CN106897370B - Picture examination expert recommendation method based on Pearson similarity and FP-Growth - Google Patents

Abstract

The invention discloses a recommendation method of image review experts based on Pearson similarity and FP-Growth, which comprises the steps of preprocessing project attributes in a comprehensive project record set, adopting a Pearson similarity method, extracting ten items of review experts with the scale closest to that of a project to be reviewed, combining the extracted experts according to branch projects of the comprehensive project to be reviewed and the research direction of the image review experts, processing a historical project review expert set by using the FP-Growth method to obtain an image review expert combined frequent item set, calculating the support degree of each expert combined set by using the combined frequent item set, and finally obtaining the expert combined set with the maximum support degree, namely the expert combined set with the highest degree of engagement, which is the expert set participating in the project to be reviewed. The method effectively recommends the image review expert combination with the highest fitting degree, improves the cooperative review efficiency of the experts, and increases the use value of the historical project review expert set data.

Description

Picture examination expert recommendation method based on Pearson similarity and FP-Growth

Technical Field

The invention belongs to the technical field of recommendation algorithms and association rule mining, and particularly relates to a recommendation method for experts in image review based on Pearson similarity and FP-Growth, which is mainly used for calculating the support degree, namely the engagement degree, of a project review expert combination so as to improve the collaborative review efficiency of experts and increase the use value of historical project review expert set data.

Background

The project examination expert recommendation algorithm has important function and significance for realizing the efficient selection of project examination experts in the project examination field. The traditional way of manually selecting the project review expert group has been unable to meet the requirements of the project review field. In recent years, researchers have proposed corresponding personalized recommendation schemes, such as content-based recommendation, collaborative filtering, association rules, utility recommendation, combined recommendation, and the like, for the needs of different recommendation systems.

The existing research bases of von willi, vermilion and the like include: wanli Feng. research of the maintenance extraction for chip degradation base on lexical cellulose. International Journal of Multimedia and Ubiotous Engineering, Vol.11, No.6(2016), pp.379-388; wanli Feng, Ying Li, Shangbing Gao, Yunyang Yan, Jianxun xue.Anovel flame edge detection algorithm via a novel active restriction model.International Journal of Hybrid Information Technology, Vol.9, No.9(2016), pp.275-282; liu jin Ling, Von Wanli. Pattern matching method based on attribute dependency [ J ]. microelectronics and computers, 2011,28(12): 167-; liu jin Ling, von Wanli, Zhang Yao red text clustering [ J ] of initializing cluster centers and reconstructing scale functions computer application research, 2011,28(11): 4115-; liu jin Ling, von Wanli, Zhang Yao red Chinese text clustering method based on rescaling [ J ] computer engineering and applications, 2012,48(21): 146-; the classification and extraction algorithm of Web science and technology news [ J ] academic newspaper of Huaiyin institute of Industrial science and technology, 2015,24(5): 18-24; lixiang, Zhu-Quanyin, collaborative clustering and scoring matrix shared collaborative filtering recommendations [ J ] computer science and exploration 2014,8(6): 751-; quanyin Zhu, Sun qun Cao. anovel Classifer-independent FeatureSelect Algorithm for Imbalanced Datases.2009, p: 77-82; quanyin Zhu, Yunyang Yan, Jin Ding, Jin Qian, the Case Study for Price extraction of Mobile PhoneShell Online.2011, p: 282-285; quanyin Zhu, Suqun Cao, Pei Zhou, Yunyang Yan, hong Zhou. Integrated print for based on Dichotomy Back filling and Disturbancework elastomer International Review on Computers and Software, 2011, Vol.6(6): 1089-; the related patents applied, published and granted by cinnabar, von willebra et al: an intelligent wireless network terminal device for monitoring the state of a refrigerated truck, namely Von Wanli, Shaohuashuai and Zhuang Jun, is CN203616634U [ P ] 2014; zhuquanhyin, Hurongjing, He Su group, peri-culture and the like, a commodity price prediction method based on linear interpolation and self-adaptive sliding windows, Chinese patent ZL201110423015.5,2015.07.01; the Chinese patent ZL 201110422274.6,2013.01.02; li Xiang, Zhu quan Yin, Hurong Lin, Zhonhang an intelligent recommendation method for cold-chain logistics stowage based on spectral clustering Chinese patent publications CN105654267A, 2016.06.08.

