Method and system for measuring and calculating global passenger flow based on mobile signaling data

Abstract

The invention relates to a method and a system for measuring and calculating global passenger flow based on mobile signaling data. The method comprises the following steps: obtaining a resident identification result of each user by adopting a preset dynamic weighting model and the mobile signaling data; performing identification and classification processing on all users to obtain a global passenger identification and classification result; carrying out passing passenger filtering processing on the global passenger identification and classification result based on a preset city minimum residence time calculation model to obtain a first global passenger result; carrying out atypical tourist identification and elimination on the first global tourist result based on a preset special user group identification model to obtain a second global tourist result; and performing sample expansion calculation on the second global tourism result according to the market share of the target operator, and performing statistics to generate a global passenger flow analysis result. According to the invention, the dynamic weighting model is introduced to identify the residence, and the passing passengers and the atypical tourists are identified and filtered, so that accurate elimination of non-tourism behaviors is realized, and the purity and accuracy of tourism passenger flow statistics are improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of data analysis and data mining, and particularly relates to a method and system for measuring global passenger flow based on mobile signaling data. BACKGROUND

[0002] With the rapid development of social economy and the improvement of residents' living standards, tourism has become an important pillar of the national economy. Accurate, real-time and comprehensive passenger flow and tourism statistical data have irreplaceable strategic value for tourism resource optimization, public service facility planning, regional economic development decision-making and emergency management in the event of an emergency. According to the current definition of tourism, tourists usually refer to people who leave their usual living environment, travel more than 10 kilometers, travel more than 6 hours, and the purpose of travel is not to obtain remuneration.

[0003] Traditional passenger flow statistical methods, such as scenic spot ticket checking gate systems, fixed-point manual questionnaire surveys and hotel check-in registration, have many inherent defects. Not only is the statistical cost high and the manpower investment huge, but more importantly, the data coverage is limited to specific physical spaces (for example, scenic spots and hotels), which cannot capture the complete movement trajectory of tourists in the global range, resulting in a serious "information island" effect of the data.

[0004] In recent years, with the popularization of mobile communication technology, using mobile terminal signaling big data for passenger flow analysis has become a highly potential emerging technology path. Mobile signaling data can record the user's space-time position at a low cost and high frequency, naturally having the advantages of global coverage, real-time continuity and large sample size, which provides the possibility to break through the bottleneck of traditional statistical methods. However, the current passenger flow identification technology based on mobile phone signaling is still in its early stages, and the existing technical solutions have obvious deficiencies in the rigor of the identification logic and the processing ability of complex scenarios, resulting in a significant discount in the accuracy and scientificity of the statistical results. The common problem of existing technical solutions is that the identification model is relatively extensive, and most of them rely on single-dimensional static rules or fixed thresholds for judgment. For example, some methods simply distinguish between residents and tourists by the length of time a user stays in a certain area. This "one-size-fits-all" rule-based classification does not strictly follow the core definitions of "usual environment" and "travel purpose" in the tourism definition, and is prone to misjudgment.

[0005] The more severe challenge is that the population flow in the real world is extremely complex, and many non-tourism purposes of travel behavior are highly similar in space-time characteristics to tourism behavior. The existing technology lacks effective identification and filtering mechanisms, resulting in serious "noise" pollution of the statistical data. The deficiencies of the existing technology are mainly reflected in the following aspects:

[0006] Misjudgment of "passenger flow": Many passengers on long-distance intercity trips may only stop briefly in a certain city. Their spatiotemporal trajectories may meet simple tourist identification rules, making it difficult to effectively identify and exclude "passing" behavior. This exaggerates the number of tourists passing through the cities and causes passenger flow data to be inflated.

[0007] Confusion of large-scale "student mobility": During specific periods such as winter and summer vacations, large-scale student migrations back to their hometowns or schools often match the basic definition of tourists in terms of travel distance and duration. Due to a lack of effective identification of student groups, they are often misjudged as tourist traffic, leading to abnormal fluctuations and distortions in the number of tourists at specific time points.

[0008] Interference from highly mobile "professional behavior": With the rise of the platform economy, the mobility patterns of new business practitioners such as ride-hailing drivers, food delivery riders, and chauffeur drivers are characterized by high frequency, wide range, and long time periods. It is impossible to effectively distinguish between professional mobility and real leisure tourism behavior, thus incorrectly including a large number of professional trajectories in tourist statistics.

[0009] In summary, existing technologies for passenger flow calculation using mobile signaling face multiple technical bottlenecks, such as the vague definition of users' "habitual environment" and the lack of accurate identification methods for various typical non-tourism travel scenarios, resulting in inaccurate passenger flow calculations. Summary of the Invention

[0010] To address the aforementioned technical problems, this invention provides a method and system for calculating passenger flow across the entire region based on mobile signaling data.

[0011] The technical solution adopted in this invention is:

[0012] Firstly, a method for calculating overall passenger flow based on mobile signaling data is provided, including:

[0013] Obtain mobile signaling data from multiple users within the target province, and use a preset dynamic weighting model and mobile signaling data to obtain the permanent residence identification result for each user;

[0014] Using a pre-set global passenger identification and classification model and the results of permanent residence identification, all users are identified and classified to obtain the global passenger identification and classification results.

[0015] Based on the preset minimum stay duration calculation model for cities, the results of the identification and classification of passengers across the entire region are filtered for transit passengers to obtain the first result of passengers across the entire region.

[0016] Based on a pre-set special user group identification model, atypical tourists are identified and eliminated from the first overall passenger results to obtain the second overall passenger results.

[0017] According to the market share of the target operator, the second global tourism result is expanded sample calculation, and the global passenger flow analysis result is generated.

[0018] Further, the expression of the preset dynamic weighting model is:

[0019] ;

[0020] Among them, represents the resident score value of user A at a specific base station B on the tth day; represents the residence duration of user A at a specific base station B on the tth day; represents the normalized reference total duration; is a preset normalization coefficient; represents the historical score value accumulated by the user A at the specific base station B on the t-1th day, is a preset time decay factor.

