CN115660296B - Automatic issuing method of non-compliance project proposal scheme based on machine learning - Google Patents

Abstract

The invention is applicable to the field of project compliance examination, and provides an automatic issuing method of an irregular project proposal scheme based on machine learning, which comprises the following steps: establishing a multi-rule integrated data base and a learning result library; automatically checking the information of the imported items; processing the inspected compliant items and non-compliant items; adopting an improved KNN algorithm to automatically give out examination comments aiming at non-compliance projects; matching compliance examination rules, training results, land mutual exclusion and land mutual adaptation rules from the obtained non-compliance examination opinions to obtain a final proposal scheme set; calculating project values for schemes in the suggested scheme set by adopting an entropy method; outputting a non-compliance project audit report. The invention realizes the automatic examination of the compliance of urban construction projects, can provide examination reports of the compliance projects and examination reports of non-compliance projects containing a plurality of proposal schemes ranked at the top in a short time, and improves the efficiency of project compliance examination and new scheme determination.

Description

Automatic issuing method of non-compliance project proposal scheme based on machine learning

Technical Field

The invention belongs to the field of project compliance examination, and particularly relates to an automatic issuing method of an irregular project proposal scheme based on machine learning.

Background

Project compliance review is a legal program that all city construction projects must pass before getting the land pre-review and site selection opinion books, and if the project fails the compliance review, the project scope needs to be adjusted and then the compliance review is performed, otherwise, the project cannot continue to advance. The same project is required to be subjected to multiple range adjustment and multiple compliance examination due to non-compliance, so that the project process is delayed, the administrative cost is wasted, and adverse effects are caused on the commercial environment and the social economy.

Existing compliance analysis requires the following steps to be taken: (1) the construction unit puts forward project compliance examination application; (2) the management department related departments respectively perform man-machine interaction to check the coincidence of project site selection and the planning of the management department of the family, and have examination comments; (3) related departments will review and comprehensive review opinions (without project non-compliance space position diagram); (4) feeding back the opinion to the construction unit; (5) a construction unit adjustment project; (6) and the process of checking … … "again forms a dead cycle of checking-non-compliance-adjustment-checking-non-compliance-adjustment … …", and related departments are needed to be abutted for a plurality of times, so that administrative cost is wasted and project progress is seriously affected.

Machine learning is a multidisciplinary cross-specialty covering probabilistic knowledge, statistical knowledge, approximate theoretical knowledge and complex algorithmic knowledge, uses a computer as a tool and aims at simulating human learning in real time, and performs knowledge structure division on existing content to effectively improve learning efficiency. How to use machine learning to improve project compliance examination efficiency, save administrative cost for natural resource management, optimize the commercial environment, and become the key work at the present stage.

Disclosure of Invention

In view of the above problems, the present invention aims to provide an automatic issuing method of an irregular project proposal based on machine learning, which aims to solve the above technical problems.

The invention adopts the following technical scheme:

a method for automatically issuing a non-compliance project proposal scheme based on machine learning comprises

Step S1, a multi-rule-in-one data base is established, and historical compliance analysis opinions are trained to obtain a learning result library;

step S2, importing project information, automatically selecting space management and control class planning and elements, selecting reference planning and elements, and automatically checking according to compliance checking rules;

s3, automatically issuing a compliance review opinion for a review compliance project, analyzing results of a review non-compliance project, and automatically issuing the non-compliance review opinion by adopting an improved KNN algorithm;

step S4, matching compliance censoring rules, training achievements in a learning achievements library, and matching land mutex rules and land inter-fit rules from the obtained non-compliance censoring opinions to obtain a final proposal set;

s5, calculating item values for schemes in the suggested scheme set by adopting an entropy method;

and S6, outputting a non-compliance project inspection report of a plurality of suggested schemes with top project value ranks.

