Too Long; Didn't Read
As a data scientist who has worked on geospatial data for more than one year, <a href="https://hackernoon.com/tagged/traffic-prediction" target="_blank">traffic prediction</a> has always been a great challenge for our team. We use a <a href="https://hackernoon.com/tagged/machine-learning" target="_blank">machine learning</a> algorithm for traffic estimation and a navigation system based on our live traffic estimated data. We have built a simple traffic estimation prediction that is used to predict navigation travel time. The output of our services is surprisingly accurate. On top of that, our travel time prediction has shown brilliant results, except in anomalies like vacation days or unpredictable changes in traffic behaviors. These anomalies happen less than 20 percent of the time but our users need more accurate prediction especially in anomalies moments, so here I will talk about the approach we took to solve the problem.