Artificial intelligence systems enabling the tasks of world football and related organizations, and facilitating the development and popularization of football

Description

[0001] ARTIFICIAL INTELLIGENCE SYSTEMS ENABLING THE TASKS OF WORLD FOOTBALL AND RELATED ORGANIZATIONS, AND FACILITATING THE DEVELOPMENT AND POPULARIZATION OF FOOTBALL

[0002] Technical Field

[0003] The invention involves artificial intelligence systems responsible for executing tasks related to world football and its affiliated organizations, encompassing refereeing and fair play principles of national football federations, discipline regulations, fair play, international relations, and collaborations, financial management of football, popularization of football and creating an impact in society, as well as technology usage and innovation in football-related tasks and activities, and is further concerned with ensuring the implementation of football rules set by regulatory bodies.

[0004] Prior Art

[0005] The origins of football trace back to ancient times. In the 2nd and 3rd centuries BC, a sport known as 'harpastum' was played in ancient Rome, yet its rules and playing field were not fully standardized. During those periods, football was played for military training or entertainment purposes and carried a more savage character. Throughout the Middle Ages, football was played with different rules in various countries. In England, a game known as 'mob football' was played as a complex and often violent competition among town or village residents. The playing fields included streets, squares, and even churchyards. The character of football during this period was utilized to pacify conflicts between local communities or to stimulate further competition. Football is not only seen as a sports game but also as a reflection of culture and society. It aids individuals in forming their identities and developing a sense of community. Fandom manifests itself through loyalty to a football club and enthusiastic matchdays. Additionally, football brings people together and encourages international relations.

[0006] World Football Federation (Federation Internationale de Football Association) emerged from the need to standardize football on the international stage and popularize it worldwide. Despite the ancient origins of football, the game being played with different rules in different countries posed challenges for organizing international matches. For these and similar reasons, World Football Federation was established on May 21, 1904, in Paris. The organization's purpose is to contribute to the development of international football, ensure the global standardization of football, and facilitate the organization of international matches. The first president of World Football Federation was the French footballer and administrator Jules Rimet. Rimet, serving as president for 33 years, made significant contributions to World Football Federation's growth. Over time, World Football Federation has organized a series of programs and tournaments to assist in the global spread of international football. The first World Cup was held in 1930, marking a significant milestone in demonstrating the immense interest in football on the international stage.

[0007] European Football Federation (Union of European Football Associations) was established to regulate the organization and management of football in Europe. In the post-World War II period, there arose a need to organize football and hold international tournaments among European countries. Therefore, European Football Federation was founded on June 15, 1954, in Basel, Switzerland. Its aim is to support the development of football in Europe, organize major tournaments such as the European Championship, and promote the spread of football in Europe. European Football Federation has made significant contributions to the growth of football in Europe and has ensured the organization of prestigious tournaments like the European Championship. Additionally, European Football Federation has organized and continues to organize major club tournaments, such as the European Football Federation Champions League, to support and enhance club football. Until today, World Football Federation and European Football Federation have signed numerous projects and programs to increase the prevalence and popularity of football worldwide. They have contributed to the development of football and have become significant figures on the international stage, playing critical roles in the international management and organization of football.

[0008] World Football Federation is an organization responsible for the international governance of football. Its core activities and responsibilities are as follows: Coordinating and supporting national football federations.

[0009] Overseeing the organization and management of international football tournaments.

[0010] Establishing and updating the rules of football.

[0011] Conducting doping controls and engaging in activities related to sports ethics. Initiating programs and projects for the development and education of football. Implementing a series of programs and initiatives contributing to the international growth and development of football, reaching children, youth, and adults, and promoting the values of the sport.

[0012] To fulfill these activities, World Football Federation executes the following tasks: Working on and implementing refereeing and fair play principles.

[0013] Determining disciplinary rules and ensuring discipline regarding fair play. Developing and collaborating on international relations.

[0014] Ensuring the financial management of football.

[0015] Promoting further dissemination of football and creating an impact in society.

[0016] Integrating the use of technology and innovation into the development of football.

[0017] These mentioned tasks stand out as common responsibilities for both World Football Federation and European Football Federation.

[0018] In conclusion, World Football Federation is one of the most authoritative and influential organizations on the international level for football. World Football Federation's worldwide membership relies on national football federations, encompassing more than 200 countries and regions. World Football Federation continues to execute various projects for the dissemination and international growth of football. Moreover, football has a significant economic impact, as World Football Federation generates substantial revenues through major television agreements, sponsorships, and commercial activities, which are then allocated towards the development of football. The establishment and development of World Football Federation have made substantial contributions to the international growth and global popularity of football. World Football Federation is widely recognized as one of the most important governing bodies for international football and continues to lead the development of world football.

[0019] In the technical realm of artificial intelligence applications in football, there are some developments. One such development in the technical field is described in patent document number 2022 / 019706, where the invention pertains to a system and method involving the creation of a database pool for match positions used in training and seminars by IFAB, World Football Federation, and affiliated federations and confederations. This system enables the real-time scanning of positions during live matches, including database matching for positions that have led to significant controversies, allowing for instant notification to the referee team through artificial intelligence in cases where a match position matches a previously experienced one. The system facilitates the communication of decisions (warnings / dismissals, penalties, offside, etc.) that need to be made for the position during the match. However, it is crucial to note that while the related patent document discusses integrating an artificial intelligence module into the VAR room's imaging system only during the duration of football matches, our patent, in the relevant section, focuses on using an artificial intelligence algorithm to verify the proper functioning of the VAR room's electronic components. Therefore, it should not be considered similar to any artificial intelligence algorithm making decisions during the live monitoring of football matches, as our patent does not involve any algorithm actively observing live matches.

[0020] Based on the known state of the technique, the purpose of the invention is primarily to carry out the functions of World Football Federation and European Football Federation, as mentioned in the explanations, using artificial intelligence algorithms. Brief Description of the Invention

[0021] One of World Football Federation's responsibilities is to improve and implement refereeing and fair play principles. The training of referees, performance evaluation, and the fair adjudication of decisions during matches have been a constant challenge for World Football Federation. Referee errors and controversial decisions resonate globally. This task requires the impartial and transparent assessment of referee performances, with the appointment of referees, assistant referees, and VAR room referees possessing appropriate qualifications for each match. In this context, the invention utilizes decision trees artificial intelligence algorithms to neutralize negative factors such as favoritism or nepotism, and to prevent instances of incompetence.

[0022] Another duty of World Football Federation is to enforce discipline and fair play rules. In other words, it is among World Football Federation's responsibilities to apply disciplinary rules among players and teams and uphold fair play principles. However, the consistent application of disciplinary rules and the encouragement of sportsmanship behavior can be challenging. Information regarding players, coaches, spectators, or referees violating disciplinary rules in football matches is typically recorded in official records and computer databases. Maintaining these records ensures the orderly and disciplined functioning of football organizations and assists in documenting any infractions. Data and documents related to violation cases are kept and, when necessary, reviewed by relevant sports federations or organizations. Subsequently, this information is used for the enforcement of disciplinary rules, determination of penalties, and ensuring accountability for the violators. These activities support the proper implementation of discipline and fair play principles in football. However, it is not always possible to verify this data. The invention proposes the use of artificial intelligence algorithms in continuous evaluation, utilizing the mentioned data to enforce disciplinary rules, reward teams and countries adhering to fair play principles, and conducting these activities based on concrete and transparent data through recommendation systems. Another duty of World Football Federation is international relations and cooperation. World Football Federation has undertaken the responsibility of collaborating with many countries and football federations worldwide. The maintenance of these collaborations can occasionally pose unique challenges due to the necessity of respecting diverse cultural and political differences. Each country has its own political and cultural contexts. Taking these diverse contexts into account, World Football Federation adapts its collaboration strategies with various country federations and teams. This requires a respectful approach to the type of political systems or cultural differences. For example, World Football Federation must respect the political system of any country and develop a collaboration strategy tailored to the specific needs within that context. Similarly, in regions with unique cultural contexts, it is necessary to consider these contexts to determine an appropriate collaboration strategy. In pursuit of this goal, the invention employs a supervised learning artificial intelligence algorithm.

[0023] Another responsibility of World Football Federation is the financial management of football. As World Football Federation needs to operate with substantial revenues and budgets, financial management is a sensitive task. It is essential to distribute revenues fairly and ensure financial stability. World Football Federation's financial management is a fundamental element for the international governance of football. By efficiently managing significant revenues, World Football Federation contributes to the development and sustainability of football. The key sources of World Football Federation's income and expenditures are as follows:

[0024] Income sources:

[0025] World Cup Revenues: World Football Federation's primary source of income is the World Cup tournaments. The World Cup generates substantial revenue through television rights, ticket sales, sponsorship agreements, and licensing deals. Since the World Cup is held every four years, each tournament ensures a significant income.

[0026] Sponsorship Agreements: World Football Federation secures substantial income through sponsorship agreements for the World Cup and other events. International companies pay large amounts to sponsor World Football Federation events and promote their brands in association with these organizations.

[0027] Television Rights: World Football Federation earns significant revenue from the television broadcasting rights of football matches. Millions of people worldwide watch World Football Federation events on their televisions or digital platforms, making television rights a major financial income source.

[0028] Licensing Agreements: World Football Federation generates income from the sale of football products, video games, and licensed merchandise worldwide. Football clubs, national teams, and players earn revenue by entering into such licensing agreements with World Football Federation.

[0029] Ticket Sales: Football enthusiasts who want to attend World Football Federation events pay substantial amounts to purchase tickets. Ticket sales for events like the World Cup are a significant source of income.

[0030] Income Amounts: World Football Federation's annual income amounts are largely dependent on the periods when World Cup tournaments are held and the durations of sponsorship agreements. Generally, World Football Federation's annual revenues reach billions of dollars. Especially during World Cup years, these revenues are much higher.

[0031] Evaluation and Distribution of Revenues: World Football Federation utilizes its earned revenues for the development of football, organizing and supporting international events. Some key points regarding the evaluation and distribution of these revenues include:

[0032] World Football Federation Operations: A portion of the revenues is used for World Football Federation's own operations and the organization of events. This includes financing the management of World Football Federation's central office, planning and executing events, and providing resources for staff and facilities.

[0033] Support for Member Countries: World Football Federation allocates a portion of its revenues to support the development of football in member countries. This involves improving football infrastructure, financing training programs, and expanding the grassroots level of football. World Football Federation Aid Projects: World Football Federation provides financial support to football projects in developing countries. These projects may include field construction, football schools, education programs, and more.

[0034] Football Development: A portion of World Football Federation's revenues is used to support the global development of football. This includes initiatives to popularize football, discover young talents, and ensure the sustainability of football through various programs.

[0035] Organizational Expenses: World Football Federation utilizes a portion of its revenues to organize and promote events. This includes operational expenses and promotional costs for events like the World Cup and other tournaments.

[0036] While World Football Federation carries a significant responsibility as the international manager of world football, managing financial operations such as revenues and expenses, it is susceptible to potential corruption in an environment with such a substantial financial turnover. The corruption scandal in 2015, which rocked World Football Federation, is a vivid example highlighting the potential for corruption in such a vast financial operation.

[0037] World Football Federation has implemented various reforms and transparency measures to combat such issues and preserve the integrity of sports. However, large-scale operations like financial transactions require careful oversight due to the influence of human factors. In this context, to effectively manage World Football Federation's financial operations and reduce costs, a combination of Audit and Business Process Automation algorithms is employed. This integrated approach tracks financial transactions while automating repetitive tasks to enhance overall efficiency. Various sources are used to increase the data source, including databases, web pages, and manually entered data. Additionally, the wealth and bank accounts of World Football Federation employees and their relatives, subject to decisions by the World Football Federation board or voluntarily sharing information, can be used as data sources. These additional data sources enhance the effectiveness of financial audits and provide valuable information to prevent potential corruption. Data Analytics and Recommendation System algorithms are also employed for this task, using the mentioned data to identify opportunities for increasing revenue and reducing expenses. The algorithms analyze financial trends, identify cost optimization opportunities, and assess risk factors. This enables World Football Federation to make financial decisions based on better-informed data and achieve sustainable financial success.