The beneficial effects of the invention are as follows: firstly, the method can realize the automatic examination of the compliance of urban construction projects, and can provide examination reports of the compliance projects and examination reports of the compliance projects containing a plurality of proposal schemes with top ranking in a short time according to examination opinions and machine learning for the compliance projects, thereby having very simple operation and greatly improving the efficiency of project compliance examination and new scheme determination; in addition, when the method is realized in detail, a unified compliance data base is formed, an improved KNN algorithm is adopted, historical examination opinions of compliance projects are grouped, non-compliance projects are changed from original traversal of all samples to traversal of samples of the belonging group, self-adaptive K values and T values are found out, the belonging category can be obtained, examination opinions are automatically generated, and examination time is greatly saved; in addition, the rules of mutual exclusion and mutual adaptation are formulated, support is provided for project proposal schemes, the weights of proposal scheme influence factors are objectively and scientifically determined by adopting an entropy method, the comprehensive scores of all schemes in the proposal scheme set are determined, subjectivity caused by manually determining the weights is avoided, and the ranking of the proposal schemes is more scientific.

Drawings

FIG. 1 is a flow chart of a method for automatically issuing a non-compliance project proposal based on machine learning provided by an embodiment of the invention;

fig. 2 is a flowchart of step S4 provided in an embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

In order to illustrate the technical scheme of the invention, the following description is made by specific examples.

As shown in fig. 1, the method for automatically issuing a non-compliance project proposal based on machine learning according to the embodiment includes the following steps:

and S1, establishing a multi-rule and one-data base, and training historical compliance analysis opinions to obtain a learning result library.

The specific process of the method is as follows:

and S11, integrating space management and control type planning, elements and reference planning, elements, and standardizing according to a unified data standard to form the multi-rule and unified data base.

The space management and control class planning comprises the following steps: three control lines (permanent basic farmland, ecological protection red line and town development boundary), land utilization overall planning, city overall planning, and increasing and decreasing hooks to implement planning and control detailed planning.

The space management and control type elements comprise: basic farmland protection area, first-stage woodland, natural protection area, first-stage drinking water source protection area, river management range, scenic spot area, second-stage drinking water source protection area and quasi-water source protection area, river and its embankment (inland river), reservoir and its engineering area, raw water pipe, water conservancy main engineering and its production area, flood storage area, woodland, forest park, wetland park, geological disaster high-risk area, infrastructure isolation area, green zone of surrounding city, suburban park, ecological travel vacation area.

The reference programming includes: and (5) special planning.

The reference elements include: airport control, city updating, and rail traffic control area.

And normalizing the data according to unified data standards of metadata information such as unified data names, data types, data structures and the like, and finally forming the multi-rule and unified data base.

And S12, collecting historical compliance analysis opinions, and performing text preprocessing on the compliance analysis opinions to obtain text corpus.

And collecting historical project compliance analysis opinions, and performing text preprocessing on the historical compliance analysis opinions by using the technologies of word segmentation, noise reduction, optimal matching and the like. If not, the following are: a permanent basic farmland is occupied, and ecological protection red lines are occupied; the landfill is at least 500 meters from the residential area, etc.

S13, constructing a training model, and selecting polar words from the text corpus to perform model training.

Polar words are words that are used for positive sentences or only for negative sentences. Through polar word model training, the accuracy of training results is continuously improved.

And S14, performing polarity judgment and classification on the compliance analysis opinions by adopting a KNN algorithm, and storing a classified learning result library.

The method integrates three control lines, land utilization overall planning, city overall planning, hook increasing and decreasing implementation planning, special planning, management and control elements and reference elements. Various data are difficult to use due to non-uniform coordinate systems, data standards, data specifications and data formats. Therefore, a unified coordinate system and a unified data standard are needed, quality inspection is carried out on all data according to a unified quality inspection rule, and the data are incorporated into a multi-rule one planning data pool for project compliance analysis after being qualified.

And S2, importing project information, automatically selecting space management and control class plans and elements, selecting reference plans and elements, and automatically checking according to compliance checking rules.

The specific process of the method is as follows:

s21, importing project information of the basic information and the accessory material information.

