https://github.com/lukehackett/clusteranalysistool

https://github.com/lukehackett/clusteranalysistool

Last synced: 1 day ago
JSON representation

• Host: GitHub
• URL: https://github.com/lukehackett/clusteranalysistool
• Owner: LukeHackett
• Created: 2012-09-07T17:21:47.000Z (over 12 years ago)
• Default Branch: master
• Last Pushed: 2013-04-24T19:38:03.000Z (over 11 years ago)
• Last Synced: 2023-03-29T00:35:46.794Z (almost 2 years ago)
• Language: C#
• Size: 15.5 MB
• Stars: 0
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
Cluster Analysis Tool

===================

###Problem

Consider a wireless product consistently dropping calls within a one mile radius 

at a particular point upon a drive test. Currently, RIM is able debug this 

product within the test field laboratory to find out if there is a hardware and/or 

software bug associated with the error. However if the results of the laboratory 

tests came back as a pass, it would have to be assumed that there is an issue 

with the mobile network. 

It is this assumption that can cause frustration for RIM's third parties (e.g. 

customers and mobile network operators), even more so if other wireless devices 

within the same area are able to communicate perfectly fine.

###Solution

To design, implement and test one (or more) clustering algorithms in order to 

cluster coordinates that are generated by cellular devices. A further comparison 

of the various devices' meta-data will be required in order to allow for a more 

specific, in depth analysis.

###User Requirements

The following objectives have been set directly by the client, RIM. These objectives 

have been outlined using the MoSCoW prioritisation technique, and are outlined 

below:

**MUST**

* Design and develop an algorithm to cluster GPS coordinates.

* Compare RAT footprints of two data sets (products or pins) with each set being 

N weeks' worth of data and highlight differences in the RAT usage/drop/fails 

along the route.

* Compare MIX_BAND (Frequency bands) footprints of two data sets and highlight 

differences in usage/drops/fails between sets.

**SHOULD**

* Compare two sets of data for call its drop/failure clusters and highlight 

differences. 

* Ability to tag drops/fails with classification attributes and even compare 

based on attributes.

* Produce a web-based interface to the results, with a map plugin allowing the 

user to view the data.

**COULD**

* Velocity differences along route.

* Plot a map for all call attempts and call ends (success/fail) for a given 

period of time, for different devices.

* Allow the user to filter the clustering down (e.g. show only call drops).

**WON'T**

* Integration with current internal RIM testing systems.