Author | Renato Boemer Translator | Sambodhi Planning | Liu Yan It may be difficult to explain machine learning to someone without a technical background.
This article was originally published on the Towards Data Science blog, authorized by the original author Renato Boemer, and translated and shared on the InfoQ Chinese website.
If you are a professional data scientist, you will often be asked a question-"What is your job?" It may be difficult to explain this problem to someone without a technical background.
The definition of machine learning given by Professor Tom Mitchell, a well-known computer scientist at Carnegie Mellon University, is
"A computer program that learns certain types of task T and performance index P from experience E, if its performance in task T (measured by P) increases with experience E."
Frankly speaking, in any informal conversation, it may be difficult to continue the conversation by citing this highly professional definition.
As a data scientist, you often need to explain technical terms to non-technical audiences. Therefore, every time I find myself explaining my work, I use the same technique my philosophy teacher used to use: football analogy. Even if people don't like football, they can link machine learning with football and the rules in some way.
Hopefully the football metaphor helps you understand or explain machine learning to others.
Players (data)
Obviously, there is no football game without players. It doesn't matter whether you are playing a professional game at Wembley Stadium or playing with friends on the street. Without players, those places are just empty football fields and streets.
For machine learning, data is like a player. Without data, nothing can be done. However, not all data sets are the same. Just like players, Ronaldo and Messi are great players. They surpassed people's expectations for a wonderful football game. But if I were to play, it would be impossible. Therefore, good players will have outstanding performance.
Similarly, there is a famous saying in data science: "garbage in, garbage out". No matter how superb your programming skills are or how profound your mathematical knowledge is, if you don't have a useful data set, your machine learning project is likely to disappoint your team.
Football Manager (data preparation)
一支足球队的成功离不开足球经理。即便拥有挑选顶级球员的豪华条件，英格兰国家足球队自 1966 年以来也再没有赢得过世界杯。足球经理负责决定谁将参加世界杯。同时，他也负责为球员提供指导，指导日常训练。这个过程很花时间，如果不能很好地完成，球队就不能为下届冠军做好准备。
据一份研究报告称，约 80% 的数据科学家会做数据准备和数据清理。数据专业人员必须将他们的数据集转化为机器学习模型可以学习的格式（例如，将数据归一化，处理空白值等）。不论对于数据科学家还是足球专业人士，这些都不是最令人兴奋的事情。
足球战术（机器学习模型）
球队要想夺冠，就必须根据每个对手的情况改变战术。举例来说，如果美国国家足球队面对四届世界冠军德国国家足球队，他们很有可能建立一个强大的防守体系。若美国队对阵冰岛足球队，则可采用强攻策略，采用不同的进攻战术。因此，一支经过良好训练的球队，只要做到战术合理，那么在 90 分钟内，很有可能进球并取得胜利。
机器学习从业者必须根据给定的特定数据集和期望的结果来决定要应用哪种算法或模型。举例来说，机器学习专业人员根据问题来选择预测模型：分类模型是关于预测标签的，而回归模型是关于预测数量的。因此，熟知哪些规则和技术是项目成功的关键。如，K- 最近邻、逻辑回归、朴素贝叶斯分类器和随机森林是一些常用的机器学习模型。
足球设备（硬件和软件）
足球在不同位置需要不同的装备和训练。举例来说，只有门将才能用手触球。因此，他们需要（特殊的）手套和独特的体能训练，而其他人则需要来回奔跑 90 分钟，并尝试用额头进球得分。另外，拥有强大赞助商的团队可以雇佣营养师、医学专家甚至数据科学家来分析表现数据。归根结底，设备和独特的专业人才能够帮助一支球队在世界杯上获得成功。
Similarly, to process a small data set (1000 rows × 5 columns) to create some graphics, these graphics can be generated on a standard laptop using Microsoft Excel, but if you want to extract data from multiple servers and process hundreds of Ten thousand rows of data requires a specific programming language Python and high-performance equipment with extraordinary computing capabilities.
Different alliances (domain expertise)
No matter where you go, there may always be people playing football. It may be children/adults, men/women, indoor/amateur, online/outdoor or amateur/professional, etc. It doesn't matter, someone is always playing. In addition, you will encounter a huge difference in skill level.
Football will not have defects due to different technical levels and game types. This is precisely the diversity and inclusiveness of football. Each technical level or type of competition can meet a particular need. Some people like to play outdoors on the grass, while others like to play with friends online. It doesn't matter, these people specialize in a certain type of football.
Machine learning is like football. Different professionals have different expertise and work in their respective fields, for example, business and corporate fields (financial markets); academic and technical fields (research and development of new algorithms in universities).
to sum up
When you are becoming a machine learning expert, you are bound to explain your work to people from different backgrounds. This simple and effective analogy can help you make it easier for them to understand machine learning. Pay attention to the audience's general impression of football, and establish an easy-to-remember connection with machine learning. I hope that now, you have an interesting analogy to explain complex topics in daily life metaphorically and popularly.