Study Of Match Correlation Analysis And Prediction Based On Historical Data Of Football Matches

The sports industry under the background of the big data era brings some new development and opportunities,especially the big data mining in the sports industry provides the technical support for industry iteration,promotes the common innovation and development of the sports industry,and accelerates the process of technology in the sports industry.Soccer as a popular item in the field of sports has received wide attention from fans at home and abroad,while data analysis and data mining in the field of soccer are also developing rapidly.Match prediction,a passion of betting and sports participants,has turned out to be one of the main hot spots of soccer business activities and a hot spot for the industry.From previous studies,it is found that betting companies’ data analysis of soccer matches mostly based on data to win,while the team side focuses on the presence of players analyzing the situation before and after the match.The analysis of the factors influencing the match results is mostly done by the traditional method of calculating the complete influencing factors to achieve the purpose of match analysis and prediction.Due to the many factors affecting the outcome of a soccer match,it is quite difficult to predict the winner or loser of a soccer match in reality.Because of this,both the business,academic and sports communities prefer to conduct research on correlation analysis of soccer matches.As a scientific evaluation method for quantitative analysis of multiple factors,the Fuzzy grey relational analysis method is very suitable for correlation analysis of soccer matches,thus it has good results for problems with complex influencing factors.The correlation analysis method in this thsis can explore the influencing factors after feature engineering in a more practical way,discover valuable information,and provide soccer participants with a proven analysis and prediction method.In this thsis,the match data from soccer website(http://op1.win007.com)is used as the data source,and the historical data of five complete seasons of serceral main European leagues from 2014-2019 by web crawling technology.The collected data are subjected to data cleaning and feature engineering in detail,and fuzzy affiliation combined with gray correlation analysis is used to perform multi-factor analysis on the soccer historical data.Finally,the extracted high-contribution features are used to predict soccer match wins and losses by using a decision model to discover influencing factors that are valuable for soccer match analysis and prediction.This is stydy contains the following aspects: first,we conducts a background survey on the current situation of data analysis of soccer matches and related sports events,and describes the data cleaning and feature engineering ideas and processes of multi-factor correlation analysis of soccer matches,and systematically introduces the application scenarios of fuzzy grey correlation analysis;Next,we through the processing and understanding of the basic match data,a set of feature engineering models are built to meet the multi-factor correlation of soccer matches,and the results of specific matches are compared and validated to ensure the generalization performance of the prediction effect;Then,the influential factors affecting soccer matches extracted from the multiple association analysis models are compared with Light GBM,XGBoost and Random-Forest decision models to make decision judgments and finally find out the set of influential features on the match results.Finally,we through the analysis of multiple influencing factors,the important features influencing the match results are unearthed and the issues that should be paid attention to in the correlation analysis of soccer matches are proposed.Compared with the traditional soccer match analysis and prediction studied,the match correlation analysis and prediction method using soccer match history data not only enables batch collection and processing of soccer match data,but also constructs a reasonable set of soccer match feature engineering models to support the analysis and prediction of match results.The final experimental results show that the prediction result of the soccer match is only 0.28% less than the prediction result using all the influencing factors with a sharp reduction of 67.5% in the set of factors after the multi-factor correlation score,and the actual prediction accuracy reaches 73.01%.The comparison of multiple correlation analysis models revealed that the influencing factors were mainly focused on the recent home and away status,odds situation,home and away offensive and defensive status and other dimensions,indicating that the recent engagement situation of the home team and the away team and the previous match against the two teams will have an impact on the outcome of the next match.While the initial odds,as the pre-match reference basis provided by the bookmaker to the fans,is also a factor that cannot be ignored for the prediction of the match outcome.