sub-activity chain and repeat steps A4-A7 for the last non-commuting activity destination to perform reassignment;

[0032] A8: repeat steps A4-A7 until the non-commuting activity destination selection in all sub-activity chains is completed.

[0033] As a preferred scheme of the urban resident non-commuting activity destination selection method based on data fusion, wherein: the average access frequency of the sub-purpose includes,

[0034] Combine the mobile phone signaling data with activity purpose label and 500m*500m grid data, calculate the total number of visits and the total number of visits in the time scale of the mobile phone signaling data with activity purpose label, and then obtain the average access frequency of the sub-purpose;

[0035] The calculation of the average access frequency of the sub-purpose includes,

[0036]

[0037] wherein F i represents the average visit frequency of the grid i, V i represents the total visit frequency of the grid i, P i represents the total number of visitors of the grid i.

[0038] As a preferred scheme of the method for selecting the non-commuting activity destination of the urban resident based on data fusion, the calculation of the comprehensive selected probability comprises,

[0039]

[0040]

[0041]

[0042] D S = D N + D V

[0043] wherein, represents the normalized value, represents the selected probability of the destination a calculated by the non-commuting activity destination selection probability prediction model, N M represents the number of the candidate destinations in the candidate destination set without the resident travel survey visit record, represents the selected probability of the destination i calculated by the non-commuting activity destination selection probability prediction model, represents the normalized value, represents the selected probability of the destination b calculated by the mobile signaling data, N V represents the number of the candidate destinations in the candidate destination set without the resident travel survey visit record, represents the selected probability of the destination i calculated by the mobile signaling data, represents the comprehensive selected probability of the kth candidate destination, represents the normalized value of the selected probability of the destination k calculated by the non-commuting activity destination selection probability prediction model, represents the normalized value of the selected probability of the destination i calculated by the non-commuting activity destination selection probability prediction model, represents the normalized value of the selected probability of the destination i calculated by the mobile signaling data, represents the normalized value of the selected probability of the destination k calculated by the mobile signaling data, D S represents the candidate destination set of the non-commuting activity destination, D M ​​D represents a destination candidate set with resident travel survey visit records V D represents a destination candidate set without resident travel survey visit records.

[0044] In a second aspect, the present application provides a data fusion based urban resident non-commuting activity destination selection system, comprising:

[0045] a feature extraction module configured to extract activity attribute features and built environment features required for urban resident non-commuting activity destination selection probability prediction based on mobile phone signaling data with activity destination labels, resident travel survey data, urban POI data, and urban 500m×500m grid data;

[0046] a model construction module configured to model the urban resident non-commuting activity destination selection probability prediction using multiple mathematical models and extract an optimal model;

[0047] a destination selection module configured to incorporate the trained urban resident non-commuting activity destination selection probability model into a non-commuting activity destination selection algorithm considering activity time constraints to realize non-commuting activity destination selection.

[0048] In a third aspect, the present application provides a device, comprising,

[0049] a processor;

[0050] a memory for storing processor executable instructions;

[0051] the processor is configured to invoke the instructions stored in the memory to execute the method described in any of the embodiments of the present application.

[0052] In a fourth aspect, the present application provides a computer readable storage medium having computer program instructions stored thereon, comprising:

[0053] the computer program instructions are executed by a processor to implement the method described in any of the embodiments of the present application.

[0054] The present application has the following beneficial effects: the present application provides a data fusion based urban resident non-commuting activity destination selection method, which models urban resident non-commuting activity destination selection by using multiple mathematical models, combining various data characteristics and resident activity spatio-temporal constraints, considering resident historical selection regularity and randomness of selection for unknown locations, to obtain a more scientific and accurate urban resident non-commuting activity destination selection model, which can more effectively control urban resident future travel distribution structure to provide more reliable urban planning decision basis. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced. Obviously, the drawings in the following description only constitute some of the embodiments of the present application, and all other drawings obtained by those skilled in the art without creative labor should fall within the protection scope of the present application.