The basic information includes item names, land attributes, item attributes, land areas. Basic information and accessory material information are imported, and the item range self-drawn by the user can also be imported. The imported project land parcel data format comprises CAD, SHP format and TXT format of natural resource homeland government edition.

S22, selecting a necessary selected space management and control class planning and elements, and selecting a reference planning and elements according to requirements.

Firstly, the system automatically selects the necessary space management and control type planning and elements, and a user selects the reference planning and elements according to the needs and then analyzes the reference planning and elements. The space management and control class planning and elements belong to the necessary options, the system is automatically selected and can not be manually changed, and the accuracy of the examination result is ensured. The reference planning and the elements belong to unnecessary options, and can be used for carrying out selective superposition analysis according to the actual situation of the land.

And S23, superposing the project information with the selected plans and elements, and analyzing the project compliance in a space superposition mode to realize automatic examination of the compliance.

And superposing the imported project information with the selected plan and elements, and carrying out superposition analysis on project compliance by adopting a spatial superposition analysis method of a geographic information system to obtain an analysis chart and analysis data.

And selecting space management and control type planning and elements, consulting the planning and the elements, combing the consistency of various land attributes with layers and elements according to the requirements of related legal and regulatory documents, and making an inspection rule. Such as: the educational facility is overlapped with three control lines, and can not be built in a permanent basic farmland range, can be built in a ecological protection red line range and can be built in a town development boundary range; the method is overlapped with the overall planning of land utilization, and accords with the construction of the final land pattern spots, the construction in the allowed construction area range, the construction in the conditional construction area range, the construction in the limited area range, the construction in the forbidden construction area and the construction in the town/village construction land range; overlapping with urban overall planning, wherein the urban overall planning is applicable to construction in a suitable construction area, is not applicable to construction in a limited construction area, is not applicable to construction in a forbidden construction area, is not applicable to construction in a urban ultraviolet range, is not applicable to construction in a urban yellow line range, is not applicable to construction in a urban green line range and is not applicable to construction in a urban blue line range; the method is overlapped with the addition and subtraction of hooks in a planning way, and cannot be constructed in the old area and the new area; overlapping with the space management and control type elements, and constructing can not be performed in case of conflict; and overlapping with a reference plan (educational facility special plan), and constructing the building by conforming the land scope and the land property.

And S3, automatically issuing the compliance review opinions for the review compliance items, analyzing results of the review non-compliance items, and automatically issuing the non-compliance review opinions by adopting an improved KNN algorithm.

The specific process of the method is as follows:

s31, constructing a compliance project and non-compliance project inspection report template, wherein the compliance project inspection report template comprises two parts of analysis details and analysis conclusions, and the non-compliance project inspection report template comprises three parts of analysis details, analysis conclusions and proposal schemes.

The analysis details comprise project land occupation space management and control planning, element and reference planning, areas and types of the elements, analysis results of various occupied/unoccupied plans and elements are listed one by one in a mode of combining figures, and whether compliance is judged according to occupied conditions and compliance inspection rules in an analysis conclusion.

S32, setting a link between the examination result and a corresponding module in the examination report template.

If the result of the project overlapping permanent basic farmland is positioned in the first row, the result of the project overlapping ecological protection red line is positioned in the second row, and the like.

S33, selecting different examination report templates according to an automatic examination result, and selecting a compliance project examination report template if the examination report templates are compliance; if not, selecting a non-compliance project inspection report template.

And S34, automatically matching a learning result library by adopting an improved KNN algorithm according to the plan of the occupation of the non-compliance project and the number and the category of the elements, and automatically issuing the examination comments.

And superposing the project, the space management and control plan and the elements as well as the reference plan and the elements, and determining the plan, the element number and the category of the project occupation according to the formulated examination rules. The method comprises the steps of automatically analyzing the plan, the element number and the category of the occupation of the non-compliance project, automatically matching a learning result library by adopting an improved KNN algorithm, and automatically issuing the examination opinion.

