Extracting the Main Trend in a Data Set: The Sequencer Algorithm

Scientists aim to extract simplicity from observations of the complex world. An important component this process is exploration data in search trends. In practice, however, tends be more an art than a science. Among all trends existing natural world, one-dimensional trends, often called sequences, are particular interest as they provide insights into simple phenomena. However, some challenging detect may expressed manners. We present Sequencer, algorithm designed generically identify main trend dataset. It does so by constructing graphs describing similarities between pairs observations, computed with set metrics and scales. Using fact that continuous lead elongated graphs, can which aspects relevant establishing global sequence. Such approach used beyond proposed optimize parameters any dimensionality reduction technique. demonstrate power Sequencer using real-world astronomy, geology well images show that, number cases, it outperforms popular t-SNE UMAP techniques. This exploratory analysis, not rely on training nor tuning parameter, has potential enable discoveries wide range scientific domains. The source code available github we online interface at \url{http://sequencer.org}.