A method and system for managing sports events of big data

Abstract

The present invention discloses a method and system for managing sports events in big data; by using management time points, management locations, management frequencies, and management totals, and adjusting the input manner of multiple time position change feature vectors corresponding to multiple time position management images through the use of multiple adjusted causal convolutional networks with random structures, to detect the features of the states of various resource categories of events corresponding to multiple event times, so as to find the average relationship features; inputting the average relationship features into the first shrinking neural network in the intermediate resource prediction network to detect the number of intermediate resource categories; and then adjusting the number of intermediate resource categories through the recorded spare resource vectors.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and more particularly, to a method and system for managing sports events with big data. Background Art

[0002] Currently, when organizing and preparing for an event, it is necessary to predict and ensure the maintenance of the venue order at the event site. However, in the actual event site, the event resources often cannot meet the requirements, so backup resources are needed. How to utilize historical event resources and backup resources to find the critical value of the resources that just meet the event requirements at this venue during event preparation is a problem. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for managing sports events with big data to solve the above problems existing in the prior art.

[0004] In a first aspect, an embodiment of the present invention provides a method for managing sports events with big data, including:

[0005] Obtaining event management information, an event resource vector, and a backup resource vector corresponding to multiple event times; the subscript of the event resource vector represents the resource category, and the value in the event resource vector represents the resource quantity; the event management information includes multiple resource categories and their corresponding management time points, management locations, management frequencies, and total management objects; the backup resource vector represents the resource quantities of multiple resource categories called outside the estimate.

[0006] Based on the event management information corresponding to multiple event times, determining the usage status of resource categories to obtain an average relationship feature; one resource category corresponds to one average relationship feature.

[0007] Obtaining an intermediate resource prediction network; the intermediate resource prediction network includes a first shrinking neural network and a first expanding neural network.

[0008] Inputting the average relationship feature into the first shrinking neural network to obtain the intermediate resource category quantity; the intermediate resource category quantity represents the critical value of the resources that can meet the event usage; multiple resource categories correspond to obtaining multiple intermediate resource category quantities.

[0009] Adjusting the intermediate resource category quantity based on the backup resource vector and the event resource vector.

[0010] Optionally, the determining the usage status of resource categories based on the event management information corresponding to multiple event times to obtain an average relationship feature includes:

[0011] Based on the event management information corresponding to multiple event times, multiple time-location management images are obtained; the length and width of the time-location management image represent geographical locations, and the height of the time-location management image represents the number of management time points; one event time corresponds to one time-location management image;

[0012] By adjusting the causal convolutional network, based on multiple time-location management images, the correlation relationship of the overall management location changes in adjacent management time points is detected to obtain the first time-location feature;

[0013] Based on the time-location management image and the first time-location feature, the feature of the usage status of the resource category is detected to obtain n relationship features; the relationship feature represents the feature of the relationship of the resources participating in the event; one resource category corresponds to one relationship feature;

[0014] Obtain m*n relationship features corresponding to m event times; average the m relationship features corresponding to one resource category to obtain the average relationship feature; the average relationship feature represents the feature of the detected resource category;

[0015] n resource categories correspond to obtaining n average relationship features.

[0016] Optionally, the step of "by adjusting the causal convolutional network, based on multiple time-location management images, the correlation relationship of the overall management location changes in adjacent management time points is detected to obtain the first time-location feature" includes:

[0017] The adjusted causal convolutional network includes a temporal convolutional network, a causal convolutional network, multiple random structures, and a fusion network;

[0018] Based on the time-location management image, multiple time-location change feature vectors and the values in the random structures are obtained; the time-location change feature vector represents the feature of the overall management location changes in adjacent management time points;

[0019] The number of input neurons of the causal convolutional network is equal to the number of random structures;

[0020] According to the values in the random structures, multiple time-location change features are input into the corresponding input neurons in the causal convolutional network; the same values in two random structures indicate that the corresponding two time-location change feature vectors are input into the same input neuron to detect their correlation relationship;

[0021] Multiple random vectors are obtained multiple times; multiple random causal features are obtained corresponding to the multiple random vectors;

[0022] The multiple random causal features are input into the fusion network to detect the relationship of the position changes at multiple times to obtain the first time-location feature.

