官术网_书友最值得收藏!

  • Deep Learning By Example
  • Ahmed Menshawy
  • 344字
  • 2021-06-24 18:52:43

Feature engineering

Feature engineering is one of the key components that contribute to the model's performance. A simple model with the right features can perform better than a complicated one with poor features. You can think of the feature engineering process as the most important step in determining your predictive model's success or failure. Feature engineering will be much easier if you understand the data.

Feature engineering is used extensively by anyone who uses machine learning to solve only one question, which is: how do you get the most out of your data samples for predictive modeling? This is the problem that the process and practice of feature engineering solves, and the success of your data science skills starts by knowing how to represent your data well.

Predictive modeling is a formula or rule that transforms a list of features or input variables (x1, x2,..., xn) into an output/target of interest (y). So, what is feature engineering? It's the process of creating new input variables or features (z1, z2, ..., zn) from existing input variables (x1, x2,..., xn). We don't just create any new features; the newly created features should contribute and be relevant to the model's output. Creating such features that will be relevant to the model's output will be an easy process with knowledge of the domain (such as marketing, medical, and so on). Even if machine learning practitioners interact with some domain experts during this process, the outcome of the feature engineering process will be much better.

An example where domain knowledge can be helpful is modeling the likelihood of rain, given a set of input variables/features (temperature, wind speed, and percentage of cloud cover). For this specific example, we can construct a new binary feature called overcast, where its value equals 1 or no whenever the percentage of cloud cover is less than 20%, and equals 0 or yes otherwise. In this example, domain knowledge was essential to specify the threshold or cut-off percentage. The more thoughtful and useful the inputs, the better the reliability and predictivity of your model.

主站蜘蛛池模板: 徐州市| 陆川县| 万盛区| 兴化市| 贵州省| 衢州市| 武陟县| 平顺县| 通城县| 宜君县| 宁强县| 沙河市| 黄骅市| 清河县| 武安市| 科尔| 周宁县| 辰溪县| 神农架林区| 陇西县| 淅川县| 淮阳县| 叙永县| 嘉黎县| 柳林县| 景宁| 平湖市| 五莲县| 威宁| 乌恰县| 陆良县| 贵港市| 八宿县| 马鞍山市| 台湾省| 衡水市| 镇安县| 平江县| 蓬溪县| 潮州市| 泰来县|