This study focuses on addressing traffic congestion and enhancing Intelligent Transportation Systems (ITS). Anomalies in data are managed through polynomial fitting, and daily and hourly traffic flow on highways is analyzed. It is observed that traffic is heaviest on weekends compared to weekdays. Probability statistics are then utilized to tackle potential errors caused by vehicles transitioning between segments, calculating leakage rates based on toll booth passage probabilities, with passenger vehicles showing the highest leakage rate. A probability model is developed to determine toll lane numbers and emergency systems at each site. Finally, a BP neural network prediction module optimizes speed limits for various segments over the next 24 hours in real-time. In conclusion, the BP neural network model excels in traffic flow prediction, significantly reducing errors compared to other models. This research provides substantial support for Intelligent Transportation Systems and related fields, vital for mitigating congestion and improving road networks.