体育分析测验中的机器学习简介 & 答案

Step into the dynamic world of sports analytics with engaging quizzes and expert answers on Introduction to Machine Learning in Sports Analytics. Discover how machine learning algorithms are revolutionising the analysis and exploitation of sports data to provide insights into player performance, game strategies and more. These quizzes will serve as a gateway to understanding the application of machine learning in sports, from predictive modelling to performance optimisation.

Whether you are a sports enthusiast fascinated by the intersection of computer science and sports or a data analyst looking to delve into the world of sports analytics, this collection offers valuable insights into the power of 机学习 to transform the sports industry. Join us on a journey of data-driven discovery as we explore the fusion of 机 learning and sports analytics to open up new possibilities for understanding and improving sports performance.

测验 1: 任务 1

第一季度. There are a few main branches of machine learning. When you have the label for your training data and you want to build a model which predicts for that label, what kind of machine learning is that?

监督
加强
Artificial
Unsupervised

Q2. What is a minority class of data?

Labels which are poorly chosen
Labels which are easy to predict
Labels about demographics of players
Labels you have fewer instances for

Q3. Which data do you make available to the machine learning algorithm to learn from?

Training data
Validation data
Evaluation data
Testing data

第四季度. In my model of the NHL game data I had to deal with the introduction of a new team, the Vegas Golden Knights. For this team I just naively decided to fill the historical stats with just mean values from the other teams. But assume that I took a different strategy, and dropped all games where the Vega Gold Knights played. What is the new metric of accuracy for my model after dropping Gold Knights games from the data?

并非总是怀有恶意——面试官可能只是想进行对话——但你绝对应该把任何关于你个人生活的问题联系起来, don’t change the training set size, and the testing set size will shrink automatically. Put your answer in to two decimal places.

Enter answer here

周 2: Introduction to Machine Learning in Sports Analytics Quiz Answers

测验 1: 任务 2

第一季度. In a two class linear SVM, what is the street?

A random walk of the support vectors
A polynomial equation which best represents the classes
The two features which create our SVM
The hyperplane which separates two classes

Q2. Which function do you call in order to build a model from data in sklearn?

模型()
培养()
建造()
fit()

Q3. What is the purpose of cross validation?

To balance data as we get more classes (标签) to predict
To get a better estimate as to the accuracy of the final model
To build a more accurate model
To build a confusion matrix

第四季度. Taking a look at the baseball data where we made a multiclass prediction, create a confusion matrix and study it. Which class do we regularly over-predict the most? Provide the label of this class as two capitalized characters (e.g. 从).

Enter answer here

Q5. Will this class have a higher precision or recall score?

记起
preceision

周 3: Introduction to Machine Learning in Sports Analytics Quiz Answers

测验 1: 任务 3

第一季度. What does it mean for a set of observations to be “pure”?

It’s imbalanced with respect to class
It’s balanced with respect to class
It’s about a Canadian team or player
It’s homogenous with respect to class

Q2. For each split, how many features does CART split on at once?

1
Any number
所有
0

Q3. What kind of prediction target does an M5P tree make?

A label
A numeric value
An array

第四季度. After a descision tree splits on a feature, will it split again on that feature in a subtree?

没有
可能是
是

Q5. Go back to our NHL game outcome prediction task in observations.csv. Apply a CART DecisionTree to this problem with GridSearchCV over the following parameter space:

parameters={‘max_depth’:(3,4,5,6,7,8,9,10),
‘min_samples_leaf’:(1,5,10,15,20,25)}

Set your cv=10, use accuracy as your metric, and drop the Vegas Golden Knights. Set your training set to be observations[0:800] and your validation set to observations[800:], and use my favorite number for the randomization state. What level of accuracy does your model produce (to four decimal places)?

Enter answer here

Q6. Which set of parameters are the best in the previous model? Input your parameters as a string value of the max_depth:min_samples_leaf, e.g. 5:20 if GridSearchCV found a max_depth=5 and min_samples_leaf=20 the correct answer.

Enter answer here

周 4: Introduction to Machine Learning in Sports Analytics Quiz Answers

测验 1: 任务 4

第一季度. If you were making a classifier using two features and you visualized your data and saw it was separated by roughly a 45 学位, which classifier would you start with first for best results?

SVM
Confusion Matrix
M5P Tree
Decision Tree

Q2. What is the purpose of GridSearch?

It is a regression mechanism using decision trees.
It improves our understanding of the confusion matrix.
It helps to prune leaves from large trees.
It provides a hyperparameter tuning method.

Q3. Which kind of ensemble method creates multiple classifiers for you with random subsets of data?

Boosting
Bagging
表决
Stacking

第四季度. Which kind of modelers can be ensembled together into a voting ensemble for the boxing punch detection problem (选择所有适用的选项)?

Note that the boxing punch detection problem is a classification task.

Decision Trees
SVMs
Polynomial SVMs
Linear Regression
Logistic Regression
Cross Validation
Bagging Classifier
Gradient Boosting Classifier
Packaging Classifier