Prediction Focused Topic Models via Feature Selection
Prediction Focused Topic Models via Feature Selection
复制标题
通过特征选择预测聚焦主题模型
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
F. Doshi
中科院分区：
文献类型：
--
作者：
Jason Ren;Russell Kunes;F. Doshi
监督主题模型通常被用来平衡预测质量和可解释性。然而，当模型（不可避免地）被错误指定时，标准方法很少同时满足这两个要求。我们介绍了一种新的方法，预测为重点的主题模型，使用的监督信号只保留词汇，提高，或至少不妨碍，预测性能。通过去除具有不相关信号的术语，主题模型能够学习与任务相关的连贯主题。我们在几个数据集上证明，与现有方法相比，以预测为中心的主题模型在保持竞争性预测的同时学习到更多连贯的主题。
Supervised topic models are often sought to balance prediction quality and interpretability. However, when models are (inevitably) misspecified, standard approaches rarely deliver on both. We introduce a novel approach, the prediction-focused topic model, that uses the supervisory signal to retain only vocabulary terms that improve, or at least do not hinder, prediction performance. By removing terms with irrelevant signal, the topic model is able to learn task-relevant, coherent topics. We demonstrate on several data sets that compared to existing approaches, prediction-focused topic models learn much more coherent topics while maintaining competitive predictions.
犯罪主题建模
影响因子：
6.1
作者：
Kuang, Da;Brantingham, P. Jeffrey;Bertozzi, Andrea L.
通讯作者：
Bertozzi, Andrea L.