Method, system, device, medium and program product for predicting passenger-carrying hot spot

Abstract

The invention provides a method, system and device for predicting a passenger-carrying hotspot, a medium and a program product. The method comprises the following steps: acquiring historical data of a taxi passenger-carrying point; performing data clustering on historical data based on a density clustering algorithm of a greedy strategy to obtain a passenger-carrying hotspot clustering result; the passenger-carrying hotspot clustering result comprises a plurality of passenger-carrying hotspots for taxi drivers to select. According to the method, data mining is performed on the sampled historical data through the density clustering algorithm of the greedy strategy, so that the problems that a traditional clustering algorithm cannot well process large-scale data and the clustering effect is poor due to insufficient priori knowledge can be avoided; meanwhile, the adaptability of a search radius and a density threshold value during clustering of a density data set with uneven density is improved, the clustering efficiency of large-scale data is improved, a passenger-carrying hotspot clustering result serves as a decision-making basis for reasonable distribution and scheduling of taxies, the scheduling efficiency and the passenger-carrying success rate are improved, and the service life of the taxies is prolonged. And the operation efficiency of the taxi can be improved.

Description

Technical Field

[0001] This disclosure relates to the field of data mining technology, and in particular to a method, system, device, medium, and program product for predicting passenger hotspots. Background Technology

[0002] With the rapid pace of urbanization, my country's cities are constantly expanding. Taxis provide convenience for people's travel, but they also exhibit inefficient operating methods with high randomness, such as driving empty or cruising for passengers, which is a waste of resources for taxis. As the scale of data continues to grow, how to extract useful information from this massive amount of data has become a research hotspot in the era of big data. Traditional clustering algorithms are commonly used to mine taxi data.

[0003] However, taxis are equipped with GPS devices that transmit information to the service desk every few seconds, resulting in a very large amount of data. In addition, the distribution of taxi passenger hotspots is not uniform due to the influence of urban natural environmental factors. Therefore, traditional clustering algorithms cannot process such large-scale data well, and the clusters formed by taxi passenger hotspot areas are irregular, all of which bring certain difficulties to clustering. Summary of the Invention

[0004] The technical problem to be solved by this disclosure is to overcome the above-mentioned defects in the prior art and provide a method, system, device, medium and program product for predicting passenger hotspots.

[0005] This disclosure solves the above-mentioned technical problems through the following technical solution:

[0006] This disclosure provides a method for predicting passenger hotspots, the method comprising:

[0007] Obtain historical data on taxi pick-up points;

[0008] The historical data is clustered using a density clustering algorithm with a greedy strategy to obtain passenger hotspot clustering results; the passenger hotspot clustering results include several passenger hotspots for taxi drivers to choose from.

[0009] Optionally, the method further includes:

[0010] The target passenger hotspot area is quickly located based on the passenger hotspot clustering results, according to the determined object parameters and the passenger hotspot clustering results.

[0011] Optionally, obtaining historical data on taxi pick-up points includes:

[0012] Obtain raw data of actual taxi pick-up points within a preset historical time period;

[0013] The original data is statistically analyzed to obtain statistical results, and the sample size is determined based on the statistical results.

[0014] Based on the stated sample size, the original data is sampled using a reservoir sampling method to obtain the sampled data;

[0015] The sample data is used as historical data for the taxi pick-up points.

[0016] Optionally, the historical data includes GPS longitude and GPS latitude;

[0017] The process of using a density-based greedy clustering algorithm to cluster historical data to obtain passenger hotspot clustering results includes:

[0018] Based on the GPS longitude and GPS latitude of each taxi pick-up point, we gradually search for areas with high density to form clusters, and select the optimal local solution for the current state at each step, starting from the initial solution.

[0019] Optionally, the step of progressively searching for high-density areas to form clusters based on the GPS longitude and GPS latitude of each taxi pick-up point, and selecting the optimal local solution for the current state at each step starting from the initial solution, includes:

[0020] For any selected taxi pick-up point, a greedy strategy is used to determine the search radius, and a circular cluster is obtained based on the search radius; wherein, the search radius satisfies the condition that there are more than a certain number of taxi pick-up points within the range of the search radius centered on the selected taxi pick-up point;

[0021] A neighborhood query is performed around the newly obtained circular cluster. For any selected unclustered taxi pick-up point, a new search radius is determined using a greedy strategy, and a new circular cluster is obtained based on the new search radius.

[0022] Repeat the previous step until no new circular clusters can be found;

[0023] The circular clusters are merged to obtain the clustered clusters;

[0024] Noise points are detected in the clusters to remove them.