Pearson product-moment correlation coefficient:

pearson product-moment correlation coefficient (Pearson product-moment correlation coefficient) is used to measure the correlation between two variables X and Y, with values between-1 and 1. In the field of natural science, this coefficient is widely used to measure the degree of correlation between two variables.

And (3) association rule algorithm:

association rule based recommendations are more common in e-commerce systems and have also proven effective in the sense that users who have purchased some items prefer to purchase others, and the primary goal of association rule based recommendations systems is to mine association rules, i.e., collections of items that are purchased by many users at the same time, and the items in these collections can be recommended to each other. Recommendation systems based on association rules generally have a higher conversion rate because when a user has purchased several items in a frequent collection, there is a higher likelihood of purchasing other items in the frequent collection. However, the association rule of the mining project set has a large calculation amount, and meanwhile, the problem of sparsity of user data also exists, so that the recommendation accuracy is reduced.

FP-Growth algorithm:

the FP-Growth algorithm is a correlation analysis algorithm proposed by Hanwein et al in 2000, and adopts the following divide-and-conquer strategy: the database providing the frequent item set is compressed to a frequent pattern tree (FP-tree), but the item set association information is still retained. The FP-tree is a special prefix tree and is composed of a frequent item head table and an item prefix tree. The FP-Growth algorithm accelerates the whole excavation process based on the structure. Compared with the Apriori algorithm in the frequent item set algorithm for mining association rules, the FP-Growth algorithm adopts a divide-and-conquer strategy to mine the database, does not generate a candidate item set, adopts the FP-Tree to store important information of the database, only needs to scan the database twice, then stores key information in a memory in the form of the FP-Tree, and avoids huge expense brought by scanning the database for many times.

Disclosure of Invention

The purpose of the invention is as follows: the conventional project review expert group is manually selected, and there is a problem in that: the selected expert group does not examine projects of similar scale, which wastes a lot of time; the selected expert group members have low conformity, which results in low project examination efficiency. Aiming at the problems of the traditional method, the invention adopts a picture review expert recommending method based on Pearson similarity and FP-Growth to recommend an examination expert group with the highest conformity for the project to be reviewed by comprehensively analyzing a historical project examination expert set and a historical comprehensive project record set.

The technical scheme is as follows: the invention provides a diagram examination expert recommending method based on Pearson similarity and FP-Growth, which comprises the following steps:

step 1: carrying out normalization pretreatment on project attributes in a record set of projects to be examined and comprehensive projects, wherein the projects to be examined and the comprehensive projects are represented by comprehensive project types, branch project types of the comprehensive project types and the project attributes, and the specific method comprises the following steps:

step 1.1: defining a comprehensive type of item, a branch type of item and an item attribute;

step 1.2: recording the maximum value and the minimum value of each item of data in the item attribute of the comprehensive item record set;

step 1.3: carrying out normalization processing on the data of the comprehensive project record set and the project attributes of the projects to be processed, wherein the specific formula is as follows:

A_norm＝(A-A_min)/(A_max-A_min)

in the formula, A_maxAnd A_minRespectively the maximum value and the minimum value of each item of data of the item attribute, A is the data before normalization, A_normIs normalized data.

Step 2: processing the normalized data set by a Pearson similarity method to obtain ten items which are closest to the scale of the items to be examined, and extracting ten items of examination experts, wherein the examination experts are represented by the types of the researched branch items and examination item records, and the specific method comprises the following steps:

step 2.1: defining a picture inspection expert data set and a reviewed project record set, wherein the picture inspection expert data is represented by an expert number and a branch project type of expert research, and the picture inspection expert data set is represented by a project number and a picture inspection expert number;

step 2.2: integrating experts in the reviewed project record set according to the project numbers to obtain an engineering project review expert set for reviewing different projects;

step 2.3: calculating the similarity of the project to be examined and each project in the comprehensive project record set, wherein the specific formula is as follows:

in the formula, sim_iFor similarity of the project to be examined to the ith project, X_jAnd Y_ijRespectively are project attribute data set elements of a project to be checked and an ith project;

and

respectively the mean values of the project attribute data of the project to be examined and the project of the ith;

step 2.4: and sequencing the similar pairs, and extracting the project numbers corresponding to the first ten projects and the corresponding review expert sets to obtain the preselected image review expert sets.