KNN (K-nearest neighbor), known as the K nearest neighbor method, is a basic classification and logistic regression algorithm in machine learning. The working mechanism is relatively simple: given a sample set, also called a sample training set, the correspondence of the data in the sample data set to the class to which it belongs is known. After inputting a new data without classification, the computer extracts the features of the data and compares them with the features in the dataset to find out K data close to the features, the most frequent classification among the K data being the classification of this new data. The algorithm does not have any model parameters and does not need to train the model parameters. It has two major drawbacks: (1) The time complexity of the calculation process is high and the efficiency is low. Because the number of the training samples of the manual historical examination opinion is large, when classifying and calculating any new sample to be measured, all training sample sets must be traversed, so that the calculation process is longer, the complexity is higher and the efficiency is low; (2) accuracy is not high. Because the value of K is not verified to be the most proper value of K, K is too small, and the classification result is easily affected by noise points; too large K, in turn, may contain too many other classes of points in the neighborhood, leading to a significant degree of inaccuracy in the results.

This embodiment makes the following two improvements: (1) And re-grouping the historical sample sets according to the number of the non-conforming layers and the elements, and only traversing all samples in the corresponding groups. (2) And obtaining a K value optimal solution by adopting an MAPE evaluation method and a prediction algorithm. The specific implementation process is as follows:

341. and grouping historical compliance analysis opinions in the learning result library according to the number of layers of the project non-conforming map.

The natural language processed historical review opinions are trained separately according to the project non-conforming layer number groupings. According to compliance analysis rules, the project needs to be subjected to superposition analysis with 45 layers and elements at most, so that historical examination opinions are divided into: a first group that does not conform to 1 layer and element; a second group of non-conforming 2 layers and elements; sequentially provided withSet 45 groups that do not correspond to 45 layers and elements. The possible cases of non-conforming layers and elements in each class of data sets are

C represents combination mathematically, n represents the total number of layers and elements, m represents the number of layers in each group where the items do not fit; m is equal to or greater than 1 and n is equal to or less than 45.

342. And determining corresponding groups according to the number of the non-compliance layers and the elements of the sample to be tested.

After the grouping, the grouping of the samples to be tested can be determined.

343. And traversing all samples in the current group, establishing a search mechanism by using a state vector, a distance measurement mode and the number K of neighbors, and searching.

The state vector is a standard for comparing the current examination result with the historical examination opinion, and generally, the factor most relevant to the sample to be tested is selected to balance the requirements of prediction precision and traversal time. By observing the classification of the censoring opinions in the history censoring opinion library.

The distance measurement mode is used for measuring the approximation degree of each historical sample in the historical database correlation group and the current sample to be measured, and Euclidean distance is used as a measurement index:

。/>

the distance between the historical sample vector and the sample vector to be measured is; />

There are u non-conforming layers/elements for the sample to be tested, < ->

S possible cases are contained for u non-conforming groups of layers; t has a value of [1,45 ]]。

344. And calculating a prediction result by using the K adjacent data in the searched current group as a test data set and other data as a training data set and using a K neighbor prediction algorithm.

345. And when the MAPE obtains the minimum value, the corresponding group and K pieces of adjacent data are the optimal values of the sample to be tested, and the optimal values are examination comments corresponding to the category to which the current non-compliant item belongs.

For more intuitive prediction accuracy and difference of the reaction model, MAPE (mean absolute percentage error) is used, smaller MAPE indicates higher accuracy, i.e. closer.

Wherein r is the number of samples; />

Is the actual value of the sample; />

Is a sample predictor.

How to predict the non-compliance examination result of the sample to be tested by using the searched K pieces of adjacent data is searched and predicted:

wherein:

predicting values for non-compliance data; />

The method comprises the steps of searching an i-th adjacent corresponding non-compliance examination result in a history examination opinion database; />

Is the distance of the current vector from the i-th nearest point.