[0025] Optionally, for any selected taxi pick-up point, determining the search radius using a greedy strategy and obtaining a circular cluster based on the search radius includes:

[0026] Taking any chosen taxi pick-up point as the core object (P), find the k closest to P (k = P). min-1) Point, P is the core object of the current search, P min is the preset threshold for the number of points within a cluster, and k is the threshold for the number of non-center points within a cluster;

[0027] For the set of all taxi pick-up points D = {d(P, P1), d(P, P2), d(P, P3), ..., d(P, P...} k Sort the elements in the array d(P, P) in descending order. k The distance (R) is the Euclidean or Manhattan distance between each point in the set and the core object (P), and the search radius (R) is determined as R = d(P, P) k );

[0028] A circular cluster with center P and radius R is obtained.

[0029] This disclosure also provides a system for predicting passenger hotspots, the system comprising: an acquisition module and a greedy clustering module;

[0030] The acquisition module is used to acquire historical data on taxi pick-up points;

[0031] The greedy clustering module is used to perform data clustering on the historical data based on a density clustering algorithm with a greedy strategy to obtain passenger hotspot clustering results; the passenger hotspot clustering results include several passenger hotspots for taxi drivers to choose from.

[0032] Optionally, the system further includes: a positioning module;

[0033] The positioning module is used to determine object parameters based on the current travel time, and to quickly locate the target passenger hotspot area based on the determined object parameters and the passenger hotspot clustering results.

[0034] Optionally, the acquisition module is further configured to acquire raw data of actual taxi pick-up points within a preset historical time period;

[0035] The acquisition module is also used to perform statistics on the raw data to obtain statistical results, and determine the number of samples to be sampled based on the statistical results;

[0036] The acquisition module is also used to perform water tank sampling on the original data according to the number of samples to obtain sampled data.

[0037] The acquisition module is also used to use the sample data as historical data of the taxi pick-up points.

[0038] Optionally, the historical data includes GPS longitude and GPS latitude;

[0039] The greedy clustering module is also used to progressively find areas with high density to form clusters based on the GPS longitude and GPS latitude of each taxi pick-up point, and select the optimal local solution in the current state at each step, starting from the initial solution.

[0040] Optionally, the greedy clustering module is further configured to determine the search radius for any selected taxi pick-up point using a greedy strategy, and obtain a circular cluster based on the search radius; wherein the search radius satisfies the condition that there are taxi pick-up points with a number threshold or higher within the range of the search radius centered on the selected taxi pick-up point.

[0041] The greedy clustering module is also used to perform neighborhood queries around the newly obtained circular cluster. For any selected unclustered taxi pick-up point, a new search radius is determined by a greedy strategy, and a new circular cluster is obtained based on the new search radius.

[0042] The greedy clustering module is also used to repeat the previous step until no new circular clusters can be found.

[0043] The greedy clustering module is also used to merge the circular clusters to obtain clusters;

[0044] The greedy clustering module is also used to detect noise points in the clusters in order to remove noise points.

[0045] Optionally, the greedy clustering module is also used to find the k closest (k=P) taxi pick-up points to P, using an arbitrarily selected taxi pick-up point as the core object (P). min -1) Point, P is the core object of the current search, P min is the preset threshold for the number of points within a cluster, and k is the threshold for the number of non-center points within a cluster;

[0046] The greedy clustering module is also used for the set of all taxi pick-up points D = {d(P, P1), d(P, P2), d(P, P3), ..., d(P, P...} k Sort the elements in the array} from largest to smallest, d(P, P...). k The distance (R) is the Euclidean or Manhattan distance between each point in the set and the core object (P), and the search radius (R) is determined as R = d(P, P) k );

[0047] The greedy clustering module is also used to obtain a circular cluster with P as the center and R as the radius.

[0048] This disclosure also provides an electronic device including a memory, a processor, and a computer program stored in the memory and for running on the processor, wherein the processor executes the computer program to implement the aforementioned method for predicting passenger hotspots.

[0049] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method for predicting passenger hotspots.

[0050] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned method for predicting passenger hotspots.

[0051] Based on common knowledge in the field, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred embodiments of this disclosure.

[0052] The positive and progressive effects of this disclosure are as follows: By sampling a large amount of historical data on taxi pick-up points using a reservoir sampling technique, effective hotspot data is obtained. A greedy density clustering algorithm is then used to mine the sampled historical data, avoiding the problems of traditional clustering algorithms' inability to handle large-scale data effectively and poor clustering results due to insufficient user prior knowledge. Furthermore, it improves the adaptability of search radius and density threshold when clustering density datasets with uneven density, increasing the clustering efficiency for large-scale data. Using the passenger pick-up hotspot clustering results as the basis for rational taxi distribution and dispatching decisions improves dispatching efficiency and passenger success rate, thereby increasing taxi operating efficiency and the economic income of taxi drivers. Attached Figure Description

[0053] Figure 1 A flowchart of a method for predicting passenger hotspots provided in Embodiment 1 of this disclosure;

[0054] Figure 2 A schematic diagram of a density-uniform dataset for a method of predicting passenger hotspots provided in Embodiment 1 of this disclosure;