[0056] Figure 1 The basic flowchart of the method for selecting the non-commuting activity destination of urban residents based on data fusion provided by the present application;

[0057] Figure 2 The feature set required for constructing the non-commuting activity destination selection probability prediction model of the method for selecting the non-commuting activity destination of urban residents based on data fusion provided by the present application;

[0058] Figure 3 The resident travel time-travel distance distribution map based on resident travel survey of the method for selecting the non-commuting activity destination of urban residents based on data fusion provided by the present application;

[0059] Figure 4 The distance distribution curve of each group of results of the method for selecting the non-commuting activity destination of urban residents based on data fusion provided by the present application;

[0060] Figure 5 The OD expected line distribution map (street level) of each group of results of the method for selecting the non-commuting activity destination of urban residents based on data fusion provided by the present application;

[0061] Figure 6 The OD quantity comparison scatter plot (street level) of the results of experimental group 1 and experimental group 2 of the method for selecting the non-commuting activity destination of urban residents based on data fusion provided by the present application. DETAILED DESCRIPTION

[0062] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the protection scope of the present application.

[0063] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without creative labor, so the present application is not limited to the specific embodiments disclosed below.

[0064] Third, the "one embodiment" or "an embodiment" referred to herein means a specific feature, structure, or characteristic under discussion. Each of the various embodiments presented in the specification are not necessarily all mutually exclusive alternatives from other embodiments presented. It is therefore not a requirement that all of the features of the application be present in one specific implementation, even if that implementation is referred to as "the embodiment."

[0065] The application is described in detail below with reference to the accompanying drawings. In the detailed description of the application, the cross-sectional views of the device structure are partially enlarged without the general scale for the convenience of explanation, and the schematic views are only examples which should not limit the scope of the application. In addition, the three-dimensional spatial dimensions of length, width and depth should be included in the actual manufacture.

[0066] In the description of the application, it should be noted that the terms "upper, lower, inner and outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the application. In addition, the terms "first, second or third" are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0067] Unless otherwise expressly specified and limited, the terms "mounting, connecting, connection" in the application should be broadly understood, for example: it can be fixed connection, detachable connection or integral connection; it can also be mechanical connection, electrical connection or direct connection, it can also be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the application can be understood according to the specific circumstances.

[0068] Embodiment 1

[0069] Reference Figure 1 For one embodiment of the application, a data fusion-based urban resident non-commuting activity destination selection method is provided, comprising:

[0070] S1: Based on mobile signaling data with activity destination labels, resident travel survey data, urban POI data and urban 500m x 500m grid data, the activity attribute features and built environment features required for urban resident non-commuting activity destination selection probability prediction are extracted.

[0071] S2: A variety of mathematical models are used to model the urban resident non-commuting activity destination selection probability prediction, and the optimal model is extracted. It should be noted that:

[0072] Based on resident travel survey data, urban POI data and 500m*500m grid data, the activity attribute features and built environment features required for non-commuting activity destination selection probability prediction are extracted, a variety of mathematical models are used to model the non-commuting activity destination selection probability prediction of urban residents, and the optimal model is extracted;

[0073] It should be noted that the various mathematical models include support vector machine (SVM) model, decision tree (DT) model and XGBoost model;

[0074] Further, the extraction of the optimal model includes,

[0075] The activity attribute features and built environment features are normalized, the maximum and minimum values of the features required for non-commuting activity destination selection probability modeling are calculated, and the original data of the required features are converted to the range of [0, 1] to obtain the normalized original feature data;

[0076] It should be noted that the calculation of the normalization processing includes,

[0077]

[0078] Wherein, The normalized original feature data is represented by x i The original feature data is represented by x max The maximum value of the original data set is represented by x min The minimum value of the original data set is represented by x

[0079] Further, the normalized features are used as the feature set X for model training, and the grid ID of the non-commuting activity in the resident travel survey data is used as the sample label set Y. The feature set X and the sample label set Y are constructed into training set and test set according to the ratio of 8:2;

[0080] Specifically, the training set and the test set are input into the support vector machine (SVM) model, the decision tree (DT) model and the XGBoost model for training, and the performance of the support vector machine (SVM) model, the decision tree (DT) model and the XGBoost model is tested by cross-validation;

[0081] Further, the performance of the support vector machine (SVM) model, the decision tree (DT) model and the XGBoost model is described by using the overall accuracy index, and the model with the highest overall accuracy index is selected to obtain the optimal model;

[0082] It should be noted that the calculation of the overall accuracy index includes,

[0083]

[0084]

[0085] wherein score represents whether the true value is correctly predicted, label Predic represents the predicted value of the label, label True represents the true value of the label, accuracy represents the accuracy, represents the number of test set samples, D test represents the test set.