And step 3: combining the extracted experts according to the branch project type of the comprehensive project of the project to be examined and the research direction of the picture examination experts to obtain all candidate combination expert sets, wherein the specific method comprises the following steps:

step 3.1: removing experts with examination tasks from a preselected image examination expert set;

step 3.2: selecting the experts with the same research branch project type as the branch project type of the project to be examined from the expert set obtained in the step 3.1, and expressing the experts according to the branch project type;

step 3.3: if the expert set obtained in the step 3.2 has the branch type of the project to be examined and has no expert, searching the expert data set for examining the branch project type from all image examination expert data sets according to the branch type of the project and adding the expert without a work task;

step 3.4: and 3.3, extracting at least one expert from each branch item type corresponding to the expert set obtained in the step 3.3, and obtaining all the candidate combination expert sets.

And 4, step 4: processing the historical project examination expert set by using an FP-Growth method to obtain a frequent item set of the picture examination expert combination;

and 5: the support degree of each candidate expert combination set is calculated by utilizing the combined frequent item set, and the expert combination set with the maximum support degree, namely the highest degree of engagement, is the expert set participating in the items to be examined, wherein the specific method comprises the following steps:

step 5.1: taking a candidate combined expert set as an example,the expert set has n experts, 1 expert is extracted from the candidate combined expert set, and the total number is

The extraction method comprises extracting 2 experts from the candidate combined expert, and sharing

The extraction mode is analogized, and n is extracted as an expert all the time, and the total is

The extraction mode is that all the extraction results are combined into a Subset set, and the number of the Subset sets is

Initializing the fitness SValue of the candidate combined expert set to be 0;

step 5.2: traversing Subset, if one extracted expert combination in Subset is in the picture expert combination frequent item set, adding the product of the frequency number in the frequent item set corresponding to the extracted expert combination and the expert number in the extracted expert combination to the degree of engagement of the candidate combination expert set in step 5.1, namely:

SValue＝SValue+f*k

in the formula, SValue is the degree of fit of the expert set of the alternative combination, f is the frequency of the frequent item set corresponding to the extracted expert combination, k is the number of experts in the extracted expert combination, and the traversal is finished, so that the final degree of fit of the expert set of the alternative combination in the step 5.1 is obtained;

step 5.3: and (5) calculating the degrees of engagement of all the candidate combination expert sets by the methods of the steps 5.1 and 5.2, wherein the candidate combination expert set with the highest degree of engagement is the expert set participating in the project to be examined.

By adopting the technical scheme, the invention has the following beneficial effects: the method effectively recommends an image review expert combination with the highest conformity by utilizing the comprehensive project record set and the historical project review expert set, improves the review efficiency and specifically comprises the following steps: the invention utilizes expert review project history records to carry out data mining, finds combination relation and contact degree among experts, adopts a Pearson similarity algorithm to obtain a historical project review expert set similar to a project to be reviewed, extracts experts without review tasks in the expert set, and combines the processed experts according to branch projects of the comprehensive project to be reviewed and the expert review direction, so that the experts contained in each combination are experts which have reviewed and are similar to the project to be reviewed. In addition, the invention creatively provides an expert combination conformity algorithm for calculating the conformity of each expert combination, the expert group with the highest conformity is the expert group of the item to be examined recommended finally, and the examination efficiency is improved.

Drawings

FIG. 1 is an overall flowchart of a diagram review expert recommendation method;

FIG. 2 is a flow diagram of a project and review expert related data preprocessing and association rules method;

FIG. 3 is a flow chart of a method for normalization processing and similarity calculation of project-related data;

FIG. 4 is a flow chart of an expert panel method;

FIG. 5 is a flowchart of a method for selecting a sum of expert groups with the highest fitness among all candidate expert combinations;

FIG. 6 is a flow chart of the adaptive fitness method for each expert combination.

Detailed Description

The present invention is further illustrated by the following examples, which are intended to be purely exemplary and are not intended to limit the scope of the invention, as various equivalent modifications of the invention will occur to those skilled in the art upon reading the present disclosure and fall within the scope of the appended claims.