[0055] Figure 3 A flowchart illustrating a specific implementation of a method for predicting passenger hotspots provided in Embodiment 1 of this disclosure;

[0056] Figure 4 A flowchart illustrating a specific implementation of step S11 of a method for predicting passenger hotspots provided in Embodiment 1 of this disclosure;

[0057] Figure 5 A flowchart illustrating a specific implementation of step S12 of a method for predicting passenger hotspots provided in Embodiment 1 of this disclosure;

[0058] Figure 6 A flowchart illustrating a specific implementation of step S121 of a method for predicting passenger hotspots provided in Embodiment 1 of this disclosure;

[0059] Figure 7 A schematic diagram illustrating the cluster discovery and merging of a method for predicting passenger hotspots provided in Embodiment 1 of this disclosure;

[0060] Figure 8 This is a schematic diagram of a method for predicting passenger hotspots provided in Embodiment 1 of this disclosure, using noise points as random initial points;

[0061] Figure 9 A flowchart illustrating a specific implementation of step S1211 of a method for predicting passenger hotspots provided in Embodiment 1 of this disclosure;

[0062] Figure 10 A schematic diagram of the clustering results of taxi passenger hotspot data, which is an example of a method for predicting passenger hotspots provided in Embodiment 1 of this disclosure;

[0063] Figure 11 A schematic diagram of the results of clustering historical passenger hotspot data of urban main roads, which is an example of a method for predicting passenger hotspots provided in Embodiment 1 of this disclosure;

[0064] Figure 12 A schematic diagram of the results of clustering passenger hotspot prediction data on urban main roads, as an example of a method for predicting passenger hotspots provided in Embodiment 1 of this disclosure;

[0065] Figure 13 This is a schematic diagram of a system for predicting passenger hotspots provided in Embodiment 2 of this disclosure;

[0066] Figure 14 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of this disclosure. Detailed Implementation

[0067] The present disclosure is further illustrated below by way of embodiments, but the present disclosure is not limited to the scope of the embodiments described herein.

[0068] The prefixes such as "first" and "second" used in this disclosure are merely for distinguishing different descriptive objects and do not limit the position, order, priority, quantity, or content of the described objects. The use of ordinal numbers and other prefixes used to distinguish descriptive objects in this disclosure does not constitute a limitation on the described objects. The description of the described objects is given in the claims or the context of the embodiments, and should not be construed as an unnecessary limitation. Furthermore, in the description of this embodiment, unless otherwise stated, "multiple" means two or more.

[0069] Example 1

[0070] Figure 1 A flowchart of a method for predicting passenger hotspots provided as an exemplary embodiment of this disclosure, the method comprising:

[0071] S11. Obtain historical data on taxi pick-up points.

[0072] S12. For historical data, a density-based clustering algorithm using a greedy strategy is used to cluster the data to obtain passenger hotspot clustering results. These results include several passenger hotspots for taxi drivers to choose from.

[0073] The historical data of taxi pick-up points can be raw data or sampled raw data to reduce the amount of data processing. The historical data is obtained by merging data from several taxis, unifying the format and measurement, and unifying the coordinate system of GPS driving records, aligning the time, and cleaning up abnormal data (such as multiple points reported by the same device within the same second) to ensure the spatiotemporal continuity and measurement consistency of the data.

[0074] Historical data within appropriate timeframes can be used as needed, such as data from several consecutive working days, several weekends, several consecutive holidays, a consecutive week, or peak hours on weekdays. Considerations include the significant differences in taxi demand between weekdays and weekends, and the obvious differences between peak and off-peak tourist seasons.

[0075] Traditional density clustering algorithms can find shaped clusters even with noisy data, but they have the following drawbacks:

[0076] (1) The radius and density threshold involved in the algorithm are generally obtained through experience. If the user's prior knowledge is insufficient, the clustering effect of the obtained clustering dataset will be relatively poor.

[0077] (2) When clustering density datasets with uneven density, the radius and density threshold in the algorithm are not very adaptable.

[0078] Figure 3 The dataset is shown to have uneven density, with two clusters C1 and C2, and two noisy points o1 and o2. According to the density clustering algorithm, if the distance between objects in C1 is used as the radius for clustering, the clustering result is only cluster C1, and the migrated objects are regarded as noise data. If the selected radius is too large, both clusters can be obtained, but o2, which is far away from cluster C2, will be classified into C1. This is because the density clustering algorithm uses a set of global variables to cluster clusters with different densities.

[0079] To solve the above problems, we can use a greedy algorithm to improve the algorithm. The idea of ​​density clustering algorithm based on greedy strategy is to obtain the global optimal solution through the local optimal solution. That is, the core of the greedy algorithm is to select the optimal solution in the current state at each step in order to achieve the global optimal solution.

