Crowd flow prediction system and method
By establishing a neural network prediction model through the data obtained from social media, the existing people's traffic prediction method is solved, and more accurate people's traffic prediction is achieved, providing objective data support for traffic and scenic spot management.
Patent Information
- Application Number
- CN202311499748.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-10
- Publication Date
- 2024-12-31
- Estimated Expiration
- 2043-11-10
AI Technical Summary
The existing methods of traffic prediction rely on experience, are subjective and have large errors, making it difficult to provide accurate prediction results.
By obtaining text data, image data and their interaction data related to the target area during the historical time period from social media, a neural network prediction model is established, and these data are used to predict future traffic. The specific steps include obtaining data from the first historical time period, obtaining the traffic of people in the second historical time period, establishing a model, and using the model to make predictions within the current time period.
It provides relatively objective and accurate flow forecast results to help traffic optimization, scenic spot management and activity planning.
Smart Images
Figure CN117592599B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of computer technology, and more specifically, to a method for predicting the flow of people. Background Art
[0002] The flow of people is crucial for traffic optimization, scenic area management, and event planning. However, the current prediction of the flow of people mainly relies on the flow of people data in the same period of previous years, or relies on simple predictions based on factors such as weather, cultural and sports activities. Obviously, this prediction method relies on human experience, is relatively subjective, and has a large deviation from the actual flow of people. Therefore, it is necessary to design a technical solution that can overcome the above defects to a certain extent. Summary of the invention
[0003] An object of the present invention is to provide a crowd flow prediction system and method, which can achieve accurate prediction of crowd flow in a target area and provide a reference for traffic optimization, scenic area management, and activity planning.
[0004] In order to achieve these purposes and other advantages of the present invention, according to one aspect of the present invention, the present invention provides a method for predicting human flow, including: S1: obtaining text data, image data and interaction data thereof related to a selected area in a first historical time period from social media; S2: obtaining the human flow of the selected area in a second historical time period, the second historical time period being later than the first historical time period; S3: establishing a neural network prediction model based on the interaction data obtained in S1 and the human flow obtained in S2; S4: obtaining text data, image data and interaction data thereof related to a target area in the current time period from social media, inputting the data into the neural network prediction model, and predicting the human flow of the target area in the future time period.
[0005] Furthermore, the interaction data includes the number of comments, forwardings, favorites, likes and readings.
[0006] Furthermore, keywords and key objects matching the selected area and the target area are determined, and text data related to the selected area and the target area are extracted based on the keywords, and image data related to the selected area and the target area are extracted based on the key objects.
[0007] Furthermore, it also includes identifying positive keywords that are positively correlated with increasing the flow of people and reverse keywords that are negatively correlated with increasing the flow of people in the text data, dividing the text data into a forward group and a reverse group according to the forward keywords and the reverse keywords, calculating a positive index using the interaction data of the text data and image data in the forward group, calculating a reverse index using the interaction data of the text data in the forward group, taking the forward index and the reverse index as weighted averages of the interaction data, and constructing a neural network prediction model using the forward index, the reverse index and the flow of people.
[0008] Furthermore, the publishing location information and publisher account information corresponding to the text data are obtained, the text data is graded, and in the process of calculating the positive index and the negative index, different coefficients are assigned to the interaction data of text data of different levels.
[0009] Furthermore, the second historical time period is 3 to 5 days later than the first historical time period.
[0010] The present invention also provides a pedestrian flow prediction system, including: a first data collection module, used to obtain text data, image data and their interaction data related to a selected area in a first historical time period from social media, and used to obtain text data, image data and their interaction data related to a target area in a current time period from social media; a second data collection module, used to obtain the pedestrian flow of a selected area in a second historical time period, the second historical time period being later than the first historical time period; a model building module, used to establish a neural network prediction model based on the obtained interaction data and the obtained pedestrian flow; a prediction module, used to use the neural network prediction model to predict the pedestrian flow of a target area in a future time period.
[0011] Furthermore, the first data collection module is also used to determine keywords and key objects matching the selected area and the target area, extract text data related to the target area based on the keywords, and extract image data related to the target area based on the key objects; the interaction data includes the number of comments, reposts, likes and readings.