[0023] Optionally, managing the image based on the time position to obtain a plurality of time position change feature vectors and values in the random structure includes:

[0024] Inputting the plurality of manager position images corresponding to the time position management image into a temporal convolutional network in sequence from the earliest time point to the latest time point to obtain time position change feature vectors;

[0025] A plurality of time position management images respectively obtain a plurality of time position change feature vectors; the number of elements in the plurality of time position change feature vectors is equal;

[0026] Randomly arranging natural numbers from 0 to the value obtained by dividing the number of elements in the time position change feature vector by 2 to obtain a first random vector;

[0027] Randomly arranging the values in the first random vector again to obtain a second random vector;

[0028] Adding the elements in the second random vector after the elements in the first random vector to obtain a random vector;

[0029] Wherein, there are two identical values in one random vector;

[0030] Inputting the two values in the random vector into the random structure in sequence; the values in the random structure represent the subscripts of the input neurons connected to the time position change feature vector in the causal convolutional network.

[0031] Optionally, the training method of the intermediate resource prediction network includes:

[0032] Obtaining training relationship features and labeled resource quantities; the training relationship features represent the features of the relationships of resources participating in events at historical time points; the labeled resource quantities represent the resource quantities of the corresponding resource categories that can exactly meet the event requirements;

[0033] Inputting the training relationship features into a first shrinking neural network to obtain training predicted resource quantities;

[0034] Inputting the predicted resource quantities into a first enlarging neural network to obtain training predicted intermediate resource features;

[0035] Training the first shrinking neural network by using the training predicted resource quantities and the labeled resource quantities through a loss function;

[0036] Training the first enlarging neural network by using the output of inputting the labeled resource quantities into the first enlarging neural network and the training relationship features through a loss function;

[0037] Through the loss function, train the first amplification neural network and the first reduction neural network with the intermediate resource features and training relationship features of the training prediction.

[0038] Optionally, based on the time position management image and the first time position feature, detecting the feature of the usage status of the resource category to obtain n relationship features, including:

[0039] According to the resource category, split the time position management image into multiple resource time position management images; one resource time position management image corresponds to one resource category;

[0040] Input the resource time position management image into the first convolutional network to detect the change of the management position corresponding to the resource category, and obtain the second time position feature;

[0041] Fuse the first time position feature and the second time position feature to obtain the fused time position feature;

[0042] Input the management frequency and the total amount of management objects corresponding to the resource category into the second neural network to detect features and obtain the second management feature;

[0043] Multiple resource categories correspond to obtaining multiple first time position features and multiple second management features;

[0044] Overlay the first time position feature and the corresponding second management feature to obtain the relationship feature;

[0045] n resource categories correspond to obtaining n relationship features.

[0046] Optionally, the event resource vector, the spare resource vector, and the number of multiple intermediate resource categories are stored in different data tables;

[0047] The event resource vector and the spare resource vector can be actively modified;

[0048] The number of intermediate resource categories cannot be actively modified.

[0049] Optionally, based on the spare resource vector and the event resource vector, adjusting the number of intermediate resource categories includes:

[0050] If the value in the spare resource vector is 0, set the resource category value to 0; the resource category value of 0 indicates that the estimated resource quantity can meet the event requirements;

[0051] If the value in the spare resource vector is greater than 0, set the resource category value to 1; the resource category value of 1 indicates that it is necessary to call the resource quantity outside the estimate to meet the event requirements;

[0052] If the resource category value is 1, the sum of the corresponding values of the spare resource vector and the event resource vector is used as the corresponding intermediate resource category quantity.

[0053] Optionally, obtaining multiple time position management images based on the event management information corresponding to multiple event times includes:

[0054] At a management time point, mark the management positions corresponding to multiple resource categories to obtain a management personnel position image;

[0055] Multiple management time points correspondingly obtain multiple management personnel position images;

[0056] Arrange the management personnel position images corresponding to multiple management time points in chronological order and superimpose them to obtain a time position management image;

[0057] Multiple event times correspondingly obtain multiple time position management images; the height of the time position management image is equal to the number of corresponding management time points.