Step 1: performing normalization preprocessing on the project attributes in the record set of the projects to be inspected and the comprehensive projects, wherein the projects to be inspected and the comprehensive projects are represented by the comprehensive project type, the branch project type of the comprehensive project type and the project attributes, and the specific steps are as shown in fig. 2:

step 1.1: definition G₁,G₂,G₃,G₄,G₅Defining B for urban water supply and drainage, architectural decoration, residential building, house building investigation and single design engineering in comprehensive project type₁,B₂,B₃,B₄,B₅,B₆,B₇Geotechnical engineering investigation, building, heating and ventilation, electricity, structure, water supply and drainage and road types in the branch project types are respectively adopted, and the following relations are satisfied:

G₁＝{B₁,B₂,B₃,B₄,B₅,B₆,B₇},G₂＝{B₁,B₂,B₃,B₄,B₅,B₆},G₃＝{B₁,B₂,B₃,B₄,B₅,B₆},G₄＝{B₁,B₂,B₄,B₅,B₆},G₅＝{B₁,B₂,B₄,B₅,B₆}

step 1.2: ProjectInfo is defined as all the synthetic project datasets, ProjectInfo { pr }₁,pr₂,...,pr_A}，pr_i＝{id_i,G_B，Ar_i，Fl_i，Hi_i，Ac_i，Co_i，Am_iIs a single composite project dataset, where a ═ Card (projectinfo), the function Card () is used to calculate the collection element number, the variable i ∈ [1, a]Variable B ∈ [1,5 ]]，id_iNumbering the items, G_B，Ar_i，Fl_i，Hi_i，Ac_i,Co_i,Am_iRespectively indicates item numbers as id_iThe comprehensive project type, the floor area, the floor number, the building height, the accounts receivable, the formula content and the consumable amount of the project;

step 1.3: defining HP as a project to be processed, wherein the type of the comprehensive project is HPType, and a project data set HPInfo is { HPType, HAR, HFl, HHi, HAc, HCo, HAm }, wherein HAR, HFl, HHi, HAc, HCo, HAm are respectively the floor area, floor number, building height, accounts receivable, formula content and consumable amount of the HP project;

step 1.4: definition of Ar_min,Fl_min,Hi_min,Ac_min,Co_min,Am_minMinimum values of Ar, Fl, Hi, Ac, Co, Am in ProjectInfo in step 1.2, Ar_max,Fl_max,Hi_max,Ac_Max,Co_max,Am_maxDefining a loop variable P for traversing the ProjectInfo in the step 1.2 according to the maximum values of Ar, Fl, Hi, Ac, Co and Am in the ProjectInfo in the step 1.2, wherein the initial value of P is 1;

step 1.5: when the circulating variable P is less than or equal to A, turning to the step 1.6; otherwise, executing step 1.8;

step 1.6: ar (Ar)_P＝(Ar_P-Ar_min)/(Ar_max-Ar_min)，Fl_P＝(Fl_P-Fl_min)/(Fl_max-Fl_min)，Hi_P＝(Hi_P-Hi_min)/(Hi_max-Hi_min)，Ac_P＝(Ac_P-Ac_min)/(Ac_max-Ac_min)，Co_P＝(Co_P-Co_min)/

(Co_max-Co_min)，Am_P＝(Am_P-Am_min)/(Am_max-Am_min) (ii) a Namely, the data in the comprehensive project record set is normalized;

step 1.7: changing to step 1.5 when P is P + 1;

step 1.8: HAr ═ Ar (HAr-Ar)_min)/(Ar_max-Ar_min)，HFl＝(HFl-Fl_min)/(Fl_max-Fl_min)，HHi＝(HHi-Hi_min)/(Hi_max-Hi_min)，HAc＝(HAc-Ac_min)/(Ac_max-Ac_min)，HCo＝(HCo-Co_min)/(Co_max-Co_min)，HAm＝(HAm-Am_min)/(Am_max-Am_min) (ii) a I.e., the normalization of the data for the item to be reviewed.

Step 2: processing the normalized data set by a pearson similarity method to obtain ten items which are closest to the size of the item to be examined, and extracting ten items of examination experts, wherein the examination experts are represented by the types of the researched branch items and examination item records, and the details are shown in fig. 3:

step 2.1: define ExtInfo ═ { ExtInfo }₁,expertInfo₂,...,expertInfo_E"is all graphic examination expert data set, expertInfo_F＝{Ma_F,B_gIs a single image review expert data set, expert all ═ Ma₁,Ma₂,...,Ma_EThe expert numbering set for all charts, where E ═ card (expertinfo), Ma_FNumbering the experts in the Picture, variable F ∈ [1, E]，g∈[1,7]，B_gIs numbered as Ma_FBranch item types studied by the panel examination experts;

step 2.2: define CenSorOpionions as a set of records of reviewed projects for panel review experts, CenSorOpionions { { id { (id)₁,Ma_C1},{id₁,Ma_C2},...,{id_A,Ma_D1},{id_A,Ma_D2H, wherein, C1, C2, D1, D2 ∈ [1, E ]]，N＝Card(CenSorOpinions)；