[0012] Furthermore, the keywords include positive keywords that are positively correlated with increasing the flow of people and reverse keywords that are negatively correlated with increasing the flow of people. The model building module is also used to divide the text data into a forward group and a reverse group according to the forward keywords and the reverse keywords, calculate the positive index using the interaction data of the text data and the image data in the forward group, calculate the reverse index using the interaction data of the text data in the forward group, take the forward index and the reverse index as the weighted average of the interaction data, and build a neural network prediction model using the forward index, the reverse index and the flow of people.
[0013] Furthermore, the first data collection module is also used to obtain the publishing location information and publisher account information corresponding to the text data, classify the text data, and assign different coefficients to the interaction data of text data of different levels.
[0014] The present invention has at least the following beneficial effects:
[0015] The present invention obtains text data, image data and their interactive data related to a selected area in a first historical time period from social media, obtains the flow of people in the selected area in a second historical time period, establishes a neural network prediction model, and uses the neural network prediction model to predict the flow of people in the target area in a future time period; the present invention provides relatively objective data and prediction methods for the prediction of flow of people, and the prediction results are accurate, which provides a basis for traffic optimization, scenic area management, and activity planning.
[0016] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 This is a flow chart of an embodiment of the present application. DETAILED DESCRIPTION
[0018] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.
[0019] It should be understood that terms such as “having”, “including” and “comprising” used herein do not exclude the existence or addition of one or more other elements or combinations thereof.
[0020] like Figure 1 As shown, the embodiment of the present application provides a method for predicting human flow, including:
[0021] S1: Acquire text data, image data and interaction data related to the selected area in the first historical time period from social media; optionally, social media may include Weibo, Xiaohongshu, etc.; optionally, the selected area includes a scenic spot, a supermarket, an administrative division, etc., such as the Yellow Crane Tower Scenic Spot, a zoo, and AEON Supermarket; the text data is the comments related to the selected area posted by users, and the image data is the photos related to the selected area posted by users; optionally, the first historical time period is generally selected to be 7 to 10 days, and a shorter or longer time will have a weaker correlation with the flow of people; the interaction data includes common interaction forms, and optionally includes the number of comments, the number of reposts, the number of favorites, the number of likes, and the number of reads;
[0022] S2: Obtain the flow of people in the selected area in a second historical time period, the second historical time period being later than the first historical time period; optionally, the second historical time period is 3 to 5 days later than the first historical time period, and the flow of people generated in the second historical time period has a strong correlation with the social media activities in the first historical time period; optionally, the flow of people can be obtained from the flow of people data released by the management department of the selected area, or the video captured by the camera in the selected area, the number of people in the video is counted, and the flow of people in the selected area is estimated in combination with the area of the selected area;
[0023] S3: Based on the interaction data obtained in S1 and the flow of people obtained in S2, a neural network prediction model is established; optionally, the neural network is a long short-term memory network LSTM, the interaction data is digitized as input, and the flow of people is used as output, and the neural network prediction model is trained; optionally, 10 to 15 areas are selected, and 3 to 4 first historical time periods are selected, with a total of 30 to 60 groups of data, which are divided into a training group and a test group, and the neural network prediction model is used after verification;
[0024] S4: obtaining text data, image data and interaction data related to the target area in the current time period from social media, inputting the data into the neural network prediction model, and predicting the flow of people in the target area in the future time period; the current time period is the same length as the first historical time period, and the future time period is the same length as the second historical time period; the target area is the area where the flow of people needs to be predicted; optionally, the interaction data in the current time period is processed in the manner of S3, and inputting the data into the neural network prediction model, so as to obtain the flow of people in the future time period;
[0025] Compared with the prior art that estimates passenger flow based on historical passenger flow, weather, and cultural and sports activities during the same period, this embodiment provides more objective data and prediction methods for passenger flow prediction, with accurate prediction results, and provides a basis for traffic optimization, scenic area management, and activity planning.
[0026] In another embodiment, keywords and key objects matching the selected area and the target area are determined, and text data related to the selected area and the target area are extracted based on the keywords, and image data related to the selected area and the target area are extracted based on the key objects;
[0027] Optionally, the keywords are words that can reflect the specificity of the selected area or target area. For example, for the Yellow Crane Tower Scenic Area, you can select keywords such as Yellow Crane Tower, Yangtze River Bridge, Baiyun Pavilion, Snake Mountain, Cui Hao, etc., and you can select the shape of the Yellow Crane Tower and the shape of the Yangtze River Bridge as key objects; by acquiring social data published within a certain geographical range, through keyword recognition and image recognition, you can get text data related to the Yellow Crane Tower Scenic Area and image data related to the Yellow Crane Tower Scenic Area; preferably, keywords and key objects can be further increased according to the cultural and sports activities held in the selected area or target area during the same period.
