Українська   Русский
DonNTU   Masters' portal

Abstract

Content

Introduction

Automating the process of compiling the sports calendar for cyclists is to use a system of support and decision-making regarding the selection of the most appropriate composition of athletes for the race.

Decision support system, DSS, Decision Support System, DSS — computer automated system, which aims to help people make decisions in difficult conditions for a full and objective analysis of objective activity [4].

Decision Support System is designed to support multiple criteria decision in a complex IT environment. Thus, under the multiple criteria it understood the fact that the results of the decisions are evaluated not by one, but by a set of multiple indicators (criteria) are considered simultaneously. Information complexity is determined by the need to accommodate a large volume of data, the processing of which without the help of modern computer technology is almost impossible. Under these conditions, the number of possible solutions are usually very large, and selecting without the full analysis can lead to serious errors.

Decision support systems DSS solves two main tasks:

To solve this problem is to use the first task — to choose the best solution among many possible.

1. Theme urgency

Now, the process of registration of the physical state of a professional athlete is a mandatory factor to achieve maximum results. This is because professional cycling is characterized by high competitiveness among the athletes that imposes high demands on the level of physical fitness.

With the development of computer technology, data characterizing the training and physical condition of the athlete are in electronic form. This makes it easy to manipulate them, make the tables and graphs for data, and make calculations for generation recommendations for athlete.

Considering the above, the development of the software application is relevant, effective and efficient solutions for the future users of the system.

2. Goal and tasks of the system

The developed system has the accounting and analytical character, with a focus on professional athletes cyclists and amateur sportsmen, with the necessary devices for reading the results of their training, which will serve as input to the system being developed.

Purpose of the system – organize information about conducted trainings athlete cyclist. This information can be used to produce sports calendar year, and for the distribution of roles in the team riders. The system should give these recommendations, thereby to perform the work of sports analyst and athletic director to help make a decision about the composition of teams for various races in the current season.

The purpose of the system to help select the most effective part of the athletes on a certain race, the team was able to show the best result, which is likely to lead to new investment sponsors.

The main objectives of the system:

3. The mathematical formulation

Input parameters are loaded into the system with bicycle computer that is connected to the PC. This data includes pulse rate, cadence, power, temperature, altitude [6]. The indicators are represented as arrays of current values at each time point measurement sensors. Also recorded the length and duration of the trip.

Primary data processing is calculating the average, maximum and minimum values for each of the input parameters for the entire distance and short intervals, for example, one kilometer away.

After these operations, it is possible to display statistical graphs and tables.

Statistical graphs and tables contain the average, maximum and minimum values for each training on such indicators as the heart rate, speed, cadence, power, wind power, set / height loss. The graph data in the form of curves plotted by points — X-axis — the date of exercise, along the axis Y — value. Since the values of the parameters will vary widely, such as displaying the value on a scale temperatures and minimum pulse will appear unclear, the Y-axis graph is separated into single areas. An example is showing in Fig. 1.

Example of display the statistical graphics

Figure 1 - Example of display the statistical graphics

Based on a few workouts, you can judge the progress or regression of the athlete in terms of fitness. The accuracy of the result will depend on the amount of training conducted. It is necessary to monitor changes in the values of the parameters heart rate, speed, cadence and power at approximately the same values of the air temperature and the set / height loss. If the results of the athlete for a few workouts deteriorate, it is possible to judge the lack of rest and the need to reduce the load.

Also by training data, system can judge the most appropriate expertise for an athlete. To do this, define the possible specializations: sprinter — a man with a better finish, which can be determined by the ability to develop a capacity of more than 700 watts, climber — a person with an average speed of the best climb over 1,000 meters per 100 kilometers, time trial — a man with a good individual result, this is determined by a relatively stable capacity and high average speed of climb is less than 500 meters per 100 kilometers. For the correct determination of specializations, necessary for system to attended by several riders, and ideally — the whole team.

We cannot say clearly refers to which specialization rider belongs, because these concepts are elements of fuzzy logic. Degree of membership is determined by the following algorithm forming membership functions [5]. The basis of the construction of the membership function of occurrence frequency response put characteristic values. Universal set each linguistic variable is the union of the supports of fuzzy sets of terms — the set of all possible values of the trait. At the stage of construction eliminated outliers — emissions. The value of the membership function μij(х′i) ∈ [0,1] is the degree of confidence with which the sample x′ the value of the i-th component equal x′i, corresponds to the j-th class.


