this is for holding javascript data
Bart van Merriƫnboer edited Approach.tex
over 10 years ago
Commit id: 5359246c8e979a30674541e272fbfdd4e885321f
deletions | additions
diff --git a/Approach.tex b/Approach.tex
index 730b7d9..4051a0d 100644
--- a/Approach.tex
+++ b/Approach.tex
...
\subsection{Modelling approach}
We can assume our job arrivals to be governed by a Poisson process with parameter $\lambda$. However, our arrival times differ depending on the time of day (e.g. less jobs at night).
We The arrivals can then
describe our arrivals be modelled as an inhomogeneous Poisson process $\lambda = f(t)$, where we need to determine $f(x)$ from the data given.
Some days are busier than others, so we normalize the given data
(dividing by the number of arrivals in 24 hours) to determine $f(t)$ and write $\lambda = R_i f(t)$ where $R_i$ is the relative business of the day.
The easiest approach is to assume $f(t)$ is piecewise linear. For a given segment we can then simply use the fact that the maximum likelihood estimator (MLE) for a homogeneous Poisson population of $n$ samples is
its average
\[
\widehat{\lambda} = \frac{1}{n}\sum_{i=1}^n k_i.