[0058] In a second aspect, an embodiment of the present invention provides a big data sports event management system, including:

[0059] An acquisition module, configured to acquire event management information, an event resource vector, and a spare resource vector corresponding to multiple event times; the subscript of the event resource vector represents the resource category, and the value in the event resource vector represents the resource quantity; the event management information includes multiple resource categories and their corresponding management time points, management positions, management frequencies, and total management objects; the spare resource vector represents the resource quantities of multiple resource categories called outside the estimate;

[0060] A resource relationship module, configured to determine the usage status of resource categories based on the event management information corresponding to multiple event times, and obtain an average relationship feature; one resource category corresponds to one average relationship feature;

[0061] An intermediate resource prediction network acquisition module, configured to acquire an intermediate resource prediction network; the intermediate resource prediction network includes a first shrinking neural network and a first enlarging neural network;

[0062] An intermediate resource category quantity detection module, configured to input the average relationship feature into the first shrinking neural network to obtain an intermediate resource category quantity; the intermediate resource category quantity represents the critical value of resources that can meet the event usage; multiple resource categories correspondingly obtain multiple intermediate resource category quantities;

[0063] An adjustment module, configured to adjust the intermediate resource category quantity based on the spare resource vector and the event resource vector.

[0064] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0065] The embodiments of the present invention also provide a method and system for managing sports events in big data.

[0066] In the present invention, management time points, management locations, management frequencies, and management totals are adopted. By using a plurality of adjusted causal convolutional networks after random structure adjustment, the input methods of a plurality of time position change feature vectors corresponding to a plurality of time position management images are adjusted to detect the feature of the state of each resource category of the events corresponding to a plurality of event times, so as to find the average relationship feature. Obtain the training relationship feature that can be used as the input of the first shrinking neural network. Train the first shrinking neural network with the labeled resource quantity. When the labeled resource quantity is input into the first enlarging neural network, use the training relationship feature as the labeled data to train the first enlarging neural network. When the training relationship feature is input into the first shrinking neural network and the first enlarging neural network, use the training relationship feature as the labeled data to train the first shrinking neural network and the first enlarging neural network. Input the average relationship feature into the first shrinking neural network in the intermediate resource prediction network to detect the intermediate resource category quantity. Then adjust the intermediate resource category quantity through the recorded spare resource vector. It achieves the technical effect of more accurately finding the critical value of the resources that can meet the needs of the events. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a flowchart of a method for managing sports events in big data provided by an embodiment of the present invention.

[0068] Figure 2 It is a schematic structural diagram of the adjusted causal convolutional network in the flowchart of the method for managing sports events in big data provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0069] The present invention will be described in detail below with reference to the accompanying drawings.

[0070] Embodiment 1

[0071] As Figure 1 shown, the embodiments of the present invention provide a method for managing sports events in big data, and the method includes:

[0072] S101: Obtain event management information, event resource vectors, and spare resource vectors corresponding to a plurality of event times; the subscripts of the event resource vectors represent resource categories, and the values in the event resource vectors represent resource quantities; the event management information includes a plurality of resource categories and their corresponding management time points, management locations, management frequencies, and management object totals; the spare resource vector represents the resource quantities of a plurality of resource categories called outside the estimate.

[0073] Among them, one resource category corresponds to one resource quantity. The resource quantity represents the quantity of the personnel or items corresponding to the resource category.

[0074] Among them, the event management information represents the data recorded during the event management at a historical time point.

[0075] Among them, as in this embodiment, the resource categories include cleaning personnel and security personnel. The event resource vector represents the resource quantities of the cleaning personnel and security personnel estimated before the event. The event resource vector for a certain time is [30, 50]. The first value represents the cleaning personnel, and the corresponding resource quantity is 30, indicating 30 cleaning personnel. The second value represents the security personnel, and the corresponding resource quantity is 50, indicating 50 security personnel.

[0076] Among them, as in this embodiment, the spare resource vector for a certain time is [3, 0], indicating that 3 cleaning personnel and 0 security personnel are dispatched. Therefore, the total number of actual cleaning personnel participating in the event is 33, and the number of security personnel is 50.

[0077] Among them, as in this embodiment, the management time point corresponding to the cleaning personnel is the cleaning time point, and the cleaning time point represents the time point when the cleaning personnel clean. The management frequency corresponding to the cleaning personnel is the cleaning frequency, and the cleaning frequency represents the area of one area by the cleaning personnel divided by the cleaning time length. The total management object quantity corresponding to the cleaning personnel is the total cleaning object quantity, and the total cleaning object quantity represents the volume of the garbage cleaned by the cleaning personnel after an event. The management time point corresponding to the security personnel is the security time point, and the security time point represents the time point when the security personnel work. The management frequency corresponding to the security personnel is the security frequency, and the security frequency represents the number of people attempting to break through the security in one area by the security personnel. The total management object quantity corresponding to the security personnel is the total security object quantity, and the total security object quantity represents the number of people watching the event protected by the security personnel after an event.