Table 1 - Assessment the degree of compliance for proposed confidence data

Assessment the degree of compliance for proposed confidence data

Functions are built using a sliding window. The size of the sliding window chosen experimentally. Fig. 2 shows an example of the term set of linguistic variable attribute "Cadence value" for a situation where the number of classes n = 5.

Term set of linguistic variable "Cadence value"

Figure 2 - Term set of linguistic variable "Cadence value"
(animation: 5 frames, fixated, 21 kilobytes)

The figure shows that the terms of membership functions are often very close or overlap, which makes it impossible to draw a conclusion on the data of only one feature, so requires an integrated evaluation of a set of attributes.

To obtain the degree of certainty of compliance proposed by way of each of the classes form vi, constructed table 1 assess the degree of certainty.

Формула расчета функции принадлежности

where μij(x) – membership function j-th term of the i-th linguistic variable, ω(w) — integrated assurance, a xij — i-th component of a recognizable image of x.

Ranking table the last column of integrated confidence and get the class vi, for which the ω(w) as much as possible. He is identified as the most appropriate way proposed x. Expert on the table 1 defines the class that best matches the original athlete. In some situations, it appears that with slightly lower degree of confidence algorithm assigns this object and other classes. This situation is explained by the following factors — the classes are not always linearly separable. The percentage of correct answers is dependent on the completeness of the training sample and the amount used for the detection of informative features.

Having data of the specialization for racer and race schedule for the upcoming season, you can make the racing calendar individual athlete. Sports Director every race exhibited a priority from 0 to 10, and depending on this value the program picks up riders in the race, the higher the priority, the more powerful rider in the category you want is selected. The strength of the driver is determined by the conditional coefficient is calculated on the basis of comparing it with the average speed of other drivers of his specialization. Each race racers require a certain specialization, unlike the stages where there are races of all types, and therefore need all the riders classification.

For correct selection of athletes for the upcoming race, you must have enough information about the race. Primarily distinguished one-day and multi-day race. For such an important stage race sportsman feature as recoverability. This is determined by the frequency and effectiveness of training, for example, if the rider has a training period lasting at least 5 days with an average heart rate over 170 beats / min — one can judge the recoverability of a good athlete, and it makes sense to recommend it for multi-day race. It is also the hallmark of stage races is the variety of stages, suggesting the need for the various categories of athletes in the team.

Each race, and in the case of the tour — the stage should be described in the system, and contain information about the duration, climb and descent, maximum gradient of the date and a short description.

There are many external factors beyond the computer analysis, so the selected athlete classification and the proposed calendar of races will not always be optimal, suggesting the need for verification of output data.

Based on the above information we can say that the output parameters will be the team racing calendar, as well as statistical tables and graphs with performance training.

Conclusion

As a result of the research was grounded the necessity of developing a computerized system for Automating the process of compiling the sports calendar for cyclists, identified its main function, analyzed the existing methods of data classification algorithm is a principle of formation of the membership function. Was described the expected functional and mathematical problem statement.

This master's work is not completed yet. Final completion: winter 2015—2016. The full text of the work and materials on the topic can be obtained from the author or his head after this date.

References

  1. А.В Леоненков. Нечеткое моделирование в среде MATLAB и fuzzyTECH. — СПб.: БХВ-Петербург, 2003. — 736 с.
  2. А.П. Ротштейн, Д.И. Кательников Идентификация нелинейных объектов нечеткими базами знаний // Кибернетика и системный анализ. — 1998. — № 5 — 122 с.
  3. С.Д. Штовба Настройка нечеткой модели по обучающей выборке с нечетким выходом // Кибернетика и системный анализ. — 2007. — № 3 — с. 55–68.
  4. Системы поддержки принятия решений / Интернет-ресурс. — Режим доступа: http://bourabai.kz/tpoi/dss.htm
  5. В.А. Козловский, А.Ю. Максимова Алгоритм распознавания, основанный на нечетком подходе // «Искусственный интеллект» — 2008 — №4 — с. 594–599.
  6. Шеннон Совндаль, Анатомия велосипедиста — Попурри, 2011. — 129 с.
  7. Обзор алгоритмов кластеризации данных / Интернет-ресурс. — Режим доступа: http://habrahabr.ru/post/101338
  8. О.И. Ларичев. Теория и методы принятия решений. — М., Логос, 2000 — 98 c.
  9. М.Ф. Каспшицкая, А.А. Провотарь Решение нечеткой задачи классификации: алгоритм и результаты вычислительного эксперимента // Компьютерная математика. — 2013, — № 2 — с. 144–152.
  10. В.А. Ибрагимов Элементы нечеткой математики — БАКУ, 2009. — 165 с.