AXA Driver Telematics Analysis Kaggle competition

This student’s project is based on telematics data provide by AXA for a Kaggle completion, and was supervised by Conceição Amado, Prof, Math Department, IST. The objective of this work is to develop an algorithm that allows profiling for each driver, namely: knowing if a driver makes short or long trips; If attends more curvy roads (e.g. urban) or rectilinear (e.g. highways); accelerating a lot when he started from a null speed; they use the same roads, etc. Answering each of these questions and put them in order is what is intended in this work to achieve a unique “fingerprint” for each of the drivers and so distinguish them from others. Principal components analysis and the Ramer–Douglas–Peucker algorithm were the methods applied to try to achieve a solution. (J. Gois, C. Amado)

image001

Stochastic models for prediction of pipe failures in water supply systems

The failure prediction process plays an important role in infrastructure asset management of urban water systems. This process aims at assessing the future behavior of a urban water network. However, failure prediction in urban water systems is a complex process, since the available failure data often present a short failure history and incomplete records. In this study, the single-variate Poisson process, the Weibull accelerated lifetime model and the linear extended Yule process, were implemented and explored in order to identify robust and simple models that combine good failure prediction results using short data history. (A. Martins, C. Amado, J.P. Leitão)

image001

Statistical evaluation of image quality measures: from visualization to quantification

 Image quality assessment plays an important role in the field of image processing. Measuring the quality of an image is a difficult process since human perception is affected by physical and psychological parameters. This work aims to present a statistical approach, according to the performance of the quality measures, in alternative to the approaches based only in the characteristics and properties of the measures or based in a subjective evaluation of image quality. Different statistical methods were used, in particular Clustering Analysis and Analysis of Variance. (P. Neves, C. Amado, E. Carrasquinha)

image001

image001

Data reconstruction of flow time series in water distribution networks

The presence of missing values in flow data severely restricts the use of these data for billing/customers’ management and network control in water distribution systems. Missing values are frequent due to problems with metering, acquisition and flow data storage. In order to reconstruct the missing data, in this project a new procedure is developed, based on a forecasting model which can accommodate the several seasonality cycles present in the data. (R. Barrela, C. Amado, D. Lolureiro)

image001

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: