Angela Wei
Visiting Research Assistant of the PhD program between 07.06.2010 and 29.08.2010.advisorIris Adäorganisational data
|
project description
Analyzing, modeling and forecasting of time series using segmentation, clustering and sequential pattern miningTime series data is a series of data points which occur in a specific order. With every different source (stocks, heart rate, sinusoidal movement, etc.), the data has unique behaviors — texture, patterns. The goal is to generate additional data of a particular sort with only a sample. New data must possess the same patterns and general attributes of the original data, basically extending the sequence from the end. Realistic generated data finds various applications, e.g. as a teaching or a testing tool, where controlled data is needed to target and operate on certain behaviors.
The prediction process is implemented as a process of sequential pattern mining. A sequential pattern is a "sequence of itemsets that frequently occurred in a specific order". Sequential Pattern Mining is "trying to find the relationships between occurrences of sequential events, to find if there exists any specific order of the occurrences" (Zhao 7). By extending association mining to account for ordering, predictions can be made. Frequent sequences of categorized data segments and their associated next element are collected. As with an association rule, when the more frequent sequences on this list appear, the next element is expected to follow.
The process can be seen as a cycle beginning with a set of data, analyzing it (segmenting, clustering, predicting clusters), and producing a new set of data. The flow is achieved in KNIME by creating new nodes and linking the new and existing nodes together.
Testing has been conducted on stock data — DAX closings for 20 years (Yahoo) — and generated periodic data.
curriculum vitae
2008-2012 | Bachelor of Arts in Computer Science at Princeton University, Princeton, NJ, USA |
2009 | Visiting researcher at Peking University, Beijing, China through University of Michigan REU exchange |