Athlete competition performance prediction model training method, prediction method, computer device, medium and program product

Abstract

The invention provides an athlete competition performance prediction model training method, a prediction method, a computer device, a medium and a program product. The method comprises the following steps: acquiring an athlete historical performance evaluation data set; acquiring an athlete popular race performance evaluation data set; integrating the athlete historical performance evaluation data set and the athlete popular competition performance evaluation data set to obtain a competition athlete performance prediction feature data set; and training a deep neural network prediction model based on a multi-layer perceptron based on the competition athlete performance prediction feature data set. The deep neural network prediction model obtained through training has high accuracy and good convergence capacity, competition performance prediction based on performance of athletes in warm-up sports is achieved, and coaches and analysts are assisted in making decisions.

Description

Technical Field

[0001] The present invention relates to the field of machine learning technology, and in particular to a model training method for predicting athlete competition performance, a prediction method, a computer device, a medium and a program product. Background Art

[0002] Traditional predictions of athletes' performance on the field are often based on experience. Since most coaches lack a background in mathematics or statistics, most current decisions, such as starting lineups and field performance predictions, still rely heavily on coaches and analysts to make predictions based on the players' status on the day of the game and their performance in pre-match warm-up exercises. However, this traditional empiricism-based prediction is a subjective prediction method, and its judgment is often affected by a variety of subjective factors, such as the club to which the athlete belongs, the competitiveness of the league, the region of the club, and discrimination bias. This prediction is not objective enough.

[0003] With the rapid development of big data and machine learning, technologies such as machine learning and data analysis have begun to be applied in the field of sports. By collecting and analyzing athletes' historical data, quantitative predictions of athletes' future performance can be achieved.

[0004] However, there are very few existing player performance prediction analyses. Many sports are team-based, and the performance of players on the field depends on factors such as their historical performance, state of the game, position on the field, starting formation, club affiliation, league competitiveness, etc., making prediction difficult.

[0005] In addition, some existing prediction models use relatively simple machine learning techniques, such as support vector machines, decision trees, and logistic regression. However, these models have low prediction accuracy and are less applicable to datasets with multiple features. Since the prediction of athlete performance involves multiple datasets with multiple features, more complex machine learning solutions are needed to predict datasets with multiple features and maintain relatively accurate prediction results. Summary of the invention

[0006] The first object of the present invention is to provide a method for training an athlete competition performance prediction model to improve the accuracy of athlete performance prediction.

[0007] A second object of the present invention is to provide a method for predicting athlete competition performance that improves the accuracy of athlete performance prediction.

[0008] The third object of the present invention is to provide a computer device for implementing the above-mentioned athlete competition performance prediction model training method and athlete competition performance prediction method to improve the accuracy of athlete performance prediction.

[0009] A fourth object of the present invention is to provide a computer program product that implements the above-mentioned athlete competition performance prediction model training method and athlete competition performance prediction method to improve the accuracy of athlete performance prediction.

[0010] In order to achieve the above-mentioned first purpose, the present invention provides a method for training an athlete's game performance prediction model, which includes the following steps: obtaining an athlete's historical performance evaluation data set; obtaining an athlete's warm-up match performance evaluation data set; integrating the athlete's historical performance evaluation data set and the athlete's warm-up match performance evaluation data set to obtain a game athlete performance prediction feature data set; based on the game athlete performance prediction feature data set, training a deep neural network prediction model based on a multi-layer perceptron.

[0011] As can be seen from the above scheme, the present invention obtains a complete performance prediction feature data set of athletes in the competition by fusing and expanding the athlete historical performance evaluation data set and the athlete warm-up performance evaluation data set, thereby establishing a multi-output model to predict the athlete performance. The present invention solves the problem that the existing prediction of athlete performance is too dependent on subjective experience and the athlete performance data is insufficient. Compared with the problem that most of the existing prediction models using simple machine learning technology can only achieve good prediction results in small-scale data sets, the present invention can be better applied to multiple data sets with multiple features, with higher accuracy and better convergence ability.

[0012] A further solution is to perform feature importance analysis on both the athlete's historical performance data and the athlete's warm-up performance data to obtain an athlete's historical performance evaluation data set and an athlete's warm-up performance evaluation data set.

[0013] It can be seen that feature importance analysis can determine the most influential features in the athlete performance history data and the athlete warm-up performance data, and eliminate irrelevant or redundant features, thereby improving the training efficiency and accuracy of the subsequent deep neural network prediction model obtained based on the acquired competition athlete performance prediction feature data set.

