System and method for determining the value of a crowd network
View Patent ↗The present invention proposes that the value of a crowd network comes from network entropy and a very well engaged crowd has more value than a poorly engaged crowd. The engagement (and the resultant value generated from the engagement) of a crowd network can be measured in terms of the entropy of the network. The entropy is calculated as a function of the probability distribution of incoming and outgoing messages, which represents the entropy or uncertainty in the activity over the network. In this way, the activity occurring over a network in terms of message exchange is translated as a measure of the value of the network. An evolutionary algorithm is presented to optimize the entropy of a network by successively changing the network topology. Results indicate that the value of a crowd network very closely relate to its small world-ness, sparsity, and connectedness.
1. A method for computing a value V C of a crowd network C, the method comprises:
assigning an initial value to V c ;
collecting statistics information representing communication activities among a total of n communication nodes in the crowd network;
constructing an adjacency matrix A, an incoming message probability matrix X, and an outgoing message probability matrix Y of the crowd network, wherein the incoming message probability matrix X={x ij |i,j<=n}, wherein x ij denotes an incoming message probability of node i to receive info from node j, and the outgoing message probability matrix Y={y ij |i,j<=n}, wherein y ij denotes an outgoing message probability of node i to transmit info to node j, and wherein the incoming message probability matrix X and the outgoing message probability matrix Y are determined at least in part based on the collected statistics information;
computing an inbound entropy H in of the crowd network using matrices A and X, and an outbound entropy H out of the crowd network using matrices A and Y;
calculating an updated value V C ′ of the crowd network using the inbound entropy H in , the outbound entropy H out , and a pre-defined weighting parameter α;
assigning the updated value V C ′ to V c if the updated value V C ′ is greater than a previous value V C ;
repeating the constructing, computing, calculating, and assigning steps until the value V C has not been increased in the last n 2 iterations; and
outputting the value V C indicating a maximum value of the crowd network C.
2. The method of claim 1 , wherein the adjacent matrix A={a ij |i,j<=n}, wherein a ij denotes an existence of a link between nodes i and j.
3. The method of claim 1 , wherein the statistics information is collected from a social network website, wherein each node is associated with a registered user, and wherein communication activities include sending and receiving messages and postings comments among the registered users.
4. The method of claim 1 , wherein the statistics information is collected from an email server, wherein each node is associated with an email user, and wherein communication activities include sending and receiving emails among the email users.
5. The method of claim 1 , wherein the statistics information is collected from a mobile phone service provider, wherein each node is associated with a mobile customer, and wherein communication activities include phone calls and text messages among the customers.
6. The method of claim 1 , wherein the statistics information is collected from an influence network server, wherein communication activities include sending and receiving messages implicitly in the influence network.
7. A method for optimizing a value V C of a crowd network C by a server computer, the method comprises:
(a) initializing a set of network parameters including a crowd network size n, an initial value V C =0, an incoming message probability matrix X, and an outgoing message probability matrix Y of the crowd network, wherein the incoming message probability matrix X={x ij |i,j<=n}, wherein x ij denotes an incoming message probability of node i to receive info from node j, and the outgoing message probability matrix Y={y ij |i,j<=n}, wherein y ij denotes an outgoing message probability of node i to transmit info to node j, and wherein the incoming message probability matrix X and the outgoing message probability matrix Y are determined at least in part based on collected statistics information;
(b) generating a random adjacency matrix A with a total number of links less than m;
(c) computing an inbound entropy H in of the crowd network using matrices A and X, and an outbound entropy H out of the crowd network using matrices A and Y;
(d) calculating an updated value V C ′ using H in , H out , and a weighting parameter α;
(e) assigning the updated value V C ′ to V C if the updated value V C ′ is greater than a previous value V C ;
(f) repeating (b) to (e) until the value V C has not been increased in the last n 2 iterations; and
(g) outputting the value V C indicating a maximum value of the crowd network C with no more than m links.
8. The method of claim 7 , wherein the adjacent matrix A={a ij |i,j<=n}, wherein a ij denotes an existence of a link between nodes i and j, and wherein a ij is applied with a uniformly randomly generated probabilty p in (b).
9. The method of claim 7 , wherein the value V C of the crowd network is substantially maximized when the total number of links (m) increases.
10. The method of claim 7 , wherein the value V C of the crowd network is substantially maximized when a clustering coefficient (c) increases.
11. The method of claim 7 , wherein the value V C of the crowd network is substantially maximized when an average shortest path length (l) decreases.
12. The method of claim 7 , wherein the value V C of the crowd network is substantially maximized when a total number of components of accessible nodes (S) decreases.
13. The method of claim 1 , wherein the initial value of V c is 0.
14. The method of claim 1 , wherein the assigning step also comprises assigning a maximum number of links m and the outputted value V c indicates a maximum value of the crowd network C with no more than m links.