CN103886073A - Coal mine information recommendation system based on collaborative filtering - Google Patents

Abstract

The invention provides a coal mine information recommendation system based on collaborative filtering. The system comprises a coal mine information collaborative filtering module, a community user group access log module and a recommendation module, wherein the coal mine information collaborative filtering module is used for analyzing access logs of users and finding out user neighbors by means of a collaborative filtering algorithm; the community user group access log module is used for managing the logs; the recommendation module is used for calculating recommended information and recommending the information to the users. By the adoption of the system, items or information, according with requirements of the users, can be recommended to the users according to hobbies and interests of the users, and personalized service is achieved. The coal mine information recommendation system provides targeted recommendations according with specific users on the basis of the hobbies and interests of the users. Because of the recommendations, user experience is enhanced, user usage convenience is improved, and work efficiency of the users is improved.

Description

Coal information commending system based on collaborative filtering

Technical field

The invention belongs to Coal Information System.

Background technology

Due to the develop rapidly of Internet, the information capacity of every field is all very huge, comprises field, colliery.Under this background, the search engine system in field, colliery has obtained studying widely and applying, and provides convenience for people find useful information.But along with improving constantly of people's Search Requirement, the deep processing of result for retrieval becomes an emphasis of this area research gradually.

The search modes of traditional " One-Size-Fit-All " can not allow user satisfied, the substitute is the personalization of result for retrieval, and replaces " people looks for information " with the pattern of " information is looked for people ".So it is very necessary that coal information commending system just seems.

Commending system general reference guides user as the system of selecting using personalized recommendation as the system of output or in extensive optional object according to personal like.The coal information commending system that the present invention proposes is according to the log content of having browsed before user, and new, that do not browse, user may be pushed to user by interested coal information automatically.

Collaborative filtering is one of the method in Technologies of Recommendation System in E-Commerce that is early used in, its core concept is that user tends to buy and has commodity or the content that the customer group of similar interests hobby is bought, and basic starting point is: user is can be by categorize interests; The interest-degree that user has comprised user to the evaluation of different information; User to the evaluation of a unknown message by similar with its similar interests user's evaluation.

In the environment of community, between user, often have identical interest, demand, expectation and motivation and go to select similar information, this situation is that the recommendation based on collaborative filtering provides the foundation.

Summary of the invention

In order to realize the present invention, a kind of coal information commending system based on collaborative filtering, Community-oriented has been proposed, this system comprises coal information collaborative filtering module, community users colony access log module and recommending module; Coal information collaborative filtering module is responsible for the access log of analysis user, utilizes the algorithm of collaborative filtering to find out user neighbour; Community users colony access log module is responsible for the management of daily record; Recommending module is responsible for the information of calculated recommendation and is recommended user; Wherein, in coal information collaborative filtering module, adopt following methods to filter:

(1) system initialization: based on the access log of user in the community of colliery, build coal mine user-colliery webpage rating matrix A, matrix A comprises s user u ₁, u ₂..., u _sto t webpage p ₁, p ₂..., p _tscoring, A is the matrix on s × t rank, s and t are natural number, the element a in matrix _ijrepresent the scoring of user i to webpage j, i and j are natural number, and 1≤i≤s, 1≤j≤t,

(2) rating matrix optimization: a t webpage obtains k bunch after K-Means clustering algorithm, is designated as c ₁, c ₂..., c _k, k is number of clusters order, k is natural number; Bunch c _mthe webpage number comprising is N _m, this N _mthe sequence number of individual webpage in original t webpage is respectively f (m, 1), f (m, 2) ..., f (m, N _m), f function is for calculating the original number of bunch webpage after cluster, i.e. sequence number in original t webpage, f (m, N _m) expression bunch c _min N _mthe original number of individual webpage, m is natural number, 1≤m≤k, 1≤N _m≤ t, the rating matrix after being optimized

wherein b _imthe element of representing matrix B,

$b_{\mathrm{im}}=\frac{1}{N_m}\underset{n=1}{\overset{N_m}{\mathrm{\Σ}}}{a}_{\mathrm{if}(m,n)};$

