Depuis les 10 dernières années, l’intelligence d’affaires et l’analytique (IA&A) sont devenues des sujets d’intérêts pour la recherche en systèmes d’information (SI) ainsi que pour les professionnels du domaine. Les initiatives en IA&A ont généré des bénéfices pour de nombreuses organisations dans plusieurs secteurs tels que la finance, les assurances, le divertissement ou encore les communications. Un domaine relativement nouveau où l’IA&A a fait son apparition est le sport compétitif. Les institutions, organisations sportives et athlètes d’élite tentent de mettre à profit l’utilisation des données et des technologies pour améliorer leurs performances, et ce, à différents niveaux. Alors que l’utilisation d’outils en IA&A dans les sports compétitifs a joui d’une médiatisation plus importante ces derniers temps, la recherche académique dans le domaine reste encore à un stade primaire. Dans cette étude, nous utiliserons deux situations dans le football universitaire pour mettre en évidence les composantes d’un cadre conceptuel de création de valeur dans les sports compétitifs. Nous présenterons une méthodologie afin d’optimiser le processus de recrutement collégial d’une organisation universitaire, ainsi qu’une autre méthodologie afin d’optimiser la prise de décision dans une situation de match précise.

Sports and technology have always co-operated to bring better and more specific sports statistics. The collection of sports game data as well as the ability to generate valuable sports statistics of it is growing. This thesis investigates the development of a sports statistics application that should be able to collect sports game data, structure the data according to suitable data models and show statistics in a proper way. The application was set to be a web application that was developed using modern web technologies. This purpose led to a comparison of different software stack solutions and web frameworks. A theoretical study of sports analytics was also conducted, which gave a foundation for how sports data could be stored and how valuable sports statistics could be generated. The resulting design of the prototype for the sports statistics application was evaluated. Interviews with persons working in sports contexts evaluated the prototype to be both user-friendly, functional and fulfilling the purpose to generate valuable statistics during sport games.

Machine learning has recently made significant progress due to modern neural network architectures and training procedures. When neural networks learn a task, they create internal representations of the input data. The specific neural network architecture, training process, and task being addressed will influence the way in which the neural network interprets and explains the patterns in the data. The goal of representation learning is to train the neural network to create representations that effectively capture the overall structure of the data. However, the process by which these representations are generated is not fully understood because of the complexity of neural network data manipulations. This makes it difficult to choose the correct training procedure in real world applications. In this dissertation, we apply representation learning to improve the performance of neural networks in three different areas: NBA movement data, material property prediction, and generative protein modeling. First, we propose a novel deep learning approach for predicting human trajectories in sporting events using advanced object tracking data. Our method leverages recent advances in deep learning techniques, including the use of recurrent neural networks and long short-term memory cells, to accurately predict the future movements of players and the ball in a basketball game. We evaluate our approach using data from the NBA's advanced object tracking system and demonstrate improved performance compared to existing methods. Our results have the potential to inform real-time decision making in sports analytics and improve the understanding of player behavior and strategy. Next, we focused on group activity recognition (GAR) in basketball. In basketball, players engage in various activities, both collaborative and adversarial, in order to win the game. Identifying and analyzing these activities is important for sports analytics as it can inform better strategies and decisions by players and coaches. We introduce a novel deep learning approach for GAR in team sports called NETS. NETS utilizes a Transformer-based architecture combined with LSTM embedding and a team-wise pooling layer to recognize group activity. We test NETS using tracking data from 632 NBA games and found that it was able to learn group activities with high accuracy. Additionally, self- and weak-supervised training in NETS improved the accuracy of GAR. Then, study an application of neural networks on protein modeling. Recent work on autoregressive direct coupling analysis (arDCA) has shown promising potential to efficiently train a generative protein sequence model (GPSM) to adequately model protein sequence data. We propose an extension to this work by adding a higher order coupling estimator to build a model called autoregressive higher order coupling analysis (arHCA). We show that our model can correctly identify higher order couplings in a synthetic dataset and that our model improves the performance of arDCA when trained on real-world sequence data. Finally, we study material property prediction. Incorporation of physical principles in a machine learning (ML) architecture is a fundamental step toward the continued development of AI for inorganic materials. As inspired by the Pauling’s rule, we propose that structure motifs in inorganic crystals can serve as a central input to a machine learning framework. To demonstrate the use of structure motif information, a motif-centric learning framework is created by combining motif information with the atom-based graph neural networks to form an atom-motif dual graph network (AMDNet), which is more accurate in predicting the electronic structures of metal oxides such as bandgaps. The work illustrates the route toward fundamental design of graph neural network learning architecture for complex materials by incorporating beyond-atom physical principles. / Computer and Information Science