[0078] S102: Based on the event management information corresponding to multiple event times, determine the usage status of the resource categories to obtain the average relationship features; one resource category corresponds to one average relationship feature.

[0079] S103: Obtain an intermediate resource prediction network; the intermediate resource prediction network includes a first shrinking neural network and a first expanding neural network.

[0080] S104: Input the average relationship features into the first shrinking neural network to obtain the intermediate resource category quantities; the intermediate resource category quantities represent the critical values of the resources that can meet the event usage; multiple intermediate resource category quantities are obtained corresponding to multiple resource categories.

[0081] S105: Adjust the number of intermediate resource categories based on the backup resource vector and the event resource vector.

[0082] Optionally, the determining the usage status of resource categories based on the event management information corresponding to multiple event times to obtain the average relationship feature includes:

[0083] Obtain multiple time-location management images based on the event management information corresponding to multiple event times; the length and width of the time-location management image represent geographical locations, and the height of the time-location management image represents the number of management time points; one event time corresponds to one time-location management image;

[0084] By adjusting the causal convolutional network, based on multiple time-location management images, detect the association relationship of the overall management location change among adjacent management time points to obtain the first time-location feature;

[0085] Based on the time-location management image and the first time-location feature, detect the features of resource categories to obtain n relationship features; the relationship feature represents the feature of the relationship of resources participating in the event; one resource category corresponds to one relationship feature;

[0086] Obtain m * n relationship features corresponding to m event times; average the m relationship features corresponding to one resource category to obtain the average relationship feature; n resource categories correspond to obtaining n average relationship features.

[0087] Wherein, n and m are natural numbers greater than 0.

[0088] Optionally, the adjusting the causal convolutional network, based on multiple time-location management images, detecting the association relationship of the overall management location change among adjacent management time points to obtain the first time-location feature includes:

[0089] The adjusted causal convolutional network includes a temporal convolutional network, a causal convolutional network, multiple random structures, and a fusion network.

[0090] Wherein, the structural schematic diagram of the adjusted causal convolutional network is as Figure 2 shown.

[0091] Wherein, in this embodiment, the temporal convolutional network is a temporal convolutional neural network (TCN), the causal convolutional network is a Causal Convolutional Neural Network (CCNN), the random structure is a vector, and the fusion network is a Fully connected neural network (FCN).

[0092] Manage the image based on the time positions to obtain multiple time position change feature vectors and values in the random structure; the time position change feature vectors represent the features of the overall management position change in adjacent management time points.

[0093] The number of input neurons of the causal convolutional network is equal to the number of the random structures.

[0094] According to the values in the random structure, input multiple time position change features into the corresponding input neurons in the causal convolutional network; the same values in two random structures indicate that the corresponding two time position change feature vectors are input into the same input neuron for detecting their correlation.

[0095] Obtain random vectors multiple times; multiple random vectors correspondingly obtain multiple random causal features.

[0096] Among them, in this embodiment, the number of times of obtaining the random vectors is 10.

[0097] Input multiple random causal features into the fusion network to detect the relationship of the position changes at multiple times, and obtain the first time position feature.

[0098] Optionally, the managing the image based on the time positions to obtain multiple time position change feature vectors and values in the random structure includes:

[0099] Input the multiple manager position images corresponding to the time position management image into the time convolutional network in sequence from the earliest time point to the latest time point to obtain time position change feature vectors.

[0100] Multiple time position management images correspondingly obtain multiple time position change feature vectors; the number of elements in multiple time position change feature vectors is equal.

[0101] Randomly arrange the natural numbers from 1 to the value obtained by dividing the number of elements in the time position change feature vector by 2 to obtain the first random vector.

[0102] Among them, in this embodiment, the number of elements in the time position change feature vector is 16, and the natural numbers between [1, 8] are randomly arranged. For example, one of the first random vectors is [1, 3, 2, 7, 4, 5, 8, 6].

