A topic modeling framework for spatio-temporal information management

Highlights：

• Propose a neural network using transfer learning techniques to enhance the framework ability to detect unrelated messages over data streams existing in twitter. We focus our attention in healthcare to present examples.

• Create automatic deep cleaning method to enhance the quality of data to perform better classification in outlier and topic detection.

摘要

•Propose a robust procedure to take a decision for selecting the best topic model. We design an adaptive framework to use gained knowledge for improving the result over time. For our case study we used four topic modeling techniques and report the result of the evaluation techniques.•Propose a neural network using transfer learning techniques to enhance the framework ability to detect unrelated messages over data streams existing in twitter. We focus our attention in healthcare to present examples.•Create automatic deep cleaning method to enhance the quality of data to perform better classification in outlier and topic detection.

论文关键词：Spatio-temporal real time analysis,Traceability,Topic modeling,Visualization,Artificial intelligent,Transfer learning

论文评审过程：Received 10 December 2019, Revised 11 June 2020, Accepted 11 June 2020, Available online 6 July 2020, Version of Record 20 October 2020.