Health and fitness apps have become ubiquitous as smart devices become a major necessity in day-to-day life. However, an obvious issue with mobile health (mHealth) apps is that a substantial portion of them lack a scientific foundation and instead utilize  experiential  stratagems.  Hence,  the  acquired  data  becomes  unreliable.  In sports, where data collection is extensive, this becomes a vital factor for success due to  the  increasing  usage  of  mHealth.  Therefore,  the  Swedish  School  of  Sport  and Health Sciences has, in collaboration with other organizations, created the Perceived Load  and  Fitness  Scale  Questionnaire.  The  purpose  of  this  questionnaire  is  to function as a marker for overtraining, and thus injury prevention and intervention will become a simpler and more efficient task. A computer software was developed for the questionnaire; however, a mobile version was required, and thus requested. Consequently, the mHealth prototype app eValuate was developed. Research, in the form of literature studies, and dissection of other apps, for additional information, contributed  to  the  development  of  it.  The  prototype  was  developed  using  the programming language Java with Android Studio as the Integrated Development Environment  and  Cloud  Firebase  Firestore  as  a  database  solution.  The  finished prototype, eValuate, had to be trialled to ensure that it satisfies the criteria. Thus, the Mobile Application Rating Scale was employed as the most appropriate means of evaluation. A small-scale study was planned to trial the prototype by utilizing this scale.  However,  due  to  unforeseen  events,  only  four  respondents  could  provide feedback. The prototype performed admirably and scored 3.8 stars out of 5 stars. Nonetheless, the testing sample is too small to draw any real conclusions.

Yes / Amid much recent interest we discuss a Variance Gamma model for Rugby Union matches (applications to other sports are possible). Our model emerges as a special case of the recently introduced Gamma Difference distribution though there is a rich history of applied work using the Variance Gamma distribution – particularly in finance. Restricting to this special case adds analytical tractability and computational ease. Our three-dimensional model extends classical two-dimensional Poisson models for soccer. Analytical results are obtained for match outcomes, total score and the awarding of bonus points. Model calibration is demonstrated using historical results, bookmakers’ data and tournament simulations.

Golf is a sport where the correct technical execution is important for performance and injury prevention. The existing feedback systems are often cumbersome and not readily available to recreational players. To address this issue, this thesis explores the potential of using 2D Human Pose Estimation as a mobile phone-based swing analysis tool. The developed system allows to identify three events in the swing movement (toe-up, top and impact) and to measure specific angles during these events by using an algorithmic approach. The system focuses on quantifying the knee flexion and primary spine angle during the address, and lateral bending at the top of the swing. By using only the wrist coordinates in the vertical direction, the developed system identified 37% of investigated events, independently of whether the swing was filmed in the frontal of sagittal frame. Within five frames, 95% of the events were correctly identified. Using additional joint coordinates and the event data obtained by the above-mentioned event identification algorithm, the knee flexion at address was correctly assessed in 66% of the cases, with a mean absolute error of 3.7°. The mean absolute error of the primary spine angle measurement at address was of 10.5°. The lateral bending angle was correctly identified in 87% ofthe videos. This system highlights the potential of using 2D Human Pose Estimation for swing analysis. This thesis primarily focused on exploring the feasibility of the approach and further research is needed to expand the system and improve its accuracy. This work serves as a foundation, providing valuable insights for future advancements in the field of 2D Human Pose Estimation-based swing analysis. / Golf är en sport där korrekt tekniskt utförande är avgörande för prestation och skadeförebyggelse. Feedbacksystem som finns är ofta besvärliga och inte lättillgängliga för fritidsspelare. För att åtgärda detta problem undersöker detta examensarbete potentialen att använda 2D mänsklig poseuppskattning som mobiltelefonsbaserat svinganalysverktyg. Det utvecklade systemet gör det möjligt att identifiera tre händelser i svingen (toe-up, top och impact) och att mäta specifika vinklar under dessa händelser genom en algoritmisk metod. Systemet fokuserar på att kvantifiera knäböjningen och primära ryggradsvinkeln under uppställningen, och laterala böjningen vid svingtoppen. Genom att endast använda handledskoordinater i vertikalriktning identifierade det utvecklade systemet 37% av de undersökta händelserna oavsett om svingen filmades från frontal- eller medianplanet. Inom fem bildrutor identifierades 95% av händelserna korrekt. Genom att använda ytterligare ledkoordinater och händelsedata som erhållits genom den tidigare nämnda algoritmen för händelseidentifiering, bedömdes knäböjningen vid uppställningen vara korrekt i 66% av fallen med en medelabsolutfel på 3.7°. Medelabsolutfelet för mätningen av primär ryggradsvinkel vid uppställningen var 10.5°. Laterala böjningen identifierades korrekt i 87% av tillfällena. Detta system belyser potentialen i 2D mänsklig poseuppskattning för svinganalys. Detta examensarbete fokuserade främst på att utforska tillvägagångssättets genomförbarhet och ytterligare forskning behövs för att utveckla systemet och förbättra dess noggrannhet. Detta arbete är grundläggande och ger värdefulla insikter för framtida forskning inom området för svinganalys baserad på 2D mänsklig poseuppskattning.

