登录查看更多内容

How can you handle categorical variables in logistic regression models?

由人工智能和领英社区提供技术支持

Logistic regression is a popular data mining technique for predicting binary outcomes, such as whether a customer will buy a product or not. However, many real-world datasets contain categorical variables, such as gender, color, or location, that cannot be directly used as predictors in logistic regression models. How can you handle categorical variables in logistic regression models? In this article, you will learn three common methods: dummy coding, effect coding, and one-hot encoding.

此文章中的业界达人

由社区从 3 条内容中精选。了解更多

Vino Duraisamy

Developer advocate @Snowflake ?? | Data & ML Engineer | Spark, Python, SQL, Airflow, Snowflake

1 Dummy coding

Dummy coding is a method of creating new binary variables, called dummy variables, for each level of a categorical variable, except for one reference level. For example, if you have a categorical variable with three levels: A, B, and C, you can create two dummy variables: D1 and D2, where D1 = 1 if the level is A, and 0 otherwise, and D2 = 1 if the level is B, and 0 otherwise. The reference level is C, which is represented by D1 = 0 and D2 = 0. You can then use D1 and D2 as predictors in your logistic regression model, instead of the original categorical variable.

添加您的观点

Vino Duraisamy

Developer advocate @Snowflake ?? | Data & ML Engineer | Spark, Python, SQL, Airflow, Snowflake
举报内容
While working with logistic regression models, you need to encode the categorical values in the independent variables. There are two popular encoding methods: 1. dummy variable coding 2. one hot encoding If the category values are nominal (for example, country codes), you should use one hot encoding. Nominal simply means that the categories are in no particular order. That is, one country code is not valued higher than the other and vice versa. If the category values are ordinal (for example, priority of a task - P0, P1 and P2), you should use dummy variable encoding. Ordinal means the categories are in a particular order and the order has a meaning attached to it. That is, P0 is urgent whereas P2 is not.

已翻译

赞

2 Effect coding

Effect coding is similar to dummy coding, but instead of using 0 and 1, it uses -1 and 1 to represent the levels of a categorical variable. For example, if you have a categorical variable with three levels: A, B, and C, you can create two effect variables: E1 and E2, where E1 = 1 if the level is A, -1 if the level is B, and 0 if the level is C, and E2 = 1 if the level is C, -1 if the level is A, and 0 if the level is B. You can then use E1 and E2 as predictors in your logistic regression model, instead of the original categorical variable. Effect coding can be useful when you want to compare the effects of each level to the overall mean effect of the categorical variable.

添加您的观点

加载更多内容

3 One-hot encoding

One-hot encoding is another method of creating new binary variables for each level of a categorical variable, but unlike dummy coding and effect coding, it does not exclude any level as a reference. For example, if you have a categorical variable with three levels: A, B, and C, you can create three one-hot variables: O1, O2, and O3, where O1 = 1 if the level is A, and 0 otherwise, O2 = 1 if the level is B, and 0 otherwise, and O3 = 1 if the level is C, and 0 otherwise. You can then use O1, O2, and O3 as predictors in your logistic regression model, instead of the original categorical variable. One-hot encoding can be useful when you do not have a natural reference level for your categorical variable, or when you want to avoid the dummy variable trap, which is a situation where the dummy variables are highly correlated with each other or with the intercept term, causing multicollinearity problems in your model.

添加您的观点

4 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?

添加您的观点

加载更多内容

Data Mining

+ 关注

给文章评分

我们借助人工智能创建了此文章。您认为这篇文章怎么样？

很棒不太好

举报此文章

查看全部

How can you handle categorical variables in logistic regression models?

1

2

3

4

1 Dummy coding

2 Effect coding

3 One-hot encoding

4 Here’s what else to consider

Data Mining

给文章评分

感谢您的反馈

更多Data Mining相关文章

更多相关阅读内容

How can you handle categorical variables in logistic regression models?

1

2

3

4

1 Dummy coding

2 Effect coding

3 One-hot encoding

4 Here’s what else to consider

Data Mining

给文章评分

感谢您的反馈

查看其他技能