ISSN 2348-1196 (print) International Journal of Computer Science and Information Technology Research ISSN 2348-120X (online) Vol. 8, Issue 3, pp: (195-200), Month: July - September 2020, Available at: www.researchpublish.com
WHO WILL WIN THE CHAMPIONSHIP ? Ahmad Lutfullah Saudi Aramco, Saudi Arabia
Abstract: As data mining techniques are becoming more essential for many life aspects, it has many significant additions to the field of sports. Using data mining, we can predict performance of teams, predict game outcomes, identify outstanding players, relate between positions and players statistics, find the most important factors in determining best players or best teams, and much more. Experts, coaches, players and even fans can benefit a lot of such effort. Keywords: Data mining; K-means clustering; Linear and Quadratic regression.
I. INTRODUCTION It is very significant for experts, teams' managers, players and fans to utilize the techniques and methodologies of data mining in the field of sports. National Basketball Association (NBA) league is one of the most well-known leagues in the whole world and hence it is more interesting to apply data mining techniques on the NBA league. The most exciting in sports is to forecast the next stage whether to expect game outcome or to predict teams' rank in the next season. Also, it is exciting to detect outstanding players depending on the statistics collected over 30 years and therefore more accurate players' ranking. Moreover, the team's managers and experts are interested in players' positions and related statistics in order to study and outline the strategies of their teams or opposing teams. Outstanding Players Detection: How can experts confirm outstanding players list? Good idea is to locate the most influencing features (statistics such as number of points or rebounds made) that classify players and then use these features to rank the players. Positions Inference: Can players' positions determine statistics? can statistics determine positions? It is very important for experts to have this information and hence form better field strategies. Ranking Forecast: How can we use historic NBA data to predict future seasons? People are interested either in predicting individual game outcome or in forecasting new season rank. In this report, I will focus on predicting the teams' ranking as well as to explore the most important statistics to determine the season winner. The next sections are organized as follows: related work discussing existing or similar solutions; followed by describing the datasets and their pre-processing; followed by the algorithms and methodologies implemented to resolve the problems; followed by showing the result and analysis.
II. RELATED WORK [1] The effort was to predict the outcomes of individual Major League Baseball (MLB) games as well as to explore the more important features of a team. The data was aggregated to have 18,717 individual games' score between 1971 and 2000. Using 25 features, the project implemented logistic regression, Naive Bayes, Support Vector Machine (SVM) and an ensemble classifier. The pre-processing of features was done in Perl and others in Matlab except for SVM that was done by SVM-light. This work achieved an accuracy of 50-58%. [2] This paper focused on the National Basketball Association (NBA) to predict and analyze the performance. Data was collected from 2007-2008 NBA seasons for both teams and players. The paper initially shows that there is no relation between the statistics and time as the season progress. Logistic regression was implemented in Matlab to predict team performance using 25 features to achieve 66.7% accuracy. Also, SVM was implemented using SVM-Light with 67.8% accuracy.
Page | 195 Research Publish Journals