[0103] Randomly arrange the values in the first random vector again to obtain the second random vector.

[0104] Among them, in this embodiment, the natural numbers between [1, 8] are randomly arranged again. For example, one of the second random vectors is [1, 3, 5, 8, 2, 7, 4, 6].

[0105] Append the elements in the second random vector after the elements in the first random vector to obtain a random vector.

[0106] Among them, as in this embodiment, the random vector is [1, 3, 2, 7, 4, 5, 8, 6, 1, 3, 5, 8, 2, 7, 4, 6].

[0107] Among them, the number of the random structures is equal to the number of elements in the first random vector and is equal to the number of elements in the second random vector. In this embodiment, the number of elements in the first random vector is 8, so the number of random structures is 8.

[0108] Among them, there are two identical values in one random vector.

[0109] Input the two values in the random vector into the random structure in sequence; the values in the random structure represent the subscripts of the input neurons connected in the causal convolutional network in the time position change feature vector.

[0110] Among them, in this embodiment, the starting subscript of the input neurons connected in the causal convolutional network is 1.

[0111] Among them, as in this embodiment, the value of the 1st random structure is [1, 3], the value of the 2nd random structure is [2, 7]…, and the value of the 8th random structure is [4, 6].

[0112] Optionally, the training method of the intermediate resource prediction network includes:

[0113] Obtain the training relationship feature and the labeled resource quantity; the training relationship feature represents the feature of the relationship of the resources participating in the event at the historical time point; the labeled resource quantity represents the quantity of the resources of the corresponding resource category that can exactly meet the event requirements.

[0114] Input the training relationship feature into the first shrinking neural network to obtain the training predicted resource quantity.

[0115] Among them, the first shrinking neural network is a fully connected neural network (FCN), and the number of output neurons of the first shrinking neural network is equal to the number of resource categories.

[0116] Input the predicted resource quantity into the first amplifying neural network to obtain the training predicted intermediate resource feature.

[0117] Among them, the first amplification neural network is a fully connected neural network (FCN). The number of input neurons of the first amplification neural network is equal to the number of resource categories, and the number of output neurons is equal to the length of the training relationship features.

[0118] The first reduction neural network is trained by using the training predicted resource quantity and the labeled resource quantity through a loss function.

[0119] Among them, the loss function is a cross-entropy loss function.

[0120] The first amplification neural network is trained by using the output of inputting the labeled resource quantity into the first amplification neural network and the training relationship features through a loss function.

[0121] Among them, in this embodiment, the cross-entropy loss function is used to obtain the loss.

[0122] The first amplification neural network and the first reduction neural network are trained by using the training predicted intermediate resource features and the training relationship features through a loss function.

[0123] Among them, in this embodiment, the cross-entropy loss function is used to obtain the loss.

[0124] Optionally, the features for detecting the usage status of resource categories based on the time position management image and the first time position feature to obtain n relationship features include:

[0125] The time position management image is split into multiple resource time position management images according to resource categories; one resource time position management image corresponds to one resource category.

[0126] Among them, in this embodiment, the position with a gray value of 1 in the time position management image represents the position of the cleaning staff. The value with a gray value of 1 is retained, and other values are deleted to obtain the resource time position management image corresponding to the cleaning staff. The position with a gray value of 2 in the time position management image represents the position of the security staff. The value with a gray value of 1 is retained, and other values are deleted to obtain the resource time position management image corresponding to the security staff.

[0127] Among them, in this embodiment, if the resource categories include cleaning staff and security staff. Then two resource time position management images are obtained. One resource time position management image corresponds to the cleaning staff, and the other resource time position management image corresponds to the security staff.

[0128] The resource time position management image is input into the first convolutional network to detect the change of the management position corresponding to the resource category and obtain the second time position feature.

[0129] Among them, the length of the second time position feature is the same as that of the first time position feature.

[0130] Among them, the first convolutional network is a Convolutional Neural Networks (CNN).

[0131] Fuse the first time position feature and the second time position feature to obtain a fused time position feature.

[0132] Among them, in this embodiment, the method of averaging the first time position feature and the second time position feature is used for fusion.

[0133] Input the management frequency corresponding to the resource category and the total amount of management objects into the second neural network to detect features and obtain a second management feature.

[0134] Among them, in this embodiment, the second neural network is a Fully connected neural network (FCN).

