Slicing the Past to Predict Future

In this research, a heuristic approach is proposed and experimented with to detect warning signs based on historical data. The paper makes use of COVID-19 positive count reported by different states of the USA for illustration of the methods. These are data in the form of several single variable time series related to a concept. The basic idea is to divide the total time period into smaller segments and examine changes within and between series in the segments. Algorithms for clustering series and concordance are used as tools of the heuristic. The approach is to observe the behavior of members of clusters in the segments and predictions are formulated based on observed changes which is heuristic part. Hidden Markov Model is used for change detection and series clustering. Concordance is used for comparing similarity in behavior of series in segments.