Skip to content

Web service for building teams for peer review sites using clustering and top trading cycles.

Notifications You must be signed in to change notification settings

shobbb/IntelligentAssignment

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

44 Commits
 
 
 
 
 
 

Repository files navigation

#Intelligent Assignment

Intelligent assignment creates teams based off of each individual users ranking of topic preference, using k-means clustering. This webservice requires an input of each users ranks [1 being most preferred and 0 indicating no preference] and unique id [pid], as well as the max team size. This also uses top trading cycles to switch members of teams who have already worked with other members on that team.

Accessing the service

###1) Access the Webservice online This service is hosted at: http://peerlogic.csc.ncsu.edu/intelligent_assignment/ [method name]. This service can be called without copying the code onto your local machine, simply make a post request to this url with one of the method names mentioned below.

###2) Run it on your local machine The service can be copied from its github repository (https://github.com/peerlogic/IntelligentAssignment). It should be deployed as a webservice; though, it will also require the python libraries flask and scipy:

-Scipy

-Flask

Methods

###Creating teams (/merge_teams): Uses K-means clustering to group users with similar topic interests. Works to eliminate competition for any single topic and increase the likelihood that each user obtains their most preferred topic.

Sample input and output:

-Input:

{"users":[{"ranks":[1,0,2,3], "pid":1023},{"ranks":[1,2,0,3], "pid":4535},{"ranks":[0,2,3,1], "pid":1363},{"ranks":[2,1,0,3], "pid":9841}],"max_team_size":4}

-Output:

{"users": [{"ranks": [1,0,2,3],"pid": 1023},{"ranks": [1,2,0,3],"pid": 4535},{"ranks": [0,2,3,1],"pid": 1363},{"ranks": [2,1,0,3],"pid": 9841}],"teams": [[1363,9841],[1023,4535]]}

###Swapping Team Members (/swap_team_members): Uses Top Trading Cycles to swap members, that have already worked with members on their team, with other teams' members. This method only swaps a max of one member per team per run. Begins by first sorting the list of available members by distance from the teams centroid and then by whether or not other members of the team have worked with them. This method requires a history of users that each user has worked with, along with the general information.

Sample input and output:

-Input:

{"users":[{"ranks":[1,0,2,3], "history":[4535,9841,9843], "pid":1023},{"ranks":[1,2,0,3], "history":[1023,9843,8542], "pid":4535},{"ranks":[0,2,3,1], "history":[3649,9841,9843], "pid":1363},{"ranks":[2,1,0,3], "history":[1363,1023,3649], "pid":9841}],"teams": [[1023,2549],[4535,9843],[1363,1867,3649],[9841,8542,7521]]} 

-Output:

{"users": [{"ranks": [1, 0, 2, 3], "pid": 1023, "history": [4535, 9841, 9843]}, {"ranks": [1, 2, 0, 3], "pid": 4535, "history": [1023, 9843, 8542]}, {"ranks": [0, 2, 3, 1], "pid": 1363, "history": [3649, 9841, 9843]}, {"ranks": [2, 1, 0, 3], "pid": 9841, "history": [1363, 1023, 3649]}], "teams": [[9841, 4535], [1023, 1363]]}

Client Code Example

import requests
import json

#Test data
data = json.dumps({"users":[{"ranks":[1,0,2,3], "history":[4535,9841,9843], "pid":1023},{"ranks":[1,2,0,3], "history":[1023,9843,8542], "pid":4535},{"ranks":[0,2,3,1], "history":[3649,9841,9843], "pid":1363},{"ranks":[2,1,0,3], "history":[1363,1023,3649], "pid":9841}],"max_team_size":2})
header = {'content-type': 'application/json'}
response = requests.post("http://127.0.0.1:5000/merge_teams",data= data,headers=header)
print response.text
#The response from merge teams can be used in swap team members if
#history was given in the body of the request to merge teams	
response = requests.post("http://127.0.0.1:5000/swap_team_members", data=response.text,headers=header)
print response.text

About

Web service for building teams for peer review sites using clustering and top trading cycles.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages