SQL-BASED APPROACH TO SIGNIFICANT INTERVAL DISCOVERY IN TIME-SERIES DATA
With time-series data, events (like turning off a light, opening garage door, turning on TV) occur with a high degree of certainty not at specific time points but within time intervals (sequence of time points). So, it is useful for applications to consider data as contiguous time points. The smallest interval that satisfies the criteria of interval-confidence (i.e., ratio of total support of participating time points and the number of days) is termed as Significant Interval (SI). Significant Interval Discovery (SID) algorithm finds SIs from time-series data.
The main focus of this thesis is on the improvement of existing SID algorithms and the design and development of new SQL-based algorithms which work directly on Relational Database Management System (RDBMS). The experiments compare the performance of the main memory SID against SQL-based SID. The larger goal of this thesis is to achieve scalability.