[0038] Another fundamental task of World Football Federation is the dissemination of football and the mission to create an impact in society. The key components of this task include:

[0039] Promoting Gender Equality: World Football Federation conducts various projects to support gender equality and promote the development of women's football. Steps are taken, such as providing financial resources for the development of women's football, establishing infrastructure, and organizing training programs for the development of women footballers.

[0040] Support for Disabled Football: World Football Federation carries out projects to popularize and support disabled football. Various initiatives, such as creating infrastructure to increase the participation of disabled footballers and organizing football training programs and tournaments for people with disabilities, aim to provide opportunities for them to engage in sports.

[0041] Education and Awareness: World Football Federation conducts educational and awareness campaigns to create positive impacts of football in society. Emphasizing values such as fair play, non-violence, and the unifying power of sports, especially among young people, encourages healthier sports engagement within the community.

[0042] Social Responsibility Projects: World Football Federation supports various social responsibility projects worldwide, believing that football can contribute to solving societal issues. Projects are financed to provide educational opportunities, offer health services, or contribute to the development of children and youth.

[0043] Popularizing Football: World Football Federation supports infrastructure projects and assists in the construction of football fields globally to popularize football. This allows more people to engage in sports and benefit from the unifying power of football within society. Through such projects and strategies, World Football Federation encourages the use of football as more than just a sport but as a positive agent for change that brings communities together. While facing different challenges in each country and community, World Football Federation collaborates globally to fulfill this mission, providing various resources and programs to enhance the societal impacts of football.

[0044] In the task of disseminating football and creating an impact in society, as seen, many activities involve tasks related to financing and infrastructure construction, which are areas covered by algorithms performing financial management tasks as described in the previous section. Tasks other than financing in the mission of disseminating football and creating an impact in society can be filtered as follows: Support for Women's Football and Gender Equality:

[0045] While World Football Federation plays a significant role in promoting gender equality and supporting women's football, it needs to evaluate whether women's football is adequately supported and assess whether the ratio of women's football to men's football is increasing. The impact of programs supporting women's football, the levels of participation of women footballers, and public interest should be part of this evaluation.

[0046] Support for Disabled Football and Societal Impacts:

[0047] World Football Federation conducts various projects to popularize and support disabled football. However, it should evaluate whether these projects genuinely increase the participation of disabled individuals and whether they act as a unifying force in societies. Additionally, these projects should be measured for potential impacts such as racism, conflicts, or injuries.

[0048] Education Opportunities and Social Responsibility Projects:

[0049] World Football Federation is committed to providing education opportunities and supporting social responsibility projects. However, it should evaluate whether these projects genuinely enable individuals from disadvantaged backgrounds to participate in education and whether issues like nepotism exist in these areas.

[0050] To assess the impacts of these mentioned tasks and provide answers to such questions, the invention utilizes decision tree algorithms. These algorithms evaluate the societal impacts and successes of these projects, taking into account different factors. Specifically, decision trees are created in different areas such as women's football, disabled football, education opportunities, and social responsibility projects. This enables a clearer understanding of the effects of these tasks on societies.

[0051] World Football Federation's mission goes beyond the success of football on the field; it also aims to make positive changes among communities. The digital evaluation of projects undertaken for this purpose helps determine better strategies and achieve better results.

[0052] Another task of World Football Federation is the use of technology and innovation. The task of utilizing technology and innovation in football governance aims to integrate technological advancements and ensure their fair use. This task represents a crucial aspect shaping the future of football and includes several key components:

[0053] VAR (Video Assistant Referee) System: World Football Federation is making significant efforts to integrate technological innovations like the VAR system into football governance. The VAR system aims to make referee decisions fairer and more accurate, particularly in critical moments such as controversial positions, offside decisions, and penalty calls. The VAR system adapts to the fast-paced nature of football, assisting in making decisions quickly.

[0054] Hawk-Eye and Other Technological Tools: World Football Federation employs technological tools like the Hawk-Eye for goal-line technology to enhance the accuracy of goal decisions. Such technological innovations are used to determine whether the ball has crossed the goal line and carry significant importance in major tournaments.

[0055] In the invention, the Support Vector Machine (SVM) algorithm is used to determine whether technological systems such as the VAR system and Hawk-Eye are functioning accurately.

[0056] Additionally, for World Football Federation's task in this regard, examining patents related to football and implementing applicable ones may be necessary. In the invention, data on patents related to football is collected by web scraping from the websites of countries' patent offices, and decision tree algorithms are used to identify patents related to football.

[0057] One of the important organizations collaborating with World Football Federation is the International Football Association Board (IF AB). IF AB, the International Football Association Board, is a global authority responsible for developing, changing, and overseeing the rules and regulations of football. IF AB determines the fundamental rules of football and works to ensure their appropriate application worldwide. In the invention, Natural Language Processing (NLP) and decision tree artificial intelligence algorithms work in an integrated manner to evaluate whether IFAB's rules are implemented, and to what extent.

[0058] Detailed Description of the Invention

[0059] The initial step of the decision tree algorithm is the data collection phase. This step begins with gathering the necessary data to evaluate referee performance. The collected data includes statistics from the past matches of each referee. Sample data encompasses factors such as cards, fouls, offside decisions, VAR usage, match results, and referee experience levels. Essentially, the following data is collected:

[0060] Match statistics: Basic match information is gathered for each match, including the match duration (in minutes), field dimensions (in meters), names of the teams playing the match, and match results.

[0061] Referee decisions: Referee decisions such as given cards (yellow and red cards), foul count, and offside decisions are quantified.

[0062] VAR usage: Decisions reviewed by the VAR system and their outcomes are recorded.

[0063] Referee experience: Data is recorded to numerically express each referee's experience, including the number of matches officiated, national and international experience (in years), education, and certifications.

[0064] Match results: The outcome and score of each match are recorded as mathematical data. Other variables: Factors such as weather conditions (temperature, wind speed), field conditions (grass type, field condition), and audience size are recorded in a mathematically representable form.

[0065] The mathematical expressions for the collected data are as follows:

[0066] Match statistics: Match duration (min): $M_field size (m): $S_b$, home team: $Home$, ayaw team: $Away$, resut: $Result$

[0067] Refree decisions: Yellow caed count: $Yellow_{card}$, red card count: $Red_{card}$, faul count: $F aul_{ count} $, pffside decision count: $Offs_{decision}$.

[0068] VAR usage: VAR usage (represented by 0 or 1):: $VAR_{usage}$. VAR Decision Result (represented by 0 or 1, whether it was changed or not): $(VAR_{result}$.

[0069] Referee experience:: Number of officiated matches: $Managed_{ match }$, national experience (Years): $National_{experience}$, international experience (Years): $ International) experience} $, education level (e.g., License or European Football Federation Referee Certificate): $Education_{level}$.

[0070] Match results: Won: \(Result_{won}\), lost: \(Result_{lost}\), draw: \(Result_{draw}\), Score (e.g., 2-1): \(Score_{home}, Skor_{away}\)

[0071] Other variables: Temperature (°C): \(Air_{temperature}\), wind speed (m / s): \(Wind_{speed}\), audience size: \(Audience_{number}\)

[0072] The obtained data is used to create an objective referee scoring model, which is fundamentally expressed as follows:

[0073] Referee performance score (\(P_h\)) calculation formula:

[0074] \[P_h = P_k + P_v - P_y\]

[0075] Here, \(P_k\) represents the performance score based on cards, \(P_v\) represents the performance score based on VAR usage, and \(P_y\) represents the experience and education-based performance score.

[0076] The performance score based on cards is calculated based on the yellow and red cards and foul counts given by referees:

[0077] \[P_k = -a \cdot Yellow_{card} - b \cdot Red_{card} - c \cdot Faul_{number}\] Here, \(a\), \(b\), and \(c\) are adjustable constants determining the impact of yellow cards, red cards, and foul counts on performance.

[0078] The performance score based on VAR usage evaluates the contribution of VAR decisions and is calculated as follows:

[0079] \[d \begin{cases} 1 & \text{if } VAR_{usage} = 1 \text{ and } VAR_{result} = \text{ "Changed"} \\ 0 & \text{ otherwise} \end{cases}\]

[0080] Here, \(d\) is a constant given when VAR decisions have a positive impact.

[0081] Experience and Education-Based Performance Score (\(P_y\)):

[0082] The experience and education levels of referees also influence performance. For instance, experienced and highly educated referees tend to have higher performance scores. This score is calculated based on experience and education levels:

[0083] \[P_y = e \cdot National_{experience} + f \cdot National_{experience} + g \cdot Education_{level}\]

[0084] Here, \(e\), \(f\), and \(g\) are constants determining the impact of experience and education levels on performance.

[0085] The second step of the decision tree algorithm involves the cleaning and preparation of the collected data. This step includes handling missing or incorrect data, converting categorical data into numerical form, and normalizing the data. The basic technical structure of this step is as follows:

[0086] Handling missing data: Missing data represents empty or null values in the dataset. Techniques for handling missing data include:

[0087] Removing missing data (\(\text{NaN}\)): Rows in the dataset with features containing missing data are excluded from the dataset.

[0088] Filling missing values with mean values (\(\text{Mean}\)): Missing values are filled with the mean of the respective feature. For example, a missing numerical data feature is filled with the mean of all other data for that feature.

[0089] Transformation of categorical data: Categorical data in the dataset, such as team names, is converted into numerical form. This involves transforming the data into a format that the decision tree algorithm can process. One-hot encoding: Categorical data is transformed into numerical form by creating a new feature for each category. Each category takes the value of 1 in its corresponding new feature, while others take the value of 0. For example, a category consisting of team names is represented with two new features: "Home" and "Away."

[0090] Data normalization: Data normalization ensures that data with different scales are in the same scale. This is important for the decision tree algorithm to function effectively. Data normalization is used to scale data to a specific range (usually between 0 and 1):

[0091] Min-max normalization: Data is rescaled based on the minimum and maximum values. Each feature is rescaled using the following formula:

[0092] \[X' = \frac{X - \min(X)} {\max(X) - \min(X)}\]

[0093] Here, \(X\) represents the original value of the feature, and \(X'\) represents the normalized value of the feature.

[0094] As a result of this step, the data is cleaned, converted into numerical form, and normalized. This enables the decision tree algorithm to perform better and prepares the dataset for analysis.

[0095] The third step of the decision tree algorithm involves the process of constructing a decision tree using the dataset. This step attempts to determine which features are the most distinguishing factors. Specifically, a decision tree is created to explain referee performance and identify important factors in match assignments. The technical and mathematical expressions of this step are as follows:

[0096] The third step of the algorithm is the decision tree construction step. The algorithm constructs a decision tree structure using the features and target variables in the dataset. At each node, it splits the dataset into subsets and makes decisions based on the best splitting criteria. The decision tree is essentially created using the following mathematical expressions:

[0097] Tree Structure ($A$): The structure of the decision tree is represented by branches starting from the root node. Each node is associated with a feature and a threshold value. For example, a node may express a rule such as "Yellow Card Count < 2." Node Splitting Criterion ($C_k$): The splitting of each node into subsets is based on a splitting criterion. In this context, Gini impurity criteria are utilized. The splitting criterion measures how homogeneous the dataset is at each node. The fundamental mathematical structure of Gini impurity can be explained as follows: Gini(D)=l-X(i=l) to (c) (pi)2 Where:

[0098] $Gini(D)$): Gini impurity value of the dataset.

[0099] $c$: Number of classes (e.g., two classes like whether a yellow card is given or not).

[0100] $p_i$: Probability of the data at the respective node belonging to the \( i \)th class. The Gini impurity value ranges from 0 to 1. A lower Gini impurity value indicates that the node is considered more homogeneous. The best split is determined by the lowest Gini impurity value. For example, if a node has data from four different classes with probabilities as mentioned, the Gini impurity calculation is as follows:

[0101] $p_l = 0.1$ (1. class)

[0102] $p_2 = 0.4$ (2. class)

[0103] $p_3 = 0.3$ (3. class)

[0104] $p_4 = 0.2$ (4. class)

[0105] Gini (D)=l-[(0.1)2+(0.4)2+(0.3)2+(0.2)2]

[0106] Gini(D)=l-[0.01+0.16+0.09+0.04]

[0107] Gini(D)=l-0.3

[0108] Gini(D)=0.7

[0109] The Gini impurity value for this node is calculated as 0.7. Nodes with lower Gini impurity values represent more homogeneous subsets and are preferred in the creation of the decision tree.