[0086] S3: Incorporate the trained urban resident non-commuting activity destination selection probability model into the non-commuting activity destination selection algorithm considering activity time constraints to realize the selection of non-commuting activity destinations. It should be noted that:

[0087] The process of the non-commuting activity destination selection algorithm considering activity time constraints includes,

[0088] A1: Based on resident travel survey data, according to commuting activities, the personal activity chain is divided into sub-activity chains with both starting point and ending point being commuting activities, and the non-commuting activity destination of the sub-activity chain is the destination to be selected;

[0089] A2: Combined with resident travel survey data and 500m×500m grid data, the travel time-travel distance (including average value, 10% quantile, 90% quantile) distribution curve is drawn with 5 minutes as an interval;

[0090] A3: Combined with mobile signaling data with activity destination label and 500m×500m grid data, the grid is classified according to the classification of urban resident non-commuting activity destinations (shopping and entertainment, business travel, and others), and the average access frequency of each grid is calculated;

[0091] A4: According to the classification results of the grid, combined with the sub-activity chain and the travel time-travel distance distribution curve, the non-commuting activity destination candidate set is generated;

[0092] A5: Extract individual activity features and built environment features, input into the optimal model of the urban resident non-commuting activity destination selection probability prediction model using multiple mathematical models, calculate the selected probability of some destinations in the non-commuting activity destination candidate set, for the grid not covered by the resident travel survey data, use the average access frequency obtained from the mobile signaling data with activity destination label instead, and recalculate the comprehensive selected probability;

[0093] A6: Select the destination candidate set with the maximum comprehensive selected probability in the non-commuting activity destination candidate set as the final selection;

[0094] A7: For a sub-activity chain, repeat steps A4 to A6 until all non-commuting activity destinations in the sub-activity chain have been assigned. Using the travel time-travel distance distribution curve, verify whether the distance from the last non-commuting activity in the sub-activity chain to the destination is within the travel distance range for that travel time. If it is within the travel distance range, proceed to the assignment of non-commuting activity destinations in the next sub-activity chain. Otherwise, remove the last non-commuting activity destination from the sub-activity chain and repeat steps A4 to A7 for the last non-commuting activity destination to be reassigned.

[0095] A8: Repeat steps A4 to A7 until all non-commuting activity destinations in the sub-activity chains have been selected;

[0096] It should be noted that the acquisition of average visit frequency per user for each category includes,

[0097] By combining mobile signaling data with activity purpose labels and 500m×500m grid data, the total number of visitors and total number of visits for each grid are calculated within the time scale of mobile signaling data with activity purpose labels, and then the average visit frequency per person for each purpose is obtained.

[0098] Specifically, the calculation of average visit frequency per user for each category includes,

[0099]

[0100] Among them, F i V represents the average visit frequency per person for grid i. i P represents the total number of visits to grid i. i This represents the total number of visitors to grid i;

[0101] It should be noted that the calculation of the overall probability of selection includes,

[0102]

[0103]

[0104]

[0105] D S =D M +D V

[0106] in, express The normalized value, N represents the probability of destination a being selected, calculated using a non-commuting activity destination selection probability prediction model. M This indicates the number of candidate destinations in the destination candidate set for which there are resident travel survey interview records. denotes the probability of destination i being selected calculated by the non-commuting activity destination selection probability prediction model, denotes the probability of destination i being selected calculated by the non-commuting activity destination selection probability prediction model, denotes the normalized value of the probability of destination i being selected calculated by the non-commuting activity destination selection probability prediction model, denotes the probability of destination i being selected calculated by the mobile phone signaling data, V denotes the number of candidate destinations in the candidate destination set without resident travel survey access records, denotes the probability of destination i being selected calculated by the mobile phone signaling data, denotes the comprehensive probability of the kth candidate destination being selected, denotes the normalized value of the probability of destination k being selected calculated by the non-commuting activity destination selection probability prediction model, denotes the normalized value of the probability of destination i being selected calculated by the non-commuting activity destination selection probability prediction model, denotes the normalized value of the probability of destination i being selected calculated by the mobile phone signaling data, denotes the normalized value of the probability of destination k being selected calculated by the mobile phone signaling data, S denotes the candidate destination set of non-commuting activity, M denotes the candidate destination set of non-commuting activity, V denotes the candidate destination set of non-commuting activity.

[0107] It should be noted that the present application provides a method for selecting the destination of non-commuting activities of urban residents based on data fusion. By using various mathematical models, combining the characteristics of various data and the spatio-temporal constraints of resident activities, the selection of non-commuting activity destinations of urban residents is modeled. While considering the historical selection regularity of residents, the randomness of the selection of unknown locations is also considered. A more scientific and accurate model for selecting the destination of non-commuting activities of urban residents can be obtained, which can more effectively control the future travel distribution structure of urban residents, and provide a more reliable basis for urban planning decisions.

[0108] The second aspect of the present application is disclosed,

[0109] The present application provides a system for selecting the destination of non-commuting activities of urban residents based on data fusion, comprising:

[0110] The feature extraction module is used for extracting the activity attribute features and built environment features required for predicting the probability of selecting the destination of non-commuting activities of urban residents based on mobile phone signaling data with activity destination labels, resident travel survey data, urban POI data, and urban 500m*500m grid data.

[0111] A model construction module is configured to model the probability of destination selection of the non-commuting activities of the urban residents by using a plurality of mathematical models, and extract an optimal model;

[0112] A destination selection module is configured to incorporate the trained probability model of destination selection of the non-commuting activities of the urban residents into a non-commuting activity destination selection algorithm considering the activity time constraint, and realize the selection of the destination of the non-commuting activities.

[0113] The third aspect of the present application,

[0114] An apparatus is provided, comprising:

[0115] A processor;

[0116] A memory for storing processor-executable instructions;

[0117] The processor is configured to invoke the instructions stored in the memory to execute the method of any one of the preceding aspects.

[0118] The fourth aspect of the present application,

[0119] A computer readable storage medium is provided, which stores computer program instructions, comprising:

[0120] The computer program instructions are executed by the processor to implement the method of any one of the preceding aspects.

[0121] The present application can be a method, apparatus, system and / or computer program product. The computer program product can include a computer readable storage medium having stored thereon computer readable program instructions that, when executed by a computer, carry out any of the aspects of the present application.

[0122] Computer readable storage media can be tangible devices that can retain and store instructions for use by an instruction execution device. Computer readable storage media can be, for example, but is not limited to, electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. Computer readable storage media can be non-transitory in nature, such as non-transitory electronic storage devices, non-transitory magnetic storage devices, non-transitory optical storage devices, non-transitory electromagnetic storage devices, non-transitory semiconductor storage devices, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of computer readable storage media includes a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or punched tape, a

[0123] Embodiment 2

[0124] Reference Figures 2-6 For the second embodiment of the present application, which is different from the first embodiment, a verification test of the method for selecting the destination of non-commuting activities of urban residents based on data fusion is provided. In order to make the purpose, technical solutions and advantages of the present application more clear and explicit, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0125] The research data selected in this embodiment is the mobile signaling data with activity destination label in Shanghai in 2018, the resident travel survey data in Shanghai in 2019, the urban POI data in Shanghai in 2020 and the 500m x 500m grid data in Shanghai;

[0126] First, the resident travel survey data, urban POI data and urban 500m x 500m grid data are fused to extract the related influence features of the urban resident non-commuting activity destination selection probability prediction model required for modeling, including activity attribute features and built environment features, and the total sample size is 11996 activities. The detailed features are as follows: Figure 2The feature values obtained are normalized and divided into a training set and a test set in a ratio of 8:2. The training set is used as the input for model training. Support vector machine (SVM) model, decision tree (DT) model and XGBoost model are used to model the probability prediction model for the destination selection of non-commuting activities of urban residents. After the model is established, the overall accuracy and F1 score are used to describe the performance of the model to obtain the optimal model. The performance of the support vector machine (SVM) model, the decision tree (DT) model and the XGBoost model is shown in Table 1. It can be found that the values of the XGBoost model on the evaluation indicators are higher than those of the support vector machine (SVM) model and the decision tree (DT) model. Therefore, the XGBoost model after training is selected as the optimal model and input into the non-commuting activity destination selection algorithm.

[0127] Table 1: Training structure of support vector machine model, decision tree model and XGBoost model.

[0128] SVM DT XGBoost Accuracy 22.21％ 51.58％ 53.17％

[0129] The trained optimal non-commuting activity destination selection probability model is considered in the non-commuting activity destination selection algorithm considering the activity time constraint, and the non-commuting activity destination selection is performed. Before the algorithm calculation, the travel time-travel distance distribution graph needs to be drawn in advance based on the resident travel survey data, as shown in Figure 3 In addition, in order to test the effect of the algorithm, one control group and two experimental groups are set up.