Then determining an adaptive K value and T value:1. the classification of the sample to be detected and the historical sample data is matched, the sample to be detected is matched into the nth class, and N=is set

The method comprises the steps of carrying out a first treatment on the surface of the 2. Sample to be tested->

Determining the value of a state vector T; 3. let t= =>

，

The method comprises the steps of carrying out a first treatment on the surface of the Is provided with->

，/>

The method comprises the steps of carrying out a first treatment on the surface of the 4. K pieces of neighborhood data +.A.in the N-th group of the history review opinion dataset are selected>

As test dataset, other +.>

The data are used as training data sets; 5. calculate->

And->

Test data set->

Average absolute error percentage: />

The method comprises the steps of carrying out a first treatment on the surface of the 6. Calculate->

And->

Average absolute error percentages for all of the following test data sets: />

。

When MAPE takes the minimum value, the predicted corresponding

And->

The optimal value of the sample to be measured in the scene is obtained. The optimal value is the examination opinion corresponding to the category to which the non-compliant item belongs.

The step finally obtains the classification category of the sample to be tested, namely the examination opinion of the non-compliance item, and finally automatically fills the examination opinion to the corresponding position according to the set link so as to achieve the purpose of automatically issuing the examination opinion. Through the improved KNN model, the groups of the samples to be traversed can be selected according to the number of the non-compliant layers and the elements, so that the time required by traversing all the samples of the original KNN model is greatly reduced, and the examination efficiency is greatly improved.

S35, according to the set links, automatic input of the inspection comments of compliance and non-compliance is achieved.

And according to the links established in the steps, realizing automatic input of the examination opinions, for example, the project occupies a permanent basic farmland, wherein the examination opinions are as follows: non-compliance, permanent basic square meter of the farmland A occupied by the pressure, and automatically inputting a graph of the occupied space range and the occupied area A square meter into a specified position; if the ecological protection red line is occupied and the permanent basic farmland is not occupied, the examination opinion is: and (3) the method is not in compliance, the square meter B is occupied by the ecological protection red line, the graph of the ecological protection red line and the square meter B are automatically input, and the superimposed graph of the project range and the permanent basic farmland and the square meter 0 are automatically input. Automatically completing the examination report of the compliance project according to the set link, clicking to generate a preview and exporting the examination report; the non-compliance project automatically outputs the censoring opinion.

And S4, matching compliance censoring rules, training achievements in a learning achievements library, and matching land mutex rules and land inter-fit rules from the obtained non-compliance censoring opinions to obtain a final proposal scheme set.

As shown in fig. 2, the specific process of this step is as follows:

s41, entering a mutual exclusion rule and a mutual adaptation rule.

The site selection rules include site mutual exclusion rules and site mutual adaptation rules. Based on the related standard specification and the requirements of various land on the surrounding environment, the land mutual exclusion and land mutual adaptation rule is obtained. For example, the education facilities and the sites are mutually suitable, the education facilities and the sites can be closely adjacent to each other when the proposal is issued, the education facilities and the sites are irrelevant to the administrative office, the education facilities and the sites are mutually exclusive to the medical and health sites, and the education facilities and the sites are far away from each other when the proposal is issued.

And S42, automatically matching the obtained non-compliance review opinions with compliance review rules, and removing the occupied planning and the space range of the elements in the review opinions according to a space superposition analysis method to obtain a first pre-proposal scheme set.

S43, matching the first pre-proposal set with training results in a learning result library, such as that the landfill is at least 500 meters away from the residential area, and excluding all proposal schemes of the landfill which are less than 500 meters away from the residential area, so as to obtain a second pre-proposal set.

And S44, matching the second pre-proposal scheme set with the mutual exclusion rule and the mutual adaptation rule to obtain a final proposal scheme set.

And S5, calculating item values for the schemes in the proposal scheme set by adopting an entropy method.

The specific process of the method is as follows:

s51, scoring the influence factors of each proposal in the proposal set according to the traffic accessibility analysis mode, the distance analysis mode, the mutual exclusivity rule of the land and the mutual adaptability rule of the land.

