Article Preview
Top1. Introduction
Football is one of the most popular sports. Predicting the results of football matches is interesting to many, from fans to punters. It is also interesting as a research problem, in part due to its difficulty, because the result of a football match is dependent on many factors, such as a team’s morale, skills, current score, etc. So even for football experts, it is very hard to predict the exact results of football matches (Owramipur 2013). However, since various types of football matches have outstanding similarities in some respects, in theory, it is possible to find the law from a large number of football matches to find a way to judge the level of victory or defeat (Guan 2021).
With the rapid development of artificial intelligence, machine learning algorithms have been widely used in real life, such as face recognition, stock price prediction, etc. Hence, how to build a football match prediction model based on machine learning algorithms and use scientific methods to solve the prediction problem has become a topic of interest to experts and scholars. Fortunately, researchers have constructed a variety of football match result prediction methods, and have achieved some results. For example, literature (Igiri 2015) used the SVM algorithm to study the factors that affecting the results of the British Championship. Lei used the logistics regression algorithm in machine learning to analyze and process the historical results of football matches, and realized the prediction of the football match results (Lei 2019). Guan also achieved effective result prediction of the Chinese Super League team by suing fuzzy neural network and extreme learning machine (Guan 2021).
However, traditional machine learning methods usually use centralized computing, which need to upload previous matches results to the cloud server for centralized processing. When there is a lot of data to be processed, the current methods will face new problems, that is, with the increase of match data, the amount of data that needs to be uploaded to the cloud server will increase significantly, which will occupy a large amount of bandwidth and cause network congestion. And if the classification tasks are all completed by the cloud server, the computing resources cannot be dynamically scheduled, which will occupy a large amount of server computing resources. In addition, in order to improve the learning ability of the classification model, it is usually necessary to increase the number of parameters. However, the increase in the amount of model parameters will cause the classification model to occupy too much computing and storage resources, which will further increase the computing pressure of the server.
In order to solve the above problems, this paper proposes a football match result prediction method based on edge computing and machine learning technology. Edge computing is a new computing method that appears to solve the problems of network congestion and cloud center computing pressure faced by cloud computing. Specifically, edge computing realizes the nearby processing of data by deploying edge computing nodes at the end close to the collection device, and these nodes have the capabilities of computing, storage and communication. And this manner can effectively alleviate network congestion, and sharing the computing tasks of the cloud center by edge nodes can effectively reduce the computing pressure of the cloud center (Wang 2020). Inspired by this, this paper splits the football match result prediction task into some sub-tasks that can be executed at multiple edge nodes, so that the task of feature extraction and classification is deployed on the edge computing nodes.
The remainder of this paper is organized as follows: Section 2 gives a summary of previous work on football result prediction. In Section 3, we describe the proposed architecture of football match result prediction method based on edge computing and machine learning technology. The experiments are provided in Section 4. Section 5 is the conclusion.