seq2vec: Analyzing sequential data using multi-rank embedding vectors

作者：

Highlights：

• Resolving challenges in sequential data analysis: high dimensionality, time variance, privacy.

• Proposing a new approach called seq2vec, encoding sequential data into a multi-rank embedding.

• Comparing the performance of seq2vec using the raw data for the purposes of clustering, classification and forecasting.

• Suggesting a specific CNN-LSTM autoencoder-based implementation of seq2vec.

摘要

•Resolving challenges in sequential data analysis: high dimensionality, time variance, privacy.•Proposing a new approach called seq2vec, encoding sequential data into a multi-rank embedding.•Comparing the performance of seq2vec using the raw data for the purposes of clustering, classification and forecasting.•Suggesting a specific CNN-LSTM autoencoder-based implementation of seq2vec.

论文关键词：Data embedding,Sequential data analysis,Event data,Deep learning,Multi-rank embedding,Vector embedding,Time series analysis

论文评审过程：Received 31 August 2019, Revised 22 July 2020, Accepted 23 July 2020, Available online 3 September 2020, Version of Record 22 September 2020.

论文官网地址：https://doi.org/10.1016/j.elerap.2020.101003