System and method for automatic generation of user-oriented homepage

1. A method for automatic generation of a user-oriented homepage, comprising the steps of:

a. collecting browsed webpage data of a user by a network server for once in a period of time for an AI server to retrieve said data from said network server or engine in a client;

b. analyzing a user preference of said user by said AI server and obtaining user network behavior characteristics using an algorithm of random-decomposing multidimensional scaling in applied statistics;

c. obtaining a correlation map from said user network behavior characteristics and a first relational network for computation, and then obtaining group characteristics of said user in said first relational network;

d. re-computing nearby relational networks based on said group characteristics of said user in said first relational network to obtain a second relational network of said user;

e. automatically generating a customized webpage by said AI server based on a webpage index of said second relational network;

f. sending said customized webpage as said user-oriented homepage to said network server by said AI server for the next time said user connects to said network server; and

g. automatically sending said customized webpage to a user terminal the next time said user connects to said network server.

2. The method of claim 1 , wherein said AI server is implemented with said algorithm of random-decomposing multidimensional scaling in applied statistics, said algorithm comprising the steps of:

Step 1: separating N users into K partially overlapping groups, the number of overlapped users must be larger than p network behavior characteristics;

Step 2: obtaining corresponding MDS coordinates of every group by using the traditional Multidimensional Scaling method to compute with respect to every group;

Step 3: obtaining a transfer function based on OR decomposing to transfer coordinates of every group into coordinates on a first coordinate system; and

Step 4: projecting an index corresponding to a webpage content in said network server to MDS coordinates based on LSI method.

3. The method of claim 1 , wherein said algorithm of random-decomposing multidimensional scaling in applied statistics comprises the following two essential conditions:

(1) The step of separating N users into K partially overlapping groups has to be random for each group to have both short distance information and long distance information; and

(2) the number of overlapped users must be larger than the number of network behavior characteristics.

4. The method of claim 2 , wherein said algorithm of random-decomposing multidimensional scaling in applied statistics employs a PCA-RFE (Principal Component Analysis-Recursive Feature Eliminated Method) feature extraction method to enhance outcome accuracy, said feature extraction method comprising the steps of:

Step 1: randomly taking a sample in a group;

Step 2: calculating a differential vector by using said taken sample in step 1;

Step 3: calculating a prime term of a difference between groups by using a PCA (Principle component analysis) method;

Step 4: eliminating one or more features having less components, this step eliminate one or more features in one time; and

Step 5: repeating the above steps until a desired number of features have been obtained.

5. The method of claim 2 , wherein said AI server is able to couple with a general network server.

6. A system for automatic generation of user-oriented homepage, said system comprising:

a plurality of AI servers, which uses the method of claim 1 to compile a customized webpage based on user requirements;

a network server coupling to said plurality of AI servers and recording users' usage data for said plurality of AI servers; and

a plurality of user terminals connecting to said network server via Internet and receiving webpage data from said network server.

7. The system of claim 6 , wherein said user terminal is a personal computer or a personal digital assistant (PDA).