[0021] Further, the mobile signaling data of multiple users in the target province is obtained, and the resident place recognition result of each user is obtained by using the preset dynamic weighting model and the mobile signaling data, including:

[0022] Obtain the mobile signaling data of multiple users in the target province;

[0023] According to the mobile signaling data, obtain the historical residence data of all users at different base stations of the target operator;

[0024] According to the historical residence data, obtain the current residence duration of different users at different base stations in the day;

[0025] Input the current residence duration into the preset dynamic weighting model to calculate the resident score value of different users at different base stations;

[0026] Compare the resident score value with the preset resident determination threshold, and identify the resident base station of each user as the resident place recognition result.

[0027] Further, the preset global passenger identification and classification model includes a province-in and province-out tourist determination sub-model and an overnight and one-day tourist classification sub-model,

[0028] Using the preset global passenger identification and classification model and the resident place recognition result, all users are identified and classified to obtain the global passenger identification and classification result, including:

[0029] Based on the province-in and province-out tourist determination sub-model and the resident place recognition result, the province-in and province-out tourists are identified;

[0030] Based on the overnight and day trip tourist classification sub-model and the resident place recognition result, overnight tourists and day trip tourists in the in-province tourists and out-of-province tourists are recognized, and the global passenger recognition classification result is obtained.

[0031] Further, based on the in-province and out-of-province tourist determination sub-model and the resident place recognition result, in-province tourists and out-of-province tourists are recognized, including:

[0032] Based on the in-province and out-of-province tourist determination sub-model and the resident place recognition result, users without a resident base station in the target province are recognized as out-of-province candidate users, and users with a resident base station are recognized as in-province users;

[0033] It is judged whether the first cumulative travel duration of the out-of-province candidate user in the target province is not less than a preset duration threshold; if the first cumulative travel duration is not less than the preset duration threshold, the out-of-province candidate user is determined to be an out-of-province tourist; if the first cumulative travel duration is less than the preset duration threshold, the out-of-province candidate user is determined to be a non-tourist;

[0034] It is judged whether the travel distance of the in-province user is not less than a preset spatial distance threshold, and whether the second cumulative travel duration is not less than a preset duration threshold, the travel distance being the straight-line spatial distance of the in-province user from the resident base station, and the second cumulative travel duration being the cumulative duration of the in-province user's travel distance being not less than the preset spatial distance threshold;

[0035] If the travel distance is not less than the preset spatial distance threshold, and the second cumulative travel duration is not less than the preset duration threshold, the in-province user is determined to be an in-province tourist;

[0036] If the travel distance is less than the preset spatial distance threshold and / or the second cumulative travel duration is less than the preset duration threshold, the in-province user is determined to be a non-tourist.

[0037] Further, based on the overnight and day trip tourist classification sub-model and the resident place recognition result, overnight tourists and day trip tourists in the in-province tourists and out-of-province tourists are recognized, and the global passenger recognition classification result is obtained, including:

[0038] Based on the overnight and day trip tourist classification sub-model and the resident place recognition result, the first cumulative residence duration of the out-of-province tourist in the target province in a preset night time window in a tourist destination area, and the second cumulative residence duration of the in-province tourist in the preset night time window in a night residence base station are obtained;

[0039] It is judged whether the first cumulative residence duration of the out-of-province tourist is not less than a preset night residence duration threshold;

[0040] If the first cumulative residence duration is not less than the preset night residence duration threshold, the out-of-province tourist is determined to be an out-of-province overnight tourist;

[0041] If the first cumulative residence duration of the out-of-province visitor is less than the preset overnight residence duration threshold, the out-of-province visitor is determined to be an out-of-province day trip visitor.

[0042] It is determined whether the second cumulative residence duration of the in-province visitor is not less than the preset overnight residence duration threshold, and whether the straight-line spatial distance between the overnight residence base station and any permanent base station is not less than the preset spatial distance threshold.

[0043] If the second cumulative residence duration is not less than the preset overnight residence duration threshold, and the straight-line spatial distance between the overnight residence base station and any permanent base station is not less than the preset spatial distance threshold, the in-province visitor is determined to be an in-province overnight visitor.

[0044] If the second cumulative residence duration is less than the preset overnight residence duration threshold and / or the straight-line spatial distance between the overnight residence base station and any permanent base station is less than the preset spatial distance threshold, the in-province visitor is determined to be an in-province day trip visitor.

[0045] According to the determination results of the overnight visitors and day trip visitors corresponding to the out-of-province visitors and the in-province visitors, a global passenger identification classification result is obtained.

[0046] Further, the global passenger identification classification result is filtered for passing-through passengers based on a preset city minimum residence duration calculation model to obtain a first global passenger result, including:

[0047] According to the global passenger identification classification result, in-province travel trajectories of each visitor among multiple cities in the target province are determined.

[0048] From the in-province travel trajectories, intermediate cities that are transfer stations or short-term stops are identified, and residence duration data in the intermediate cities are collected.

[0049] The residence duration data is visually analyzed to verify whether it presents a bimodal distribution statistical characteristic, in which one peak corresponds to short residence of passing-through passengers, and the other peak corresponds to longer residence of non-passing-through passengers.

[0050] A statistical or machine learning algorithm is used to automatically find and determine a critical point that can best segment the bimodal distribution in the bimodal distribution statistical characteristic, and a duration value corresponding to the critical point is taken as a minimum residence duration threshold of the intermediate city.

[0051] The preset city minimum residence duration calculation model with the minimum residence duration threshold is used to filter passing-through passengers in each city based on the global passenger identification classification result to obtain the first global passenger result.