For example, a traffic accessibility analysis mode is adopted, the accessibility degree of the road traffic around each proposal in the proposal set is calculated, and the traffic accessibility influence factor in each proposal is scored according to the set scoring rule; if the reaching degree is more than or equal to 90 percent, dividing into 10 points; the reachable degree is more than or equal to 70% and less than 90%, and the number is divided into 8; the reachable degree is more than or equal to 50% and less than 70%, and the time is divided into 5 points; the reachable degree is more than or equal to 30 percent and less than 50 percent, and the number is divided into 3; less than 30%, and the score is 0.

The scores of other influencing factors are also set with reference to this rule. Similarly, the score of each scheme under each influence factor is calculated respectively.

S52, determining the weight of each influence factor by adopting an entropy method.

And determining weights occupied in comprehensive scores of the recommended schemes by adopting an entropy method, such as traffic accessibility, distance analysis, land mutex, and the like, comprehensively scoring all schemes in the recommended scheme set, and sequencing the schemes according to the score from high to low, wherein the higher the score is, the better the comprehensive condition of the specified scheme is.

And S53, calculating the comprehensive scores of all the schemes in the proposal scheme set according to the scores and the weights, namely calculating the project values, and sorting according to the sizes of the project values.

And S6, outputting a non-compliance project inspection report of a plurality of suggested schemes with top project value ranks.

And outputting the schemes with the comprehensive scores of the top three schemes, setting the schemes with the same scores for the schemes with the same scores, and keeping the schemes with the same scores in the schemes with the top three ranks.

The invention realizes the compliance of automatically checking urban construction projects, and can provide a plurality of compliance proposal schemes for non-compliance projects in a short time according to manual checking opinion and machine learning. Taking a landfill as an example, the traditional compliance examination needs to take 5 examination staff in different departments respectively, each person takes about 10 minutes, 1 person needs to be put into the project scheme for 1 time for about 0.5 days, and if the adjusted project is still not compliant, the time is accumulated according to the time; the invention can complete the compliance examination and the compliance proposal with the efficiency increased to 1 minute, greatly improves the efficiency of project compliance examination and new proposal determination, reduces repeated adjustment and examination caused by project non-compliance, greatly improves examination efficiency and saves administrative cost.

In summary, the invention unifies data standards of all space planning, elements and referential planning and elements to form a unified compliance examination data base, and a construction unit does not need to butt up a main department of related planning one by one to examine whether projects are in compliance, so that the time cost is greatly saved, and the commercial environment is optimized; in addition, an improved KNN algorithm is adopted to group manual examination opinions of historical compliance analysis projects, the non-compliance projects are changed from original traversal of all samples to traversal of samples of the belonging group, self-adaptive K values and T values are found out, the belonging category can be obtained, examination opinions are automatically generated, examination time is greatly saved, and examination efficiency is improved; thirdly, creatively formulates compliance inspection rules of all the planning layers and elements to be overlapped in compliance analysis, and space management and control type planning/element automatic selection reduces interference caused by artificial factors; for the reference planning, the element examination personnel can automatically analyze and output the result only by selecting according to the need; fourth, the invention makes rules of mutual exclusion and mutual adaptation in land, and provides support for project proposal scheme; fifth, the invention adopts the entropy method to objectively and scientifically determine the weight of the evaluation index of the proposal, and combines the scoring rule to determine the comprehensive score of all proposal proposals in the proposal set, thereby avoiding subjectivity caused by manually determining the weight and leading the ranking of the proposal to be more scientific.

The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.