Step 2.3: for id in the CenSorOpionins dataset in step 2.2_iAnd performing row-column conversion on the Ma data items in the same data subset to obtain an engineering project examination expert set: ExpertJoin ═ ExpertJoin { (ExpertJoin)₁,expertJoin₂,...,expertJoin_ATherein, expertJoin_b＝{{Ma_H,...,

Ma_IIs numbered id_bPr of_bProject review expert set, variables H, I ∈ [1, E]，b∈[1,A]；

Step 2.4: a loop variable R is defined to traverse all the synthetic project data sets ProjectInfo, X ═ HAr, HFl, HHi, HAcc, HCo, HAm } in step 1.2,

sim_Rfor the synthetic item pr in ProjectInfo in step 1.2_RSimilarity with the item HP to be processed, Sim, is a similarity set, wherein R ∈ [1, A]R is given an initial value of 1, id_RFor a single general item pr_RItem number of (1), Sim is given an initial value of

Step 2.5: when the circulating variable R is less than or equal to A, executing the step 2.6; otherwise, turning to step 2.9;

step 2.6: y ═ Ar_R,Fl_R,Hi_R,Ac_R,Co_R,Am_RAnd (c) the step of (c) in which,

step 2.7:

wherein, X_r1，Y_r1Respectively, the r1 th data items in X and Y,

respectively represents the average value of the elements in X and Y,

respectively, the average values of the elements in X and Y, Sim-Sim ∪ { id }_R,sim_R}；

Step 2.8: making R ═ R +1, go to step 2.5;

step 2.9: obtaining Sim { { id₁,sim₁},{id₂,sim₂},...,{id_A,sim_AGet the ordered similarity set Sim_i＝{{id_j1,a_j1},{id_j2,a_j2},...,{id_jA,a_jA} in which a is_j1≥a_j2≥...≥a_jA，{id_jt，a_jt}∈Sim，jt，j1，j2，jA∈[1，A]，SimProject＝{{id_j1,a_j1},{id_j2,a_j2},...,{id_j10,a_j10}}；

Step 2.10: defining Forecast as a pre-selected image review expert set and assigning an initial value as

Defining a cyclic variable V, which is used for traversing SimProject in the step 2.9, wherein an initial value of V is 1, and a variable T is defined as a preselected image review expert number;

step 2.11: when the circulation variable V is less than or equal to 10, executing the step 2.12; otherwise, go to step 2.14;

step 2.12: let the project review expert collect expert join_jVNumber item id_jVThe set of censored experts of (a),

expert set for pre-selection picture examination

Step 2.13: changing V to V +1, and turning to step 2.11;

step 2.14: obtaining a preselected image review expert set Forecast ═ { Ma_m1，Ma_m2，...，Ma_mnIn which, Ma_miFor the ith data item, Ma, in the pre-selected panel review expert set Forecast_mi∈ExpertAll，mi∈[1,E]。

And step 3: combining the extracted experts according to the branch project type of the comprehensive project of the project to be examined and the research direction of the picture examination experts to obtain all candidate combination expert sets, which are specifically shown in fig. 4:

step 3.1: defining Work as a set of experts in image review with review task, Work { Ma }_u1,Ma_u2,...,Ma_unAnd pre-selecting a Forecast-Work expert set, wherein Ma_uiFor the ith data item in Work, Ma_ui∈ExpertAll，ui∈[1,E]；

Step 3.2: defining generalized item types G_N2＝{E₁,E₂,...,E_Z}，

GN2 ═ { E1, E2., EZ } is to be involved in the generalized item type G_N2A set of reviewing panel experts for review, wherein EJ is a to-be-participated branch item type E_JThe expert panel of the examination, EJ is given an initial value of

G_N2HPtype, i.e. generalized item type G_N2For the integrated item type HPtype, Z ═ Card (G) of the item to be processed HP in step 2.1_N2)，Z∈[5,7]，J∈[1,Z]；

Step 3.3: define a cyclic variable Num1, Num2 is used to traverse GN2 in step 3.2 and Forecast in step 3.1, respectively, and are both assigned an initial value of 1, Num3 ═ card (Forecast), E_Num1Concentrate the Num1 branch item type, Ma, for the GN2 panel expert in step 3.2_Num2Numbering the Num2 diagram examination experts in Forecast in the step 3.1;

step 3.4: when the loop variable Num1 is less than or equal to Z, executing the step 3.5; otherwise, turning to step 3.17;

step 3.5: when the loop variable Num2 is not more than Num3, executing the step 3.6; otherwise, turning to step 3.10;

step 3.6: let B_Num4Is number Ma_Num2Branch item types of expert research, { Ma_Num2:B_Num4∈ ExpertInfo, in which, Num4 ∈ [1,7 ]]；