[0052] Further, the preset special user group identification model includes a campus student identification sub-model and a professional user identification sub-model,

[0053] The first global passenger result is subjected to non-typical passenger identification and elimination based on a preset special user group identification model, and a second global passenger result is obtained, including:

[0054] The resident base station position of each passenger in the first global passenger result is matched with a preset list of base station geographic information of colleges and universities in the whole province, and the passenger whose resident base station position is in the list of base station geographic information of colleges and universities in the whole province is identified as a potential student; the real-name age information, campus exclusive communication package handling information and DPI data of the potential student are obtained, and the student identity of the potential student is subjected to multi-dimensional cross verification, and a campus student group is identified;

[0055] The daily average use time and the monthly active days of the professional application of each passenger in the first global passenger result are obtained, and the passenger whose daily average use time and monthly active days both reach the corresponding threshold is marked as a professional user; whether the travel trajectory of the professional user meets the passenger space-time condition is judged, if yes, the professional user is determined as a professional user passenger; if not, the professional user is determined as a professional user non-passenger;

[0056] The campus student group and the professional non-passenger are eliminated from the first global passenger result, and the second global passenger result is obtained.

[0057] In the second aspect, a system for measuring global passenger flow based on mobile signaling data is provided, including:

[0058] A resident place identification module is configured to obtain mobile signaling data of multiple users in a target province, and obtain a resident place identification result of each user by using a preset dynamic weighting model and the mobile signaling data;

[0059] A global passenger identification and classification module is configured to identify and classify all users by using a preset global passenger identification and classification model and the resident place identification result, and obtain a global passenger identification and classification result;

[0060] A passing passenger filtering module is configured to filter passing passengers from the global passenger identification and classification result based on a preset city minimum residence time calculation model, and obtain a first global passenger result;

[0061] A non-typical passenger identification module is configured to identify and eliminate non-typical passengers from the first global passenger result based on a preset special user group identification model, and obtain a second global passenger result;

[0062] A sample expansion calculation module is configured to calculate the second global passenger result based on the market share of a target operator, and generate a global passenger flow analysis result.

[0063] The present application has the following beneficial effects:

[0064] The system acquires mobile signaling data from multiple users within the target province. A pre-defined dynamic weighting model and the mobile signaling data are used to identify the permanent residence of each user. A pre-defined province-wide passenger identification and classification model, along with the permanent residence identification results, are used to identify and classify all users, resulting in province-wide passenger identification and classification results. Based on a pre-defined minimum stay duration calculation model for each city, the province-wide passenger identification and classification results are filtered for transit passengers, yielding the first province-wide passenger result. Based on a pre-defined special user group identification model, the first province-wide passenger result is used to identify and remove atypical tourists, yielding the second province-wide passenger result. The second province-wide tourism result is then expanded based on the target operator's market share, generating a province-wide passenger flow analysis result. By introducing a dynamic weighting model to identify permanent residence and identifying and filtering transit passengers and atypical tourists, the system achieves precise removal of non-tourism behaviors, improving the purity and accuracy of tourism passenger flow statistics. Attached Figure Description

[0065] Figure 1 This is a flowchart of the method for calculating the overall passenger flow based on mobile signaling data according to the present invention;

[0066] Figure 2 This is a structural diagram of the system for calculating passenger flow across the entire region based on mobile signaling data, as per the present invention. Detailed Implementation

[0067] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0068] like Figure 1 As shown, this embodiment of the invention provides a method for calculating the overall passenger flow based on mobile signaling data, including:

[0069] 101. Obtain mobile signaling data of multiple users within the target province, and use a preset dynamic weighting model and mobile signaling data to obtain the permanent residence identification result of each user;

[0070] In this embodiment, the core data is the mobile signaling data of all users in the target province and the target operator, which includes key information such as the user's anonymized ID, the latitude and longitude of the connected base station, and timestamps. This data forms the basis for constructing the user's high-precision spatiotemporal trajectory. Based on the mobile signaling data, the historical dwell time data of all users at different base stations under the target operator is obtained. By analyzing the historical dwell time data, the current dwell time of different users at different base stations on the same day can be obtained. Combined with the latitude and longitude of the connected base station, the activity trajectory of each user can be determined.

[0071] The auxiliary data includes a list of university base stations, DPI data, number segment dimension table and IMSI country code, and real-name registration table;

[0072] University base station list: a list that accurately marks the geographic location of base stations within the province's higher education institutions (such as teaching buildings, dormitory areas, libraries, and canteens), which is the key geographic fence basis for the campus student identification sub-model;

[0073] DPI data: Deep Packet Inspection (DPI) data, which can record users' usage behavior of specific applications (APP), such as usage duration, active period, and specific function page accessed, is a powerful tool for accurately identifying professional users (such as driver APP) and assisting in verifying student identity (such as learning APP);

[0074] Number segment table and IMSI country code: The number segment table stores the accurate mapping relationship between the first 7 digits of the mobile phone number and the home province; the country code in IMSI (International Mobile Subscriber Identity) (the country code for domestic users is "460") is used to directly and efficiently distinguish between domestic and overseas users. Both of them provide authoritative basis for the determination of the source of tourists from other provinces and overseas;

[0075] Real-name table: associated with the age, opening location, and other desensitization information of users registered at the operator, providing important support for the age filtering (such as 17-28 years old) and permanent residence determination of the campus student model;

[0076] Due to the inevitable inclusion of various types of noise in the original signaling data, if not effectively processed, it will seriously affect the accuracy of subsequent trajectory analysis and behavior identification, therefore, data preprocessing is also needed, the specific steps include:

[0077] (1) Basic cleaning: delete records with missing key fields (such as user ID, timestamp, base station location), and efficiently remove duplicate records with no location change produced by the same base station;

[0078] (2) Drift data filtering: identify and remove "drift points" caused by abnormal base station signals or cross-zone switching delays, drift points appear as users making long-distance displacement that is geographically impossible in a very short time (i.e. instantaneous speed far exceeds the physical limit of normal transportation tools) on the data, by setting dynamic speed and distance thresholds, this kind of severely distorted error data can be effectively filtered;