[0014] A further solution is to obtain historical performance data of athletes; pre-process the historical performance data of athletes to obtain a historical performance data set; train a first prediction model based on the historical performance data set, and the first prediction model is used to output the probability of the athlete starting in the next game; perform feature importance analysis on the historical performance data set in combination with the first prediction model to optimize the historical performance data set, and obtain the athlete's historical performance evaluation data set.

[0015] A further solution is to obtain the athlete's warm-up performance data; pre-process the athlete's warm-up performance data to obtain a warm-up performance data set; train a second prediction model based on the warm-up performance data set, and the second prediction model is used to output a classification prediction of whether the athlete will score or not in the next game; perform feature importance analysis on the warm-up performance data set in combination with the second prediction model to optimize the warm-up performance data set, and obtain the athlete's warm-up game performance evaluation data set.

[0016] A further solution is that after obtaining the warm-up performance data set, before training the second prediction model based on the warm-up performance data set, the following steps are included: extracting features from the warm-up performance data set, and performing feature importance analysis to screen out influential features; determining data labels, and readjusting the warm-up performance data set based on the data labels.

[0017] It can be seen from this that before training the second prediction model, the characteristics of the data set can be preliminarily screened and the data labels can be determined to improve the efficiency and accuracy of subsequent model training.

[0018] A further solution is to analyze the influential features through the K-means clustering algorithm to determine the data labels.

[0019] It can be seen that cluster analysis can better assist in determining data labels.

[0020] A further solution is to collect the athlete's warm-up performance data through a device with data collection function worn by the athlete.

[0021] It can be seen that obtaining warm-up performance data directly from the device worn by athletes is an efficient and easy-to-implement collection method.

[0022] In order to achieve the above-mentioned second purpose, the present invention provides a method for predicting athlete's competition performance, wherein: the above-mentioned deep neural network prediction model is obtained; the athlete's warm-up data before the game is obtained; and the target predicted performance result corresponding to the athlete's warm-up data before the game is output according to the deep neural network prediction model.

[0023] It can be seen from the above scheme that the present invention obtains the performance of athletes in warm-up exercises, quantitatively predicts the performance and game trends of athletes based on the deep neural network prediction model, and provides data support for the prediction results to a certain extent, so that coaches and analysts can more intuitively and objectively observe the entire game and respond in time.

[0024] In order to achieve the third purpose mentioned above, the present invention provides a computer device, including a processor and a memory, wherein: a computer program is stored in the memory, and when the computer program is executed by the processor, the above-mentioned athlete competition performance prediction model training method or the above-mentioned athlete competition performance prediction method is implemented.

[0025] In order to achieve the fourth purpose mentioned above, the present invention provides a computer program product, including computer instructions, wherein: when the computer instructions are executed by a processor, the above-mentioned athlete competition performance prediction model training method or the above-mentioned athlete competition performance prediction method is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of determining an athlete's historical performance evaluation data set in an embodiment of the athlete's game performance prediction model training method of the present invention.

[0027] Figure 2 It is a flow chart of determining an athlete warm-up match performance evaluation data set in an embodiment of an athlete match performance prediction model training method of the present invention.

[0028] Figure 3 It is a flow chart of determining a deep neural network prediction model in an embodiment of the athlete competition performance prediction model training method of the present invention.

[0029] Figure 4 It is a schematic diagram of the structure of a neural network of a multilayer perceptron used in a deep neural network prediction model in an embodiment of the athlete competition performance prediction model training method of the present invention.

[0030] Figure 5 It is a flow chart of an embodiment of a method for predicting athlete's game performance of the present invention.

[0031] Figure 6 This is an example diagram of applying a deep neural network prediction model to perform predictions in an embodiment of the method for predicting athlete competition performance of the present invention.

[0032] The present invention is further described below in conjunction with the accompanying drawings and embodiments. DETAILED DESCRIPTION

[0033] The athlete competition performance prediction model training method of the present invention is based on the athlete competition performance prediction model training method to train a deep neural network prediction model. The athlete competition prediction method of the present invention inputs the collected warm-up data of the athlete before the competition into the above-mentioned deep neural network prediction model to obtain a prediction result.