[0080] This clustering method, incorporating a greedy approach, divides data points into clusters based on density. It doesn't require pre-specifying the number of clusters, automatically discovers cluster structures within the data, and can handle irregularly shaped clusters. It primarily forms clusters by progressively searching for high-density regions. In density clustering, a greedy strategy can be used to select high-density regions as the starting point for clustering and gradually expand these regions. This ensures that each step from the initial solution yields an optimal local solution. Therefore, the greedy strategy adaptively finds an appropriate radius for each directly density-reachable circular cluster, and then merges circular clusters with common points to obtain the final cluster.

[0081] In this embodiment, a reservoir sampling technique is used to sample a large amount of historical data on taxi pick-up points to obtain effective hotspot data. A greedy density clustering algorithm is then used to mine the sampled historical data. This avoids the problems of traditional clustering algorithms not being able to handle large-scale data well and the poor clustering effect caused by insufficient prior knowledge of users. At the same time, it improves the adaptability of search radius and density threshold when clustering density datasets with uneven density, and improves the clustering efficiency of large-scale data. The results of passenger pick-up hotspot clustering are used as the basis for the decision-making of rational distribution and scheduling of taxis, which improves scheduling efficiency and passenger pick-up success rate, thereby improving the operating efficiency of taxis and increasing the economic income of taxi drivers.

[0082] In one embodiment, refer to Figure 3 The methods also include:

[0083] S13. Determine the object parameters based on the current travel time, and quickly locate the target passenger hotspot area based on the passenger hotspot clustering results according to the determined object parameters.

[0084] The object parameter represents the radius of the target area centered on the current taxi location. A smaller object parameter results in a smaller target area, making it easier to quickly locate passenger hotspots within the target area. However, this also means fewer passenger hotspots will be detected, allowing taxi drivers to quickly decide on one or more target hotspots and plan routes to them. Conversely, a larger object parameter results in a larger target area, increasing the time required to locate passenger hotspots within the target area. However, this also means more passenger hotspots will be detected, providing taxi drivers with more potential target hotspots and more route options to them.

[0085] During peak travel periods, when identifying and guiding taxi drivers to pick up passengers, the smallest object parameter can be used to quickly locate the demand hotspots for taxis, thus providing drivers with more accurate information on potential passenger pick-up areas. During off-peak travel periods, a larger object parameter can be used to expand the demand hotspots for taxis, providing drivers with more precise location-based passenger pick-up services.

[0086] In one embodiment, refer to Figure 4 Step S11 includes:

[0087] S111. Obtain the raw data of actual taxi pick-up points within a preset historical time period.

[0088] S112. Perform statistical analysis on the original data to obtain statistical results, and determine the sample size based on the statistical results.

[0089] S113. Based on the sample size, the original data is sampled in a reservoir to obtain the sampled data.

[0090] S114. Use sample data as historical data for taxi pick-up points.

[0091] The storage format of the raw data is shown in Table 1.

[0092] Table 1

[0093]

[0094] The principle of water tank sampling is as follows:

[0095] Let the total amount of data be n. We draw k unique samples from this dataset. The probability of each sample being drawn is... The steps for sampling from a water storage tank are as follows:

[0096] (1) Construct a water storage tank with a capacity of k. Then extract the first i data from the total data and put them into the water storage tank.

[0097] (2) Starting from the (i+1)th data point in the total number of data points, proceed according to probability. Determine whether this data should replace the data in the reservoir:

[0098]

[0099] Where: P final_selected P represents the probability that the object will ultimately be selected and placed in the reservoir. now_selected P represents the probability that the object will be selected when it appears. not_replacedLet represent the probability that an object is not replaced in each replacement. Let _i_ be the i-th (i≤k) data point in the original dataset, with a probability of 1 for being placed in the reservoir. Starting from the (i+1)-th data point, the probability that data _i_ might be replaced is... The probability of being selected is calculated using the above formula. When i = n, the probability of being selected is...

[0100] (3) Iterate through all the data in the original dataset and randomly select k data points to use as sample data.

[0101] In one embodiment, historical data includes GPS longitude and GPS latitude.

[0102] Reference Figure 5 Step S12 includes:

[0103] S121. Based on the GPS longitude and GPS latitude of each taxi pick-up point, gradually find areas with high density to form clusters, and select the optimal local solution in the current state at each step, starting from the initial solution.

[0104] With the GPS longitude and GPS latitude of each taxi pick-up point, the coordinates of each taxi pick-up point in the same coordinate system can be determined. Based on the determined coordinates, the distance between any two taxi pick-up points can be calculated. This distance can be Euclidean distance, Manhattan distance, or other defined distances.

[0105] In one embodiment, refer to Figure 6 Step S121 includes:

[0106] S1211. For any selected taxi pick-up point, a greedy strategy is used to determine the search radius, and a circular cluster is obtained based on the search radius. The search radius satisfies the condition that there are at least a certain number of taxi pick-up points within the range of the selected taxi pick-up point as the center.

