Sort facebook friend list, closest friend first.

gistfile1.txt

An algorithm to sort a person's friend list, closest friends first.

Problem: Given the publicly available information on facebook, how to
sort the friend list in an order that list closest/most relevant friends
first?

We are assuming the following things are available (in no particular order):
1. Location information: All the locations
2. Schools/University attended
3. Birthday
4. Applications used
5. Jobs
6. Groups (Common)
7. Mutual friends
8. Common interests
9. Photos (common)
10. Statuses - Mendtioned a friend.
11. Likes, Comments, Wall Posts
12. Netwroks
13. Age

Let there be person_A whose friend list we have to sort. Let the friends be
friend_< i > where 1 < i < 1000.

Using the information above, we will solve the problem of finding relevance
of friends.

Method:
-------

We loop over all friends, and we calculate a parameter, called relevance, for
each friend. To calculate this, we sum over quantities like w*f, where w is
the relative weigth for a certain parameter and f is the measure of its quantity.

The weigth is calculated dynamically, that is weights depend on the combination 
of different parameters. We have a weigth calculating function for each
parameter. The logic for each function is explained in the algorithm.

f is a number that represents the measure of a certain parameter. For example,
it might represent the number of mutual friends or the number of common 
groups.

We are taking person_A and friend_< i > to be objects which will be used to
access the relevant data.


Logic for weigth function:
--------------------------

    The weigth for any parameter will be calculated using a weight function
    for each particular parameter.

    The weight function will be different for each parameter. Also, depending
    on the parameter, the weights computed will be different. For example,
    certain parameters, such as number of mutual friends will be given more
    weight, than say, common interests of persons. The weight functions will
    take care of this.

    * We see if both persons are at same location/have been at same locations.
    Also, same location with presence at same time increases this weight. Also,
    depends on how many locations the persons have in common.

    * We assess if both persons have attended same schools/college. This will be given high
    priority because, given two people went to the same school/college,
    it is highly likely that they will be closer friends. Accrodingly, we assign a
    better/worse weight.

    * We assess if in both are in the same facebook group and also see the
    number of members in the group. For a group with more members, we'll
    assign a lower weight and for groups with fewer members, we assign a
    higher weight.

    * Determine if the networks are same.

    * We check if a friend likes my posts, and also check if I like his
    posts, that is if this goes both ways, then, depending on the mutual
    interaction, we can assign a weight.

    * We see the number of pictures in which the person and his friend has been
    tagged together.

    * We determine how many times I have mentioned a friend in my status updates and
    vice - versa.

    * We analize comment conversations, that is, number of comments made by both
    persons on the same status update. Also, if say, the person and his friend
    make successive comments, or if they make many comments in a short period
    of time, that is a very good indicator of 'closeness'.

    * We check how many time both persons post on each other's walls.

    * Common facebook applications used, common interests. It may happen that
    even though people who do not appear to be good friends may be good friends
    in real life.


Rough Algorithm:
----------------

friend_list = [ friend_1, friend_2, ..... , friend_1000 ]
relevance_list = []

relevance is a list where we will store the relevance numbers.

for friend in friend_list:
    relevance = 1

    w1 = weight_mutual_friends(person_A, friend)
    Similarly, calculate weights for each of the parameter.

    Then, claculate relevance as:

    Summation( w_i*f_i)
    relevance_list.append(relevance)

Now, using this relevance_list, we can sort the friend list.