OBJECTIVE The purpose of this study was to determine how kinematic, big data can be evaluated using computational, comprehensive analysis of movement parameters in a diverse population. METHODS Retrospective data was collected, cleaned, and reviewed for further analysis of biomechanical movement in an active population using 3D collinear resistance loads. The active sample of the population involved in the study ranged from age 7 to 82 years old and respectively identified as active in 13 different sports. Moreover, a series of exercises were conducted by each participant across multiple sessions. Exercises were measured and recorded based on 6 distinct biometric movement parameters: explosiveness, velocity, power, deceleration, braking, consistency, endurance, and range of motion. Analysis and data visualization portrayed how 3D collinear resistance load impacted specific muscles and performance metrics. RESULTS The model with the highest accuracy rate was Naive Bayes and Fast Large Margin at 58.3% for future predictions considering impact for specific muscles, movement parameters, and performance metric data. The data visualization involved a proof-of-concept human-computer interface and presented each component in relation to one another within the active population database, movement parameters, and performance metrics. DISCUSSION Understanding the findings regarding 3D collinear resistance sets a precedence for future development for the active population and research in the sports analytics field. Additionally, the visual proof of concept interface promotes future development for a diverse, active population.

[pt] Em meio ao avanço expressivo da tecnologia e às evoluções contínuas observadas no ramo de inteligência artificial, esta última se mostrou ter potencial para ser aplicada a diferentes setores da sociedade. No contexto de extrema competitividade e relevância crescente nos esportes mais famosos ao redor do mundo, o basquete se apresenta como um esporte interessante para a aplicação de mecanismos de apoio à decisão capazes de aumentar a eficácia e consistência de vitórias dos times nos campeonatos. Diante desse contexto, este estudo propõe o desenvolvimento de sistemas de apoio à decisão baseados em modelos de redes neurais e k-Nearest Neighbors (kNNs). O objetivo é avaliar, para cada substituição durante um jogo de basquete, qual grupo de jogadores em quadra, conhecido por quinteto, apresenta mais chances de ter uma maior vantagem sobre o adversário. Para tal, foram treinados modelos para classificar, ao final de uma sequência de posses de bola, a equipe que conseguiria vantagem, e prever a magnitude dessa vantagem. A base de dados foi obtida de partidas do Novo Basquete Brasil (NBB), envolvendo estatísticas de jogadores, detalhes de jogo e contextos diversos. O modelo apresentou uma acurácia de 76,99 por cento das posses de bola nas projeções de vantagem entre duas equipes em quadra, demonstrando o potencial da utilização de métodos de inteligência computacional na tomada de decisões em esportes profissionais. Por fim, o trabalho ressalta a importância do uso de tais ferramentas em complemento à experiência humana, instigando pesquisas futuras para o desenvolvimento de modelos ainda mais sofisticados e eficazes na tomada de decisões no âmbito esportivo. / [en] In light of the recent significant growth in technological capabilities andthe observed advancements in the field of computational intelligence, the latterhas demonstrated potential for application in various sectors of society. Inthe context of extreme competitiveness and increasing relevance in the mostfamous sports around the world, basketball presents itself as an interestingsport for the application of decision-support mechanisms capable of enhancingthe efficacy and consistency of team victories in championships. In this context,this study proposes the development of decision-support systems, such asneural networks and k-Nearest Neighbors (kNNs). The goal is to evaluate, foreach substitution during a match, which group of players in the field, knownas lineup, presents the most probability to be superior to their opponent. Forthis, models were trained to predict, during a sequence of possessions, theteam that would have advantage and the magnitude of this advantage. Thedatabase was obtained from Novo Basquete Brasil (NBB) matches, involvingplayers statistics, match details and different contexts.. The model achieved anaccuracy of 76,99 percent in projections of superiority between the playing lineups,demonstrating the potential of using computational intelligence methods indecision-making applied to professional sports. Finally, the study highlightsthe importance of using such tools in conjunction with human experience,encouraging future research for the development of even more sophisticatedand effective models for decision-making in the sports field.

