Tuesday 22 November 2011

Frequency of events analysis

Different users have a list of their own activities during a day. These activities might have a chance to regconise as a pattern. To analyse this, we take the events which annotated from the real users during the FigureEnergy experiment to check the frequency of the type of events per day in term of appearance and energy usage. The results are shown as follows:

Figure 1. Frequency of events for user ecenergy22

Figure 2. Frequency of events for user ecenergy23

Figure 3. Frequency of events for user ecenergy24

Figure 4. Frequency of events for user ecenergy25

Figure 5. Frequency of events for user ecenergy30

Figure 6. Frequency of events for user ecenergy33

Figure 7. Frequency of events for user ecenergy34

Figure 8. Frequency of events for user ecenergy36
Posted by at

4 comments:

  1. enrico23 November 2011 at 07:48

    Hi Henry,

    Thank you for posting this.

    A minor thing: these are not time series, so it does not make sense to use a line plot. In these cases you should use bars (see http://www.mathworks.co.uk/help/techdoc/ref/bar.html )

    Turning to less mundane issues: what are the two colours? Are they week-day and week-end again? If so, how did you calculate things with the week-end gaps?

    Did you read again about Poisson distributions, yet? Do you see how you could relate this frequency data to those distributions? And especially their parameter estimation and prediction?

    I know right now you are working with very little data, so I do not expect very good results, but if you develop the methods (and more important the understanding) it should be easy to plug more data later..

    Can you please also try to plot these frequencies for all users together?

    ReplyDelete
  2. Henry Ngoc Cuong Truong23 November 2011 at 16:45

    Yes I need to turn them into the bar graph.

    I use multi-dimension y scale axes. As you can see from the graph, each colour represents for different frequency. The blue line is for the frequency of appearances of events. The green line is for the frequency of event usage.

    Yes, I think the frequency of events could be related to Poisson distributions. Additionally, FigureEnergy data can be divided into two themes: labels, and usage. I now try to set up a list of threshold of usage, then I might try to apply Poisson to estimate the probability of the threshold which could be at the given time step t'.

    ReplyDelete
  3. enrico24 November 2011 at 21:11

    Hi Henry,

    I do not understand what you mean by "event appearance" and "event usage" -- can you pleas explain?

    > I now try to set up a list of threshold of usage,
    > then I might try to apply Poisson to estimate
    > the probability of the threshold which could
    > be at the given time step t'.

    Sounds good! I look forward to see this.

    ReplyDelete
  4. Henry Ngoc Cuong Truong25 November 2011 at 11:11

    The event usage is actually the everage consumption for the specific event type.

    ReplyDelete

Subscribe to: Post Comments (Atom)