Hey all.
The rate of mobile user activity has been considered as exponential distribution in much of the literature. How is that? Following are my own understanding.
Mobile users' (read cellular users) activity is a random process, usually not following previous history (hence memorylessness).
One cannot certainly predict number of active/deactive mobile nodes in a given state of time.
These are typical exponential distribution property.
If I need to predict number of cellular nodes connected to a device (parent node) since the start of time (t=0). How can I predict them?
I assume that the 'parent node' must keep some state of memory.
What proper distribution should I follow ?
Thanks
Shan
Hi Shan,
In my opinion, while is may seem impossible to predict the number of active users in a certain cell or connected to a particular BS, Big Data collected over a significant period of time may assist with the fairly accurate estimation of the total number of users connected in particular intervals and then Inferential Statistics(Hypothesis Testing) may be applied to draw some scaled conclusions. For instance, during weekdays when people are going to and back from work, they are driving/taking the train/bus on the same route every and are going through the same cells at pretty much the same time intervals. At work, they are staying within the same cell or are being connected to the same BS for 6 to 8 hours. There are companies that analyze all these patterns and are able to accurately estimate the variations so that the telco to invest in extra capacity where needed. There is definitely a significant degree of uncertainty with the users that are not following the pattern described before, however I think the variance is not that significant.(just a personal assumption).
Having said that, I personally believe that the assumption about the distribution being random (memoryless -exponential) is not correct as the arrival of a significant percentage of users is "known" . I hope that helps to some extent.
Some helpful links:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.11.9293&rep=rep1&type=pdf
Quick answer: age demographics. Teen years and 20s, heavy use. Otherwise, not so much. ;)
Well going by traffic statistics, you can work with an average value over a time period. It is however safer to break down your result into different average values for different time periods of the day. Usually, there are relatively silent hours (midnight-early morning), busy hours (7pm-11pm), and moderate traffic hours (11am-6pm) for each day.
There is already existing algorithm for generating traffic statistics in Telcos and the values generated over time at different periods of the day tend to have a defined pattern. The traffic pattern may however be distorted in exceptional circumstances like a huge conference or competition that will naturally result in traffic spikes.
As for deriving number of users, existing equations that estimate average traffic per user can be applied. To be fair, your result will at best be an informed estimate. I hope this helps.
https://www.researchgate.net/publication/305627093_Exploring_the_Dimensions_of_Mobile_Banking_Service_Quality_Implications_for_the_Banking_Sector
Article Exploring the Dimensions of Mobile Banking Service Quality: ...
- Badges
- Science topic
- Similar topics
- Engineering
- Telecommunications Engineering