[0107] S1212. Perform a neighborhood query around the newly obtained circular cluster. For any selected unclustered taxi pick-up point, determine a new search radius using a greedy strategy, and obtain a new circular cluster based on the new search radius.

[0108] S1213. Repeat the previous step until no new circular clusters can be found.

[0109] S1214. Merge circular clusters to obtain clustered clusters.

[0110] S1215. Perform noise point detection on the clusters to remove noise points.

[0111] First, a greedy strategy is used to determine the search radius, and a circular cluster is obtained based on the search radius. Specifically, if each point in the dataset can find Pmin points near its radius R, then cluster discovery and merging are performed. Figure 7 As shown.

[0112] In density-based clustering algorithms using a greedy strategy, clusters can be obtained regardless of whether the initial point is a noise point or a kernel object, thus ensuring the algorithm's robustness. For any chosen point P, the greedy strategy selects the k closest points (k = Pmin-1) to P, where P is the kernel object of the search. For the set D = {d(P, P1), d(P, P2), d(P, P3), ..., d(P, P...}, ... k Sort the elements in the array} from largest to smallest, d(P, P...). k ) is the Euclidean distance (or other defined distance, such as Manhattan distance or Minkowski distance) between each point in the set and the core object, and the search radius is R = d(P, P k ).

[0113] After obtaining the search radius R, a circular cluster with center P and radius R can be obtained. Different clusters may contain the same points, so they need to be merged to adjust the clustering results.

[0114] for Figure 7 The system contains four circular clusters, each with a different radius. As shown in the diagram above, clusters C3 and C4 share a common point, so they need to be merged.

[0115] Then, a neighborhood query is performed around the cluster.

[0116] Density clustering algorithms traverse the neighborhood of a randomly selected core object P without an index. In this case, the algorithm's execution efficiency is relatively low. If the dataset contains n objects, the time complexity of density clustering is O(n^2). 2 ).

[0117] After merging clusters, all objects within each cluster need to be traversed and marked as "clustered." In density-based clustering algorithms using a greedy strategy, neighborhood queries are used to improve execution efficiency. For Figure 7 The algorithm expands the kernel by searching for objects not marked as "already clustered" in the neighborhood of cluster C3, finding cluster C4, and so on, until no cluster with a reasonable and effective direct density can be found within the neighborhood. Neighborhood lookup effectively avoids the time-consuming and useless operations caused by blindly traversing the dataset, resulting in a time complexity of O(nlogn), significantly improving the algorithm's execution efficiency.

[0118] Next, noise points are detected, which can identify unreasonable clusters.

[0119] Although a randomly selected initial point is an outlier or noisy point, it is possible to find the k nearest points to form a cluster. However, the logical relationship of this cluster is incorrect, resulting in an unsatisfactory final clustering result. This paper detects noisy points through relative density, using the average distance d between different points as the standard for measuring relative density. Therefore:

[0120]

[0121] In the above formula: P i P is any data point in cluster C. k_nestest(i) Is with P i For the k nearest points, d(·) is the Euclidean distance (or other defined distance). The smaller the value of d, the greater the density around the point.

[0122] If an outlier point P becomes a kernel object under certain conditions, then the average distance dis between point P and the midpoint of its neighborhood can be calculated. avg Since noise points reach a cluster density directly during the search, they will assign edge points from the nearest high-density cluster to their own cluster. Since the density of noise point clusters is relatively low, the radius R will be much larger than the average distance between other points within the cluster, i.e., to satisfy dis... avg <<R indicates that point P is an outlier noise point. For example... Figure 8 As shown, the radius R of point P is greater than the average distance dis between the remaining points in the cluster. avg The difference is much larger, indicating that point P is an outlier noise point.

[0123] In one embodiment, refer to Figure 9 Step S1211 includes:

[0124] S12111. Taking any randomly selected taxi pick-up point as the core object (P), find the k closest to P (k = P). min -1) Point, P is the core object of the current search, P min is the preset threshold for the number of points within a cluster, and k is the threshold for the number of non-center points within a cluster.

[0125] S12112. For the set of all taxi pick-up points D = {d(P,P1),d(P,P2),d(P,P3),...,d(P,P...}} k Sort the elements in ),...} in descending order, d(P,P) kThe distance () is the Euclidean or Manhattan distance between each point in the set and the core object (P). The search radius (R) is determined as R = d(P, P) k ).

[0126] S12113. A circular cluster with center P and radius R is obtained.

[0127] If Euclidean distance is used as the standard for distance measurement, then all distance data will use Euclidean distance; if Manhattan distance is used as the standard for distance measurement, then all distance data will use Manhattan distance.

[0128] Preferably, Manhattan distance can be used as the standard for distance measurement, which is expressed as follows:

[0129] D(i,j)=|x i -x j |+|y i -y j |

[0130] Where i(x) i ,y i ) and j(x j y j Let be the coordinates of two points in space, and D(i,j) represent the sum of the linear distances between these two points in terms of orientation. As can be seen from the formula, the Manhattan distance is easier to calculate than the Euclidean distance, which can improve the computational efficiency.

