User preference modeling method using fuzzy networks

• US 6,338,051 B1
• Filed: 04/29/1999
• Issued: 01/08/2002
• Est. Priority Date: 04/30/1998
• Status: Expired due to Term

• Current Assignee: Korea Advanced Institute of Science and Technology, Samsung Electronics Co. Ltd.
• Original Assignee: National Institute of Advanced Industrial Science and Technology (Government of Japan), Samsung Electronics Co. Ltd.
• Inventors: Chung, Sun-wha, Lee, Kwang-hyung, Kang, Ho-seok, Yoon, Joo-young, Sung, Kyoung-a
• Primary Examiner(s): Davis, George B.
• Assistant Examiner(s): Booker, Kelvin
• Application Number:
  US09/301,317
• Time in Patent Office: 985 Days
• Field of Search: 706/2, 706/12, 706/13, 706/14
• US Class Current: 706/2
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

A user preference modeling method using fuzzy networks. The user preference modeling method includes the steps of: (a) changing a user modeling structure into a fuzzy network structure in which a plurality of layers including one or more graphs with one or more nodes are stacked; (b) when information is input from a user, searching a node directly associated with the input information on the fuzzy networks, and calculating a new preference for the node with a predetermined equation; (c) calculating connection strengths among each node in a graph to which the node belongs according to the new preference obtained in step (b) and calculating a new preference for each node of the graph according to the connection strengths; (d) when a node of the graph to which the node searched in step (b) belongs is a macro node of a graph of a lower layer, and a node is defined as the macro node if a graph of a lower layer defines sub-regions of the node, transferring a first message as preference change information from the macro node to the graph of the lower layer; (e) when the graph to which the node searched in step (b) belongs has a macro node in an upper layer, transferring a second message to the macro node, as preference change information for all nodes of the graph to which the node belongs; (f) when a graph receives the first message from a macro node, calculating a new preference for all nodes in the graph that has received the first message, and when a node of the graph that has received the first message is a macro node of a graph of a lower layer, transferring a first message as preference change information to the graph of the lower layer; and (g) when a node receives the second message from a graph of a lower layer, calculating a new preference for the node that has received the second message and performing steps (c) through (e) to other nodes.

Claims

1. A user preference modeling method using fuzzy networks, comprising the steps of:

(a) changing a user modeling structure into a fuzzy network structure in which a plurality of layers including one or more graphs with one or more nodes are stacked;

(b) when information is input from a user, searching a node directly associated with the input information on the fuzzy networks, and calculating a new preference for the node with a predetermined equation;

(c) calculating connection strengths among each node in a graph to which the node belongs according to the new preference obtained in step (b) and calculating a new preference for each node of the graph according to the connection strengths;

(d) when a node of the graph to which the node searched in step (b) belongs is a macro node of a graph of a lower layer, and a node is defined as the macro node if a graph of a lower layer defines sub-regions of the node, transferring a first message as preference change information from the macro node to the graph of the lower layer;

(e) when the graph to which the node searched in step (b) belongs has a macro node in an upper layer, transferring a second message to the macro node as preference change information for all nodes of the graph to which the node belongs;

(f) when a graph receives the first message from a macro node, calculating a new preference for all nodes in the graph that has received the first message, and when a node of the graph that has received the first message is a macro node of a graph of a lower layer, transferring a first message as preference change information to the graph of the lower layer; and

(g) when a node receives the second message from a graph of a lower layer, calculating a new preference for the node that has received the second message and performing steps (c) through (e) to other nodes.

2. The method of claim 1, wherein assuming that 1, j and k are positive integers, F′

(m_ijk) represents a new preference for a node m_ijk after the preference for the node is changed, F(m_ijk) represents a preference for the node m_ijk before the change, Diff(m_ijk) represents the difference in the preference before and after the change, defined as Diff(m_ijk)=F′

(m_ijk)−

F(m_ijk), m_ijk′

represents other nodes in a graph G_ij to which the node m_ijk belongs and α

represents an experimental constant, connection strengths W(m_ijk, m_ijk′

) between the node m_ijk and the nodes m_ijk′

, are calculated by an equation given by $W<br><br>\left(m_{\mathrm{ijk}},m_{{\mathrm{ijk}}^{\mathrm{\prime <br><br>}}}\right)=\mathrm{\alpha <br><br>}·<br><br>\mathrm{Min}<br><br>\left\{\frac{F<br><br>\left(m_{{\mathrm{ijk}}^{\mathrm{\prime <br><br>}}}\right)}{F^{\mathrm{\prime <br><br>}}<br><br>\left(m_{\mathrm{ijk}}\right)},\frac{F^{\mathrm{\prime <br><br>}}<br><br>\left(m_{\mathrm{ijk}}\right)}{F<br><br>\left(m_{{\mathrm{ijk}}^{\mathrm{\prime <br><br>}}}\right)}\right\}$

and a new preference F(m_ijk′

) for the nodes m_ijk′

is calculated by an equation given by

3. A computer readable medium having embodied thereon a computer program for the user preference modeling using fuzzy networks according to claim 2.