Claims (1)

1. A machine learning based method for automatically issuing a proposal of an irregular project, which is characterized by comprising the following steps:

step S1, a multi-rule-in-one data base is established, and historical compliance analysis opinions are trained to obtain a learning result library;

step S2, importing project information, automatically selecting space management and control class planning and elements, selecting reference planning and elements, and automatically checking according to compliance checking rules;

s3, automatically issuing a compliance review opinion for a review compliance project, analyzing results of a review non-compliance project, and automatically issuing the non-compliance review opinion by adopting an improved KNN algorithm;

step S4, matching compliance censoring rules, training achievements in a learning achievements library, and matching land mutex rules and land inter-fit rules from the obtained non-compliance censoring opinions to obtain a final proposal set;

s5, calculating item values for schemes in the suggested scheme set by adopting an entropy method;

s6, outputting a non-compliance project inspection report of a plurality of suggested schemes with top project value ranking;

the specific process of the step S1 is as follows:

s11, integrating space management and control class planning, elements and reference planning, elements and standardizing according to unified data standards to form a multi-rule and integrated data base;

s12, collecting historical compliance analysis opinions, and performing text preprocessing on the compliance analysis opinions to obtain text corpus;

s13, constructing a training model, and selecting polar words from the text corpus to perform model training;

s14, polarity judgment and classification are carried out on the compliance analysis opinions by adopting a KNN algorithm, and a classified learning result library is stored;

the specific process of the step S2 is as follows:

s21, importing project information of basic information and accessory material information;

s22, selecting a necessary space management and control type plan and elements, and selecting a reference plan and elements according to requirements;

s23, superposing the project information with the selected planning and elements, and carrying out superposition analysis on project compliance in a space superposition mode to realize automatic examination of the compliance;

the specific process of the step S3 is as follows:

s31, constructing a compliance project and non-compliance project inspection report template, wherein the compliance project inspection report template comprises two parts of analysis details and analysis conclusions, and the non-compliance project inspection report template comprises three parts of analysis details, analysis conclusions and proposal schemes;

s32, setting a link between the examination result and a corresponding module in the examination report template;

s33, selecting different examination report templates according to an automatic examination result, and selecting a compliance project examination report template if the examination report templates are compliance; if not, selecting a non-compliance project examination report template;

s34, automatically matching a learning result library by adopting an improved KNN algorithm according to the plan of the occupation of the non-compliance project and the number and the category of the elements, and automatically giving out examination comments;

s35, according to the set links, automatic input of the inspection comments of compliance and non-compliance is realized;

the specific process of step S34 is as follows:

341. grouping historical compliance analysis opinions in a learning result library according to the number of layers of the project non-conforming map;

342. determining corresponding groups according to the number of the non-compliance layers and the elements of the sample to be tested;

343. traversing all samples in the current group, establishing a search mechanism by using a state vector, a distance measurement mode and the number K of neighbors, and searching;

344. using the K adjacent data in the searched current group as a test data set, using other data as a training data set, and calculating a prediction result by using a K neighbor prediction algorithm;

345. when the MAPE obtains the minimum value, the corresponding group and K pieces of adjacent data are the optimal values of the sample to be tested, and the optimal values are examination comments corresponding to the category to which the current non-compliant item belongs;

the specific process of the step S4 is as follows:

s41, entering a land mutual exclusion rule and a land mutual adaptation rule;

s42, automatically matching the obtained non-compliance review opinions with compliance review rules, and removing space ranges of occupied plans and elements in the review opinions according to a space superposition analysis method to obtain a first pre-proposal scheme set;

s43, matching the first pre-proposal scheme set with training results in a learning result library to obtain a second pre-proposal scheme set:

s44, matching the second pre-proposal scheme set with the land mutual exclusion rule and the land mutual adaptation rule to obtain a final proposal scheme set;

the specific process of the step S5 is as follows:

s51, scoring the influence factors of each scheme in the proposal scheme set according to a traffic accessibility analysis mode, a distance analysis mode, a land mutual exclusion rule and a land mutual adaptation rule;

s52, determining the weight of each influence factor by adopting an entropy method;

and S53, calculating the comprehensive scores of all the schemes in the proposal scheme set according to the scores and the weights, namely calculating the project values, and sorting according to the sizes of the project values.

Images