[0110] Decision Rules ($R$): The branches under each node are expressed with decision rules. For example, rules like "If Yellow Card Count < 2, go to the left branch; otherwise, go to the right branch" may exist.

[0111] Decision Tree Evaluation: The created decision tree is used to explain features and target variables in the dataset. The performance of the decision tree is evaluated using metrics such as accuracy, precision, and recall, especially for classification problems.

[0112] Decision Tree Pruning: After creating the decision tree, observations are made to check for overfitting or oversimplification issues. Adjustments can be made to the tree, branches can be pruned, or parameters can be tuned to address these problems.

[0113] As a result of this step, a decision tree is created and used to explain referee performance and important factors in match assignments. The decision tree represents relationships and decisions within the dataset, becoming a model that can be used for analysis.

[0114] The fourth step of the decision tree algorithm involves the data set splitting process during the construction of the tree based on a specific feature. This step determines which features will be used in each node and how the data will be best split at each node, enabling the growth of the tree.

[0115] Feature Selection: The selection of which feature to use for splitting the data set at each node is crucial. Feature selection is performed using the Gini impurity criterion. The following steps are taken for feature selection:

[0116] Evaluation of all features: All features in the current data set are considered.

[0117] Evaluation based on the splitting criterion: The splitting criterion is calculated for each feature.

[0118] Selection of the best feature: The feature with the lowest splitting criterion is chosen as the feature to be used in the node.

[0119] Data set splitting: The selected best feature is used to split the data set into subsets. Each subset is separated based on the selected feature value. This step is performed at each node of the tree, and the subsets are used to create new nodes.

[0120] Node Creation: After each splitting process, new nodes are created. Each node is associated with a feature and its value. For example, a node such as "Weather Condition = 'Rainy'" can be created. In this example, the node uses the weather condition feature to separate data with the value 'Rainy.' In other words, the "Weather Condition" feature in this node represents rainy days. The decision tree grows with various nodes using different features and values, dividing the data set into more homogeneous subsets. This function is applied to nodes created by all parameters.

[0121] Node Splitting Control: The tree-growing process continues until a certain condition is met (e.g., maximum depth or minimum sample size). Node splittings are stopped when the specified conditions are satisfied.

[0122] This step plays a crucial role in structuring the decision tree. Splitting each node separates the data set into more homogeneous subsets, enhancing the performance of the decision tree.

[0123] The fifth step of the decision tree algorithm is the decision and classification step. In this step, when the decision tree is complete, data progresses from the root of the tree downwards. At each node, it is compared with a specific feature value, and a branch is selected based on this value. In this way, classifications are made for the performance of each referee and match assignments.

[0124] Decision Path: Each example in the data set follows a decision path starting from the root node of the tree. While following the decision path, it encounters a feature value at each node and progresses based on the value of that feature. For example: At the root node: "Match Result = 'Won'"

[0125] At the second node: "Referee Experience (Years) < 5"

[0126] At the third node: "Yellow Card Count < 2"

[0127] This example is a data example following a decision path, and at the end of this example, a classification is made at a specific node.

[0128] Classification: Each data example is classified by the decision tree into a class or outcome. For example, for a referee's performance, classification results such as "High Performance," "Medium Performance," or "Low Performance" can be obtained. These classifications are determined at the end of the decision tree based on the feature values provided in a specific node and the journey.

[0129] This step represents the interpretation phase of classifying data based on the rules created by the decision tree.

[0130] The sixth step of the algorithm is the evaluation and adjustment of the model. In this step, the generated decision tree is tested to assess how well it fits the data. The accuracy and performance of the model are observed. If the model exhibits low performance, adjustments are made to the tree size or other parameters in an attempt to enhance the model. The basic structure of this step is as follows:

[0131] Model testing: The created decision tree is evaluated on a separate test dataset. The test dataset is used to measure the model's performance, comparing the model's output ($\hat{y}$) with the actual labels ($y$).

[0132] Performance measurement: The model's performance is evaluated using various criteria, including:

[0133] Accuracy = Correct Predictions / Total Data Count

[0134] Precision: (Precision = TP / (TP+FP)

[0135] Recall: Recall=TP / (TP+FN)

[0136] These criteria are calculated using mathematical expressions involving True Positives ($TP$), False Positives ($FP$), and False Negatives ($FN$).

[0137] Model improvement: If the model shows low performance, adjustments can be made to the tree size or other parameters. Attention is particularly paid to overfitting or underfitting issues, and necessary adjustments are made, such as modifying the maximum depth of the tree or splitting criteria.

[0138] Model evaluation: The final version of the model is re-evaluated on the test dataset. The impact of changes on the model's performance is observed, and necessary adjustments are made until a satisfactory result is achieved.

[0139] This step aims to assess how well the model works mathematically after its creation and make improvements if necessary.

[0140] The seventh step of the decision tree algorithm is the match assignments and performance evaluation step. The generated decision tree provides guidance on which referees should be assigned to each match. Additionally, referees' performance is regularly evaluated, and these evaluations are used to update the decision tree. This ensures fair match assignments, taking into account the merit of referees. The sub-steps forming this step are essentially as follows:

[0141] Match assignments: The decision tree provides guidance on which referees should be assigned to each match. Starting from the tree's root, an appropriate node is selected based on the match's characteristics and the referees' performance. This node determines which referees should be assigned to the match. Performance evaluation: Referees' performance is regularly assessed at defined intervals. These evaluations are based on predetermined performance criteria, focusing on factors such as the accuracy of referee decisions, the use of VAR, the frequency of issuing cards, and similar factors. Performance evaluations help objectively measure referees' current performance.

[0142] Updating the decision tree: Referees' performance evaluations are used to update the decision tree. If a referee's performance is low or does not meet certain criteria, the decision tree takes these pieces of information into account to make more suitable decisions in future match assignments. This ensures fair match assignments by considering referees' merit and performance.

[0143] This step aims not only to guide match assignments using the decision tree but also to continuously monitor and improve referees' performance.

[0144] These steps describe the use of decision trees in the assessment of referee performance and match assignment processes. Decision trees offer an objective and data-driven approach, ensuring fair and unbiased referee assignments. This system provides significant advantages in championships organized by the World Football Federation.

[0145] Neutrality of referee decisions: The decision tree algorithm relies on objective criteria for referee assignments, providing equal opportunities to every referee. This eliminates the influence of negative factors such as favoritism, personal relationships, or nepotism on referee assignments. Each referee is evaluated and assigned based on their performance.

[0146] Improvement of referee performance: The decision tree algorithm regularly monitors and evaluates referees' performance, assisting them in becoming better. Performance evaluations help identify deficiencies and can be used to implement training or development programs as needed.

[0147] Prevention of scandals and controversies: Scandals or controversies arising from referee decisions can be a significant issue for sports organizations. The decision tree algorithm makes decisions based on objective criteria, helping prevent controversies and legal processes related to referee decisions. This way, the World Football Federation can maintain the reputation of sports organizations. Data-driven performance measurement: The decision tree algorithm provides performance measurement based on data. This includes objective data such as cards, fouls, offside decisions, VAR usage, and match results. Referees' performance is mathematically evaluated using this data.

[0148] In conclusion, using the decision tree algorithm can help the World Football Federation manage referee performance in a fair and transparent manner in championships. This approach allows sports organizations to avoid scandals and controversies, enhance the credibility of football, and gain the trust of fans. The method stands out as a robust tool to ensure fair and exciting football worldwide.

[0149] The first step of the recommendation system algorithm is the data collection and preparation phase. This step involves gathering and preparing the data required for the recommendation system algorithm. The process of creating a dataset for the algorithm to operate is carried out using data obtained from various sources related to football matches and disciplinary rules.

[0150] Data sources and collection: In this sub-step, sources of data related to football matches and disciplinary rules are identified. These sources include match compositions, game statistics, referee decisions, violation reports, and other relevant sources. The data collection process encompasses the following types of data for each match:

[0151] Match information: Basic match details such as date, time, location, home team, and away team information are collected.

[0152] Game statistics: Events that occur during the match, player statistics, goal count, foul count, offside decisions, cards, and score-related game statistics are recorded. Referee decisions: Decisions made by referees, including yellow cards, red cards, free kicks, and penalties, are included in the dataset.

[0153] Violation reports: Reports related to violations of disciplinary rules, including player, coach, or spectator violation reports, are obtained.

[0154] Formatting and structuring data: After the data collection process, the obtained data is transformed into an appropriate format and structure. This step ensures that the data becomes processable and facilitates the algorithm's access to the data. Data cleaning: Missing or erroneous entries in the dataset are identified and corrected. For example, appropriate values are assigned in place of missing data, and meaningless entries are cleaned.

[0155] Data transformation: Categorical data (such as team names) is transformed into numerical values, making the data analyzable.

[0156] Data normalization: Numerical data undergoes normalization to ensure that all numerical features are on the same scale, making the data comparable and processable.

[0157] Creating data structures: Data is transformed into a structure that the algorithm can process, organized in the form of tables or matrices.

[0158] This step facilitates the algorithm's access to data and processing in later steps.

[0159] The second step of the recommendation system algorithm is the data feature engineering step. This step involves transforming the dataset into a format that the model can understand, addressing missing data, and creating new features. It essentially consists of the following sub-steps:

[0160] The numerical transformation of categorical data: Categorical data is converted into processable numerical values by the model. This transformation is achieved through the One-Hot Encoding method, where each categorical feature is transformed into a vector consisting of Os and Is.

[0161] $Category_i$: Indicates the membership status in the second category (1 : If the sample belongs to this category, 0: Otherwise).

[0162] Handling missing data: Techniques are employed to address missing data, such as removing missing data, or filling them with mean, median, or nearest neighbor values.

[0163] $Missing_i$: The $i$-th instance of missing data (1: If it is missing, 0: Otherwise).

[0164] Creating new features: New features are introduced to add more meaning to the dataset. This is achieved by performing operations between features or using specific combinations. For example, a new feature can be created by mathematically summing two features:

[0165] $New_feature_i = Featurel_i + Feature2_i$ Such operations are used to provide additional information to the dataset.

[0166] Scaling features: Feature scaling is applied to ensure that different features are on the same range or scale. Scaling is performed using the mean and standard deviation of features.

[0167] $Scaled_feature_i = \frac{Feature_i - Mean}{Standard\_Deviation}$

[0168] This step plays a critical role in enhancing the quality of the dataset and improving the model's performance.

[0169] The third step of the algorithm is the creation of the recommendation system. In this step, the recommendation system model is built using the data. The model is utilized to reward behaviors compliant with disciplinary rules and detect those violating the rules. The basic sub-steps are as follows:

[0170] Data representation: Data used for the recommendation system model is represented in the form of a user-item interaction matrix.

[0171] \[R \in \mathbb{R}A{m \times n}\]

[0172] Here, \(R_{ij }\) represents the interaction of user \(i\) with item \(j\), where \(m\) is the number of users, and \(n\) is the number of items."

[0173] Selection of recommendation model: Two fundamental types are considered for selecting the recommendation system model: collaborative filtering (CF) and content-based (CB) models. For the task described in the invention, collaborative filtering is employed.

[0174] Collaborative Filtering (CF) Model: Recommendations are generated by calculating user similarity and item similarity. Cosine similarity metric is used for user similarity, denoted by $Sim(u, u')$, and item similarity is similarly calculated, denoted by $Sim(i, i')$.

[0175] Model training: The selected recommendation model is trained using training data. The collaborative filtering model is trained using user similarity matrices and interaction data.

[0176] $P_{u} = R_u \cdot Sim(u, u')$: Computes the weighted ratings of a user with similar users.

[0177] $P_i = RAT_i \cdot Sim(i, i')$: Computes the weighted ratings of an item with similar items. Generating recommendations: After training, the model creates recommendations for users. The recommendation process is used to select items that users have not interacted with or rated before. Mathematically, the probability scores of users for items are calculated, and items with the highest scores are recommended.

[0178] $R_{ui} = P_{ui}$: Computes the ratings for each item for the user.

[0179] $\hat{R}_{ui} = \text{arg}\max_i R_{ui}$: Recommends the item with the highest rating to the user.

[0180] These steps contribute to the creation of the recommendation system model, encouraging users to adhere to disciplinary rules and promoting fair play principles in football organizations.

