An Application of IoT Cloud Network: Traffic Hotspots Prediction

With millions and millions of intelligent vehicles connect to the GPS and Internet, IoT cloud network becomes an important medium to provide support and coordination by analyzing the big data. In this study, an application of predicting real-world traffic hotspots based on historical traffic information is introduced. The traffic road networks were retrieved from real-world maps provided by OpenStreetMap, whereas the vehicles were generated by Sumo traffic simulation. A parser program was designed to process the massive simulation results by sampling the road segments and formulating the nearby roads into subgraphs using Dijkstra’s algorithm as well as determining the traffic hotspots by investigating the location and velocity of individual cars at each time step and their waiting time on the roads. A neural network based on TensorFlow was then trained to predict whether if the traffic hotspots could occur when the road coordinates are entered. With future improvement, traffic hotspots’ prediction based on IoT cloud network could be useful for traffic route planningU of I OnlyUndergraduate senior thesis not recommended for open acces