Sports analytics is an area that is growing at a rapid rate. It can be described as the use of data and analytics to gain an advantage in sports. It can include scouting, recovery, tactics and so on. This thesis uses data about association football passes to analyse players passing attributes. The dataset that has been used is acquired from the company Football Analytics Sweden AB and is produced by a subcontractor. The dataset includes event data for all games played in the Swedish division 1 during the 2020 season. In total there are 48 different types of events, but only the category passes accurate are analysed. Accurate passes are represented by x and y coordinates for the start and end point of the pass. The thesis's intention is the study how passes can be categorized in order to differentiate different players passing attributes, and how to identify types of players in regard to passing from that categorization. The categorization has been made possible by dividing the football pitch into different zones. The zones have been chosen on sports science and football grounds. Every pass is assigned a category based on which zone the pass starts in and what zone the pass ends up in. The proportion of passes a player hits for every category is then calculated and is used to perform factor analysis. The factor analysis identifies 13 underlying factors that can describe players different passing attributes. These factors have been interpreted and given names. The factors describe both general attributes as well as more specific attributes. Factor scores are then used to compare and identify which or what type of passing player a particular player is. / Sports analytics är ett begrepp som används mer och mer. I allmänhet kan det beskrivas som användningen av data och dataanalys för att få en fördel inom sport. Det kan vara allt från scouting (rekrytering av spelare), återhämtning och taktik. Den här studien använder information om fotbollspassningar för att undersöka fotbollsspelares passningsegenskaper. Datamaterialet som använts är erhållet från företaget Football Analytics Sweden AB och är framtaget av en underleverantör. Datamaterialet innehåller händelsedata för alla matcher som spelades i division 1 under säsongen 2020. Totalt finns det 48 olika typer av händelser, men endast händelsen lyckade passningar tas i beaktande. Lyckade passningar representeras av x och y koordinater för start- och slutpunkten för passningen. Studiens syfte är att undersöka hur passningar kan kategoriseras för att göra det möjligt att differentiera olika spelares passningsförmågor och hur man kan identifiera typ av passningsspelare utifrån kategoriseringen. Kategorisering av passningar utfördes med hjälp av en zonuppdelning av fotbollsplanen. Zonuppdelning är grundad på sportvetenskapliga och fotbollsmässiga grunder.  Varje passning tilldelas en kategori beroende på vilken zon passningen startar och slutar i. Andelen av passningar en spelare slår inom varje kategori beräknas, för att sedan användas i en faktoranalys. Faktoranalysen identifierar bakomliggande faktorer som påverkar en spelares andelar av passningar inom varje kategori. Totalt identifierades 13 bakomliggande faktorer och dessa kan beskriva en spelares passningsegenskaper. Dessa faktorer har verklighetstolkas och namngetts. Faktorerna beskriver både allmänna egenskaper och mer specifika egenskaper.  Faktorpoäng som beräknats från dem bakomliggande faktorerna används sedan för att jämföra och identifiera vilka eller vilken typ av passningsspelare en spelare är.

Football analytics is a rapid growing area which utilizes conventional data analysis and computational methods on gathered data from football matches. The results emerging out of this can give insights of performance levels when it comes to individual football players, different teams and clubs. A difficulty football analytics struggles with daily is to translate the analysis results into actual football qualities and knowledge which the wider public can understand. In this master thesis we therefore take on the ball event data collected from football matches and develop a model which classifies individual football player’s playing styles, where the playing styles are well known among football followers. This is carried out by first detecting the playing positions: ’Strikers’, ’Central midfielders’, ’Outer wingers’, ’Full backs’, ’Centre backs’ and ’Goalkeepers’ using K-Means clustering, with an accuracy of 0.89 (for Premier league 2021/2022) and 0.84 (for Allsvenskan 2021). Secondly, we create a simplified binary model which only classifies the player’s playing style as "Offensive"/"Defensive". From the bad results of this model we show that there exist more than just these two playing styles. Finally, we use an unsupervised modelling approach where Principal component analysis (PCA) is applied in an iterative manner. For the playing position ’Striker’ we find the playing styles: ’The Target’, ’The Artist’, ’The Poacher’ and ’The Worker’ which, when comparing with a created validation data set, give a total accuracy of 0.79 (best of all positions and the only one covered in detail in the report due to delimitations).  The playing styles can, for each player, be presented visually where it is seen how well a particular player fits into the different playing styles. Ultimately, the results in the master thesis indicates that it is easier to find playing styles which have clear and obvious on-the-ball-actions that distinguish them from other players within their respective position. Such playing styles, easier to find, are for example "The Poacher" and "The Target", while harder to find playing styles are for example " The Box-to-box" and "The Inverted". Finally, conclusions are that the results will come to good use and the goals of the thesis are met, although there still exist a lot of improvements and future work which can be made.  Developed models can be found in a simplified form on the GitHub repository: https://github.com/Sommarro-Devs/avatar-playing-style. The report can be read stand-alone, but parts of it are highly connected to the models and code in the GitHub repository.