Road network matching shenanigans — The purpose of this article is to complement and correct the previous one on the same subject. In that article, I presented an approach to reconstructing missing map-matched data from the Extended Vehicle Energy Dataset¹ (EVED) [1]. My technique explored the mathematical properties of triangles to find the lost data…

Triangles have mighty properties for geospatial queries — Triangles are shapes with many practical geometric properties. In this article, I illustrate using such properties when performing opportunistic optimizations while solving a particular geospatial problem: the recovery of missing map-matched information. I started exploring the Extended Vehicle Energy Dataset¹ (EVED) [1] a while ago to search for compelling geospatial…

How can we quickly find overlapping trajectories? — While traveling through space, an object describes a trajectory. We can think about a trajectory as a function of time that outputs positions in space. Conceptually, trajectories are continuous functions, although we pragmatically use their discrete versions. A discrete trajectory is a time-ordered collection of points in space where we…

This article explains how to use optimization to perform stratified K-Fold cross-validation on a grouped dataset — Cross-validation is a common resampling technique to get more mileage from your dataset. The procedure involves taking repeated independent samples from the original dataset and fitting them to the desired model. Cross-validation is helpful for model selection, as it provides better generalization performance estimates than the holdout method. …

This article explains how to estimate travel times using known speed vectors indexed by quadkeys — How long is your road trip going to last? If you are like me, you go to Google maps or Here maps and query the route directly. You will not only get the fastest or shortest paths, but you will also get estimates of your travel times. …

How can we determine the equivalent DBSCAN ε parameter for HDBSCAN-generated clusters? — HDBSCAN is a hierarchical density-based clustering algorithm that works under simple assumptions. At a minimum, it only requires the data points to cluster and the minimum number of observations per cluster. The algorithm accepts a distance matrix if the data has a non-obvious associated distance metric. Like its predecessor, DBSCAN…

This article explains how to perform a stratified split of a grouped dataset into train and validation sets — One of the most frequent steps on a machine learning pipeline is splitting data into training and validation sets. It is one of the necessary skills all practitioners must master before tackling any problem. The splitting process requires a random shuffle of the data followed by a partition using a…

Learn how to create custom tiles for your interactive maps — Interactive maps are now a staple of our everyday digital life. We use them to learn of our whereabouts, plan the next trip, or review our past travels. In a professional setting, maps became priceless tools for all manner of businesses in planning, operations, and analytics. An interactive map displays…