[0181] The fourth step of the recommendation system algorithm is the training and evaluation of the model, with one of the sub-steps being the evaluation metrics. The performance of the model is assessed using various metrics to measure how well it operates. In this context, the average absolute error metric is employed for this purpose.

[0182] Average Absolute Error (MAE): Measures how much deviation the model's predictions show from the actual values. MAE takes the average of the absolute values of each prediction error.

[0183] \[ MAE = \frac{l }{n} \sum_{i=l }A{n} |y_i - \hat{y }_i| \]

[0184] This metric aids in quantitatively assessing the performance of the model and is used to enhance its calibration.

[0185] Cross-validation: This sub-step is utilized to reliably evaluate the model's performance. The training data is divided into training and validation sets. The model training and evaluation process is repeated multiple times, each time using different data partitions. This is crucial to test the model's generalization ability.

[0186] This step encompasses the evaluation of the recommendation system model's training and performance. The ability of the model to provide recommendations for users and items is measured and improved through this evaluation process.

[0187] Recommendation system algorithm's fifth step is the application and monitoring of recommendations. This stage encompasses the processes of monitoring, implementing, and tracking disciplinary and fair play rules following the creation of the recommendation system model.

[0188] Monitoring Disciplinary and Fair Play Rules:

[0189] Data Monitoring: The data monitoring process begins by utilizing the recommendations generated by the model. This process involves tracking how the recommendations and predictions calculated by the model align with real-world events. Data monitoring is used to determine whether possible rules are being violated.

[0190] Definition of Rules: Disciplinary and fair play rules required for various football organizations, primarily including FIFA and UEFA, are defined. These rules regulate the behavior of athletes, teams, referees, and spectators. Violation of these rules is among the situations documented and detected by the recommendation system.

[0191] Detection of Rule Violation: The recommendation system algorithm continuously monitors user and item ratings calculated by the model. If data indicating a violation of rules by players, teams, or other stakeholders is recorded, the recommendation system detects such situations.

[0192] Implementation of Recommendations:

[0193] Incentivizing Rule-Compliant Behavior: The recommendation system algorithm is utilized to reward athletes, teams, or other stakeholders adhering to the rules. Rewarding may involve offering advantages to those exhibiting rule-compliant behavior over a specific period. These behaviors contribute to promoting sportsmanship.

[0194] Monitoring Situations of Rule Violation: The recommendation system algorithm monitors situations where rules are violated and creates records for further investigation of such cases. It provides information to organizations when necessary steps need to be taken against violators.

[0195] Monitoring Model Performance:

[0196] Performance Metrics: The performance of the recommendation system algorithm model is evaluated using specific metrics. These metrics include factors such as the accuracy, specificity, precision, and user satisfaction of recommendations. Model Updates: Performance monitoring allows continuous improvement of the model. The model can be updated with new data, or hyperparameters can be adjusted, facilitating the generation of better and more reliable recommendations.

[0197] Discipline penalties and fair play incentives in football organizations aim to ensure consistency and fairness in every match and football event. The recommendation system algorithm contributes to fulfilling these crucial tasks. It provides a transparent framework for players, coaching staff, team managers, spectators, and fans, ensuring that everyone is confident that the rules and fair play principles of the sport are correctly enforced.

[0198] This system facilitates the identification of players or teams violating disciplinary and fair play rules and the application of necessary sanctions when needed. However, it also rewards individuals and teams adhering to the rules, promoting sportsmanship and commitment to fair play principles. In this way, the recommendation system not only encourages compliance with rules in football organizations but also helps keep the spirit of fair play alive.

[0199] Moreover, this system leaves no room for doubt in the minds of football players, coaching staff, team managers, spectators, and fans. It demonstrates that decisions regarding discipline penalties and fair play incentives are made entirely fairly, bringing more transparency to football organizations and enhancing the credibility of the sport. Thus, all stakeholders can fully enjoy the beauty of football without any reservations.

[0200] In the invention, a supervised machine learning artificial intelligence algorithm is employed to implement the World Football Federation's international Relations and Cooperation mission.

[0201] The first step of the algorithm is the data collection and preparation phase. This step involves gathering data related to relationships and collaborations with different countries and football federations. The data is obtained from various sources such as history, diplomatic relations, football organizations, and previous collaboration examples. The data sources (S) represent the number of accessible or required data sources by the World Football Federation. Each source is identified with an index (1 < i < S). The data types (T) form a set that encompasses the types of data to be collected. Data types can include text (such as diplomatic correspondences), numerical (such as statistical figures), categorical (such as types of political regimes), or other formats. Each data type is identified with an index (1 < j < T). The collected data (D): Data collected from each source is represented within a dataset. This is represented as D[i, j] referring to specific data points associated with a particular source and data type. Data sampling (N): It is often impractical to collect all data, so data sampling is conducted, with N specifying the size of this sampling. Data storage (V): The collected data can be stored in a database or another storage format, ensuring access to data and maintaining data integrity for subsequent processes. This system initiates the foundational database creation and data preparation processes for the subsequent data analysis and modeling stages.

[0202] The next sub-step of the data collection phase is the data transformation step. This step involves transforming different data types into standard data structures, as data can come in the form of text, numbers, categories, and other formats. Therefore, transforming the data into a standard format is necessary to process and analyze the data. The components of this sub-step are as follows:

[0203] Data type transformation (T[i, j]): T[i, j] indicates the transformation of the data type for D[i, j]. For example, text data can be transformed into numerical data, or numerical data can be converted into categorical labels.

[0204] Data standardization (S[i, j]): Some data may exist on different scales. Data standardization is used to transform these data into similar scales.

[0205] Handling missing data (M[i, j]): Data sets may contain missing values. This substep addresses and handles missing data by filling them in or processing them appropriately.

[0206] Feature engineering (E[i, j]): This sub-step involves creating new features or variables. Feature engineering enables better analysis of data and improvement of results.

[0207] Data structure transformation (V[i, j]): Data is transformed into an appropriate data structure to make it processable. The second step of the algorithm is the feature engineering step. Although this step serves a different scenario in supervised learning algorithms compared to the recommendation system algorithm mentioned earlier, it will not be reiterated here to avoid unnecessary extension of the patent document, as it fundamentally includes the same mathematical structure. Essentially, this step performs tasks such as making the dataset more manageable, addressing missing data, transforming categorical data appropriately, and creating new features to enhance the model's performance.

[0208] The third step of the supervised learning algorithm, "labeling and target definition," is a crucial step for the success of the project. This step explains how the data is labeled and how the targets are determined. Its sub-steps are fundamentally as follows:

[0209] Labeling: In this step, examples in the dataset need to be labeled to determine with which countries or football federations collaboration should be established. Labels are created based on previous collaboration successes, political factors, and other determining criteria. These labels for examples are part of the classification problem. Mathematically, a label (usually 0 or 1) representing the class to which each example belongs is assigned.

[0210] This process can be represented mathematically with an equation like the following:

[0211] Yi=f(Xi)

[0212] Here:

[0213] Yi represents the target variable with label 0 or 1 for the \(i\)-th example.

[0214] Xi is a vector containing the features of the \(i\)-th example. f represents the function that performs labeling. This function makes decisions using previous collaboration successes, political factors, and other data points.

[0215] Target Definition:

[0216] Yi=g(Xi)

[0217] Here: represents the target variable for the \(i\)-th example (e.g., "Initiate new collaboration" or "Strengthen existing collaboration").

[0218] Xi is a vector containing the features of the \(i\)-th example. g represents the function that determines these targets. This function guides the process of target definition based on the goals and priorities of the project.

[0219] This step provides a more detailed and mathematical expression of how labeling is done and how project goals are defined. Labeling and target definition for each example are fundamental components of the learning process in a machine learning algorithm.

[0220] The fourth step in the supervised machine learning artificial intelligence algorithm is the 'Creation of Training and Test Data.' In this stage, data is appropriately partitioned into training and test subsets for the machine learning model's training and evaluation. This step is crucial for assessing the overall performance of the machine learning model and detecting issues such as overfitting or underfitting. The fundamental mathematical structure of this step is as follows:

[0221] Creation of Training and Test Data: In this phase, the dataset is randomly divided into two subsets, namely training data and test data.

[0222] Training Data: This subset is utilized for initiating the learning process of the model. Training data constitutes a significant portion of the dataset. Mathematically, training data can be expressed as follows:

[0223] Dtraining={(Xi,Yi)}for i=l,2, .. ,,Ntraining

[0224] Here:

[0225] \[D_{\text{train}}\] represents the training dataset.

[0226] $(X_i, Y_i)$ represents the $i$-th example and its corresponding target label. $N_{\text{train}}$ represents the number of examples in the training dataset.

[0227] Test Data: This is a subset of data used to evaluate the model's performance and test its generalization ability. Test data constitutes a smaller portion of the dataset. Mathematically, test data can be expressed as follows:

[0228] Dtest={(Xi,Yi)}for i=l ,2, .. ,,Ntest

[0229] Here:

[0230] $D_{\text{test}}$ represents the test dataset.

[0231] $(X_i, Y_i)$ represents the $i$-th example and its corresponding target label. $N_{\text{test}}$ represents the number of examples in the test dataset. This step determines the data partitions required for the training and testing of the model. Ensuring that data is randomly selected is a good practice to ensure fairness in the learning and testing processes.

[0232] The fifth step of the supervised learning algorithm is the 'Model Selection' step. In this context, the decision tree algorithm is employed in the invention.

[0233] For the management of international relations and cooperation at the World Football Federation, collaboration with countries in the database is essential. Decision-making regarding which countries to collaborate with and determining the level of collaboration is crucial. For this purpose, the decision tree algorithm in supervised machine learning artificial intelligence is employed.

[0234] Problem Type: In this step, countries are classified based on collaboration levels, such as "Low Collaboration," "Medium Collaboration," and "High Collaboration." Data Type: Diplomatic correspondence, archives of official sports organizations, media sources, and social media platforms are the sources of data in this step. This data comes in various formats, including text, numbers, and categories.

[0235] Model Complexity: The decision tree algorithm is known for its simplicity and high explainability. Additionally, parameters such as the depth of the tree can be adjusted to control the model's complexity.

[0236] Performance Metrics: Metrics such as accuracy, precision, or Fl score can be used to evaluate the model's performance.

[0237] Model Parameters: Adjusting the model's parameters, including classification levels and the depth of the decision tree, helps the model better fit the data. Parameters are tailored to the data and problem type for improved model performance.

[0238] In summary, this step employs the decision tree algorithm to make better decisions in determining the countries the World Football Federation will collaborate with and the levels of collaboration. This contributes to creating data- driven strategies in international relations and cooperation management. The technical and mathematical structure of the decision tree algorithm in the section on implementing the principles of refereeing and fair play by the World Football Federation using artificial intelligence has already been explained and is not reiterated here for brevity.

[0239] The 6th step of the supervised learning algorithm is the "Model Training" phase. This stage involves the learning of the selected model using training data. In the context of this invention, this model is utilized by the World Football Federation to accurately predict the countries for collaboration and their collaboration levels. The process aims for the model to understand the data, learn the relationships between input features and output labels, and make predictions. The sub-steps that constitute this step are fundamentally as follows:

[0240] Data Feeding: Before commencing model training, the training data serves as the input for the model. Each data example comprises features and target variables. Features encompass various data types, such as the intensity of diplomatic correspondences associated with countries, participation in official sports organizations, media coverage, etc. Target variables represent collaboration levels.

[0241] Data Set Splitting: The data set is divided into two subsets: training data and test data.

[0242] Training Data:

[0243] $X_{train}$: Matrix containing the features of the training data.

[0244] $y_{train}$: Vector containing the target variables of the training data.

[0245] Test Data:

[0246] $X_{test}$: Matrix containing the features of the test data.

[0247] $y_{test}$: Vector containing the target variables of the test data.

