This period’s research lecture is given by Dewi Peerlings, member of the KE (Quantitative Economics) department since 2018, and currently finishing her PhD. She will present her current research area: Road Sensor Traffic Flow Density Estimation Using Neural Networks. If you’re interested in a preview of the lecture, Dewi summarised it below:

Explaining traffic flow is important for planning purposes and for developing intelligent transport systems. Such estimations often are based on road sensor data which contain missing and erroneous data points due to sensor malfunctioning. We propose to apply a non-parametric probability density function (PDF) estimation based on Neural Networks (NN) to pre-process road sensor data and to report the properties of the traffic flow distribution. We extend the literature in two ways. First, we propose a density estimation method instead of the standard point estimation method as to account for raw data and estimation uncertainty as well as to report the probability of e.g. extreme events in traffic flow. Second, we design NNs to specifically allow for correlations between adjacent road sensors. We apply the proposed method to data obtained from highways in The Netherlands belonging to different sensors on the same highway section and coded as time series of vehicle counts. We show that the proposed model captures properties of these data and can be used as a pre-processing and density estimation method.

The lecture will take place at SBE, TS53 C-1.03, at 19:00.

The evening will end with drinks at the Preuverij. 🍻