[0028] In another embodiment, it also includes identifying positive keywords that are positively correlated with increasing the flow of people and reverse keywords that are negatively correlated with increasing the flow of people in the text data, dividing the text data into a positive group and a reverse group according to the positive keywords and the reverse keywords, calculating a positive index using the interaction data of the text data and the image data in the positive group, calculating a reverse index using the interaction data of the text data in the positive group, taking the positive index and the reverse index as weighted averages of the interaction data, and constructing a neural network prediction model using the positive index, the reverse index and the flow of people;
[0029] Optionally, the positive keywords are words with positive evaluation, such as beautiful scenery, fun, good service, good atmosphere, etc., and the reverse keywords are words with negative evaluation, such as crowded, difficult parking, expensive, poor experience, etc.; the positive keywords and reverse keywords are used to divide the text data into two groups, a positive group and a reverse group. The positive group will increase the flow of people, while the reverse group will reduce the flow of people. The image data will increase the flow of people by default; the weighted average is used to calculate the interaction data of the positive group and the image data to obtain the positive index, and the interaction data of the reverse group is calculated to obtain the reverse index; optionally, the positive index and the reverse index are used to calculate the interaction data of the reverse group. The indicators are the weighted averages of the number of comments, forwardings, favorites, likes and readings. The number of comments has the greatest influence and the largest weight. The number of forwardings and likes have similar influences and their weights are slightly lower than the number of comments. The number of readings has the smallest weight. For example, taking Weibo as an example, the weights of the number of comments, forwardings, likes and readings are 0.5, 0.2, 0.2 and 0.1 respectively. Preferably, the judgment matrix can be combined to determine the weights in actual applications. Optionally, a neural network prediction model is established with positive indicators and reverse indicators as input and traffic as output.
[0030] In another embodiment, the publishing location information and publisher account information corresponding to the text data are obtained, the text data is graded, and in the process of calculating the positive index and the negative index, different coefficients are assigned to the interaction data of text data of different levels;
[0031] This embodiment aims to reduce the interference of low-quality and low-relevance social information on the results by grading text data. Generally speaking, the closer the publishing location is to the selected area or target area and the better the publishing account is, the higher the authenticity of the information is and the greater the correlation with the flow of people. Conversely, there is more low-quality and low-relevance information. Specifically, while obtaining the text data, the location information and publisher account information corresponding to the text data are also obtained, and the location information and publisher account information are used to grade the text data. The publisher account level can be replaced by the Weibo account level or the number of fans, and the publishing location information can be obtained from the location information of the social information. Optionally, five distance ranges are divided from near to far according to the distance from the center of the selected area or target area, and the publisher account is divided into five distance ranges according to the publisher account. The account level or number of fans is divided into multiple five account level intervals; multiple levels are divided according to the combination of distance range and account level interval, and coefficients are given. For example, if it is in the nearest first distance range and in the first interval with the highest account level interval, a coefficient of 1 is given; for example, if it is in the first distance range and the account level is in the second interval or in the second distance range and the account level is in the first interval, a coefficient of 0.9 is given, and so on. A coefficient is assigned to each text data; preferably, the specific coefficient can be determined using a judgment matrix; this embodiment sorts the text data by publishing location information and publisher account information, highlights high-quality and high-relevance social information, reduces the interference of low-quality and low-relevance social information on the traffic prediction results, and improves the accuracy of traffic prediction.