[0131] Step S1212, "For any selected unclustered taxi pick-up point, determine a new search radius using a greedy strategy, and obtain a new circular cluster based on the new search radius," is similar to step S1211. Step S1211 involves arbitrarily selecting one taxi pick-up point from all historical taxi pick-up points as the core object (P), at which point all taxi pick-up points are unclustered. Step S1212 involves arbitrarily selecting one taxi pick-up point from all unclustered taxi pick-up points in historical data as the core object (P), removing clustered taxi pick-up points from the query traversal range, and ensuring that clustered taxi pick-up points cannot be used as the center to form a new circular cluster. The specific steps are not elaborated here.

[0132] The following is an example of a method for implementing predictive passenger hotspots.

[0133] The test data used in this paper comes from the GPS driving records of 8,000 taxis in a certain city from December 4th to 8th, 2023. Historical data was used because real-time GPS data is unique and not publicly available. Each taxi transmits its data to the service desk every 30-70 seconds, and the GPS data is stored in text format. The five days selected are all weekdays; the first four days are used for data mining, and the fifth day is used for prediction data to verify the effectiveness of the algorithm. Weekends were not used because the surge in travel on weekends causes significant fluctuations in taxi GPS data, which would compromise the general applicability of the mining results. The format of the raw GPS data storage is shown in Table 1.

[0134] Uploaded GPS data includes: 184051, 1, 1, 20231205080137, 116.0955505, 39.9377313, 80, 304, 1. Based on the characteristics of urban roads, this paper uses Manhattan distance as the standard for distance measurement, expressed as follows:

[0135] D(i,j)=|x i -x j |+|y i -y j |

[0136] Where i(x) i y i ) and j(x j y j Let be the coordinates of two points in space, and D(i,j) represent the sum of the linear distances between these two points in terms of orientation. As can be seen from the formula, the Manhattan distance is easier to calculate than the Euclidean distance, which can improve the computational efficiency.

[0137] After merging GPS driving record data from December 4th to 7th, 2023, the sample size was determined through statistical analysis. Then, a reservoir technique was used for sampling. Finally, a density-based clustering algorithm based on a greedy strategy was applied to the sampled data for clustering. The clustering results are shown below. Figure 10 As shown in the figure, the clustering data is the latitude and longitude of the selected taxis (samples) at a certain GPS time.

[0138] pass Figure 10It can be seen that taxi passenger hotspot data in urban areas is very dense. Due to urban planning, road design, and other factors, the clustered taxi passenger hotspot data exhibits linear, zigzag, and cross-shaped distributions. During clustering, relatively scattered passenger points are deleted, while those that are retained indicate a higher probability of user demand for taxis. These points can then be used to guide taxi drivers to these locations via a control center. During peak travel periods, the minimum object parameter is used to quickly locate taxi demand hotspot areas, providing drivers with more accurate information about potential passenger hotspots. During off-peak travel periods, a larger object parameter can be used to expand the taxi demand hotspot areas, providing drivers with more precise passenger location services.

[0139] After mining and clustering historical data, the morning rush hour passenger traffic hotspot data for December 8, 2023, is predicted to illustrate the effectiveness of this example in prediction. This paper uses the main roads of the city as an example for prediction, and the clustering results of their historical passenger traffic hotspot data are as follows: Figure 10 As shown.

[0140] Figure 10 The data on taxi passenger hotspots during the morning rush hour were clustered. Therefore, the parameter used was the smallest object. After removing the relatively scattered noise points, 30 clusters were obtained, which are the passenger hotspot areas. In the process of analyzing the hit rate of passenger hotspot areas, the hotspot areas formed by historical data were represented by gray shading. The results of clustering the morning rush hour on December 8, 2023 were obtained.

[0141] like Figure 11 and Figure 12 As shown, the predicted clustered passenger hotspot areas basically match the actual areas, with only a small amount of data outside the clusters, which demonstrates the effectiveness of the algorithm in this paper.

[0142] Clustering and statistically predicting all taxi hotspot data in the urban area yielded a coverage hit rate of 83.51%, while the coverage hit rate of the taxi hotspot area discovery method based on functional zone division was 73.98%. This demonstrates that, based on reservoir sampling, the density clustering algorithm using a greedy strategy can effectively identify taxi hotspot areas and maintain a high hit rate in discovering and predicting passenger demand. This provides a good basis for the rational distribution and real-time dispatch of taxis, helping users solve the problem of difficulty in hailing a taxi.

[0143] Example 2

[0144] Corresponding to the aforementioned method embodiments for predicting passenger hotspots, this disclosure also provides embodiments of systems for predicting passenger hotspots.

