Extracting Hidden Patterns Within Road Accident Data Using Machine Learning Techniques

Road accidents may not be stopped altogether, but can be reduced. Driver emotions such as sad, happy, and anger can be one reason for accidents. At the same time, environment conditions such as weather, traffic on the road, load in the vehicle, type of road, health condition of driver, and speed can also be the reasons for accidents. Hidden patterns in accidents can be extracted so as to find the common features between accidents. This paper presents the results of the framework from the research study on road accident data of major national highways that pass through Krishna district for the year 2013 by applying machine learning techniques into analysis. These datasets collected from police stations are heterogeneous. Incomplete and erroneous values are corrected using data cleaning measures, and relevance attributes are identified using attribute selection measures. Clusters that are formed using K-medoids, and expectation maximization algorithms are then analyzed to discover hidden patterns using a priori algorithm. Results showed that the selected machine learning techniques are able to extract hidden patterns from the data. Density histograms are used for accident data visualization.