[0135] Multiple resource categories respectively obtain multiple first time position features and multiple second management features;

[0136] Superimpose the first time position feature and the corresponding second management feature to obtain a relationship feature.

[0137] Among them, after superimposing the first time position feature on the corresponding second management feature, the length of the relationship feature is equal to the sum of the lengths of the first time position feature and the corresponding second management feature.

[0138] n resource categories respectively obtain n relationship features.

[0139] Among them, n is a natural number greater than 0.

[0140] Optionally, the event resource vector, the backup resource vector, and the quantities of multiple intermediate resource categories are stored in different data tables;

[0141] The event resource vector and the backup resource vector can be actively modified.

[0142] Among them, the active modification means manually logging in to the data table for modification.

[0143] The quantities of the intermediate resource categories cannot be actively modified.

[0144] Among them, the quantities of the intermediate resource categories can only be predicted by the intermediate resource prediction network and adjusted through the backup resource vector and the event resource vector.

[0145] Optionally, adjusting the number of intermediate resource categories based on the spare resource vector and the event resource vector includes:

[0146] If the value in the spare resource vector is 0, set the resource category value to 0; a resource category value of 0 indicates that the predicted resource quantity can meet the event requirements.

[0147] Among them, the initial value of the resource category value is 0.

[0148] If the value in the spare resource vector is greater than 0, set the resource category value to 1; a resource category value of 1 indicates that it is necessary to call resources outside the prediction to meet the event requirements;

[0149] If the resource category value is 1, use the sum of the corresponding value in the spare resource vector and the corresponding value in the event resource vector as the corresponding number of intermediate resource categories.

[0150] Optionally, obtaining multiple time-location management images based on the event management information corresponding to multiple event times includes:

[0151] At a management time point, mark the management locations corresponding to multiple resource categories to obtain an administrator location image; among them, the administrator location image represents the image after marking the locations of multiple personnel or items for managing the event.

[0152] Among them, as in this embodiment, the resource categories include cleaning personnel and security personnel. Mark the location of the personnel corresponding to the cleaning personnel as 1 at the management time point, and mark the location of the personnel corresponding to the security personnel as 2 at the management time point.

[0153] Among them, the administrator location image is a grayscale image.

[0154] Multiple administrator location images are obtained corresponding to multiple management time points;

[0155] Arrange the administrator location images corresponding to multiple management time points in chronological order and superimpose them to obtain a time-location management image;

[0156] Multiple time-location management images are obtained corresponding to multiple event times; the height of the time-location management image is equal to the number of corresponding management time points.

[0157] Among them, the time-location management image is a three-dimensional image.

[0158] Embodiment 2

[0159] Based on the above-mentioned big data-based sports event management method, an embodiment of the present invention also provides a big data-based sports event management system, and the system includes:

[0160] An acquisition module for acquiring event management information, an event resource vector, and a spare resource vector corresponding to multiple event times; the subscript of the event resource vector represents the resource category, and the value in the event resource vector represents the resource quantity; the event management information includes multiple resource categories and their corresponding management time points, management locations, management frequencies, and total management objects; the spare resource vector represents the resource quantities of multiple resource categories called outside the estimate.

[0161] A resource relationship module for determining the usage status of resource categories based on the event management information corresponding to multiple event times to obtain average relationship features; one resource category corresponds to one average relationship feature.

[0162] An intermediate resource prediction network acquisition module for acquiring an intermediate resource prediction network; the intermediate resource prediction network includes a first shrinking neural network and a first enlarging neural network.

[0163] An intermediate resource category quantity detection module for inputting the average relationship feature into the first shrinking neural network to obtain the intermediate resource category quantity; the intermediate resource category quantity represents the critical value of resources that can meet event usage; multiple resource categories correspond to obtain multiple intermediate resource category quantities.

[0164] An adjustment module for adjusting the intermediate resource category quantity based on the spare resource vector and the event resource vector.

[0165] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The structures required to construct such systems will be apparent from the above description. In addition, the present invention is not directed to any particular programming language. It should be understood that the content of the present invention described herein can be implemented using various programming languages, and the descriptions of specific languages above are for disclosing the best mode of the present invention.

[0166] In the specification provided herein, a large number of specific details are set forth. However, it can be understood that the embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures, and technologies have not been shown in detail so as not to obscure the understanding of this specification.