[0248] Feature Matrix: Matrices representing different information in the data set. Each feature matrix contains a row and column for each feature. For training data: $X_{train} = \begin{bmatrix}

[0249] X_{train} [1, 1] & X_{train}[l,2] & \cdots & X_{train}[l,m] \

[0250] X_{train}[2,l] & X_{train}[2,2] & \cdots & X_{train}[2,m] \

[0251] \vdots & \vdots & \ddots & \vdots \

[0252] X_{train}[n,l] & X_{train}[n,2] & \cdots & X_{train}[n,m] \end{bmatrix}$ For Test Data:

[0253] $X_{test} = \begin{bmatrix}

[0254] X_{test} [1, 1] & X_{test} [1,2] & \cdots & X_{test}[l,m] \

[0255] X_{test} [2, 1] & X_{test} [2,2] & \cdots & X_{test}[2,m] \

[0256] \vdots & \vdots & \ddots & \vdots \

[0257] X_{test} [k, 1] & X_{test} [k,2] & \cdots & X_{test}[k,m]

[0258] \end{bmatrix}$

[0259] Here, $n$ represents the number of training examples, $k$ the number of test examples, and $m$ the number of features.

[0260] Target Variable Vector: Target variables contain the model's learning outputs or the values it attempts to predict. Vectors are expressed as follows:

[0261] For Training Data:

[0262] $y_{train} = \begin{bmatrix} y_{train}[l] \ y_{train}[2] \

[0263] \vdots \ y_{train}[n] \ end{bmatrix}$

[0264] For Test Data:

[0265] $y_{test} = \begin{bmatrix} y_{test}[l] \ y_{test}[2] \

[0266] \vdots \ y_{test}[k]

[0267] \end{bmatrix}$

[0268] These expressions illustrate the fundamental components and decomposition of the data. The model learns from these data to make predictions.

[0269] Learning Process: The algorithm attempts to understand the relationships between data using training data. In this process, decision tree algorithms start from the root node, dividing the data and making decisions in the subnodes to reach a conclusion. The decision tree learns to predict collaboration levels by splitting the data into smaller subgroups.

[0270] Fitting to Data: The algorithm adapts to the training data, creating a structure that best fits the training data. It forms a structure that connects input features to target variables, aiming to find the optimal model parameters for predicting the fundamental target data. This is an optimization problem, and the method used in the invention is the least squares method.

[0271] The fitting step adjusts the model's parameters to minimize the loss function (L) and measure how much the model's predictions deviate from actual data:

[0272] L(0) = S(yi - f(xi; 6))A2

[0273] Where:

[0274] L(9) is the loss function.

[0275] 9 represents the model's parameters. yi is the real target value from the dataset. f(xi; 9) is the value predicted by the model and depends on the parameters \(\theta\).

[0276] S represents the sum and takes the sum over all examples in the dataset.

[0277] Loss Function: The model compares predictions to real values and calculates a loss function that measures how well the model performs. The goal is tominimize this loss.

[0278] Optimization Algorithm: The model uses an optimization algorithm to minimize the loss function. In this case, the decision tree algorithm constructs decision trees to perform this optimization.

[0279] Model Validation: During the training process, the model evaluates its performance using a validation dataset. This dataset consists of examples that the model has not seen before. Evaluating performance on unseen data helps detect overfitting or underfitting issues.

[0280] When the model training process is complete, it provides predictions on which countries the World Football Federation should collaborate with and the levels of collaboration. These predictions are generated based on data and developed as a result of the model's learning process. In this way, the World Football Federation successfully takes the step of model training to make more data-driven, knowledge-based, and effective decisions in international relations and collaboration management.

[0281] The 7th step of the supervised learning algorithm is the "Model Evaluation" phase. This step is crucial for objectively measuring and evaluating the model's performance. It consists of the following sub-steps:

[0282] Use of Test Data: Independent test data, not used during the training phase, is employed to evaluate the model's performance. Test data comprises examples that the model has not encountered before.

[0283] Selection of Performance Metrics: Appropriate metrics are selected to measure the model's performance. These metrics may vary depending on the problem type and objectives. In this invention, a combination of precision and recall metrics is used to predict collaboration levels.

[0284] Evaluation of Model Performance: The model makes predictions on test data and compares these predictions to actual values. The model's ability to make correct classifications is evaluated based on the selected metrics.

[0285] Comparative Evaluation: The model's performance using different metrics is examined. The balance between precision and recall reflects the model's ability to predict collaboration levels. The balance between precision and recall is measured using the Fl score metric:

[0286] F1=(2-P-R) / (P+R)

[0287] Where:

[0288] P (Precision) represents the ratio of true positive predictions to the total predictions.

[0289] R (Recall) represents the ratio of true positive predictions to the total true positives.

[0290] When developing a model to predict collaboration levels for the World Football Federation, striking a balance is essential. The model's goal is not only to avoid missing collaboration opportunities but also to avoid false collaboration opportunities. Cross-Validation: Cross-validation is used to assess the model's stability and generalization ability. It involves splitting the data into different parts and evaluating the model's performance on different test-training splits.

[0291] The supervised learning algorithm developed in this invention provides the World Football Federation with an effective tool for data analysis and decision-making in international relations and collaboration management. The algorithm's application steps include data collection and preparation, data transformation, feature engineering, labeling and target determination, creating training and test data, model selection, model training, and model evaluation. With this algorithm, the World Football Federation can create data-driven strategies, better predict which countries to collaborate with, and enhance collaboration levels. Additionally, the model can make decisions using a variety of data sources, such as diplomatic correspondences, sports organization participation, and other factors. The use of the algorithm allows the organization to manage international relations more effectively and make data-driven decisions, contributing to improved performance and efficiency. Furthermore, the model is objectively evaluated using performance metrics such as accuracy, precision, and Fl score. This evaluation is crucial to measure the model's reliability and the effectiveness of the decision tree algorithm in predicting collaboration levels based on data. In conclusion, the developed supervised learning algorithm provides a robust tool for enhancing the World Football Federation's international collaboration management, fostering more knowledge-based and data-driven decision-making.

[0292] The fundamental structure of algorithms performing the financial duties of the World Football Federation is as follows:

[0293] Step 1 : Data Collection and Organization

[0294] The initial step involves gathering and organizing the necessary data to analyze the financial data and business processes of the World Football Federation. This data includes financial reports, transaction history, budget information, personnel details, and other financial documents.

[0295] Identification of Data Sources: To examine and audit the financial operations of the World Football Federation, data sources to be utilized are determined. Among these sources are financial reports, account summaries, budget documents, financial transaction records, personnel data, contracts, and other financial documents.

[0296] Step 1.1 : Identification and Scope Definition of Data Sources

[0297] Financial Statements:

[0298] Fundamental sources for evaluating the financial health and performance of the World Football Federation include financial reports such as the balance sheet, income statement, and cash flow statement.

[0299] 1.1.2. Account Summaries:

[0300] Summaries of the World Football Federation's bank accounts, including balances and transaction history, are utilized to monitor bank activities.

[0301] 1.1.3. Budget Documents:

[0302] Annual and projection budgets, income and expense estimates, and budget performance reports provide key reference points for assessing the financial goals and management of the World Football Federation.

[0303] 1.1.4. Financial Transaction Records:

[0304] Records from the systems used by the World Football Federation to record financial transactions include detailed transaction data such as invoice details, payment instructions, and information related to revenue sources.

[0305] 1.1.5. Personnel Data:

[0306] Data related to the financial operations of World Football Federation employees, including salaries, personal details, and information related to financial transactions and personnel management, such as hiring and termination details.

[0307] 1.1.6. Contracts:

[0308] Contracts that impact or determine the financial transactions of the World Football Federation, such as sponsorship agreements and licensing contracts, are essential sources for financial analysis.

[0309] 1.1.7. Other Financial Documents: This category encompasses other crucial documents related to the financial operations of the World Football Federation. Documents such as regulations, accounting policies, and tax returns fall into this group.

[0310] 1.1.8. Scope Definition of Data Sources:

[0311] Detailed information about the type of data contained in each data source is specified. For example, financial reports encompass the assets, liabilities, revenues, and expenses of the World Football Federation, while budget documents cover annual budget predictions. At this stage, the scope and function of each data source are clearly determined.

[0312] This sub-step defines in detail the data sources to be used to monitor and audit the financial operations of the World Football Federation and the information they contain. This forms the basis for the subsequent data collection and organization processes.

[0313] 1.2. Data Collection Process:

[0314] The data collection process is initiated from the identified data sources. This process is carried out through methods such as automated data retrieval systems, access to databases, scanning software, or manual data entry.

[0315] Step 1.2: Data Collection Process and Access to Data Sources

[0316] 1.2.1. Automated Data Retrieval Systems:

[0317] Automated data retrieval systems are used to collect the financial data of the World Football Federation. These systems automatically access predefined data sources and retrieve updated data at specific intervals. Data that is continuously updated, such as bank account summaries and transactions, is obtained through this method. Mathematically, these systems connect to data sources through APIs (Application Programming Interface) and queries. Preferably, a Python program performs the data retrieval operations in the invention. The basic structure of the mentioned program code is as follows: import requests import j son import matplotlib.pyplot as pit

[0318] # Necessary information for accessing the API api url = "https: / / api.worldfootballfederationfmance.com" # Example API URL api key = "API KEY GOES HERE" # API key obtained from the World Football Federation

[0319] # Sending a request to the API headers = {

[0320] "Authorization": f 'Bearer {api_key}"

[0321] } response = requests.get(api_url, headers=headers)

[0322] # Retrieving data from the API in JSON format if response. status code == 200: data = response.json() else: data = None

[0323] # Processing and analyzing the data if data:

[0324] # Example 1 : Overview of the data total records = len(data) average value = sum(record["value"] for record in data) / total records max value = max(data, key=lambda record: recordfvalue"]) min value = min(data, key=lambda record: recordfvalue"]) print("Data Overview:") print(f 'Total Record Count: {total records}") print(f Average Value: {average_value:.2f}") print(f 'Highest Value: {max_value['value']} (Date: {max_value['date']})") print(f 'Lowest Value: {min_value['value']} (Date: {min_value['date']})")

[0325] # Example 2: Selecting records that meet a specific criterion selected records = [record for record in data if record["criterion"] == "specific_value"]

[0326] # Example 3: Visualizing the data x = [record["x"] for record in data] y = [recordfy"] for record in data] plt.scatter(x, y) plt.xlabelf'X Axis") plt.ylabelf'Y Axis") plt.title("Data Distribution") plt.show()

[0327] # Example 4: Data filtering and sorting filtered data = [record for record in data if recordfvalue"] > 50] sorted data = sorted(data, key=lambda record: recordfdate"]) print("Records Meeting a Specific Criterion:") print(selected records) print("Result of Data Filtration:") print(filtered data) print("Result of Data Sorting:") print(sorted data)

[0328] 1.2.2. Database Access:

[0329] Financial data of the World Football Federation is typically stored in a centralized database. This step ensures secure access to databases. Specific data is retrieved using SQL queries and prepared for analysis. SQL queries select particular columns and / or tables from the database, transferring this data to a DataFrame.

[0330] 1.2.3. Web Scraping:

[0331] Financial transaction data of the World Football Federation is sometimes published on websites or different platforms. In such cases, web scraping tools are employed to scan websites and automatically extract relevant financial data. Technically, software used for web scraping analyzes HTML documents, accesses specific tags or data fields, and retrieves this information. Essentially, a web scraping program is structured as follows; import requests from bs4 import BeautifulSoup import sqlite3

[0332] # Sending an HTTP request to the website url = 'https: / / www.worldfootballfederationfmancialreport.com' # Example website address response = requests. get(url)

[0333] # Checking the HTTP response if response. status code == 200:

[0334] # Analyzing the HTML content soup = Beautiful Soup(response.text, 'html. parser')

[0335] # Example: Extracting data with a specific tag financial data = soup.find('div', class_='financial-data') if financial data:

[0336] # Creating a connection to the database or using an existing one conn = sqlite3.connect('WorldFootballFederation_financial_data.db')

[0337] # Creating a cursor to operate on the database cursor = conn.cursor()

[0338] # Processing the data and writing it to the database financial data text = financial_data.text.strip() cursor. execute(" INSERT INTO financial data (data) VALUES (?)",

[0339] (financi al data text, ))

[0340] # Saving changes and closing the connection conn.commit() conn.close() print("Financial data has been saved to the database.") else: print("Financial data not found.") else: print("Page could not be loaded. HTTP status code:", response. status code)

[0341] 1.2.4. Manual Data Entry:

[0342] Manual data entry may be necessary to access certain specific data. This is particularly relevant in cases involving special contracts, contract details, or manually maintained records. The process of manual data entry must be carried out with precision and should include controls to minimize the possibility of any errors.