(3) neighbor searching: user u _va bunch sequence number set of marking is M _v, user u _wa bunch sequence number set of marking is M _w, user u _vand u _wa bunch sequence number set of all marking is M, and v and w are natural number, 1≤v≤s, 1≤w≤s, user u _vand u _wbetween similarity be $\mathrm{sim}(u_v,u_w)=\frac{{\mathrm{\Σ}}_{z\∈M}(b_{\mathrm{vz}}-{\stackrel{\‾}{b}}_v)({\stackrel{\‾}{b}}_{\mathrm{wz}}-{\stackrel{\‾}{b}}_w)}{\sqrt{{\mathrm{\Σ}}_{z\∈M_v}(b_{\mathrm{vz}}-{\stackrel{\‾}{b}}_v)}\sqrt{{\mathrm{\Σ}}_{z\∈M_w}(b_{\mathrm{wz}}-{\stackrel{\‾}{b}}_w)}},$ Wherein with

represent respectively user u _vand u _wto the average score of webpage, try to achieve each user u _ineighbour user in user's space, that sort from high to low by similarity gathers U _i;

(4) recommend: according to U _idraw set U _iin all users, by scoring from high to low, the collections of web pages P that browses _i, user u _ithe collections of web pages of having browsed is P' _i, set (P _i-P' _i) in before L coal information webpage be user u _iinterested content.

Adopt after above system, can, according to user's hobby, recommend to meet project or the information of its requirement, can realize personalized service.After introducing commending system, can, according to the hobby of coal mine user, user may interested coal information formation be recommended to user with the form of list.Can allow like this user in less range of information, search own interested content, whole information browse process is more quick, becomes " information is looked for people " from the mode of original " people looks for information ".The interest preference of coal information commending system based on each user, provides the recommendation targetedly that meets specific user, and this recommendation has strengthened the convenience that the user of user's experience, raising uses, and has also improved user's work efficiency.

Brief description of the drawings

Fig. 1 is system architecture schematic diagram of the present invention.

Embodiment

Coal information commending system coal information collaborative filtering module based on collaborative filtering, Community-oriented, community users colony access log module, recommending module.Wherein coal information collaborative filtering module is responsible for the access log of analysis user, utilizes the algorithm of collaborative filtering to find out user neighbour; Community users colony access log module is responsible for the management of daily record; Recommending module is responsible for the information of calculated recommendation and is recommended user.Coal information collaborative filtering module is the emphasis of native system, and its key step is:

(1) system initialization: this step is the access log based on user in the community of colliery mainly, builds coal mine user-colliery webpage rating matrix, is designated as A, supposes that this matrix comprises s user u ₁, u ₂..., u _sto t webpage p ₁, p ₂..., p _tscoring, A is the matrix on s × t rank, s and t are natural number, the element a in matrix _ijrepresent the scoring of user i to webpage j, i and j are natural number, and 1≤i≤s, 1≤j≤t.Matrix A can be expressed as:

(2) rating matrix optimization: under normal circumstances, user's score data is fewer, and along with the continuous growth of webpage quantity t, cause rating matrix A extremely sparse, the precision that final impact is recommended.In view of this, first the present invention does further optimization to rating matrix A, and the method for optimization is that webpage is carried out to cluster, and cluster process adopts K-Means algorithm.

Suppose that t webpage can obtain k bunch after using K-Means clustering algorithm, is designated as c ₁, c ₂..., c _k, k is number of clusters order, k is natural number.Suppose a bunch c _mthe webpage number comprising is N _m, this N _mthe sequence number of individual webpage in original t webpage is respectively f (m, 1), f (m, 2) ..., f (m, N _m), f function is for calculating the original number of bunch webpage after cluster, i.e. sequence number in original t webpage, f (m, N _m) expression bunch c _min N _mthe original number of individual webpage, m is natural number, 1≤m≤k, 1≤N _m≤ t.

Through above-mentioned optimizing process, the rating matrix B after can being optimized.