[0032] The embodiment of the present application also provides a crowd flow prediction system, including:
[0033] The first data collection module is used to obtain text data, image data and interaction data related to the selected area in the first historical time period from social media, and to obtain text data, image data and interaction data related to the target area in the current time period from social media; optionally, social media may include Weibo, Xiaohongshu, etc.; optionally, the selected area includes a scenic spot, a supermarket, an administrative division, etc., such as the Yellow Crane Tower scenic spot, a zoo, and AEON Supermarket; the text data is the comments posted by users related to the selected area, and the image data is the photos posted by users related to the selected area; optionally, the first historical time period is generally selected to be 7 to 10 days, and a shorter or longer time will have a weaker correlation with the flow of people; the interaction data includes common interaction forms, and optionally includes the number of comments, the number of reposts, the number of likes, and the number of reads;
[0034] A second data collection module is used to obtain the flow of people in the selected area in a second historical time period, and the second historical time period is later than the first historical time period; a model building module is used to establish a neural network prediction model based on the obtained interaction data and the obtained flow of people; optionally, the second historical time period is 3 to 5 days later than the first historical time period, and the flow of people generated in the second historical time period has a strong correlation with the social media activities in the first historical time period; optionally, the flow of people can be obtained from the flow of people data released by the management department of the selected area, or the video captured by the camera in the selected area, the number of people in the video is counted, and the flow of people in the selected area is estimated in combination with the area of the selected area;
[0035] A prediction module is used to use the neural network prediction model to predict the flow of people in a target area in a future time period; the current time period is the same length as the first historical time period, and the future time period is the same length as the second historical time period; the target area is the area where the flow of people needs to be predicted; optionally, the interaction data in the current time period is processed in the same way as the interaction data in the first historical time period, and input into the neural network prediction model to obtain the flow of people in the future time period.
[0036] In another embodiment, the first data collection module is also used to determine keywords and key objects matching the selected area and the target area, and extract text data related to the target area based on the keywords, and extract image data related to the target area based on the key objects; the interaction data includes the number of comments, the number of reposts, the number of favorites, the number of likes and the number of reads; optionally, the keywords are words that can reflect the specificity of the selected area and the target area, such as the Yellow Crane Tower Scenic Area, you can choose the keywords such as Yellow Crane Tower, Yangtze River Bridge, Baiyun Pavilion, Snake Mountain, Cui Hao, etc., and you can choose the shape of the Yellow Crane Tower and the shape of the Yangtze River Bridge as key objects; by acquiring social data published within a certain geographical range, through keyword recognition and image recognition, you can get the text data related to the Yellow Crane Tower Scenic Area and the image data related to the Yellow Crane Tower Scenic Area; preferably, the keywords and key objects can be further increased according to the cultural and sports activities held in the selected area and the target area during the same period.
[0037] In another embodiment, it also includes identifying positive keywords that are positively correlated with increasing the flow of people and reverse keywords that are negatively correlated with increasing the flow of people in the text data, and the model building module is also used to divide the text data into a forward group and a reverse group according to the forward keywords and the reverse keywords, calculate the positive index using the interaction data of the text data and the image data in the forward group, calculate the reverse index using the interaction data of the text data in the forward group, take the forward index and the reverse index as the weighted average of the interaction data, and use the forward index, the reverse index and the flow of people to build a neural network prediction model; optionally, the forward keywords are words for positive evaluation, such as beautiful scenery, fun, good service, good atmosphere, etc., and the reverse keywords are words for negative evaluation, such as crowded, difficult parking, expensive, poor experience, etc.; use the forward keywords and the reverse keywords to divide the text data into a forward group and a reverse group, and calculate the reverse index using the interaction data of the text data in the forward group. The data are divided into two groups, a positive group and a negative group. The positive group will increase the flow of people, while the negative group will reduce the flow of people. The image data will increase the flow of people by default. The weighted average is used to calculate the interaction data of the positive group and the image data to obtain the positive index, and the interaction data of the negative group is calculated to obtain the negative index. Optionally, the positive index and the negative index are both weighted averages of the number of comments, the number of reposts, the number of likes and the number of readings. The number of comments has the greatest influence and the largest weight. The number of reposts and the number of likes have similar influences and the weights are slightly lower than the number of comments. The number of readings has the smallest weight. For example, the weights of the number of comments, the number of reposts, the number of likes and the number of readings are 0.5, 0.2, 0.2 and 0.1 respectively. Preferably, the weights can be determined in combination with the judgment matrix in actual applications. Optionally, a neural network prediction model is established with the positive index and the negative index as input and the flow of people as output.