[0079] (3) Eliminate ping-pong switching data: ping-pong switching refers to the short, high-frequency, and no actual location movement between base stations to maintain optimal communication quality when users are at the edge or overlapping area of two or more adjacent base station signal coverage, this phenomenon will produce a large number of false mobile trajectory points, the elimination algorithm is as follows:

[0080] Set a fine time window threshold (e.g. 30 seconds or 1 minute) to detect local trajectory sequences;

[0081] Within the time window, detect whether there is a sequence pattern of repeated visits to a reference base station (such as A-B-A or A-B-C-A);

[0082] If such a ping-pong sequence is detected, the equivalent geographic center position is calculated using the weighted average of the residence time, and the single equivalent position point closer to the user's true residence center is used to replace the entire ping-pong sequence, and the calculation formula of the equivalent longitude and latitude is:

[0083] ;

[0084] ;

[0085] In the formula, represents the weighted average of the longitude, represents the weighted average of the latitude, is the residence time of the th signaling point, is the longitude of the th signaling point, is the latitude of the th signaling point;

[0086] The preset dynamic weighting model aims to accurately and dynamically identify the "customary environment" (i.e. the residence) of the user, thereby effectively distinguishing between local residents and visitors, and is the basis for all subsequent analysis. The core is the innovative dynamic weighting score algorithm of the resident base station, which is different from the traditional method of only statically counting the frequency or duration. This algorithm introduces a time decay mechanism, giving the model the ability to dynamically adapt to changes in user behavior patterns.

[0087] The expression of the preset dynamic weighting model is:

[0088] ;

[0089] Among them, represents the residence score value of user A at specific base station B on day t; represents the residence duration of user A at specific base station B on day t; represents the normalized total duration, for example, the period of 0:00-6:00 at night, is 6 hours, used for normalizing the residence contribution of the day; is a preset normalization coefficient, used to adjust the influence weight of the residence duration of the day on the total score, generally defaulting to 0.8; represents the historical score value accumulated by the user A at the specific base station B on day t-1, presetting a time decay factor, The presence of the time decay factor ensures that the historical scores decay exponentially over time, meaning that recent behaviors (such as a stay yesterday) have a higher weight than distant behaviors (such as a stay a month ago), allowing the model to update its residence judgment more quickly when the user experiences a lifestyle change (such as moving or changing jobs), effectively avoiding the excessive influence of historical "old" data and greatly improving the sensitivity and reality of the model.

[0090] The residence score value is compared with a preset residence determination threshold to identify the residence base station of each user, and the residence base station is taken as the residence recognition result.

[0091] The specific residence determination mainly includes three types:

[0092] Residence determination: the mobile signaling data of the user in the last 45 days is selected, and the trajectory of the night rest period from 0:00 to 6:00 is analyzed, and the residence score value of each appearing base station is iteratively calculated using the expression of the preset dynamic weighting model; for the residence, the preset residence determination threshold is specifically the residence threshold; for the work place, the preset residence determination threshold is specifically the work place threshold; generally, the residence is the most concentrated place of the user's trajectory, so the size of the residence threshold should be greater than the work place threshold; if the residence score value of a base station exceeds the residence threshold (for example, 300 points), the base station with the highest score among all base stations with a residence score value exceeding the residence threshold is determined as the residence base station of the user;

[0093] Work place determination: the mobile signaling data of the last 45 days is also selected, and the trajectory of the typical work period from 9:00 to 11:00 and 14:00 to 17:00 is analyzed, and if the residence score value of a base station exceeds the work place threshold (for example, 200 points), the base station with the highest score is determined as the work place base station of the user;

[0094] Third residence determination: in order to more comprehensively depict the "customary environment" of the user, in addition to the residence and work place determination, the trajectory of the user on weekends and statutory holidays is further analyzed to find other regular activity places (such as the parents' home regularly visited, the lovers' residence, etc.) with a higher score and a geographical distance greater than 10 kilometers from the former two places as the third residence base station of the user;

[0095] The determination results of the above residence base station, work place base station and third residence base station are integrated to form the residence recognition result corresponding to each user.

[0096] Since the passenger flow estimation is aimed at all users in the current target province, users who have appeared in the target province in recent time have little or even no historical residence data in the target province, and thus cannot exist in the permanent base station.

[0097] 102, using the preset global passenger identification and classification model and the permanent residence identification result, identifying and classifying all users to obtain a global passenger identification and classification result;

[0098] In this embodiment, the preset global passenger identification and classification model includes a province-in and province-out tourist determination sub-model and an overnight and day tour tourist classification sub-model.

[0099] The process of determining province-in and province-out tourists is as follows:

[0100] Based on the province-in and province-out tourist determination sub-model and the permanent residence identification result, users without a permanent base station in the target province are identified as province-out candidate users, and users with a permanent base station are identified as province-in users; this is the first hurdle to distinguish between local residents and tourists from other places.

[0101] It is determined whether the first cumulative travel duration of the province-out candidate user in the target province is not less than a preset duration threshold (specifically set to 6 hours), the first cumulative travel duration being the sum of the current residence durations of the province-out candidate user in different base stations in the target province on the same day; if the first cumulative travel duration is not less than the preset duration threshold, the province-out candidate user is determined to be a province-out tourist; if the first cumulative travel duration is less than the preset duration threshold, the province-out candidate user is determined to be a non-tourist; the purpose is to filter out short-term transits or short-term stays for non-tourism purposes.

[0102] It is determined whether the travel distance of the province-in user is not less than a preset spatial distance threshold and whether the second cumulative travel duration is not less than a preset duration threshold, the travel distance being the straight-line spatial distance of the province-in user from the permanent base station, and the second cumulative travel duration being the cumulative continuous duration of the travel distance of the province-in user being not less than the preset spatial distance threshold; the preset spatial distance threshold and the preset duration threshold are the space-time conditions of passenger behavior, for example, a user leaving the permanent base station more than 10 kilometers away for more than 6 hours is considered to be on a tour.