4. The method of claim 1, wherein assuming that 1, j and k are positive integers, m_ijk′

represents other nodes in a graph G_ij to which the node m_ijk belongs and G_(i+1)j′

is the graph at a lower level, the first message of step (d), G_msg(m_ijk′

, G_(i+1)j′

), has information about the difference in preference Diff(m_ijk′

) before and after the preference for node m_ijk′

changes, which is expressed as

5. The method of claim 3, wherein assuming that 1, j, j′

, k and k′

are positive integers, n(G_ij) represents the total number of nodes in a graph G_ij and F(m_ijk) represents preference for a node m_ijk, when in step (f) the graph G_ij receives a first message G_msg(m<sub>(i−

1)j′

k′</sub>

, G_ij) from a macro node m<sub>(i−

1)j′

k′</sub>

of the graph G_ij, a new preference F′

(m_ijk) for all nodes m_ijk belonging to the graph G_ij is calculated by an equation given by $F^{\mathrm{\prime <br><br>}}<br><br>\left(m_{\mathrm{ijk}}\right)=\mathrm{Min}<br><br>\left\{1.0,F<br><br>\left(m_{\mathrm{ijk}}\right)+\frac{1}{n<br><br>\left(G_{\mathrm{ij}}\right)}·<br><br>G_{\mathrm{msg}}<br><br>\left(m_{\left(i·<br><br>1\right)<br><br>j^{\mathrm{\prime <br><br>}}<br><br>k^{\mathrm{\prime <br><br>}}},G_{\mathrm{ij}}\right)\right\}$

and when a node m_ijk is a macro node of a graph G_(i+1)j′

of a lower layer, a first message G_msg(m_ijk, G_(i+1)j′

) is calculated using the difference in preference Diff(m_ijk) before and after the change, by an equation given by

6. A computer readable medium having embodied thereon a computer program for the user preference modeling using fuzzy networks according to claim 5.

7. A computer readable medium having embodied thereon a computer program for the user preference modeling using fuzzy networks according to claim 3.

8. The method of claim 1, wherein assuming that 1, j and k are positive integers, Diff(m_ijk) is the difference in preference before and after the preference for a node m_ijk is changed, m_(i−

1)jk′

is the macro node to m_ijk, and n(G_ij) is the total number of nodes in the graph G_ij, the second message of step (e), M_msg(G_ij, m<sub>(i−

1)j′

k′</sub>

), has information about the maximum value in degree of change for all nodes of a graph G_ij to which the node m_ijk belongs, which is expressed as

9. The method of claim 8, wherein assuming that 1, j, j′

and k′

are positive integers, and F(m_ijk) represents the current preference for a macro node m_ijk, when in step (g) the macro node m_ijk receives a second message M_msg(G_(i+1)j′

, m_ijk) from a graph G_(i+1)j′

of a lower layer, a new preference F′

(m_ijk) of the macro node m_ijk is calculated by an equation given by

10. A computer readable medium having embodied thereon a computer program for the user preference modeling using fuzzy networks according to claim 9.

11. A computer readable medium having embodied thereon a computer program for the user preference modeling using fuzzy networks according to claim 8.

12. A computer readable medium having embodied thereon a computer program for the user preference modeling using fuzzy networks according to claim 1.

14. A computer readable medium having embodied thereon a computer program for the adaptive web directory service using fuzzy networks according to claim 10.

13. A method of serving an adaptive web directory using fuzzy networks, comprising the steps of:

(a) changing a web directory structure on a web server into a fuzzy network structure in which a plurality of layers including one or more graphs with one or more nodes are stacked;

(b) when information is input by a user, searching a node on the fuzzy networks, the node being directly associated with the input information, and calculating a new preference for the node;

(c) calculating connection strengths for all nodes of a graph to which the node searched in step (b) belongs, based on the new preference obtained in step (b), and calculating new preferences for each node of the graph in consideration of connection strengths;

(d) when a node of the graph to which the node searched in step (b) belongs is a macro node of a graph of a lower layer, wherein a node is defined as the macro node if a graph of a lower layer defines sub-regions of the node, transferring a first message as preference change information from the macro node to the graph of the lower layer;

(e) when the graph to which the node searched in step (b) belongs has a macro node in an upper layer, transferring a second message to the macro node as preference change information for all nodes of the graph to which the node belongs;

(f) when a graph receives the first message from a macro node, calculating a new preference for all nodes in the graph that has received the first message, and when a node of the graph that has received the first message is a macro node of a graph of a lower layer, transferring a first message as preference change information to the graph of the lower layer;

(g) when a node receives the second message from a graph of a lower layer, calculating a new preference for the node that has received the second message and performing steps (c) through (e) to other nodes; and

(h) restructuring the web directory according to the user preference to provide a web directory structure adaptable to the user preference characteristic.