Kunal Udawat - Data Science (FB5) Semesterprojekt

Modelling GPS Data and Visualizing on a map

Most trekking guides gives us an average time taken to complete the trek. However this time can vary greatly between people based on how fast or slow their trekking speed is. Usually average trek times are calculated using basic formulas used by each country, resulting in irrelevant times for most people. How can we use data science to predict trek times more accurately?

The answer is not so complex and with very limited data parameters, it is possible to predict accurate trek times. Our goal is to calculate trek times for individuals based on their personal trekking speed, fitness etc. The implementation of the data analysis, charts and predictions were done using exploratory data analysis, correlation, clustering and regression analysis. The data was sourced from Hikr, which has publicly available GPS data on various treks across European countries. The GPS data includes various details like speed, time taken, altitude in relation to the trails a user has completed. We started out with processing the GPS data and extract important features like altitude, length, ascent and descent data. Compared to real values the predicted moving times showed to be accurate. We also generated maps using the python library Folium to visualize GPS data on a map. The map can be enhanced with multiple features like min and max times. The study can be further enhanced when trekkers are grouped according to their speeds and users can quickly find which group (beginner, intermediate or advanced) they belong to. Thereby improving trek planning and avoiding unnecessary delays and dangers of under-estimating tougher trek routes.