计算机科学
原始数据
离群值
分类
滞后
背景(考古学)
过程(计算)
数据挖掘
特征工程
特征(语言学)
机器学习
人工智能
工业工程
工程类
深度学习
情报检索
计算机网络
古生物学
语言学
哲学
生物
程序设计语言
操作系统
作者
Maxime C. Cohen,Paul-Emile Gras,Arthur Pentecoste,Renyu Zhang
出处
期刊:Springer series in supply chain management
日期:2022-01-01
卷期号:: 13-27
标识
DOI:10.1007/978-3-030-85855-1_2
摘要
This chapter covers several important pre-processing steps. Before implementing a demand prediction method, it is crucial to process the raw data in order to extract as much predictive power as possible from the different features available in the data. We discuss how to deal with missing data and how to test for outliers in the context of demand prediction. We then cover various concepts related to feature engineering for demand prediction, such as accounting for time effects and constructing lag-price variables. We end this chapter by discussing the practice of scaling features, and how to sort and export the resulting processed dataset. Each step is illustrated using the accompanying dataset.
科研通智能强力驱动
Strongly Powered by AbleSci AI