[0167] Each component embodiment of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components in the device according to the embodiments of the present invention. The present invention can also be implemented as a device or device program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

Claims

1. A method for managing sports events of big data, characterized in that, Including: Obtain the event management information, event resource vector, and spare resource vector corresponding to multiple event times; the subscript of the event resource vector represents the resource category, and the value in the event resource vector represents the resource quantity; the event management information includes multiple resource categories and their corresponding management time points, management locations, management frequencies, and total management objects; The spare resource vector represents the resource quantities of multiple resource categories called outside the estimate; Based on the event management information corresponding to multiple event times, determine the usage status of the resource category to obtain the average relationship feature; One resource category corresponds to one average relationship feature; Obtain an intermediate resource prediction network; the intermediate resource prediction network includes a first shrinking neural network and a first enlarging neural network; Input the average relationship feature into the first shrinking neural network to obtain the intermediate resource category quantity; the intermediate resource category quantity represents the critical value of the resources that can meet the event usage; multiple resource categories correspond to obtain multiple intermediate resource category quantities; Adjust the intermediate resource category quantity based on the spare resource vector and the event resource vector; The training method of the intermediate resource prediction network includes: Obtain the training relationship feature and the labeled resource quantity; the training relationship feature represents the feature of the relationship of the resources participating in the event at the historical time point; the labeled resource quantity represents the resource quantity of the corresponding resource category that can exactly meet the event demand; Input the training relationship feature into the first shrinking neural network to obtain the training predicted resource quantity; Input the predicted resource quantity into the first enlarging neural network to obtain the training predicted intermediate resource feature; Train the first shrinking neural network by using the training predicted resource quantity and the labeled resource quantity through a loss function; Train the first enlarging neural network by using the output of inputting the labeled resource quantity into the first enlarging neural network and the training relationship feature through a loss function; Train the first enlarging neural network and the first shrinking neural network by using the training predicted intermediate resource feature and the training relationship feature through a loss function.

2. The method for managing sports events of big data according to claim 1, wherein The determining the usage status of the resource category based on the event management information corresponding to multiple event times to obtain the average relationship feature includes: Based on the event management information corresponding to multiple event times, obtain multiple time-location management images; the length and width of the time-location management image represent the geographical location, and the height of the time-location management image represents the number of management time points; one event time corresponds to one time-location management image; Adjust the causal convolutional network, and based on multiple time-location management images, detect the correlation relationship of the overall management location change in adjacent management time points to obtain the first time-location feature; Based on the time-location management image and the first time-location feature, detect the feature of the usage status of the resource category to obtain n relationship features; the relationship feature represents the feature of the relationship of the resources participating in the event; one resource category corresponds to one relationship feature; Obtain the m*n relationship features corresponding to m event times; average the m relationship features corresponding to one resource category to obtain the average relationship feature; the average relationship feature represents the detected feature of the resource category; n resource categories respectively obtain n average relationship features.

3. The method for managing sports events of big data according to claim 2, wherein Adjusting the causal convolutional network to manage an image based on multiple time positions, and detecting the correlation relationship of the overall management position change in adjacent management time points to obtain the first time position feature, including: The adjusted causal convolutional network includes a temporal convolutional network, a causal convolutional network, multiple random structures, and a fusion network; Based on the image managed by the time position, obtaining multiple time position change feature vectors and values in the random structure; the time position change feature vector represents the feature of the overall management position change in adjacent management time points; The number of input neurons of the causal convolutional network is equal to the number of the random structures; According to the values in the random structure, input multiple time position change features into the corresponding input neurons in the causal convolutional network; the same value in two random structures means inputting the corresponding two time position change feature vectors into the same input neuron to detect their correlation relationship; Obtaining random vectors multiple times; multiple random vectors respectively obtain multiple random causal features; Inputting multiple random causal features into the fusion network to detect the relationship of the position changes at multiple times, and obtaining the first time position feature.