[0343] 1.2.5. Infrastructure of the Data Collection Process:

[0344] This process encompasses mathematical and software elements such as data extraction methods, query structuring, access to databases, and the principles of operation of scanning software. Mathematically, the structures of queries used in data extraction processes define data transformations and filtering. On the software side, programming languages and libraries such as Python, R, or Java can be employed. As illustrated in the example, the fundamental structure can utilize Python for web scraping using the Beautiful Soup and Requests libraries. Similarly, software tools like JDBC (Java Database Connectivity) or SQLAlchemy can be utilized for database access.

[0345] This step involves the technical knowledge and software required to successfully obtain and analyze the financial data of the World Football Federation.

[0346] 1.3. Definition of Data Types:

[0347] The types of collected data are identified, including numerical data, text data, date-time information, and categorical data, among other types.

[0348] 1.3.1. Numerical Data:

[0349] Numerical data comprises quantifiable values that can undergo mathematical operations. In the financial data of the World Football Federation, numerical data includes values such as revenue (R), expenses (E), ticket sales revenue (T), and the value of sponsorship agreements (S). Generally, these data are represented as processable numbers and are used in statistical analyses. For example, the sum of revenue and expenses can be expressed as Total Revenue (TR) = R1 + R2 + ... + Rn, Total Expenses (TE) = El + E2 + ... + En.

[0350] 1.3.2. Date-time Data:

[0351] Date-time data represents a specific point in time and is processed through time series analysis. In the financial data of the World Football Federation, date-time data can include transaction dates, payment processes, and organizational dates. Such data is utilized for sorting based on time periods, conducting time series analysis, or comparing date ranges. For example, a comparison of revenues for a specific period: 2023 Revenue (RG) > 2022 Revenue (RG-1).

[0352] 1.3.3. Categorical Data:

[0353] Categorical data expresses information belonging to a specific category. In the financial data of the World Football Federation, categorical data includes types of sponsors (e.g., clothing brands, technology companies), types of organizations (e.g., World Cup, Confederation Cup), and payment methods (e.g., credit card, bank transfer). These types of data are analyzed using frequency tables, probability distributions, and categorical analyses. The conversion of numerical data, date-time data, and categorical data into formats suitable for artificial intelligence algorithms has been discussed in previous algorithm models and is not elaborated on in this section.

[0354] 1.3.4. Text Data:

[0355] Text data consists of words or sentences and is processed using natural language processing (NLP) techniques. The financial documents of the World Football Federation may include text data such as explanations, contract texts, and staff notes. Organizing and analyzing text data is crucial for text mining. For example, word frequency analysis can be used to calculate the frequency of specific terms in a document.

[0356] The basic structures and principles of the NLP and data mining algorithms mentioned in this subsection are as follows:

[0357] Natural Language Processing (NLP): NLP is a set of algorithms and techniques used to analyze and understand text data. The basic steps are as follows:

[0358] Cleaning text data: The process of organizing text data involves removing unnecessary characters, spaces, and punctuation. Text normalization steps are applied, meaning differences in uppercase and lowercase are ignored.

[0359] Parsing the text document: The text document is broken down into sentences and words. This step is important for understanding the structure of the text.

[0360] Tokenization: Identifying words or word clusters (tokens). For example, the sentence "World Football Federation World Cup" is tokenized into "World Football Federation," "World," and "Cup." Removing stop words: Words that do not carry much meaning in the context of language (stop words) are removed. Examples of stop words include "and," "or," "a," "this," and similar words.

[0361] Identification of Word Stems and Roots: Techniques known as stemming and lemmatization are employed to find the root or base forms of words. For example, the word "running" is stemmed to "run."

[0362] Word Frequency Analysis: It involves calculating the frequency of words in a document. This helps determine how often a word appears and its importance within the document.

[0363] Word Vectorization: Methods are used to transform words into numerical vectors. Algorithms such as "word2vec" or "GloVe" are particularly used to capture the semantic similarity of words.

[0364] Text Mining: Text mining aims to extract information from large text datasets. It fundamentally includes the following steps:

[0365] Creation of document clusters: Similar documents related to a specific topic are grouped.

[0366] Combined word frequency: The frequency of specific terms in document clusters is calculated to determine the importance of these terms within the document clusters.

[0367] Document classification: Document clusters are categorized into specific classes (e.g., financial documents, explanations, contract texts).

[0368] Topic modeling: It is used to identify latent topics within document clusters. The LDA (Latent Dirichlet Allocation) algorithm is used for this purpose in this context.

[0369] 1.3.5. Mathematical and Software Infrastructure:

[0370] Mathematical and software tools are used for processing data types. For numerical data, Python's NumPy and Pandas libraries are employed. NLP libraries are used for processing text data and natural language processing. DateTime library in Python is used for processing date-time data, while Pandas categorical data type is used for categorical data.

[0371] 1.4. Data Quality Control: Control mechanisms are implemented to ensure data quality during the data collection process. This step ensures that the data is complete, accurate, and up-to- date. The appropriateness and reliability of data from data sources are evaluated. This sub-step is critical for ensuring the accuracy, completeness, and timeliness of the financial data of the World Football Federation in the data collection process. The fundamental operational processes are as follows:

[0372] 1. Data Sources: The set of data sources used in the data collection process is specified and represented as D_k. Here, k represents the number of data sources. D_k = {d_l, d_2, ..., d_n}_k

[0373] Each data source D_k contains the data d i.

[0374] 2. Data Quality Assessment: A data quality assessment is conducted for each data source, represented by an evaluation metric based on criteria such as accuracy, completeness, and timeliness.

[0375] Q_k = {q_l, q_2, ..., q_n}_k

[0376] Each q_i represents a data quality value for the data source D_k.

[0377] 3. Data Quality Control Function: The data quality control process is represented by a function that assesses the quality of data coming from various sources. This function evaluates and checks the suitability of each data source.

[0378] (D_k) -> {q_l, q_2, ..., q_n}_k

[0379] When the data source D_k is provided to the control function, each data quality value q_i is calculated as a result.

[0380] 4. Data Quality Improvement: If the data quality of a data source D_k does not exceed a certain threshold value (e.g., q_i < threshold), it is necessary to improve or correct the data from that source.

[0381] Improve(D k) -> D_k' (if improvement is needed)

[0382] The improvement function can return the data from the source D_k in a corrected or enhanced form, represented as D_k'.

[0383] 5. Result: As a result of the data quality control process, each data source is filled with updated data and undergoes data quality assessments. This prepares the data for analysis and processing of financial transactions. The technical structure of data quality control involves evaluating and, if necessary, improving data sources based on specific quality metrics. This process is a critical step in ensuring the reliability and quality of financial data.

[0384] 1.5. Data Organization and Standardization:

[0385] Collected data is organized and standardized to ensure consistent formatting and the removal of unnecessary or conflicting information. Data merging and linkage are facilitated to make the data cohesive.

[0386] 1.6. Data Storage and Backup:

[0387] Organized data is securely stored and backed up. Appropriate data storage methods and backup procedures are implemented to ensure the security and integrity of the data.

[0388] 1.7. Data Access Controls:

[0389] Data access controls are applied to allow only authorized individuals to access collected data. This is done to protect the privacy and security of the data.

[0390] 1.8. Data Documentation:

[0391] 1.9. Data Monitoring and Revision:

[0392] Detailed documentation is created for the collected and organized data. This documentation includes information on how the data is collected, organized, stored, and accessed.

[0393] 1.10. Data Monitoring and Revision:

[0394] The processes of data collection and organization are continuously monitored and revised as needed. Changes or errors in the data are quickly identified and corrected.

[0395] 1.11. Data Security:

[0396] - The security and confidentiality of data are ensured. Data is protected against unauthorized access and processed in accordance with data security policies.

[0397] These mentioned steps encompass the technical procedures that need to be followed to reliably and systematically prepare and process the financial data of the World Football Federation. This constitutes the initial and fundamental step in the financial audit and analysis process.

[0398] Audit Algorithm: The audit algorithm examines the collected data to detect financial inconsistencies, errors, and signs of corruption. This step is particularly important for auditing the large volume of financial transactions of the World Football Federation. The algorithm identifies inconsistencies, logical errors, and other abnormalities among financial documents. The basic steps of the algorithm are as follows:

[0399] Data Analysis and Processing: This stage involves analyzing the cleaned data. The algorithm primarily uses the following mathematical expressions to examine the content of financial documents: Statistical Summaries:

[0400] Mean (Average): p_i = (1 / n) * ^(cleaned data) i

[0401] Variance: cA2_i = (1 / n) * ^((cleaned dataj - p_i)A2)_i

[0402] Standard Deviation:

[0403] Correlation Analysis:

[0404] The calculation of correlation between two financial variables, such as income and expenses, is obtained mathematically as follows: p(Income, Expenses) ! = [^((cleaned lncomej - p lncome i) * (cleaned ExpensesJ - p_Expenses_i))_i] / [n * c lncome i * c Expenses i] c) Z-Score Calculation:

[0405] Comparing each data point with a specific mean and standard deviation: Z(IncomeJ)_i = (cleaned lncomej - p lncome i) / o lncome i

[0406] Abnormality and Outlier Detection:

[0407] The algorithm detects abnormalities in financial documents through statistical and mathematical analyses. For example, Z-scores exceeding a certain threshold or correlations not conforming to specific criteria indicate errors or outliers. The fundamental technical structure of this step is as follows: a. **Z-Score Analysis:**

[0408] A specific financial indicator is denoted as X. This indicator is characterized by the mean (p) and standard deviation (c) within the population. The corresponding z-score Z is calculated as follows:

[0409] Z = (X - p) / o At this stage, a specific threshold value (z_threshold) is evaluated. If |Z| > / threshold, X is considered abnormal. b. Correlation Analysis:

[0410] The correlation coefficient (r) between two financial indicators, X and Y, is calculated as follows: r = S((Xi - pX) * (Yi - pY)) / [S(Xi - pX)A2 * S(Yi - pY)A2]A0.5

[0411] When a specific correlation threshold (r threshold) is adopted, if |r| < r threshold, the correlation between X and Y is considered abnormal. c. Outlier Analysis:

[0412] The Inner Quartile Range (IQR) method is employed to identify outliers in the dataset. For a financial indicator X, the IQR is calculated as follows:

[0413] IQR = Q3 - QI

[0414] Here, QI represents the first quartile of the data, and Q3 represents the third quartile. The lower limit of IQR is considered as QI - (1.5 * IQR), and the upper limit is considered as Q3 + (1.5 * IQR). If an observation falls outside these limits, it is considered an outlier.

[0415] Logic Errors: The algorithm uses logical expressions to detect logic errors in financial documents. For instance, it raises an alarm in the presence of balance errors, negative amounts, or mathematically inconsistent expressions. The fundamental mathematical structure is as follows:

[0416] Detection of Balance Errors

[0417] To compute the ending balance (E) using the starting balance (B) and transaction movements (I), the following mathematical relationship is utilized:

[0418] E = B + SI

[0419] Here, E represents the ending balance, B is the starting balance, and I represents transaction movements related to the account. This mathematical expression elucidates the calculation of the ending balance for each account. If any account violating this relationship is identified, a balance error may exist.

[0420] Detection of Negative Amounts: To check whether the total amount of transaction movements (SI) for a specific account is negative, the following mathematical expression is used: II < 0

[0421] SI represents the total amount of transaction movements associated with the relevant account. If this total is negative, there may be a logic error in the Account.

[0422] Detection of mathematical contradictions: Special equations can be employed to identify contradictions using mathematical expressions in financial statements. For instance, considering an expression like A + B = C, the following expression can be used to check if it leads to contradictions:

[0423] A + B ^ C

[0424] If this equality is violated, it indicates a mathematical contradiction and is considered a logical error.

[0425] The mentioned mathematical expressions are fundamental tools used to detect logical errors in financial documents. When the relevant conditions are met, the system automatically generates warnings or alarms, enabling users to quickly identify and rectify issues.