[0103] If the travel distance is not less than the preset spatial distance threshold and the second cumulative travel duration is not less than the preset duration threshold, the province-in user is determined to be a province-in tourist.

[0104] If the travel distance is less than the preset spatial distance threshold and / or the second cumulative travel duration is less than the preset duration threshold, the province-in user is determined to be a non-tourist.

[0105] After the above division of in-province users and out-of-province users into tourists / non-tourists, further analysis is performed on in-province tourists and out-of-province tourists to determine whether they overnight in the target province, and the process of classifying overnight and day-trip tourists is as follows:

[0106] The overnight and day-trip tourist classification sub-model aims to further classify the stay behavior of tourists, and provide key data support for in-depth analysis of tourism consumption potential and accommodation demand. The core rule is based on the fine analysis of user nighttime space-time behavior, and different logic is used for in-province and out-of-province tourists.

[0107] Based on the overnight and day-trip tourist classification sub-model and the resident place identification result, the first cumulative residence duration of out-of-province tourists in the target province within a preset nighttime time window is obtained, and the second cumulative residence duration of in-province tourists in the target province within a preset nighttime time window is obtained.

[0108] It is determined whether the first cumulative residence duration of out-of-province tourists is not less than a preset nighttime residence duration threshold. If the first cumulative residence duration is not less than the preset nighttime residence duration threshold, the out-of-province tourists are determined to be overnight tourists. If the first cumulative residence duration of out-of-province tourists is less than the preset nighttime residence duration threshold, the out-of-province tourists are determined to be day-trip tourists. Specifically, since out-of-province tourists naturally meet the requirement of long distance from their usual environment, the determination logic is relatively direct. If they have continuous residence records in the tourism destination area for more than 5 hours in the nighttime period (0:00-7:00), they are determined to be overnight tourists. Otherwise, they are determined to be day-trip tourists.

[0109] determine whether the second cumulative residence duration of the intraprovincial tourist is not less than a preset night residence duration threshold value and whether the straight-line spatial distance between the night residence base station and any regular base station is not less than a preset spatial distance threshold value; if the second cumulative residence duration is not less than the preset night residence duration threshold value and the straight-line spatial distance between the night residence base station and any regular base station is not less than the preset spatial distance threshold value, the intraprovincial overnight tourist is determined; if the second cumulative residence duration is less than the preset night residence duration threshold value and / or the straight-line spatial distance between the night residence base station and any regular base station is less than the preset spatial distance threshold value, the intraprovincial day tourist is determined; specifically, for intraprovincial tourists, in order to avoid misjudging the late-night social activities and entertainment of local residents as overnight tourism, stricter double constraints are set, first, the straight-line distance between the residence base station of the tourist and all the regular residence places (residence places, work places, and third regular residence places) identified by the tourist is greater than 10 km, to ensure that the tourist has really left the daily "customary environment"; second, the residence duration in the residence place meeting the distance condition is also required to meet the condition of continuous residence for 5 hours or more at night, only when both the time and space conditions are met, the tourist can be determined as an "overnight tourist"; otherwise, the tourist is classified as a "day tourist", and the differentiated and rigorous logic greatly improves the accuracy of classification.

[0110] According to the determination results of the overnight tourists and day tourists of the intraprovincial tourists and the extraprovincial tourists, the global passenger identification classification results are obtained.

[0111] 103, based on a preset city minimum residence duration calculation model, the global passenger identification classification results are filtered to obtain first global passenger results;

[0112] In this embodiment, the preset city minimum residence duration calculation model is a core innovative solution proposed for the industry pain point of misjudgment of "passing-by passenger flow";

[0113] The core idea is to abandon the static filtering rules of the whole country, and dynamically calculate a "minimum residence duration" threshold value for each city, and any tourist whose total residence time in the city is lower than the threshold value is regarded as a passing-by passenger rather than a real tourist and is removed, and the minimum residence duration threshold value is dynamic, and is dynamically generated according to the transportation hub status of different cities, the tourism resource endowment, and the data distribution of different time types (holidays / weekends / working days), so that the method can accurately analyze the passenger flow mode in a specific holiday and output the tourist distribution in the period;

[0114] According to the global passenger identification classification results, the intraprovincial tourism trajectories of each tourist in multiple cities in the target province are determined.

[0115] Identify the intermediate cities as transfer stations or short stays from the provincial tourism trajectory, and determine the length of stay data of tourists in the intermediate cities through historical residence data;

[0116] Visualize the length of stay data and verify whether it presents a bimodal distribution statistical characteristic, in which one peak corresponds to the short stay of passing-by tourists, and the other peak corresponds to the longer stay of non-passing-by tourists;

[0117] Use statistical or machine learning algorithms to automatically find and determine the critical point that can best segment the bimodal distribution, and take the length value corresponding to the critical point as the minimum length of stay threshold of the intermediate city; There are two methods for calculating the minimum length of stay threshold, which are variance analysis of maximum inter-group difference and K-Means clustering analysis, which will be described below.

[0118] Variance analysis of maximum inter-group difference method: sliding window scanning strategy is adopted to find a best segmentation point, so that the difference between the two groups of data (passing-by tourists group and multi-city tourism group) after segmentation is the largest, specifically, all possible length segmentation points are traversed, the inter-group sum of squares (SSB) after the data is divided into two groups is calculated, and the segmentation point that makes SSB the largest is the optimal threshold , and the objective function is:

[0119] ;

[0120] Wherein, and are the sample sizes of the two groups before and after the segmentation point, and are the mean values of the corresponding groups;

[0121] K-Means clustering analysis method: the classic K-Means unsupervised clustering algorithm (set K=2) is used to automatically cluster the length of stay data into two categories. The numerical upper limit of the cluster (center value is smaller) representing "passing-by tourists" is another candidate threshold ;

[0122] Threshold fusion: to balance the emphasis of the two methods and enhance the stability of the results, the final city minimum length of stay threshold is the average of the two: ;

[0123] Using the preset city minimum length of stay calculation model of the minimum length of stay threshold to filter the passing-by tourists in each city of the global passenger identification and classification result, and obtain the first global passenger result.