Step 3.7: when B is present_Num4＝＝E_Num1When it is, i.e. number Ma_Num2If the branch item type of expert research and the Num1 branch item types of GN2 image expert panel are collected, executing step 3.8; otherwise, turning to step 3.9;

step 3.8 GN2 in step 3.2 examine the Num1 data item ENum1 ═ ENum1 ∪ Ma in the expert set_Num2；

Step 3.9: turning to step 3.5 with Num2 ═ Num2+ 1;

step 3.10: when in use

If so, executing step 3.11; otherwise, turning to step 3.16;

step 3.11: defining a cyclic variable c for the traversal step1.3 ExpertInfo, ExpertInfo of the c-th data item_c＝{Ma_cTy, wherein ty is the number Ma_cC, assigning an initial value of 1 to the branch item type of expert review;

step 3.12: when the loop variable c is less than or equal to E, executing the step 3.13; otherwise, executing step 3.16;

step 3.13: when in use

And ty-E_Num1If yes, executing step 3.14; otherwise, executing step 3.15;

step 3.14 ENum1 ═ ENum1 ∪ Ma_c；

Step 3.15: c is made to be c +1, and then the step 3.12 is carried out;

step 3.16: turning to step 3.4 with Num1 ═ Num1+ 1;

step 3.17: get GN2 ═ { E1, E2., EZ }, EJ ═ Ma ═_J1,Ma_J2,...,Ma_Jnu}，

nu＝Card(EJ)，J∈[1，Z]

Step 3.18: defining ExportCom as a graph review expert combination set of all alternative review HPs, and defining Com as a graph review expert combination set of one alternative review HP;

step 3.19: define Com_N3＝{Q₁，Q₂，...，QN₅}，ExportCom＝{Com₁，Com₂，...，Com_N6}，S_N3Is Com_N3With a support of SC ═ S₁,S₂，...，S_N6Is a set of support degrees, where Q_N7Represents Com_N3N7 th Picture experts number, Q_N7EN7 is the N7 th data item in GN2 in step 3.17, 1. ltoreq. N7. ltoreq.Z for any element in EN7,

n5 ≦ Z, N3 ≦ N6, N3 assigned an initial value of 1, defined End as the panel review expert set for the HP project in final review step 2.1, and End assigned an initial value of 1

And 4, step 4: processing the historical project examination expert set by using an FP-Growth method to obtain a frequent project set of image examination expert combinations, specifically: processing the engineering project examination expert set ExpertJoin in the step 2.5 by using an association rule method FP-Growth to obtain a frequent item set relationship of all the image examination expert combinations, wherein relationship is { { relationship₁:fr₁}，{relationt₂:fr₂}，...，{relationt_M:fr_M} where, relative t_X1＝{r₁,r₂,...,r_j}，r_j∈ ExpertAlll, 1. ltoreq. j. ltoreq.E, variable M ═ card (relationship), X1 ∈ [1, M ≦ E]，H1∈[1,E]，fr_x1Represents relationship t_X1The frequency of (c).

Step 5, the method flow of the expert group with the highest degree of engagement in all the candidate expert groups goes from step 51 to step 5.8, which is specifically shown in fig. 5:

step 5.1: n3 in step 3.19 is used to traverse all candidate portfolio experts set ExportCom in step 3.19, and N6 in step 3.19 is the subset number of ExportCom;

step 5.2: when N3 is less than or equal to N6,

step 5.3: com in step 3.19_N3Assign to step X₁ExpertHandle from step 5.4.1 to step 5.4.14, Rel from Relationt to step 5.4;

step 5.4: performing step X₁I.e., step 5.4.1 through step 5.4.14;

step 5.5: step X₁That is, steps 5.4.1 through 5.4.14 execute the result SValue assignment to S_N3，S_N3Element N3 in SC in step 3.19;

step 5.6: n3 ═ N3+ 1;

step 5.7: order S_N4For the maximum value in SC, Com_N4Has a support degree of S_N4Wherein, N4 ∈ [1, N6 ]]；