[0034] Example of athlete competition performance prediction model training method:

[0035] This embodiment is implemented by executing a computer program, firstly obtaining the athlete performance history data and the athlete warm-up performance data, then performing feature importance analysis on the athlete performance history data to obtain the athlete historical performance evaluation data set, and performing feature importance analysis on the athlete warm-up performance data to obtain the athlete warm-up performance evaluation data set. Then, the athlete historical performance evaluation data set and the athlete warm-up performance evaluation data set are integrated to obtain the competition athlete performance prediction feature data set. Finally, based on the competition athlete performance prediction feature data set, a deep neural network prediction model based on a multi-layer perceptron is trained.

[0036] This embodiment is described by using the case of predicting the performance of athletes in a football match. It can be understood that in other embodiments, the present invention can also be applied to predicting the performance of athletes in basketball, volleyball, rugby, tennis and other sports.

[0037] See also Figure 1 ,Obtaining the athlete’s historical performance evaluation dataset includes the following steps:

[0038] S11: Obtain athlete performance history data.

[0039] The player performance history data is the player's performance data in the past period of time. In this embodiment, the player performance history data includes but is not limited to: game date, home and guest team information, the player's goals, discipline records, assists, corner kicks, passing and touchdowns, touchdowns, rushing yards, rushing touchdowns, starting lineups, player positions, game wins and losses, and other information.

[0040] Optionally, the athlete performance history data may be derived from sports information provided by a sports portal.

[0041] S12: Preprocess the athlete performance historical data to obtain a historical performance data set.

[0042] S13: Train a first prediction model based on the historical performance dataset.

[0043] The historical performance data set is divided into a training set and a test set, and a first prediction model is established and trained. The first prediction model outputs the probability of each player starting in the next game.

[0044] Optionally, the first prediction model can also classify the players according to their positions or characteristics, and output the predicted probability of the players' starting positions in the next game. For the classification of football players, the classification according to the positions of the players includes but is not limited to defenders, midfielders, forwards, etc.

[0045] For the evaluation indicators of the first prediction model, the root mean square error (RMSE) and the mean absolute error (MAE) are used.

[0046] S14: Perform feature importance analysis on the historical performance data set in combination with the first prediction model to optimize the historical performance data set and obtain a historical performance evaluation data set of athletes.

[0047] By performing feature importance analysis on the historical performance dataset, we can find the most influential feature on the output of the first prediction model, and then use the most influential feature on the output of the first prediction model as the input of the first prediction model, and the first prediction model outputs the probability of each athlete starting in the next game. The historical performance dataset is adjusted according to the most influential feature on the output of the first prediction model, and the athlete historical performance evaluation dataset can be obtained.

[0048] See also Figure 2 ,Getting the athlete warm-up match performance evaluation dataset includes the following steps:

[0049] S21: Obtain athlete warm-up performance data.

[0050] Among them, the athlete's warm-up performance data includes but is not limited to the athlete's running distance, maximum speed, number of sprints, explosive power, real-time heart rate, heart monitoring, athlete health, injuries in career and monitoring of injuries in other parts of the body.

[0051] In this embodiment, the warm-up performance data of the athletes is collected by a device with data collection function worn by the athletes, specifically, the data is collected during the warm-up exercise before each game.

[0052] The device with data collection function can be a device based on Internet of Things (IoT) technology, for example, professional intelligent equipment such as sports jackets, sports shoes, sports watches, etc. with data collection function.

[0053] S22: Preprocess the athlete's warm-up performance data to obtain a warm-up performance data set.

[0054] S23: Extract features from the warm-up performance dataset and perform feature importance analysis to select influential features.

[0055] S24: Determine data labels, and readjust the warm-up performance data set according to the data labels.

[0056] Among them, the influential features determined in step S23 are analyzed by the K-Means algorithm, the metric distance is set to the Euclidean distance, and the optimal K value is selected by combining the inflection point method of the sum of squared deviations within the cluster and the silhouette coefficient method in the prior art.

[0057] By analyzing the athletes' movement trajectories and technical performances collected by devices with data collection functions worn by athletes, the accuracy and success rate of the technology are used as important indicators to measure the level of athletes. At the same time, the performance of the home and guest teams can also be considered, and a multi-dimensional event axis can be referred to to label the data and obtain a relatively complete data set, that is, the readjusted warm-up performance data set.

[0058] S25: Training a second prediction model based on the readjusted warm-up performance dataset.

[0059] The warm-up performance data set is divided into a training set and a test set, a second prediction model is established, and an evaluation index is determined to train the second prediction model. The output of the second prediction model is a classification prediction of whether each player will score or not in the next game.

