this is for holding javascript data
Juan de Monasterio edited section_Ongoing_work_The_CDR__1.tex
almost 8 years ago
Commit id: c50b1bf9e3b5cfc0fc2fd239c499ee6d8955535e
deletions | additions
diff --git a/section_Ongoing_work_The_CDR__1.tex b/section_Ongoing_work_The_CDR__1.tex
index 81c53c3..1753b40 100644
--- a/section_Ongoing_work_The_CDR__1.tex
+++ b/section_Ongoing_work_The_CDR__1.tex
...
\]
\subsection{Model Attributes}
Setting the five month timespan limitation, CDRs are being processed to extract features at the user level.
For every individual, Setting the
top ten most used antennas five month timespan limitation, CDRs are
logged along with being processed to extract features at the
amount user level. The quality of
use in each one. Users are tagged as 'epidemic' if their home antenna is in the
classification will rely heavily on the
risk area and 'exposed' if any of ability to characterize the
antennas used are in user and his communication pattern as differentiating as possible. In general, the
epidemic zone. A features constructed reflect calling and mobility
diameter is also processed from the radius of the convex hull defined patterns. Differentiating by the
user's logged antennas. This length is representative of time they were done during the
radius of influence of that individual week and
we are expecting to see a high correlation between a high mobility-diameter and high long-term movements of people. tagging the action or object if it is epidemic.
Calling information is aggregated and binned according to the hour The model's first version consists of the
day and during the weekends. For every user, the duration and individual count of these calls are processed, differentiating between calls made to vulnerable and non-vulnerable users. following features:
Finally, \begin{itemize}
\item Antennas: The top ten most used antennas with the number of uses. From this, users were tagged as 'epidemic' if their home antenna is in the epidemic area and 'exposed' if any of the ten antennas logged is in the risk area.
\item Mobility diameter: The user's logged antennas define a convex hull in space and the radius of the hull is taken to be as the mobility diameter. This length is representative of the area of influence of that individual. We are expecting that these feature be correlated with long-term migrations.
\item Neighbours: From the social
information processed graph built from the
data include CDRs we extracted the total
amount count of
epidemic, exposed neighbours in the communicaction graph and
the total
neighbours count of epidemic neighbors.
\item Calls: The total time and count of calls made during the five month period is aggregated per user. This information is also segmented according to the hour of the day that
any given user interacts with over the
timeperiod. calls were made and whether they were made during the weekends. Special care was taken with calls placed to and from vulnerable users and aggregated accordingly.
\end{itemiize}
\subsection{Supervised Algorithms}
Based on