Step 5.8: get the expert set End of the final examination HP project ═ K₁,K₂,...,K_ZI.e. End ═ Com_N4Word ∪ End, wherein,

step 5.4: calculating the support degree of each candidate expert combination set by each expert combination self-adaptive fitting degree method for the frequent item set, wherein the expert combination set with the highest support degree, namely the highest fitting degree, is the expert set participating in the item to be examined, and is shown in fig. 6 specifically:

step 5.4.1: define the expert combined set expert handle ═ { Ma ═ of graph review₁,Ma₂,...,Ma_NuSValue is the support of ExpertHandle, and all image experts combine frequent item sets Rel { { Rel { (Rel) }₁：f₁},{rel₂：f₂},...,{rel_M1：f_M1Num ═ card (experthandle), M1 ═ card (rel), SValue is assigned an initial value of 0;

step 5.4.2: define Subset ═ { Sub₁,Sub₂,...,Sub_Nu}，Sub₁＝{Su1₁,Su1₂,...,Su1_n1}，Su1_n1＝{d_kh}，Sub₂＝{Su2₁,Su2₂,...,Su2_n2}，Su2_n2＝{d_ki,d_kj}，Sub_Nu＝{SuNu₁}，SuNu₁＝{d_k1,d_k2,...,d_kNuIn which d is_kh,d_ki,d_kj,d_k1,d_k2,...,d_kNu∈ExpertHandle，

That is, Subset is an expert extracted from ExpertHandle and all the combined results, Sub₁To arbitrarily extract n1 ═ Nu combined result set, Sub, of 1 expert from ExpertHandle₂Consisting of 2 experts arbitrarily extracted from ExpertHandle

Combining result sets, Sub_NuNu experts are extracted from the ExpertHandle to form only one combined result set;

step 5.4.3: defining a loop variable index1 for traversing the Subset, wherein index1 is assigned an initial value of 1;

step 5.4.4: when the loop variable index1 is not more than Nu, executing the step 5.4.5; otherwise, go to step 5.4.14;

step 5.4.5: defining a loop variable index2 for traversing Sub_index1Therein, Suindex1_index2To be a slave Sub _index12 index sets extracted from the database, wherein an initial value of index2 is 1;

step 5.4.6: when the cyclic variable

If so, go to step 5.4.7; otherwise, go to step 5.4.13;

step 5.4.7: define a loop variable index3 for traversing Rel, defining { Rel_index3:f_index3The Rel index3 sets, wherein the variable index3 is given as an initial value of 1;

step 5.4.8: when the loop variable index3 is less than or equal to M1, go to step 5.4.9; otherwise, go to step 5.4.12;

step 5.4.9: when Suindex1_index2＝rel_index3If so, go to step 5.4.10; otherwise, go to step 5.4.11;

step 5.4.10: SValue + f_index3Index1, namely, the value of SValue is updated to the value of SValue plus the product of the specified expert combined frequency and the number of the expert combined experts;

step 5.4.11: turning to step 5.4.8 when index3 is equal to index3+ 1;

step 5.4.12: turning to step 5.4.6 when index2 is equal to index2+ 1;

step 5.4.13: turning to step 5.4.4 when index1 is equal to index1+ 1;

step 5.4.14: SValue is obtained.

The Pearson similarity method is characterized in that data analysis is carried out on a data set subjected to project attribute preprocessing, an FP-Growth method is used for processing a historical project examination expert set to obtain a frequent item set of image examination expert combinations, and the expert combination conformity method is used for calculating the support degree of each expert combination according to the frequent item set, namely the expert combination conformity degree.

Carrying out association rule mining on 65536 historical project examination expert records by a PF-Growth method to obtain a frequent item set of a graph examination expert combination; carrying out data compression and pretreatment on 20061 comprehensive project records, and extracting ten project examination experts with the scale closest to that of the project to be examined by adopting a Pearson similarity method, so that the extracted experts are similar to the project to be examined after examination; compared with the artificially recommended expert combination result, the similarity of the method in practical application reaches 82.13%, and the adoption rate reaches 97.25%.