[0060] S26: Perform feature importance analysis on the warm-up performance data set in combination with the second prediction model to optimize the warm-up performance data set, and obtain an athlete warm-up match performance evaluation data set.

[0061] By performing feature importance analysis on the re-adjusted warm-up performance dataset, we can find the most influential features on the output of the second prediction model, and then use the most influential features on the output of the second prediction model as the input of the second prediction model. The output of the second prediction model is the classification prediction of whether each athlete will score or not in the next game. The warm-up performance dataset is adjusted according to the most influential features on the output of the second prediction model, and the athlete warm-up performance evaluation dataset can be obtained.

[0062] See also Figure 3 , training a deep neural network prediction model based on a multi-layer perceptron includes the following steps:

[0063] S31: Obtain a dataset of athletes’ historical performance evaluation.

[0064] S32: Obtain a dataset of athlete warm-up match performance evaluation.

[0065] The athlete's historical performance evaluation data set is derived from the above step S14, and the athlete's warm-up match performance evaluation data set is derived from the above step S26.

[0066] S33: Integrate the athlete's historical performance evaluation dataset and the athlete's warm-up match performance evaluation dataset to obtain a performance prediction feature dataset for the competition athletes.

[0067] Among them, key factors such as athlete technical statistics and competition conditions are expanded with existing feature tags, including expanding data from different data sources through data links to build a more complete and comprehensive data set. In this way, a complete feature data set for predicting the performance of competition athletes is integrated.

[0068] S34: Based on the performance prediction feature dataset of the players in the competition, a deep neural network prediction model based on a multi-layer perceptron deep neural network is trained.

[0069] The prediction feature data set of the athletes' performance in the competition is divided into a training set and a test set, a deep neural network prediction model is established, the evaluation index of the deep neural network prediction model is determined, and the deep neural network prediction model is trained. The deep neural network prediction model is used to predict the performance of each athlete in the competition.

[0070] The deep neural network prediction model is a deep neural network multi-output prediction model based on a multi-layer perceptron, in which a series of indicators for measuring model performance, such as accuracy, recall, root mean square error (RMSE) and F1 value (F1 Score), can be used as evaluation indicators of the model.

[0071] See also Figure 4 The deep neural network prediction model is based on a neural network with multi-layer perceptrons, wide inputs, and residual connections. The input layer of the deep neural network prediction model contains continuous features, categorical features, and binary feature layers. In order to make the model more capable of learning, the parsable model width is wide, and the deep neural network prediction model has multiple hidden layers. The prediction of the athlete's performance is a multi-output prediction, so the model is a multi-output structure.

[0072] Athlete competition performance prediction method embodiment:

[0073] See also Figure 5 This embodiment is based on the deep neural network prediction model of the above embodiment, and specifically includes the following steps:

[0074] S41: Obtain a deep neural network prediction model.

[0075] S42: Acquire warm-up data of athletes before competition.

[0076] S43: Outputting a target predicted performance result corresponding to the athlete's pre-match warm-up data according to the deep neural network prediction model.

[0077] Therefore, by collecting the warm-up data of athletes before the game, it is possible to predict the performance of athletes in the game. Coaches can comprehensively consider and deploy appropriate starting lineups before the game based on various prediction results.

[0078] See also Figure 6 , the figure shows an example of applying a deep neural network prediction model to make predictions. Figure 6 The middle left arrow indicates a good state, the flat arrow indicates a general state, and the downward arrow indicates a bad state. In this way, before the game, the pre-game warm-up data collected by the equipment with data collection function worn by the athletes is input into the deep neural network prediction model to quickly know the player's state level, and then displayed to the coach in a similar graphical user interface to assist the coach in making decisions.

[0079] In summary, the present invention obtains a complete prediction feature data set of the performance of athletes in the competition by fusing and expanding the data set of athletes' historical performance evaluation and the data set of athletes' warm-up performance evaluation, thereby establishing a multi-output model to predict the performance of athletes. The present invention solves the problem that the existing prediction of athlete performance is too dependent on subjective experience and insufficient athlete performance data. Compared with the problem that most of the existing prediction models using simple machine learning technology can only achieve good prediction results in small-scale data sets, the present invention can be better applied to multiple data sets with multiple features, with high accuracy and good convergence ability. The deep neural network prediction model trained by the present invention has high accuracy and good convergence ability, and can predict the performance of the game based on the performance of the athletes in warm-up exercises, assisting coaches and analysts in making decisions.