[0130] Control group 1: the result reflected by the true value, as shown in Figure 4 , Figure 5 .

[0131] Experimental group 1: the selection result obtained by only using the trained optimal non-commuting activity destination selection probability model, without combining the non-commuting activity destination selection algorithm considering the activity time constraint, as shown in Figure 4 , Figure 5 , Figure 6 .

[0132] Experimental group 2: the trained optimal non-commuting activity destination selection probability model is considered in the non-commuting activity destination selection algorithm considering the activity time constraint, as shown in Figure 4 , Figure 5 , Figure 6 .

[0133] From the results, it can be found that compared with experimental group 1, experimental group 2 has better results for the reference object in control group 1. The travel distance kernel density distribution curve of experimental group 2 is more similar to that of control group 1. In theFigure 5 (b) in the street-level OD expected line reacted in, the experimental group 1 exists the selection condition of part "hot spot", but this condition does not exist in the experimental group 2, and the experimental group 2 is closer to the result of the control group 1.

[0134] In addition, Figure 6 The more detailed comparison of the street-level OD in the experimental group can find that the distribution of the scatter points is more concentrated to the central axis compared to the experimental group 1, which also indicates the effectiveness of the method for selecting the non-commuting activity destination of the urban residents based on the data fusion provided by the application. The method not only considers the historical selection regularity of the residents' travel, but also considers the randomness of the selection of the unknown place, gives the ability to visit the unknown place, and enhances the generalization ability of the model.

[0135] It should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A method for selecting destinations for non-commuting activities of urban residents based on data fusion, characterized in that, include: Based on mobile phone signaling data with activity purpose labels, resident travel survey data, urban POI data, and urban 500m×500m raster data, we extract the activity attribute features and built environment features required for predicting the probability of urban residents' non-commuting activity destination selection. Multiple mathematical models were used to model the probability prediction of destination selection for non-commuting activities of urban residents, and the optimal model was extracted. The trained probability model for choosing destinations for non-commuting activities of urban residents is incorporated into the non-commuting activity destination selection algorithm that takes into account the activity time constraints, so as to realize the selection of destinations for non-commuting activities. The process of the non-commuting activity destination selection algorithm that considers activity time constraints includes: A1: Based on the resident travel survey data, the personal activity chain is divided into sub-activity chains with both the starting point and the ending point being commuting activities, and the non-commuting activity destinations of the sub-activity chains are the destinations to be selected. A2: Combining the aforementioned resident travel survey data and 500m×500m raster data, plot the distribution curve of travel time-travel distance at 5-minute intervals; A3: Combining the mobile signaling data with activity purpose tags and the 500m×500m grid data, classify the grid according to the non-commuting activity purpose classification of urban residents, and calculate the average access frequency per person for each grid sub-purpose. A4: Based on the classification results of the grid, combined with the sub-activity chain and the travel time-travel distance distribution curve, generate a candidate set of destinations for non-commuting activities; A5: Extract individual activity characteristics and built environment characteristics, input them into the optimal model of the urban residents' non-commuting activity destination selection probability prediction model that uses multiple mathematical models, calculate the selection probability of some destinations in the candidate set of non-commuting activity destinations, and for grids not covered by the residents' travel survey data, use the per capita access frequency obtained from the mobile phone signaling data with activity purpose labels to replace them, and recalculate the comprehensive selection probability. A6: Select the destination candidate set with the highest overall probability of being selected from the non-commuting activity destination candidate set as the final selection; A7: For a sub-activity chain, repeat steps A4 to A6 until all non-commuting activity destinations in the sub-activity chain have been assigned. Based on the travel time-travel distance distribution curve, verify whether the distance from the last non-commuting activity in the sub-activity chain to the destination is within the travel distance range for that travel time. If it is within the travel distance range, proceed to the assignment of non-commuting activity destinations in the next sub-activity chain. Otherwise, remove the last non-commuting activity destination in the sub-activity chain and repeat steps A4 to A7 for the last non-commuting activity destination to be reassigned. A8: Repeat steps A4 to A7 until all non-commuting activity destinations in the sub-activity chains have been selected; The acquisition of the average access frequency per user for each category includes, Combining the mobile signaling data with the activity purpose tag and the 500m×500m grid data, the total number of visitors and the total number of visits for each grid are calculated within the time scale of the mobile signaling data with the activity purpose tag, thereby obtaining the average visit frequency per person for each purpose. The calculation of the average access frequency per user for each category includes, Among them, F i V represents the average visit frequency per person for grid i. i P represents the total number of visits to grid i. i This represents the total number of visitors to grid i; The calculation of the overall selection probability includes, in, express The normalized value, N represents the probability of destination a being selected, calculated using a non-commuting activity destination selection probability prediction model. M This indicates the number of candidate destinations in the destination candidate set for which there are resident travel survey interview records. This represents the probability of destination i being selected, calculated using a non-commuting activity destination selection probability prediction model. express The normalized value, N represents the probability of destination b being selected, calculated using mobile signaling data. V This indicates the number of candidate destinations in the candidate destination set for which there are no resident travel survey interview records. This represents the probability of destination i being selected, calculated using mobile signaling data. This represents the overall probability of the k-th alternative destination being selected. This represents the normalized value of the probability of destination k being selected, calculated using a non-commuting activity destination selection probability prediction model. This represents the normalized value of the probability of destination i being selected, calculated using the non-commuting activity destination selection probability prediction model. This represents the normalized value of the probability of destination i being selected, calculated using mobile signaling data. D represents the normalized value of the probability of destination k being selected, calculated using mobile signaling data. S D represents the alternative destinations for non-commuting activities. M D represents the set of destination candidates with resident travel survey records. V This indicates a set of destinations for which there are no resident travel survey interview records.

