Correlating Temporal Rules to Time-Series Data With Rule-Based Intuition

摘要

Analysts are frequently confronted with time-series data. A simple form is magnitude (or count) and time frame, whether the data is number of e- mails sent, number of cell phones called, purchases made by volume or cost, or a variety of other time-derived data. Studying the temporal dimension of data allows analysts more opportunities to find relational ties and trends in data, classify or group like activity, and even help narrow the search space of massively complex and large datasets. This thesis presents a new approach called the Rule Based Intuition (RBI) system that can evaluate time-series data by finding the best fitting rule, from a repository of known rules, to quickly infer information about the data. This approach is most applicable for analysts viewing large sets of data who wish to classify or correlate data from users' temporal activity.