Convolutional Neural Networks on Multichannel Time Series of Smartphone Applications for Gender or Age Range Classification
Convolutional Neural Networks on Multichannel Time Series of Smartphone Applications for Gender or Age Range Classification
复制标题
DOI：
10.1109/iiai-aai50415.2020.00109
复制
发表时间：
2020-09
期刊：
影响因子：
--
通讯作者：
Hiromi Kondo;Fumiyo N. Kondo
中科院分区：
文献类型：
--
作者：
Hiromi Kondo;Fumiyo N. Kondo
在这项研究中，我们开发了一种使用卷积神经网络(CNN)的性别和年龄范围的分类方法。我们使用智能手机应用程序的时间序列训练了一个架构--连接包含性别和年龄范围的数据集的持续时间和人口统计数据。这些输入是时间序列所有应用类别的大数据形式，自动从拥有智能手机的个人的单一来源小组收集。性别和年龄范围的人口统计数据被用于预测。为了识别个人的人口统计特征，本研究提取了移动网站访问活动的有效特征。我们提出了一种基于原始输入的特征学习模型来解决基于深度卷积神经网络的性别和年龄范围的分类问题，这是数字营销中一项重要但具有挑战性的任务。对于只能获得网络应用程序日志，而无法获得性别和年龄范围的人口统计数据的公司来说，这项研究可能提供了一种可能的解决方案，可以应用于数字营销。
In this study, we developed a classification method for gender and age ranges using convolutional neural networks (CNNs). We trained an architecture using a time series of smartphone application-connecting durations and demographics of datasets containing gender and age ranges. The inputs were in the form of big data for all application categories of the time series, automatically collected from a single source panel of individuals with smartphones. The demographics of gender and age ranges were used for prediction. To identify the demographics of individuals, this study extracted effective features of mobile website access activities. We proposed a feature learning model from the raw inputs to solve the problem of classifying gender and age ranges using a deep convolutional neural network, which is an important but challenging task in digital marketing. For companies that can obtain only web application logs but not the demographics of gender and age ranges, this research may provide a possible solution that can be applied in digital marketing.