[0124] 104, based on the preset special user group identification model, the first global passenger result is identified and excluded from the non-typical tourists, and the second global passenger result is obtained;

[0125] In this embodiment, considering the systematic interference of the regular large-scale flow of student groups during the summer and winter vacations on the statistics of tourist flow, and the difficulty in distinguishing the working behavior of high-mobility professional groups (such as online car drivers and delivery riders) from real tourism behavior, a special user group recognition model is preset to recognize and exclude atypical tourists;

[0126] The preset special user group recognition model includes a campus student recognition sub-model and a professional user recognition sub-model;

[0127] The resident base station location of each tourist in the first global passenger result is matched with the preset college base station geographic information list in the whole province; the tourists whose resident base station location is in the preset college base station geographic information list in the whole province are identified as potential students;

[0128] The real-name age information, campus exclusive communication package handling information and DPI data of the potential students are obtained, and the student identity of the potential students is verified in multiple dimensions to identify the campus student group. Specifically, users aged 17-28 years old covering the main age range of undergraduates, master's and doctoral students are selected, and teachers, family members and surrounding residents whose ages do not match are excluded. Through the business data of the operator, users who have handled campus exclusive packages (such as campus cards of dynamic zones) or summer vacation-oriented traffic preferential packages are accurately identified. Through DPI data, users who frequently use specific learning APPs (such as Superstar, CNKI, Learning, Rain Classroom, etc.) are identified to support their student identity from the behavior level. The communication circle can also be supplemented to identify the recalled students whose features are not obvious (such as not using campus packages). If a user has frequent communication records with multiple (such as ≥3) potential students determined in the previous steps in the near future, it will also be listed as a potential student for further verification.

[0129] The daily use time and monthly active days of each tourist's professional class application in the first global passenger result are obtained, and the users who frequently and intensively use specific professional class APPs are labeled as "professional users" by using DPI data. The threshold is dynamic and consistent with the industry characteristics, for example: the daily use of the driver side / app side APP order page is ≥1 hour, and the monthly active days are ≥20 days.

[0130] It is judged whether the travel trajectory of the professional user meets the passenger space-time condition, and the passenger space-time condition is more than 10 kilometers away from the resident base station and the travel time is more than 6 hours;

[0131] In order to perform one-step verification, it is also necessary to check the professional APP usage time of the professional user on the day of travel. If the daily usage time is very low (such as <1 hour), it is preliminarily judged that the user may be in a rest or vacation state on that day, and there is a possibility of tourism;

[0132] The user's residence duration at the destination is compared with the dynamically calculated minimum residence duration threshold of the city, and if the residence duration is greater than or equal to the threshold, it is finally determined that the professional user's this time is a real tourism behavior, and is regarded as a professional user passenger; otherwise, it is still regarded as a professional behavior or a passing behavior, and is regarded as a non-professional user passenger;

[0133] It should be noted that the method system of the present application has high dynamic self-adaptive ability. For example, during the Spring Festival, National Day and other major holidays, the threshold of part of the model can be automatically or manually adjusted by experts to adapt to abnormal behavior patterns. For example, the use duration threshold of the professional APP in the professional user identification sub-model is appropriately relaxed (from 1 hour to 1.5 hours) to cope with the short-term impact of order fluctuations on the driver's work mode during the holiday, and to ensure the robustness of the model.

[0134] The campus student group and the professional non-passenger are removed from the first global passenger result to obtain a second global passenger result.

[0135] 105, according to the market share of the target operator, the second global tourism result is expanded, and the global passenger flow analysis result is generated by statistics.

[0136] In this embodiment, in order to fundamentally solve the problem of inaccurate tourist source statistics, a more scientific differentiated expansion method by province is adopted, and the specific steps are as follows:

[0137] Data preparation: Obtain the latest mobile operator market share authoritative data of each province in China;

[0138] Provincial expansion calculation: for tourist sample quantities from different provinces, instead of using the unified national market share, the exclusive market share of the corresponding source province is used for expansion, and the global tourist quantity calculation formula of province O is:

[0139] ;

[0140] Summarize the national tourist quantity: accurately add the expansion results of all provinces to obtain the global total tourist quantity:

[0141] ;

[0142] Wherein: is the tourist source province index, represents the mobile market share of province O;

[0143] After the fine processing, identification, filtering and calculation of the above steps 101-105, a series of rich, deep and high-timeliness analysis products can be generated in the result output layer, including not only the conventional indicators such as the macro total tourist statistics, the tourist composition in and out of the province, the overnight and day trip tourist ratio, but also the deep analysis reports of specific time dimensions according to specific needs, specifically including: annual tourist analysis report, monthly passenger flow monitoring report, holiday special analysis report, multi-dimensional visual chart;

[0144] The annual tourist analysis report can macroscopically present the annual passenger flow trend and the distribution of hot spots in the province; the monthly passenger flow monitoring report provides data support for monthly marketing and capacity allocation; the holiday special analysis report accurately outputs the tourist source and destination distribution during a specific holiday, providing decision basis for holiday tourism management and emergency response; the multi-dimensional visual chart includes but is not limited to city passenger flow heat map, proportion of students and professional users in tourists, etc., providing intuitive and multi-dimensional decision support for tourism management departments and related industries.