Wherein b _imthe element of representing matrix B,

$b_{\mathrm{im}}=\frac{1}{N_m}\underset{n=1}{\overset{N_m}{\mathrm{\Σ}}}{a}_{\mathrm{if}(m,n)};$

(3) neighbor searching: based on the rating matrix B after optimizing, can find out the user of same or similar interest, between user, the calculating of similarity adopts improved cosine similarity based method to carry out.If user is u _va bunch sequence number set of marking is M _v, user u _wa bunch sequence number set of marking is M _w, user u _vand u _wa bunch sequence number set of all marking is M, and v and w are natural number, 1≤v≤s, 1≤w≤s, user u _vand u _wbetween similarity be:

$\mathrm{sim}(u_v,u_w)=\frac{{\mathrm{\Σ}}_{z\∈M}(b_{\mathrm{vz}}-{\stackrel{\‾}{b}}_v)({\stackrel{\‾}{b}}_{\mathrm{wz}}-{\stackrel{\‾}{b}}_w)}{\sqrt{{\mathrm{\Σ}}_{z\∈M_v}(b_{\mathrm{vz}}-{\stackrel{\‾}{b}}_v)}\sqrt{{\mathrm{\Σ}}_{z\∈M_w}(b_{\mathrm{wz}}-{\stackrel{\‾}{b}}_w)}},$

Wherein

with

represent respectively user u _vand u _wto the average score of webpage.

According to above-mentioned similarity calculating method, can be in the hope of each user u _ineighbour user in user's space, that sort from high to low by similarity gathers U _i.

(4) recommend: according to the user u obtaining _ineighbour user gather U _i, can draw set U _iin all users, by scoring from high to low, the collections of web pages P that browses _i, suppose user u _ithe collections of web pages of having browsed is P' _i, set (P _i-P' _i) in before L coal information webpage may be user u _iinterested content, can be used as content recommendation and recommends user u _i.

This system can, according to user's hobby, recommend to meet project or the information of its requirement, is a kind of individuation service system.After introducing commending system, can, according to the hobby of coal mine user, user may interested coal information formation be recommended to user with the form of list.So just can allow user in less range of information, search own interested content, whole information browse process is more quick, becomes " information is looked for people " from the mode of original " people looks for information ".The interest preference of coal information commending system based on each user, provides the recommendation targetedly that meets specific user, and this recommendation has strengthened the convenience that the user of user's experience, raising uses, and has also improved user's work efficiency.

Claims (1)

1. the coal information commending system based on collaborative filtering, this system comprises coal information collaborative filtering module, community users colony access log module and recommending module; Coal information collaborative filtering module is responsible for the access log of analysis user, utilizes the algorithm of collaborative filtering to find out user neighbour; Community users colony access log module is responsible for the management of daily record; Recommending module is responsible for the information of calculated recommendation and is recommended user; Wherein, in coal information collaborative filtering module, adopt following methods to filter:

(1) system initialization: based on the access log of user in the community of colliery, build coal mine user-colliery webpage rating matrix A, matrix A comprises s user u ₁, u ₂..., u _sto t webpage p ₁, p ₂..., p _tscoring, A is the matrix on s × t rank, s and t are natural number, the element a in matrix _ijrepresent the scoring of user i to webpage j, i and j are natural number, and 1≤i≤s, 1≤j≤t,

(2) rating matrix optimization: a t webpage obtains k bunch after K-Means clustering algorithm, is designated as c ₁, c ₂..., c _k, k is number of clusters order, k is natural number; Bunch c _mthe webpage number comprising is N _m, this N _mthe sequence number of individual webpage in original t webpage is respectively f (m, 1), f (m, 2) ..., f (m, N _m), f function is for calculating the original number of bunch webpage after cluster, i.e. sequence number in original t webpage, f (m, N _m) expression bunch c _min N _mthe original number of individual webpage, m is natural number, 1≤m≤k, 1≤N _m≤ t, the rating matrix after being optimized

wherein b _imthe element of representing matrix B,

(3) neighbor searching: user u _va bunch sequence number set of marking is M _v, user u _wa bunch sequence number set of marking is M _w, user u _vand u _wa bunch sequence number set of all marking is M, and v and w are natural number, 1≤v≤s, 1≤w≤s, user u _vand u _wbetween similarity be

wherein

with

represent respectively user u _vand u _wto the average score of webpage, try to achieve each user u _ineighbour user in user's space, that sort from high to low by similarity gathers U _i;

(4) recommend: according to U _idraw set U _iin all users, by scoring from high to low, the collections of web pages P that browses _i, user u _ithe collections of web pages of having browsed is P' _i, set (P _i-P' _i) in before L coal information webpage be user u _iinterested content.

Images