2. The method for selecting destinations for non-commuting activities of urban residents based on data fusion as described in claim 1, characterized in that: Modeling the probability prediction of destination selection for non-commuting activities of urban residents includes... Based on resident travel survey data, urban POI data, and 500m×500m raster data, the activity attribute features and built environment features required for predicting the probability of non-commuting activity destination selection are extracted. Multiple mathematical models are used to model the probability prediction of non-commuting activity destination selection for urban residents, and the optimal model is extracted. The various mathematical models include support vector machine model, decision tree model and XGBoost model.

3. The method for selecting destinations for non-commuting activities of urban residents based on data fusion as described in claim 2, characterized in that: The extraction of the optimal model includes, The activity attribute features and built environment features are normalized, the maximum and minimum values ​​of the required features for the probability modeling of destination selection for non-commuting activities are calculated, and the original data of the required features are transformed to the range of [0,1] to obtain the normalized original feature data. The normalization process includes the following calculations: in, x represents the normalized original feature data. i x represents the original feature data. max x represents the maximum value of the original dataset. min This represents the minimum value of the original dataset; The normalized features are used as the feature set X for model training, and the grid IDs of non-commuting activities in the resident travel survey data are used as the sample label set Y. The feature set X and the sample label set Y are used to construct a training set and a test set in an 8:2 ratio. The training set and the test set are input into the support vector machine model, decision tree model, and XGBoost model for training, and the performance of the support vector machine model, decision tree model, and XGBoost model is tested by cross-validation.

4. The method for selecting destinations for non-commuting activities of urban residents based on data fusion as described in claim 3, characterized in that: The extraction of the optimal model also includes, The overall accuracy of the indicators is used to describe the performance of the support vector machine model, decision tree model, and XGBoost model. The model with the highest overall accuracy of the indicators is selected to obtain the optimal model. The calculation of the overall accuracy of the indicator includes, Where, score indicates whether the true value was correctly predicted, and label... Predic The predicted value for the label. True The true value of the label is represented by "accuracy". D represents the number of samples in the test set. test This represents the test set.

5. A destination selection system for non-commuting activities of urban residents based on data fusion, using the destination selection method for non-commuting activities of urban residents based on data fusion as described in any one of claims 1 to 4, characterized in that, include: The feature extraction module is used to extract the activity attribute features and built environment features required for predicting the probability of urban residents' non-commuting activity destination selection based on mobile phone signaling data with activity purpose labels, resident travel survey data, urban POI data, and urban 500m×500m raster data. The model building module is used to model the probability prediction of destination selection for non-commuting activities of urban residents using multiple mathematical models, and to extract the optimal model. The destination selection module is used to incorporate the trained probability model of urban residents' non-commuting activity destination selection into the non-commuting activity destination selection algorithm that takes into account the activity time constraints, so as to realize the selection of non-commuting activity destinations.

6. A device, characterized in that, The device includes, processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method described in any one of claims 1 to 4.

7. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 4.

Images