Car accidents prediction based on cloud computing in data analysis

Abstract:
By using cloud computing and machine learning to analyze traffic accident data in the United States, this paper explores the practicability of cloud computing in data analysis and the possibility of using this model for traffic accident early warning. The experiment first stores the data in Hadoop File System and reads it out as Spark dataframe, and then uses Spark SQL for preliminary analysis and preprocessing. Then, traffic accident predictions are made using four machine learning algorithms in Spark MLlib. After comparing with the original data, logistic regression, random forest, and Gradient Boosted Trees all performed well, and the accuracy rate can reach more than 78% although the data itself is biased slightly, confirming that this model can be used to predict traffic accidents.