[0080] Computer device embodiment:

[0081] The computer device of this embodiment includes a processor and a memory. The memory stores a computer program. When the processor executes the computer program, the above-mentioned athlete competition performance prediction model training method embodiment or athlete competition performance prediction method embodiment is implemented.

[0082] The computer device may include but is not limited to a processor and a memory. Those skilled in the art will appreciate that the computer device may include more or fewer components, or a combination of certain components, or different components, for example, the computer device may also include input and output devices, network access devices, buses, etc.

[0083] For example, the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microcontroller or any conventional processor, etc. The processor is the control center of a computer device, and uses various interfaces and lines to connect various parts of the entire computer device.

[0084] The memory can be used to store computer programs and / or modules. The controller realizes various functions of the computer device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory. For example, the memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound receiving function, a sound conversion to text function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, text data, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Med ia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0085] If the module integrated in the computer device of the above embodiment is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the implementation of the above-mentioned athlete competition performance prediction model training method embodiment or athlete competition performance prediction method embodiment of all or part of the process can also be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the controller, the steps of the above-mentioned athlete competition performance prediction model training method embodiment or athlete competition performance prediction method embodiment can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The storage medium can include: any entity or device that can carry computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electrical carrier signals and telecommunications signals.

[0086] Computer program product embodiment:

[0087] The computer program product of this embodiment includes computer instructions, which are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the various steps of the above-mentioned athlete competition performance prediction model training method embodiment or athlete competition performance prediction method embodiment.

Claims

1. A method for training an athlete's game performance prediction model, characterized in that: The following steps are involved: Obtain a dataset of historical performance evaluations of athletes; Obtain the athlete warm-up match performance evaluation dataset; Integrate the athlete historical performance evaluation data set and the athlete warm-up match performance evaluation data set to obtain a match athlete performance prediction feature data set; Based on the performance prediction feature dataset of the competition athletes, a deep neural network prediction model based on a multi-layer perceptron is trained.

2. The athlete competition performance prediction model training method according to claim 1, characterized in that: The following steps are also included: Feature importance analysis is performed on both the athlete performance history data and the athlete warm-up performance data to obtain the athlete history performance evaluation data set and the athlete warm-up performance evaluation data set.

3. The athlete competition performance prediction model training method according to claim 2, characterized in that: The following steps are also included: Access historical athlete performance data; Preprocessing the athlete performance historical data to obtain a historical performance data set; training a first prediction model based on the historical performance data set, the first prediction model being used to output a probability of a player starting in a next game; The historical performance data set is subjected to feature importance analysis in combination with the first prediction model to optimize the historical performance data set, thereby obtaining the athlete historical performance evaluation data set.

4. The athlete competition performance prediction model training method according to claim 2, characterized in that: The following steps are also included: Obtain athlete warm-up performance data; Preprocessing the athlete's warm-up performance data to obtain a warm-up performance data set; Training a second prediction model based on the warm-up performance data set, the second prediction model is used to output a classification prediction of whether the athlete will score or not score in the next game; The warm-up performance data set is subjected to feature importance analysis in combination with the second prediction model to optimize the warm-up performance data set, thereby obtaining the athlete warm-up match performance evaluation data set.

5. The athlete competition performance prediction model training method according to claim 4, characterized in that: After obtaining the warm-up performance data set, before training the second prediction model according to the warm-up performance data set, the following steps are included: Extracting features from the warm-up performance data set, and performing feature importance analysis to screen out influential features; Data labels are determined, and the warm-up performance data set is readjusted according to the data labels.

6. The athlete competition performance prediction model training method according to claim 5, characterized in that: The influential features are analyzed by the K-means clustering algorithm to determine the data labels.

7. The athlete competition performance prediction model training method according to any one of claims 4 to 6, characterized in that: The athlete's warm-up performance data is collected through a device with data collection function worn by the athlete.

8. A method for predicting athlete competition performance, characterized in that: Obtain the deep neural network prediction model described in any one of claims 1 to 7; Get the warm-up data of athletes before the competition; Output the target predicted performance result corresponding to the athlete's pre-match warm-up data according to the deep neural network prediction model.

9. A computer device comprising a processor and a memory, characterized in that: The memory stores a computer program, which, when executed by the processor, implements the athlete competition performance prediction model training method described in any one of claims 1 to 7 or the athlete competition performance prediction method described in claim 8.

10. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by the processor, the athlete competition performance prediction model training method described in any one of claims 1 to 7 or the athlete competition performance prediction method described in claim 8 is implemented.

Images