Towards a Predictive Model Architecture for Current or Emergent Pandemic Situations

摘要

This paper presents ideas towards building predictive models of the socio-economic interactions of autonomous population units (APUs). APU interaction models (AIM) can then be used to predict current or emergent pandemic situations. Our envisaged AIM system will comprise two fundamental components: (i) an existing generic discrete event simulator (DES) which will be adapted to socioeconomic interactions of an APU, and (ii) a new data mining toolset (DMT) which will be integrated to the simulation toolset. In a part of our early work, we will identify and filter the data that flows into the AIM system. The DMT will consolidate digitized data (e.g., satellite imagery) with data from public and private news sources including human observation. The DES together with historical data generated from the DMT will form an initial model of the socio-economic interactions of an APU relative to a pandemic, which is used to get a prediction of the existing conditions. The DMT is then employed to explore the data and discover previously unknown, valid patterns and relationships pertaining to the spread of infectious disease. The model will iteratively compare the predicted data to the real-world DMT data until an accurate predictive model of the pandemic is obtained.