[0426] Consistency and fraud detection: The audit algorithm uses specific rules and filters to detect inconsistencies and potential signs of fraud in financial documents. For example, variations in amounts for similar transactions or unexpected financial movements are evaluated at this stage. The basic mathematical structure of this step is as follows: a. Detection of variations in amounts for similar transactions: Similar transactions should have similar amounts. Mathematically, this expectation is expressed as: E(|T1 - T2|) < 5

[0427] Here, T1 and T2 represent the amounts in two similar transactions. If the difference (|T 1 - T2|) exceeds a certain threshold (5), it is considered an indication of inconsistency. b. Detection of unexpected financial movements: Financial movements should follow a specific pattern or expected behavior. The mathematical expression for this behavior is:

[0428] H(x) = y Here, H represents a mathematical function representing a transaction, x is the input of the transaction, and y is the expected result. If H(x) y, it indicates an unexpected financial movement. c. Use of special filters: Special filters are used to analyze financial documents and identify transactions that match specific patterns or conditions. Mathematically, these filters are expressed as:

[0429] F(x) = z

[0430] Here, F represents a special filter operation, x is the input, and z is the filter result. If F(x) = z is not satisfied, it triggers a warning or alarm by the special filter.

[0431] The mentioned mathematical expressions are fundamental tools used to detect inconsistencies and potential signs of fraud in financial documents.

[0432] Results and alarms: Anomalies and errors detected by the algorithm are reported, and alarms are triggered for the user. The mathematical expressions that the algorithm can use to detect anomalies in a document are expressed as follows: Alarm_i = {Anomaly_l, Anomaly_2, ..., Anomaly_m}_i

[0433] Each Anomaly represents different types of anomalies in the document.

[0434] The audit algorithm goes through these mentioned stages to detect errors and anomalies in financial documents. These detections enhance the reliability of financial transactions and provide accurate and reliable data to financial managers. Step 3 : Business Process Automation

[0435] Business process automation is used to automate repetitive tasks in the financial transactions of the World Football Federation. For example, processes such as monitoring, reporting, or approval of specific financial transactions are automated in this step. This is crucial to increase business efficiency and reduce human errors.

[0436] Business process automation utilizes data obtained and organized in the data collection and organization step, and its fundamental steps are as follows: Decision-making: Business process automation makes decisions based on the analysis results. Decisions such as whether a specific transaction should be approved or not, or what to do when an error is detected, are automatically made. Decisions are based on predefined rules and policies. This step primarily consists of the following sub-steps: a- Decision-making rule (DAF): Represents the decision-making rule or function used in business process automation. DAF automatically makes decisions to approve or reject a transaction using transaction data and predefined rules. Mathematically, DAF is expressed as follows:

[0437] DAF(x, K) = {Approve, Reject}

[0438] Here:

[0439] DAF represents the decision-making rule. x represents the transaction data.

[0440] K represents predefined rules or policies.

[0441] If DAF(x, K) is "Approve," the transaction is approved; if "Reject," the transaction is rejected.

[0442] The mentioned mathematical expression represents the decision-making rule used in analyzing a specific transaction and helps in making decisions based on transaction data and predefined rules. b- Decision function (D): A function representing the decision-making rule is used in business process automation. This function automatically makes decisions to approve or reject a transaction based on the analysis results. Mathematically, this function can be expressed as follows:

[0443] D(x) = {Approve, Reject}

[0444] Here, D(x) represents the decision made based on the processing data (x). If D(x) is "Approve," the transaction is approved, and if it is "Reject," the transaction is declined. c-Rules and policies (R): The decision function operates based on specific rules and policies (R). These rules are founded on predefined business processes, costs, security requirements, and similar factors. Mathematically, rules and policies are expressed as follows:

[0445] D(x) = {Approve, Reject} (if R(x) condition is satisfied)

[0446] Here, R(x) indicates whether a particular rule or policy is satisfied based on the processing data (x). In summary, the business process automation algorithm analyzes a specific transaction using mathematical decision functions. These functions automatically decide to approve or reject the transaction based on predefined rules and policies, providing a mathematical foundation for managing business processes more effectively and systematically.

[0447] Action implementation: Transactions are processed according to decisions. For example, if a transaction is approved, the automation approves and records it. If an error is detected, the automation implements a specific correction process. The underlying structure of this step is as follows:

[0448] Action function (AF): Represents the action function used in the action implementation step. The AF performs actions and records results based on decisions. Mathematically, AF can be expressed as:

[0449] AF(D, I) = R

[0450] Here:

[0451] AF represents the action function.

[0452] D represents the set of decisions. D = {DI, D2, ..., Dn}

[0453] I represents the processing data.

[0454] R represents the results of the actions.

[0455] If AF(D, I) = R, the action function performs specific actions on I based on decisions in D and records the results as R.

[0456] The mentioned mathematical expression represents how actions are implemented and results are recorded in the business process automation algorithm based on decisions.

[0457] Monitoring and reporting: The actions occurring in the business process automation algorithm are continuously monitored and recorded. Additionally, reports are generated periodically or as needed. These reports include information such as the status of financial transactions, transaction history, and performance. The fundamental mathematical structure of this step is as follows:

[0458] Monitoring function (MF): Represents the monitoring function used in business process automation. Mathematically, MF is expressed as:

[0459] MF(t, I) = M Here:

[0460] - MF represents the monitoring function.

[0461] - 1 represents the time.

[0462] -I represents the processing data.

[0463] -M represents the monitoring results.

[0464] The expression MF(t, I) = (M) produces results (M) by monitoring the processing data (I) at a specific time (t) Monitoring results provide information about the effectiveness of business processes and the status of data.

[0465] Reporting function (RF): Represents the reporting function used in business process automation. Mathematically, RF is expressed as:

[0466] RF(I) = R

[0467] Here:

[0468] RF represents the reporting function.

[0469] -I represents the processing data.

[0470] -R represents the generated report.

[0471] The expression RF(I) = R produces reports ( R ) using specific processing data (I). These reports include information such as the status of financial transactions, transaction history, and performance.

[0472] The mentioned mathematical expressions define the fundamental concepts of monitoring and reporting in business process automation. The monitoring function monitors processing data at a specific time, while the reporting function generates reports using processing data. This is crucial for observing business processes and evaluating performance.

[0473] Regular updates and improvements: The business process automation algorithm is regularly updated and improved. These updates aim to respond to new requirements, address errors, and enhance efficiency.

[0474] Data analytics algorithm: The data analytics step is used to analyze the data obtained by audit and business process automation algorithms in more detail. This step includes advanced analytical methods to identify financial trends, income and expenditure items, cost optimization opportunities, and risk factors. Additionally, data mining and machine learning techniques are used in this step. The fundamental steps of the algorithm are as follows:

[0475] Data compilation (DC): The first step involves gathering and organizing financial data obtained based on the audit and business process automation results. Mathematically, this step is expressed as:

[0476] DC = X(D) for D in Data

[0477] Here:

[0478] DC represents the compiled data.

[0479] X symbolizes the summation.

[0480] D signifies different data sources.

[0481] Data represents the dataset encompassing all data sources.

[0482] This expression illustrates the aggregation of data obtained from all sources.

[0483] Data Cleaning and Correction (DCC): In the data analytics process, the stage of data cleaning and correction is utilized to identify and rectify missing or erroneous data. Mathematically, this step can be expressed as follows:

[0484] DCC = Clean(Data)

[0485] Here:

[0486] DCC denotes the cleaned data.

[0487] Clean represents the cleaning process.

[0488] This expression indicates that the data has been cleaned and corrected.

[0489] The fundamental program codes that outline the working structure of this substep are as follows: import pandas as pd

[0490] # Load the dataset data = pd.read csv('data.csv')

[0491] # Detect missing data missing data = data.isnull()

[0492] # Fill missing data with mean values data.fillna(data.mean(), inplace=True)

[0493] # Correct erroneous data dataf'erroneous column'] = data['erroneous_column'] .apply(correction function)

[0494] # Save the cleaned data to another file data.to_csv('cleaned_data.csv', index=False)

[0495] The mentioned mathematical framework and the fundamental code example constitute the foundation of the data cleaning and correction process. The cleaned data is then prepared for analytical processes to enhance data quality and ensure the reliability of analytical results. This stage is a critical step in improving data quality and making analysis results trustworthy.

[0496] Data Analytics Function (DAF): The data analytics function performs financial analytical processes using cleaned data. For instance, a function can be employed to analyze income and expenditure trends. Mathematically, this step is expressed as follows:

[0497] DAF = Analyze(CleanedData)

[0498] Here:

[0499] -DAF represents the results of the analysis.

[0500] -(Analyze denotes the analysis function.

[0501] This expression indicates that cleaned data is analyzed, yielding results.

[0502] Risk Assessment (RA): It is a mathematical process used to identify risk factors. This step is crucial for calculating and managing financial risks. Its fundamental mathematical structure is as follows:

[0503] RA = CalculateRisk(CleanedData)

[0504] Here:

[0505] -RA represents the results of risk assessment.

[0506] -CalculateRisk represents the risk calculation function.

[0507] This expression illustrates the calculation of financial risk factors.

[0508] The data analytics step describes, in mathematical terms, the in-depth examination of financial data to gain further insights. This plays a critical role in the formulation of financial decisions and strategies.

[0509] Recommendation System Algorithm:

[0510] The recommendation system algorithm determines opportunities for increasing revenue and reducing expenses based on data analytics results. The recommendation system suggests specific actions or policy changes to financial decision-makers. For example, it provides recommendations such as revising sponsorship agreements, adjusting ticket prices, or reevaluating supplier selection to reduce costs. Since the recommendation system algorithm is extensively detailed in overseeing the discipline and fair play rules of the World Football Federation, it will not be reiterated here. The same algorithm is applied to FIFA's financial tasks, utilizing and organizing data analytics algorithms to fulfill this responsibility.

[0511] Implementation and Monitoring:

[0512] The final step involves implementing the recommendations provided by the recommendation system algorithm and monitoring the results. The World Football Federation's finance team can make decisions based on these recommendations and update business processes or policies as needed. Simultaneously, the effectiveness of the measures taken is monitored, and measurable results of financial improvements are tracked.

[0513] This integrated model, as mentioned, enhances the transparency of the World Football Federation's financial operations, providing a powerful tool to reduce the potential for corruption and enhance financial efficiency. This approach assists the World Football Federation in managing its financial decisions more efficiently, accurately, and reliably.

[0514] The decision tree algorithm is employed by the World Football Federation for the mission of promoting football and creating an impact in society. The fundamental working principle of the algorithm is as follows:

[0515] Decision Tree Algorithm: The decision tree algorithm is a machine learning technique used for making various decisions by solving a complex problem in the form of a chain of simple decisions. The algorithm consists of a series of decision nodes and branches connecting these nodes. Each node represents a feature or data, and each branch contains a decision or result.

[0516] Data Collection: The initial step is the collection of relevant data for the examined problem, such as gender equality, disabled football, education and awareness, and social responsibility projects. Data is obtained from the World Football Federation's databases, the computers of World Football Federation employees, web pages, and manually entered data. These data sources contain the information necessary for the decision tree algorithm to perform its analysis.

[0517] Feature Selection: The decision tree algorithm is utilized to determine which features (characteristics of data points) are effective for the given problem. Features are factors that are crucial for a specific task. For example, in the task of gender equality, features like financial resources or infrastructure development for the advancement of women's football may be important.

[0518] Decision Tree Construction: The algorithm analyzes the data to determine the decisions that best separate the data points and the threshold values. For instance, to assess the effectiveness of projects supporting women's football, the algorithm evaluates the impact of financial resources on the ratio of women's football to men's football.

[0519] Evaluation: The created decision tree is used to make decisions regarding a specific task or problem. The algorithm guides the data point down a series of decision nodes, starting from the root node and leading to the result node.

[0520] Interpretation of Results: The results of the decision tree are interpreted, and the effectiveness or success of a particular activity is assessed. For example, the decision tree analysis may be employed to determine whether a specific educational program increases the participation of socially marginalized individuals in the field of education.

[0521] Strategy and Improvement: Based on the results, strategies are developed, and improvements are made. The decision tree results provide guidance for designing more effective projects and activities.

[0522] In this manner, the effectiveness of non-financial tasks (such as supporting women's football, supporting disabled football, education and awareness, and social responsibility projects) in the mission of promoting football and creating an impact in society is analyzed using the decision tree algorithm. It allows for the determination of which steps yield better results.

