APT: Approximate Period Detection in Time Series

Period detection from time series is an important problem with many real-world applications such as weather forecast, stock market predictions, electrocardiogram analysis, periodic disease outbreak. In this work, we present a novel approximate period detection method for time series. The simplicity of our algorithm and its adaptability for high dimensional datasets using renowned tools and techniques such as locality sensitive hashing (LSH) and MapReduce (using the Hadoop framework for example) make it easier to implement for practical purposes. We performed experiments to compare our technique with a classic period detection technique and two state-of-the-art techniques in terms of accuracy and noise-resilience.