Using Machine Learning to Aid in Data Classification: Classifying Occupation Compatibility with Highly Automated Vehicles
Using Machine Learning to Aid in Data Classification: Classifying Occupation Compatibility with Highly Automated Vehicles
复制标题
DOI：
10.1177/1064804620923193
复制
发表时间：
2020-05
期刊：
影响因子：
--
通讯作者：
A. Kamaraj;John D. Lee
中科院分区：
文献类型：
--
作者：
A. Kamaraj;John D. Lee
数据分类是人为因素研究的核心，而手动数据分类既繁琐又容易出错。监督学习使分析师能够通过手动对一些案例进行分类来训练算法，然后让该算法对许多案例进行分类。然而，算法通常无法利用人类的洞察力。为了解决这个问题，我们通过无监督学习和数据可视化来增强监督学习。无监督学习突出了潜在的分类错误，解释了潜在的分类，并识别了值得手动分类的其他情况。我们使用职业信息网络数据库将职业分类为可能在自动驾驶车辆中执行的任务来说明这一点。
Data classification is central to human factors research, and manual data classification is tedious and error prone. Supervised learning enables analysts to train an algorithm by manually classifying a few cases and then have that algorithm classify many cases. However, algorithms often fail to leverage human insight. To address this, we augment supervised learning with unsupervised learning and data visualization. Unsupervised learning highlights potential classification errors, explains the underlying classification, and identifies additional cases that merit manual classification. We illustrate this using the Occupational Information Network database to classify occupations as having tasks that might be performed in an automated vehicle.