[0145] Figure 13 This is a schematic diagram of a system for predicting passenger hotspots, provided as an exemplary embodiment of the present disclosure. The system includes: an acquisition module 1 and a greedy clustering module 2.

[0146] Module 1 is used to obtain historical data on taxi pick-up points.

[0147] Greedy Clustering Module 2 is used to cluster historical data using a density-based greedy clustering algorithm to obtain passenger hotspot clustering results. The passenger hotspot clustering results include several passenger hotspots for taxi drivers to choose from.

[0148] In this embodiment, a reservoir sampling technique is used to sample a large amount of historical data on taxi pick-up points to obtain effective hotspot data. A greedy density clustering algorithm is then used to mine the sampled historical data. This avoids the problems of traditional clustering algorithms not being able to handle large-scale data well and the poor clustering effect caused by insufficient prior knowledge of users. At the same time, it improves the adaptability of search radius and density threshold when clustering density datasets with uneven density, and improves the clustering efficiency of large-scale data. The results of passenger pick-up hotspot clustering are used as the basis for the decision-making of rational distribution and scheduling of taxis, which improves scheduling efficiency and passenger pick-up success rate, thereby improving the operating efficiency of taxis and increasing the economic income of taxi drivers.

[0149] In one embodiment, the system further includes a positioning module 3.

[0150] The positioning module 3 is used to determine the object parameters based on the current travel time, and to quickly locate the target passenger hotspot area based on the determined object parameters and the passenger hotspot clustering results.

[0151] In one embodiment, the acquisition module 1 is further configured to acquire raw data of actual taxi pick-up points within a preset historical time period.

[0152] The acquisition module 1 is also used to perform statistics on the raw data to obtain statistical results, and to determine the number of samples to be sampled based on the statistical results.

[0153] Module 1 is also used to perform reservoir sampling on the original data according to the sample quantity to obtain the sampled data.

[0154] The acquisition module 1 is also used to use sample data as historical data of taxi pick-up points.

[0155] In one embodiment, historical data includes GPS longitude and GPS latitude.

[0156] Greedy clustering module 2 is also used to find high-density areas to form clusters based on the GPS longitude and GPS latitude of each taxi pick-up point, and select the optimal local solution in the current state at each step starting from the initial solution.

[0157] In one embodiment, the greedy clustering module 2 is further configured to determine the search radius for any selected taxi pick-up point using a greedy strategy, and obtain a circular cluster based on the search radius. The search radius satisfies the condition that there are at least a threshold number of taxi pick-up points within the search radius centered on the selected taxi pick-up point.

[0158] Greedy clustering module 2 is also used to perform neighborhood queries around the newly obtained circular cluster. For any selected unclustered taxi pick-up point, a new search radius is determined through a greedy strategy, and a new circular cluster is obtained based on the new search radius.

[0159] Greedy clustering module 2 is also used to repeat the previous step until no new circular clusters can be found.

[0160] Greedy clustering module 2 is also used to merge circular clusters to obtain clustered clusters.

[0161] Greedy clustering module 2 is also used to detect noise points in clusters in order to remove noise points.

[0162] In one embodiment, the greedy clustering module 2 is also used to find the k closest to P (k = P) using an arbitrarily selected taxi pick-up point as the core object (P). min -1) Point, P is the core object of the current search, P min is the preset threshold for the number of points within a cluster, and k is the threshold for the number of non-center points within a cluster.

[0163] Greedy clustering module 2 is also used for the set of all taxi pick-up points D = {d(P, P1), d(P, P2), d(P, P3), ..., d(P, P...} k Sort the elements in the array} from largest to smallest, d(P, P...). k The distance () is the Euclidean or Manhattan distance between each point in the set and the core object (P). The search radius (R) is determined as R = d(P, P) k ).

[0164] Greedy clustering module 2 is also used to obtain circular clusters with P as the center and R as the radius.

[0165] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs.

[0166] Example 3

[0167] Figure 14 This is a schematic diagram of the structure of an electronic device according to an example embodiment of the present disclosure. The electronic device includes a memory, a processor, and a computer program stored in the memory and used to run on the processor. When the processor executes the computer program, it implements the method for predicting passenger hotspots as described in any of the above embodiments. Figure 14 The electronic device 90 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments disclosed herein.

[0168] like Figure 14 As shown, the electronic device 90 can be manifested as a general-purpose computing device, such as a server device. The components of the electronic device 90 may include, but are not limited to: at least one processor 91, at least one memory 92, and a bus 93 connecting different system components (including memory 92 and processor 91).

[0169] Bus 93 includes a data bus, an address bus, and a control bus.

[0170] The memory 92 may include volatile memory, such as random access memory (RAM) 921 and / or cache memory 922, and may further include read-only memory (ROM) 923.

[0171] The memory 92 may also include a program tool 925 (or utility) having a set (at least one) program module 924, including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0172] The processor 91 executes various functional applications and data processing by running computer programs stored in the memory 92, such as the method for predicting passenger hotspots provided in any of the above embodiments.