[0038] In another embodiment, the first data collection module is further used to obtain the publishing location information and publisher account information corresponding to the text data, classify the text data, and assign different coefficients to the interaction data of text data of different levels; specifically, while obtaining the text data, the location information and publisher account information corresponding to the text data are also obtained, and the text data is classified using the location information and publisher account information. The publisher account level can be replaced by the Weibo account level or the number of fans, and the publishing location information can be obtained from the location information of the social information; optionally, five distance ranges are divided from near to far according to the distance from the center of the selected area or the target area, and multiple five distance ranges are divided according to the publisher account level or the number of fans. Account level interval; according to the combination of distance range and account level interval, multiple levels are divided and coefficients are given. For example, if it is in the nearest first distance range and in the first interval with the highest account level interval, a coefficient of 1 is given; for example, if it is in the first distance range and the account level is in the second interval or in the second distance range and the account level is in the first interval, a coefficient of 0.9 is given, and so on. Each text data is assigned a coefficient; preferably, the specific coefficient can be determined using a judgment matrix; this embodiment sorts the text data by publishing location information and publisher account information, highlights high-quality and high-relevance social information, reduces the interference of low-quality and low-relevance social information on the pedestrian flow prediction results, and improves the accuracy of the pedestrian flow prediction results.
[0039] The number of devices and processing scales described here are used to simplify the description of the present invention. Applications, modifications and variations of the people flow prediction system and method of the present invention are obvious to those skilled in the art.
[0040] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.
Claims
1. A method for predicting human traffic, characterized in that: include: S1: obtaining text data, image data and interaction data related to the selected area in the first historical period from social media; S2: Obtaining the flow of people in the selected area in a second historical time period, where the second historical time period is later than the first historical time period; S3: Based on the interaction data obtained in S1 and the traffic volume obtained in S2, a neural network prediction model is established; S4: obtaining text data, image data and interaction data related to the target area in the current time period from social media, inputting the data into the neural network prediction model, and predicting the flow of people in the target area in the future time period; The interaction data includes the number of comments, forwardings, favorites, likes and readings; Determine keywords and key objects matching the selected area and the target area, extract text data related to the selected area and the target area according to the keywords, and extract image data related to the selected area and the target area according to the key objects; The method also includes identifying positive keywords that are positively correlated with increasing foot traffic and reverse keywords that are negatively correlated with increasing foot traffic in the text data, dividing the text data into a forward group and a reverse group according to the forward keywords and the reverse keywords, calculating a forward index using the interaction data of the text data and the image data in the forward group, calculating a reverse index using the interaction data of the text data in the reverse group, taking the forward index and the reverse index as weighted averages of the interaction data, and constructing a neural network prediction model using the forward index, the reverse index, and the foot traffic.
2. The method for predicting the flow of people according to claim 1, characterized in that: The publishing location information and publisher account information corresponding to the text data are obtained, the text data is graded, and in the process of calculating the positive index and the negative index, different coefficients are assigned to the interaction data of text data of different levels.
3. The method for predicting the flow of people according to claim 1, characterized in that: The second historical time period is 3 to 5 days later than the first historical time period.
4. A people flow prediction system, characterized in that: include: A first data collection module, used to obtain text data, image data and interaction data thereof related to the selected area in a first historical time period from social media, and used to obtain text data, image data and interaction data thereof related to the target area in a current time period from social media; A second data collection module, used to obtain the flow of people in the selected area in a second historical time period, the second historical time period being later than the first historical time period; A model building module, for building a neural network prediction model based on the interaction data obtained in the first historical time period and the flow of people obtained in the second historical time period; A prediction module, used to predict the flow of people in a target area in a future time period using the neural network prediction model; The first data collection module is further used to determine keywords and key objects matching the selected area and the target area, extract text data related to the target area based on the keywords, and extract image data related to the target area based on the key objects; the interaction data includes the number of comments, forwardings, favorites, likes, and readings; The keywords include positive keywords that are positively correlated with increasing the flow of people and reverse keywords that are negatively correlated with increasing the flow of people. The model building module is also used to divide the text data into a forward group and a reverse group according to the forward keywords and the reverse keywords, calculate the forward index using the interaction data of the text data and the image data in the forward group, calculate the reverse index using the interaction data of the text data in the reverse group, take the forward index and the reverse index as the weighted average of the interaction data, and build a neural network prediction model using the forward index, the reverse index and the flow of people.
5. The crowd flow prediction system according to claim 4, characterized in that: The first data collection module is also used to obtain the publishing location information and publisher account information corresponding to the text data, classify the text data, and assign different coefficients to the interaction data of text data of different levels.