Claims (4)

1. A method for recommending image review experts based on Pearson similarity and FP-Growth is characterized by comprising the following steps:

step 1: carrying out normalization pretreatment on project attributes in a record set of projects to be inspected and comprehensive projects, wherein the projects to be inspected and the comprehensive projects are represented by comprehensive project types, branch project types of the comprehensive project types and the project attributes;

step 2: processing the normalized data set by a Pearson similarity method to obtain ten items which are closest to the scale of the items to be examined, and extracting examination experts of the ten items, wherein the examination experts are represented by the types of the researched branch items and examination item records;

and step 3: combining the extracted experts according to the branch project type of the comprehensive project of the project to be examined and the research direction of the image examination experts to obtain all candidate combination expert sets;

and 4, step 4: processing the historical project examination expert set by using an FP-Growth method to obtain a frequent item set of the picture examination expert combination;

and 5: the support degree of each candidate expert combination set is calculated by utilizing the combined frequent item sets, and the expert combination set with the maximum support degree, namely the highest degree of engagement, is the expert set participating in the items to be examined, wherein the calculation method comprises the following steps:

step 5.1: taking an expert set of alternative combination as an example, the expert set has n experts, 1 expert is extracted from the expert set of alternative combination, and the expert set has the total

The extraction method comprises extracting 2 experts from the candidate combined expert, and sharing

The extraction mode is analogized, and n experts are extracted all the time, and the total is

The extraction mode is that all the extraction results are combined into a Subset set, and the number of the Subset sets is

Initializing the fitness SValue of the candidate combined expert set to be 0;

step 5.2: traversing Subset, if one extracted expert combination in Subset is in the picture expert combination frequent item set, adding the product of the frequency number in the frequent item set corresponding to the extracted expert combination and the expert number in the extracted expert combination to the degree of engagement of the candidate combination expert set in step 5.1, namely:

SValue＝SValue+f*k

in the formula, SValue is the degree of fit of the expert set of the alternative combination, f is the frequency of the frequent item set corresponding to the extracted expert combination, k is the number of experts in the extracted expert combination, and the traversal is finished, so that the final degree of fit of the expert set of the alternative combination in the step 5.1 is obtained;

step 5.3: and (5) calculating the degrees of engagement of all the candidate combination expert sets by the methods of the steps 5.1 and 5.2, wherein the candidate combination expert set with the highest degree of engagement is the expert set participating in the project to be examined.

2. The expert recommendation method for image review based on pearson similarity and FP-Growth according to claim 1, wherein the specific method in step 1 is as follows:

step 1.1: defining a comprehensive type of item, a branch type of item and an item attribute;

step 1.2: recording the maximum value and the minimum value of each item of data in the item attribute of the comprehensive item record set;

step 1.3: carrying out normalization processing on the data of the comprehensive project record set and the project attributes of the projects to be processed, wherein the specific formula is as follows:

A_norm＝(A-A_min)/(A_max-A_min)

in the formula, A_maxAnd A_minRespectively the maximum value and the minimum value of each item of data of the item attribute, A is the data before normalization, A_normIs normalized data.

3. The expert recommendation method for image review based on pearson similarity and FP-Growth according to claim 1, wherein the specific method in step 2 is:

step 2.1: defining a picture inspection expert data set and a reviewed project record set, wherein the picture inspection expert data is represented by an expert number and a branch project type of expert research, and the picture inspection expert data set is represented by a project number and a picture inspection expert number;

step 2.2: integrating experts in the reviewed project record set according to the project numbers to obtain an engineering project review expert set for reviewing different projects;

step 2.3: calculating the similarity of the project to be examined and each project in the comprehensive project record set, wherein the specific formula is as follows:

in the formula, sim_iFor similarity of the project to be examined to the ith project, X_jAnd Y_ijRespectively are project attribute data set elements of a project to be checked and an ith project;

and

respectively the mean values of the project attribute data of the project to be examined and the project of the ith;

step 2.4: and sequencing the similar pairs, and extracting the project numbers corresponding to the first ten projects and the corresponding review expert sets to obtain the preselected image review expert sets.

4. The expert recommendation method for image review based on pearson similarity and FP-Growth according to claim 1, wherein the specific method in step 3 is:

step 3.1: removing experts with examination tasks from a preselected image examination expert set;

step 3.2: selecting the experts with the same research branch project type as the branch project type of the project to be examined from the expert set obtained in the step 3.1, and expressing the experts according to the branch project type;

step 3.3: if the expert set obtained in the step 3.2 has the branch type of the project to be examined and has no expert, searching the expert data set for examining the branch project type from all image examination expert data sets according to the branch type of the project and adding the expert without a work task;

step 3.4: and 3.3, extracting at least one expert from each branch item type corresponding to the expert set obtained in the step 3.3, and obtaining all the candidate combination expert sets.

Images