Large amounts of data and information is generated for different purposes every day in the traffic system. There is an increased interest within the Swedish Transport Administration (STA) in using this data for planning of maintenance, traffic management and strategy work. In this report the first steps towards such a system is developed by the process of defining a business objective, collecting data, understand the data, prepare the data, create a model and evaluate the results. All these different steps were important in the performed study. To find a good case creating business value required lots of discussions and interviews with key figures at STA. The case investigated is to predict the traffic situation on four road segments in Gothenburg based on two years of data for the traffic situation, weather, and road situation including accidents and road works.

The data for primarily weather and traffic are not collected for the purpose of being used for this application. This is one reason for that data is missing from some of the data sets for different time periods. One conclusion from the project is that data analysts must be included not only in the data analyze phase, but also in the data collection phase to achieve good results.

Different methods for creating data driven models are evaluated and compared based on the two year period of data available. It is found that linear regression performs better than tree-based classification and prediction method regarding performance, while the tree-based method more clearly can create understanding for what variables that correlate to the traffic situation.

The methods for developing models based on the data used in this project are generic and are possible to be used when larger data sets are available. Additional data sources, such as events in the city and building works may also be included in such analysis. Furthermore, it is found valuable to have the possibility to develop models on a local computer based on a smaller data set, and make the final computations based on the larger data sets in a cloud based solution.