4. The method for managing sports events of big data according to claim 3, characterized in that Based on the image managed by the time position, obtaining multiple time position change feature vectors and values in the random structure, including: Sequentially inputting multiple manager position images corresponding to the time position management image into the temporal convolutional network from the earliest time point to the latest time point to obtain time position change feature vectors; Multiple time position management images respectively obtain multiple time position change feature vectors; the number of elements in multiple time position change feature vectors is equal; Randomly arranging natural numbers from 0 to the value obtained by dividing the number of elements in the time position change feature vector by 2 to obtain the first random vector; Randomly arranging the values in the first random vector again to obtain the second random vector; Adding the elements in the second random vector after the elements in the first random vector to obtain a random vector; Among them, there are two identical values in one random vector; Sequentially inputting the two values in the random vector into the random structure; the value in the random structure represents the subscript of the input neuron connected to the time position change feature vector in the causal convolutional network.

5. The method for managing sports events of big data according to claim 2, wherein, Based on the time position management image and the first time position feature, detecting the feature of the usage status of the resource category to obtain n relationship features, including: Splitting the time position management image into multiple resource time position management images according to the resource category; one resource time position management image corresponds to one resource category; Inputting the resource time position management image into the first convolutional network to detect the change of the management position corresponding to the resource category, and obtaining the second time position feature; Fusing the first time position feature and the second time position feature to obtain a fused time position feature; Inputting the management frequency and the total amount of management objects corresponding to the resource category into the second neural network to detect features and obtain the second management feature; Multiple resource categories respectively obtain multiple first time position features and multiple second management features; Overlay the first time position feature and the corresponding second management feature to obtain a relationship feature; n resource categories respectively obtain n relationship features.

6. The method for managing sports events of big data according to claim 1, wherein The event resource vector, the spare resource vector, and the number of multiple intermediate resource categories are stored in different data tables; The event resource vector and the spare resource vector can be actively modified; The number of intermediate resource categories cannot be actively modified.

7. The method for managing sports events of big data according to claim 1, characterized in that, Adjusting the number of intermediate resource categories based on the spare resource vector and the event resource vector includes: If the value in the spare resource vector is 0, set the resource category value to 0; a resource category value of 0 indicates that the resource quantity within the estimate can meet the event requirements; If the value in the spare resource vector is greater than 0, set the resource category value to 1; a resource category value of 1 indicates that the resource quantity outside the estimate needs to be called to meet the event requirements; If the resource category value is 1, use the sum of the corresponding value in the spare resource vector and the corresponding value in the event resource vector as the corresponding number of intermediate resource categories.

8. The sports event management method for big data according to claim 2, wherein Obtaining multiple time position management images based on the event management information corresponding to multiple event times includes: At a management time point, mark the management positions corresponding to multiple resource categories to obtain a management personnel position image; Multiple management time points respectively obtain multiple management personnel position images; Stack the management personnel position images corresponding to multiple management time points in chronological order to obtain a time position management image; Multiple event times respectively obtain multiple time position management images; the height of the time position management image is equal to the number of corresponding management time points.

9. A sports event management system for big data, characterized in that, Including: An acquisition module for acquiring the event management information, the event resource vector, and the spare resource vector corresponding to multiple event times; the subscript of the event resource vector represents the resource category, and the value in the event resource vector represents the resource quantity; the event management information includes multiple resource categories and their corresponding management time points, management positions, management frequencies, and total management objects; the spare resource vector represents the resource quantity of multiple resource categories called outside the estimate; A resource relationship module for discriminating the usage status of resource categories based on the event management information corresponding to multiple event times to obtain an average relationship feature; 1 resource category corresponds to 1 average relationship feature; An intermediate resource prediction network acquisition module for acquiring an intermediate resource prediction network; the intermediate resource prediction network includes a first shrinking neural network and a first enlarging neural network; [[ID= ​ ​ Obtain the training relationship features and the number of labeled resources; the training relationship features represent the features of the relationships of the resources participating in the event at the historical time point; the number of labeled resources represents the number of resources of the corresponding resource category that can exactly meet the event requirements; Input the training relationship features into the first shrinking neural network to obtain the training predicted resource quantity; Input the predicted resource quantity into the first enlarging neural network to obtain the training predicted intermediate resource features; Train the first shrinking neural network by using the training predicted resource quantity and the number of labeled resources through a loss function; Train the first enlarging neural network by using the output of inputting the number of labeled resources into the first enlarging neural network and the training relationship features through a loss function; Train the first enlarging neural network and the first shrinking neural network by using the training predicted intermediate resource features and the training relationship features through a loss function.

Images