[0173] Electronic device 90 can also communicate with one or more external devices 94 (e.g., keyboard, pointing device, etc.). This communication can be performed through input / output (I / O) interface 95. Furthermore, electronic device 90 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public network, such as the Internet) via network adapter 96. As shown, network adapter 96 communicates with other modules of electronic device 90 via bus 93. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with electronic device 90, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.

[0174] It should be noted that although several units / modules or sub-units / modules of the electronic device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0175] Example 4

[0176] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting passenger hotspots provided in any of the above embodiments.

[0177] The readable storage medium may be more specifically adopted, including but not limited to: portable disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0178] Example 5

[0179] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the method for predicting passenger hotspots as described above.

[0180] The program code for executing the computer program product of this disclosure can be written in any combination of one or more programming languages, and the program code can be executed entirely on a user device, partially on a user device, as a stand-alone software package, partially on a user device and partially on a remote device, or entirely on a remote device.

[0181] While specific embodiments of this disclosure have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of this disclosure is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of this disclosure, but all such changes and modifications fall within the scope of protection of this disclosure.

Claims

1. A method for predicting passenger hotspots, characterized in that, The method includes: Obtain historical data on taxi pick-up points; The historical data is clustered using a density clustering algorithm with a greedy strategy to obtain passenger hotspot clustering results; the passenger hotspot clustering results include several passenger hotspots for taxi drivers to choose from.

2. The method for predicting passenger hotspots as described in claim 1, characterized in that, The method further includes: The target passenger hotspot area is quickly located based on the passenger hotspot clustering results, according to the determined object parameters and the passenger hotspot clustering results.

3. The method for predicting passenger hotspots as described in claim 1, characterized in that, The acquisition of historical data on taxi pick-up points includes: Obtain raw data of actual taxi pick-up points within a preset historical time period; The original data is statistically analyzed to obtain statistical results, and the sample size is determined based on the statistical results. Based on the stated sample size, the original data is sampled using a reservoir sampling method to obtain the sampled data; The sample data is used as historical data for the taxi pick-up points.

4. The method for predicting passenger hotspots as described in claim 1, characterized in that, The historical data includes GPS longitude and GPS latitude; The process of using a density-based greedy clustering algorithm to cluster historical data to obtain passenger hotspot clustering results includes: Based on the GPS longitude and GPS latitude of each taxi pick-up point, we gradually search for areas with high density to form clusters, and select the optimal local solution for the current state at each step, starting from the initial solution.

5. The method for predicting passenger hotspots as described in claim 4, characterized in that, Based on the GPS longitude and GPS latitude of each taxi pick-up point, the system progressively searches for high-density areas to form clusters. Starting from the initial solution, it selects the optimal local solution for the current state at each step, including: For any selected taxi pick-up point, a greedy strategy is used to determine the search radius, and a circular cluster is obtained based on the search radius; wherein, the search radius satisfies the condition that there are more than a certain number of taxi pick-up points within the range of the search radius centered on the selected taxi pick-up point; A neighborhood query is performed around the newly obtained circular cluster. For any selected unclustered taxi pick-up point, a new search radius is determined using a greedy strategy, and a new circular cluster is obtained based on the new search radius. Repeat the previous step until no new circular clusters can be found; The circular clusters are merged to obtain the clustered clusters; Noise points are detected in the clusters to remove them.

6. The method for predicting passenger hotspots as described in claim 1, characterized in that, For any selected taxi pick-up point, a greedy strategy is used to determine the search radius, and a circular cluster is obtained based on the search radius, including: Taking any chosen taxi pick-up point as the core object (P), find the k closest to P (k = P). min -1) Point, P is the core object of the current search, P min is the preset threshold for the number of points within a cluster, and k is the threshold for the number of non-center points within a cluster; For the set of all taxi pick-up points D = {d(P,P1),d(P,P2),d(P,P3),...,d(P,P...} k Sort the elements in the array ),...} in descending order, d(P, P... k The distance (R) is the Euclidean or Manhattan distance between each point in the set and the core object (P), and the search radius (R) is determined as R = d(P, P) k ); A circular cluster with center P and radius R is obtained.

7. A system for predicting passenger hotspots, characterized in that, The system includes: an acquisition module and a greedy clustering module; The acquisition module is used to acquire historical data on taxi pick-up points; The greedy clustering module is used to perform data clustering on the historical data based on a density clustering algorithm with a greedy strategy to obtain passenger hotspot clustering results; the passenger hotspot clustering results include several passenger hotspots for taxi drivers to choose from.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and for running on the processor, characterized in that, When the processor executes the computer program, it implements the method for predicting passenger hotspots as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method for predicting passenger hotspots as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method for predicting passenger hotspots as described in any one of claims 1-6.

Images