[0145] The beneficial effects achieved by the embodiments of the present application are:

[0146] By introducing a dynamic weighted score model with time decay to identify the resident place, the method is more scientific and sensitive than the traditional static frequency statistical method, can adapt to user behavior changes, and can construct a special and multi-dimensional feature fusion filtering model for various typical noise scenes such as passing-by passenger flow, student flow and professional behavior, and realize accurate elimination of non-tourism behavior, thereby fundamentally improving the purity and accuracy of tourism passenger flow statistics;

[0147] Not only can the tourists be identified, but also multi-dimensional portraits can be made, such as accurately distinguishing between tourists from inside and outside the province, overnight / day trip behavior, and effectively distinguishing real tourism behavior from special groups such as students and professional drivers;

[0148] The city minimum residence time threshold is dynamically calculated based on local and real-time massive data distribution, and can be adaptively adjusted according to different scenes such as holidays and working days, so as to flexibly adapt to the changes of passenger flow mode in different regions and different times, and has stronger scientificity, robustness and universality;

[0149] The market share of each province is used for differential expansion, which solves the distortion problem of tourist source structure caused by the traditional single expansion factor, so that the analysis result of tourist source is closer to the real market situation.

[0150] Based on the method for measuring the global passenger flow based on mobile signaling data described in the above embodiments, the system for measuring the global passenger flow based on mobile signaling data is described by embodiments.

[0151] AsFigure 2 As shown, the embodiment of the present application provides a system for measuring global passenger flow based on mobile signaling data, comprising:

[0152] The resident place identification module 201 is configured to acquire mobile signaling data of multiple users in a target province, and obtain resident place identification results of each user by using a preset dynamic weighting model and the mobile signaling data.

[0153] The global passenger identification and classification module 202 is configured to identify and classify all users by using a preset global passenger identification and classification model and the resident place identification results, and obtain global passenger identification and classification results.

[0154] The passing passenger filtering module 203 is configured to perform passing passenger filtering processing on the global passenger identification and classification results based on a preset city minimum residence time calculation model, and obtain first global passenger results.

[0155] The atypical passenger identification module 204 is configured to perform atypical passenger identification and elimination on the first global passenger results based on a preset special user group identification model, and obtain second global passenger results.

[0156] The sample expansion calculation module 205 is configured to perform sample expansion calculation on the second global passenger results according to a market share of a target operator, and generate global passenger flow analysis results.

[0157] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0158] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks

[0159] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow Figure 1 The flow or flows and / or blocks Figure 1 The flow or flows and / or blocks

[0160] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions that execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 The flow or flows and / or blocks Figure 1 The flow or flows and / or blocks

[0161] The above merely provides an embodiment of the present application, but is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall fall within the scope of the claims of the present application.

Claims

1. A method for calculating passenger flow across an entire region based on mobile signaling data, characterized in that, include: Obtain mobile signaling data of multiple users within the target province, and use a preset dynamic weighting model and the mobile signaling data to obtain the permanent residence identification result of each user; Using a pre-defined global passenger identification and classification model and the permanent residence identification results, all users are identified and classified to obtain global passenger identification and classification results. Based on the preset minimum stay duration calculation model for cities, the transit passenger filtering process is applied to the overall passenger identification and classification results to obtain the first overall passenger result. Based on a preset special user group identification model, atypical tourists are identified and eliminated from the first overall passenger results to obtain the second overall passenger results. Based on the market share of the target operator, the second all-for-one tourism results are expanded and calculated to generate all-for-one passenger flow analysis results.

2. The method for calculating the entire area passenger flow based on mobile signaling data according to claim 1, characterized in that, The expression for the preset dynamic weighted model is: ； Among them, the This represents the score of user A's permanent presence at a specific base station B on day t; This represents the duration of user A's stay at the specific base station B on day t; The reference total duration is the normalized baseline; The preset normalization coefficients; This represents the historical score value that user A has accumulated at the specific base station B on day t-1. This is the preset time decay factor.

3. The method for calculating the entire area passenger flow based on mobile signaling data according to claim 2, characterized in that, The process of acquiring mobile signaling data from multiple users within a target province, and using a preset dynamic weighting model and the mobile signaling data to obtain the permanent residence identification result for each user, includes: Acquire mobile signaling data from multiple users within a target province; Based on the mobile signaling data, historical dwell data of all users at different base stations under the target operator are obtained; Based on the historical dwell time data, the current dwell time of different users at different base stations within the same day can be obtained; The current dwell time is input into the preset dynamic weighted model to calculate the dwell score of different users on different base stations; The resident score is compared with a preset resident determination threshold to identify the resident base station of each user, and the resident base station is used as the resident location identification result.

4. The method for calculating the entire area passenger flow based on mobile signaling data according to claim 3, characterized in that, The preset global tourist identification and classification model includes sub-models for determining tourists from within and outside the province, as well as sub-models for classifying overnight and day-trip tourists. The process employs a pre-defined global passenger identification and classification model and the results of the permanent residence identification to identify and classify all users, resulting in global passenger identification and classification results, including: Based on the sub-model for determining tourists from within and outside the province and the results of the permanent residence identification, tourists from within and outside the province are identified. Based on the overnight and day-trip tourist classification sub-model and the permanent residence identification results, overnight tourists and day-trip tourists among the tourists within the province and the tourists outside the province are identified, and the overall tourist identification and classification results are obtained.