[0523] In the detailed description section, details regarding the duty of improving and implementing refereeing and fair play principles, one of the activities of the World Football Federation, have not been provided here since the decision tree algorithm has been elaborately explained in that context.

[0524] At the invention, Support Vector Machines (SVM) algorithm is employed as part of the responsibilities of the World Football Federation, specifically in the task of disseminating football and assessing the functionality of electronic structures of technological systems such as VAR (Video Assistant Referee) and Hawk Eye for their intended purposes. Support Vector Machines (SVM) is a machine learning algorithm used for classification and regression problems. The SVM algorithm aims to find a hyperplane in the multidimensional feature space to classify data and maximize the margin between data points of two classes. The algorithm operates in the following manner:

[0525] The first step involves preparing the training dataset, consisting of examples representing specific situations labeled as 'security system operating status.' Each example is a vector containing information from different sensors.

[0526] The SVM algorithm represents data in a multidimensional feature space, where data examples are represented by features extracted from sensor data.

[0527] The SVM algorithm attempts to find a hyperplane to separate data examples into two classes, maximizing the margin between the two classes. The margin is the distance between the hyperplane and the closest data points.

[0528] To maximize the margin, the algorithm utilizes support vectors, which are the data points closest to the hyperplane and determine the margin. The hyperplane is positioned to balance among these support vectors.

[0529] The margin is calculated during the process of finding the equation of the hyperplane. The SVM employs mathematical methods to find the optimal hyperplane, maximizing the margin.

[0530] If the data is not linearly separable, the algorithm uses a technique called 'kernel trick' to make it linearly separable. This is achieved by mapping the data to a higher-dimensional space or using a different kernel function. This enables the data to become linearly separable.

[0531] The 'kernel trick' is a commonly used technique in machine learning and data analytics, especially when working with algorithms like Support Vector Machines (SVM). Some algorithms, including SVM, use the kernel trick to handle non- linearly separable datasets by transforming the data into higher-dimensional spaces, making them linearly separable. This enhances computational efficiency and improves the model's performance.

[0532] The kernel trick involves using a function that expands the space by altering the features of the data instead of directly computing the dot product. In this expanded space, the data becomes linearly separable, allowing the use of algorithms capable of linear separation.

[0533] In kernel trick, the dot product operation holds significant importance. The dot product of two vectors can be expressed as follows:a’b=E (i=l (n)) aibi

[0534] Here, (a) and (b) are two vectors and are n-dimensional vectors.

[0535] The SVM algorithm uses kernel functions to transform data points into highdimensional spaces to make them linearly separable. In other words, instead of representing the data directly in the expanded space, it calculates the inner product in this space using kernel functions. Kernel functions effectively transform the features of data in the real space to the features of data in the expanded space. The following kernel functions are used:

[0536] Linear Kernel function: K(a,b)=a-b This function treats the data in the same dimension and ensures linear separability in the original feature space.

[0537] Polynomial Kernel function: K(a,b)=(ya-b+r)d Here, (y) is a scaling factor, (r) is an extra term, and (d) is the degree of the polynomial. This function enhances linear separability by moving the data to a higher-dimensional space.

[0538] Radial Basis Function Kernel (RBF Kernel): K(a,b)=exp (- y||a- b ||2) Here, (y) is a scaling factor, and Ila— bll represents the distance between two points. This kernel function uses an RBF transformation to move the data to high-dimensional spaces.

[0539] The kernel trick allows calculating the inner product in the expanded space without explicitly computing it, offering the advantage of reducing computation costs and enabling operations in a space where the dataset is linearly separable. Using the mentioned kernel trick allows the SVM algorithm to successfully handle more complex and linearly non-separable datasets. This way, the algorithm can classify or perform regression on data more accurately.

[0540] Testing and Training: After completing the training process, the SVM algorithm can be used to classify new data. New data samples are assigned to a class by comparing them with the hyperplane learned during training.

[0541] Error Correction and Adjustment: After the training and testing stages, hyperparameters are adjusted to improve the algorithm's performance, and error correction processes are carried out to enhance the accuracy of the algorithm.

[0542] In conclusion, the SVM algorithm aims to find a hyperplane to separate data into two classes, maximizing the margin. The plane alignment method is used to make linearly non-separable data linearly separable. In the invention, the SVM algorithm is equipped with state-of-the-art VAR system, HAWK Eye system, and similar technological structures to continuously monitor the operation status and automatically correct errors.

[0543] The fundamental working principle of decision trees algorithm used by the World Football Federation in detecting patents related to football in this task is as follows:

[0544] Data Preparation: The initial step involves the preparation and organization of text data collected through web scraping. Text mining and machine learning algorithms perform better with clean and structured data. To achieve this, documents are organized, and irrelevant characters, spaces, and features outside the text content are removed.

[0545] Labeling: Labels are added to these documents to determine whether patent documents are related to football or not. Documents containing specific terms or concepts are labeled as "football-related," while others are marked as "irrelevant." Feature Engineering: Using text mining techniques to analyze documents, textbased features such as term frequencies, keywords, or specific concepts are extracted.

[0546] Training and Validation: The prepared data is used to train the decision tree model. The model learns to distinguish between football-related and irrelevant documents using training data. The model's performance is then evaluated by testing it with validation data.

[0547] Results Evaluation: The decision tree model can identify football-related patents. The model's success is assessed using metrics such as accuracy, precision, recall, and others.

[0548] Update: As patents are continually added and updated, the algorithm is regularly updated to ensure its relevance.

[0549] The decision tree algorithm provides an intuitive and transparent approach for classifying documents. It also helps understand which features are more decisive in document classification.

[0550] In the detailed description section, the activities of the International Football Association Board (IF AB), focused on improving and implementing refereeing and fair play principles, are explained in detail. The decision tree algorithm and the web scraping method in the financial management section of the World Football Federation are not detailed here.

[0551] IF AB is a global authority that establishes the fundamental rules of football and works to ensure their appropriate application worldwide. Presently, observers from the World Football Federation or other football organization officials evaluate the performances of referees in football matches, creating observation reports. These reports are uploaded to a digital platform and stored electronically, facilitating easier sharing, archiving, and accessibility for future analyses or evaluations. Using the digital data mentioned in the invention, NLP and decision tree algorithms reveal whether IFAB's rules are applied and to what extent. The fundamental working principles of the mentioned algorithms are as follows:

[0552] Data Collection and Preparation: The first step is to collect text data from observation reports and transform this data into a format that can be processed by the NLP algorithm. This step involves text analysis, word extraction, and text editing processes.

[0553] Text Analysis: The NLP algorithm conducts text analysis to examine the text data in reports and extract meaningful information from the text. This involves identifying important text features related to referee decisions and game rules. Data Features and Feature Engineering: The text features obtained by NLP are transferred to the decision tree algorithm as data features. The process called feature engineering is carried out to further process and make the extracted information usable.

[0554] Decision Tree Model: The decision tree algorithm works on these data features to classify referee decisions. For instance, it can use different features and rules to determine whether a referee decision is "correct" or "incorrect."

[0555] Scoring and Evaluation: The decision tree algorithm is utilized to evaluate the performance of referees. It is employed to determine the extent to which IF AB rules are applied and which referees exhibit better performance.

[0556] Results and Reporting: Integrated NLP and decision tree algorithms present results and analyses related to referee performance. These results can be presented in reports and used as guidance for IF AB or football organizations.

[0557] This integrated approach assists in objectively monitoring referee performance and evaluating the application of game rules. Additionally, the text processing capability of the NLP algorithm provides an advantage in extracting important information from referee reports and using it in conjunction with decision tree algorithms.

[0558] The NLP algorithm is detailed in the "Financial Responsibilities of the World Football Federation" section, while the decision tree algorithm is explained in the "Developing and Implementing Refereeing and Fair Play Principles" section. Due to this detailed explanation in those sections, further details are not provided here. The artificial intelligence algorithms mentioned in the brief and detailed descriptions of the invention, as well as the functions and metrics used in these algorithms, should not be considered binding. The purpose of the invention is to carry out the tasks of the World Football Federation and organizations affiliated with the World Football Federation through artificial intelligence algorithms.

[0559] The invention can be used in the confederations of the World Football Federation, the European Football Federation, the African Football Federation, the Asian Football Federation, the South American Football Federation, the Confederation of North, Central America and Caribbean Association Football, and the Oceania Football Federation, as well as in the football federations and clubs of countries. As the tasks of all these institutions are largely similar, and specifying each institution every time would cause confusion, in the patent file, only "World Football Federation" is written each time.

Claims

CLAIMS1. Artificial intelligence systems performing the functions of world football and its affiliated organizations, contributing to the development and global spread of football, are characterized by: incorporating artificial intelligence algorithms responsible for tasks such as refereeing and adherence to fair play principles, disciplinary rules, international relations and cooperation, financial management of football, promotion of football and creating an impact in society, as well as tasks related to the use of technology and innovation, including the responsibilities of IF AB.

2. Artificial intelligence systems responsible for the functions of world football and its affiliated organizations, contributing to the development and global spread of football, are characterized by: utilizing decision tree artificial intelligence algorithms to perform the refereeing and fair play principles tasks mentioned in Requirement 1, preventing favoritism, nepotism, and incompetence based on the data generated by the algorithm.3.Artificial intelligence systems responsible for the functions of world football and its affiliated organizations, contributing to the development and global spread of football, are characterized by: utilizing recommendation system artificial intelligence algorithms to perform the disciplinary rules and fair play tasks mentioned in Requirement 1, rewarding teams and countries adhering to fair play principles based on the data generated by the algorithm, continuously evaluating activities through the algorithm, and ensuring that results are carried out in a concrete, data-driven, and transparent manner.

4. Artificial intelligence systems responsible for the functions of world football and its affiliated organizations, contributing to the development and global spread of football, are characterized by: utilizing supervised learning artificial intelligence algorithms to perform the international relations and cooperation tasks mentioned in Requirement 1, determining appropriate cooperation strategies by taking into account specific cultural contexts in regions with such contexts based on the data generated by the algorithm.

5. Artificial intelligence systems responsible for the functions of world football and its affiliated organizations, contributing to the development and global spread of football, are characterized by: integrating data analytics, financial audit, business process automation, and recommendation system artificial intelligence algorithms that work in a cohesive manner to perform the financial management tasks mentioned in Requirement 1. This involves identifying opportunities to increase revenue and reduce expenses based on the results generated by the algorithms, analyzing financial trends, defining opportunities for cost optimization, evaluating risk factors, making financial decisions based on concrete data, and achieving sustainable financial success.

6. Artificial intelligence systems responsible for the functions of world football and its affiliated organizations, contributing to the development and global spread of football, are characterized by: utilizing decision tree artificial intelligence algorithms to perform the football promotion and societal impact tasks mentioned in Requirement 1. This involves effectively evaluating the outcomes of women's football, disabled football, education opportunities, and social responsibility projects based on the data generated by the algorithm.

7. Artificial intelligence systems responsible for the functions of world football and its affiliated organizations, contributing to the development and global spread of football, are characterized by: implementing Support Vector Machines (SVM), web scraping, and decision tree artificial intelligence algorithms to fulfill the technology usage and innovation tasks mentioned in Requirement 1. This involves, based on the data generated by the algorithms, detecting the proper functioning of electronic components of technological systems such as the Video Assistant Referee (VAR) and Hawk Eye systems, as well as examining and reporting on patents related to football worldwide.

8. Artificial intelligence systems responsible for the functions of world football and its affiliated organizations, contributing to the development and global spread of football, are characterized by: examining and reporting on the effective use of football rules set by IFAB (International Football Association Board), as mentioned in Requirement 1. This is accomplished by integrating and analyzing field match observation reports recorded in digital format by observers from organizations such as the World Football Federation, using Natural Language Processing (NLP) and decision tree artificial intelligence algorithms.

9. Artificial intelligence systems responsible for the functions of world football and its affiliated organizations, contributing to the development and global spread of football, are characterized by: the ability for the tasks and artificial intelligence solutions mentioned in the requirements to be applicable and utilized in the World Football Federation, European Football Federation, African Football Federation, Asian Football Federation, South American Football Federation, North, Central American and Caribbean Football Confederation, and Oceania Football Federation confederations, as well as in national football federations and football clubs around the world.

Images