5. The method for calculating the entire area passenger flow based on mobile signaling data according to claim 4, characterized in that, The identification of tourists from within and outside the province based on the provincial and out-of-province tourist determination sub-model and the permanent residence identification results includes: Based on the provincial and out-of-province tourist determination sub-model and the permanent residence identification results, users without permanent base stations in the target province are identified as out-of-province candidate users, and users with permanent base stations are identified as in-province users. Determine whether the first cumulative travel time of the candidate user from outside the province within the target province is not less than a preset time threshold; If the first cumulative travel time is not less than the preset time threshold, then the candidate user from outside the province is determined to be a tourist from outside the province. If the first cumulative travel time is less than the preset time threshold, then the candidate user from outside the province is determined to be a non-tourist. Determine whether the travel distance of the user within the province is not less than a preset spatial distance threshold, and whether the second cumulative travel time is not less than the preset time threshold. The travel distance is the straight-line spatial distance of the user within the province from the permanent base station, and the second cumulative travel time is the cumulative duration during which the travel distance of the user within the province is not less than the preset spatial distance threshold. If the travel distance is not less than the preset spatial distance threshold and the second cumulative travel time is not less than the preset time threshold, then the user within the province is determined to be a tourist within the province. If the travel distance is less than the preset spatial distance threshold and / or the second cumulative travel time is less than the preset duration threshold, then the user within the province is determined to be a non-tourist.

6. The method for calculating the entire area passenger flow based on mobile signaling data according to claim 5, characterized in that, Based on the overnight and day-trip tourist classification sub-model and the permanent residence identification results, overnight and day-trip tourists are identified among tourists from within the province and tourists from outside the province, resulting in a comprehensive tourist identification and classification result, including: Based on the overnight and day-trip tourist classification sub-model and the permanent residence identification results, the first cumulative stay duration of the out-of-province tourist in the tourist destination area within the target province within the preset nighttime time window, and the second cumulative stay duration of the in-province tourist at the nighttime base station within the preset nighttime time window are obtained. Determine whether the first cumulative stay duration of the out-of-province tourist is not less than a preset nighttime stay duration threshold; If the first cumulative stay duration is not less than the preset nighttime stay duration threshold, then the out-of-province tourist is determined to be an out-of-province overnight tourist; If the first cumulative stay duration of the out-of-province tourist is less than the preset nighttime stay duration threshold, then the out-of-province tourist is determined to be a one-day out-of-province tourist. Determine whether the second cumulative stay duration of the tourists within the province is not less than the preset nighttime stay duration threshold, and whether the straight-line spatial distance between the nighttime base station and any of the permanent base stations is not less than the preset spatial distance threshold; If the second cumulative stay duration is not less than the preset nighttime stay duration threshold, and the straight-line spatial distance between the nighttime base station and any of the permanent base stations is not less than the preset spatial distance threshold, then the tourist from within the province is determined to be an overnight tourist from within the province. If the second cumulative stay time is less than the preset night stay time threshold and / or the straight-line spatial distance between the night stay base station and any of the permanent base stations is less than the preset spatial distance threshold, then the tourist from within the province is determined to be a tourist on a one-day tour within the province. Based on the determination results of overnight tourists and day-trip tourists corresponding to tourists from outside the province and tourists from within the province, the identification and classification results of tourists across the entire region are obtained.

7. The method for calculating the entire area passenger flow based on mobile signaling data according to claim 6, characterized in that, The transit passenger filtering process is applied to the overall passenger identification and classification results based on the preset minimum stay duration calculation model for prefecture-level cities to obtain the first overall passenger result, including: Based on the overall passenger identification and classification results, determine the intra-provincial travel trajectory of each tourist among multiple cities in the target province; Identify intermediate cities that serve as transit points or short-term stops from the travel routes within the province, and collect data on the length of stay in these intermediate cities. Visual analysis of the dwell time data is performed to verify whether it exhibits a bimodal distribution statistical characteristic, in which one peak corresponds to the short dwell time of passing passengers and the other peak corresponds to the longer dwell time of non-passing passengers. Using statistical or machine learning algorithms, the critical point that best segments the bimodal distribution statistical features is automatically found and determined, and the duration value corresponding to the critical point is used as the minimum residence duration threshold of the intermediate city. The minimum stay duration calculation model for each city is used to filter the transit passengers in each city based on the minimum stay duration threshold of the whole region passenger identification and classification results, so as to obtain the first whole region passenger result.

8. The method for calculating the entire area passenger flow based on mobile signaling data according to claim 7, characterized in that, The preset special user group identification model includes a campus student identification sub-model and a professional user identification sub-model. The step of identifying and eliminating atypical tourists based on a preset special user group identification model to obtain a second set of overall passenger results includes: The permanent base station locations of each tourist in the first global passenger results are matched with a pre-set list of geographical information of university base stations in the province; tourists whose permanent base station locations are in the pre-set list of geographical information of university base stations in the province are identified as potential students; the real-name age information, campus-exclusive communication package application information and DPI data of the potential students are obtained, and the student identity of the potential students is cross-verified in multiple dimensions to identify the campus student group. Obtain the average daily usage time and monthly active days of each tourist's occupational application in the first global passenger results, and mark those whose average daily usage time and monthly active days both reach the corresponding threshold as occupational users; determine whether the travel trajectory of the occupational user meets the spatiotemporal conditions of a passenger. If it does, determine that the user is an occupational user passenger; if it does not, determine that the user is an occupational user non-passenger. The campus student group and the professional non-travelers are removed from the first global traveler results to obtain the second global traveler results.

9. A system for calculating passenger flow across an entire region based on mobile signaling data, characterized in that, include: The permanent residence identification module is used to acquire mobile signaling data of multiple users within the target province, and to obtain the permanent residence identification result of each user by using a preset dynamic weighting model and the mobile signaling data. The whole-domain passenger identification and classification module is used to identify and classify all users using a preset whole-domain passenger identification and classification model and the permanent residence identification results, and to obtain the whole-domain passenger identification and classification results. The transit passenger filtering module is used to perform transit passenger filtering on the whole-area passenger identification and classification results based on a preset city minimum stay duration calculation model to obtain the first whole-area passenger result. The atypical passenger identification module is used to identify and remove atypical tourists from the first overall passenger results based on a preset special user group identification model to obtain the second overall passenger results. The expansion calculation module is used to expand the second all-area tourism results based on the market share of the target operator and statistically generate all-area passenger flow analysis results.

Images