KR20220076066A - Method and system for commercialization of fire insurance non-face-to-face reverse auction O2O service using AI building fire risk index model - Google Patents

Abstract

The present invention predicts the building risk based on an AI model learned by fusion of building unit attribute information and fire information when a customer inputs insurance target building information, calculates cheap and reasonable fire insurance premiums for each insurance company, and presents them to customers. It relates to a fire insurance premium non-face-to-face reverse auction service system and method utilizing AI building fire risk model, which provides a contract brokerage service.
The system according to the present invention is a user terminal that enters insurance target building information, selects one of building fire insurance according to a number of non-life insurance companies according to a fire damage insurance brokerage service customized for each building, and concludes a fire insurance contract non-face-to-face ; and collects insurance target building-related data according to the insurance target building information input from the user terminal, sets a data set by purifying and fusing the collected data, and predicts fire and fire risk of the insurance target building based on the data set An insurance brokerage server that calculates and presents fire insurance premiums for each insurance company after learning deep learning by calculating may include

Description

{Method and system for commercialization of fire insurance non-face-to-face reverse auction O2O service using AI building fire risk index model}

The present invention relates to a fire insurance non-face-to-face reverse auction service system and method utilizing an artificial intelligence (AI) building fire risk model. Geo AI technology is applied to predict fire risk for each building, and when the user enters information on the building to be insured, a service that brokers insurance contracts is provided by calculating cheap and reasonable fire insurance premiums for each non-life insurance company and presenting them to users It relates to a fire insurance non-face-to-face reverse auction service system and method using AI building fire risk model.

In general, fire is one of the unavoidable disasters in modern life, and fire insurance is insurance to prepare for such disasters.

Among these fire insurance, in particular, state-owned buildings, educational facilities, department stores, markets, medical facilities, entertainment venues, lodging establishments, factories, apartment complexes, or other special buildings with a building area of 3,000 square meters or more as prescribed by Presidential Decree where many people work or reside. There is fire insurance for buildings.

Because special buildings are used and visited by a large number of people, insurance is compulsory within 30 days of inspection after construction is completed, and owners of special buildings are liable for damages under the Civil Code for fire disasters in their buildings.

In the past, insurance subscription and management service methods using the Internet made it possible to secure customers, purchase and manage Internet insurance online and offline, but for fire disasters in special buildings, it is difficult to purchase insurance because the configuration is insufficient. The procedure for special building fire insurance can be complicated.

In addition, another conventional method is a real estate non-life insurance contract brokerage method using the Internet, through which the user accesses the real estate non-life insurance brokerage server through the Internet network with a computer to check the user's credit status and the right of the real estate to be insured. Although it is easy to sign a real estate non-life insurance contract, the fire accident compensation insurance for special buildings using the Internet is excluded because the fire department, which checks whether the building complies with the fire law, and the damage assessment agency, which is in charge of the damage assessment of the special building after a fire occurs, are excluded. There was a problem in that it could not be applied.

As a prior art document, there is Korean Patent Application Laid-Open No. 10-2001-0110888 (published on December 15, 2001), which describes a method for insurance subscription and management service using the Internet.

It is an object of the present invention to solve the above-mentioned problems, based on the AI model learned and convergence of all building unit property information and fire information in the country, by applying Geo AI technology outside the weight-based statistics-based fire risk Using an AI building fire risk model that predicts the risk of building with insurance and provides a contract brokerage service by calculating a cheap and reasonable fire insurance premium when the user presents information on the building subject to insurance, and presenting the premium calculated by multiple non-life insurers To provide a fire insurance non-face-to-face reverse auction service system and method.

The fire insurance non-face-to-face reverse auction service system utilizing the AI building fire risk model according to an embodiment of the present invention for achieving the above object, by inputting insurance target building information, a user terminal that selects one of the building fire insurance according to the insurance company and concludes a fire insurance contract in a non-face-to-face manner; According to the insurance target building information input from the user terminal, data related to the building subject to insurance is collected, the collected data is refined and fused to set a data set, and based on the data set, the fire prediction and fire risk of the building subject to insurance are calculated. An insurance brokerage server that calculates and presents fire insurance premiums for each insurance company after calculating and learning deep learning, and concludes a building fire insurance contract non-face-to-face according to the selection of the user terminal, thereby providing a customized fire non-life insurance brokerage service for each building; and an insurance database for storing member information of the user terminal, information on the building fire insurance, and insurance contract information according to the building fire insurance contract.

In addition, a credit rating agency server that provides personal credit information from the insurance brokerage server to the insurance brokerage server in response to a personal credit inquiry request of the user terminal; At the request of the insurance brokerage server, the court registration server for providing registration data and registration information related to the insurance target building according to the insurance target building information; In response to the request of the insurance brokerage server, according to the insurance target building information according to the insurance target building-related probate contract information, the official price of the building standard site, and a server for providing tax information; In response to the request of the insurance brokerage server, a certified authentication authority server to verify the identity of the user of the user terminal and to perform a legal digital signature; In response to the request of the insurance brokerage server, a fire service server that provides information on the fire of the building subject to insurance, or whether or not the firefighting law is complied with; a national spatial information portal server that provides GIS building integrated information, individual public land price, and building age information to the insurance brokerage server; and a road name address guide server for providing road name addresses and building information of the insurance target building.

In addition, the insurance brokerage server obtains GIS building integrated information, individual public land price, and building age information from the national spatial information portal server, or obtains a road name address and building information from the road name address guide server, or a building energy DB and a data collection unit that acquires building energy information and fire information from the fire DB, and purifies and fuses the acquired data to generate a fusion data set; a data pre-processing unit for deriving a risk grade for the data collected through the data collection unit, and processing public fire safety data including fire data from the fire DB and processing it as building unit and open data fusion data; A model that generates a fire prediction model for each building by learning DNN-based fire accident information, creates a building fire risk model to which a building fire risk class is applied through clustering technique, and learns the generated fire prediction model and building fire risk model study department; and when receiving insurance target building information from the user terminal, calculate the fire insurance discount rate and fire risk class based on the fire prediction model and the building fire risk model, and based on this, fire insurance premiums according to multiple non-life insurance companies and a service platform providing unit that calculates and presents to the user terminal, and provides a non-face-to-face reverse auction service for building fire insurance premiums customized for each building by signing a building fire insurance contract non-face-to-face according to the selection of the user terminal.

In addition, the insurance database may be configured as Post GIS capable of storing spatial information and data types.

In addition, the service platform providing unit provides a Web GIS-based web service based on an open source-based Web GIS service architecture, and implements the spatial information as a web serviceable Geo server, and an Open Layers client Visualization and functions are provided using the library, and the business process is divided into Function Area, Function, Process, and Unit Process according to the business model, and fire risk rating information is provided. A business function decomposition menu can be provided by deriving a unit process for the system for

The task function division menu includes member management for member registration, modification, and withdrawal, group management for group creation, editing, and deletion, manager member management for member approval and rejection, and adding and deleting members to and from the group Permission management menu including administrator group management; Fire risk level search and search menu for map-based building level search, area search, address (building) search, and fire risk level search; Fire safety inspection priority inquiry menu for regional inspection priority map-based inquiry, inspection priority list inquiry by region; and a fire safety inspection management menu related to fire safety inspection data registration, correction, deletion, and inquiry.

In addition, the model learning unit performs under-sampling with respect to the fire prediction model at a ratio of 1:4 based on a fire accident in the convergence data set, and then converts it into a training data form of Tensorflow. A sequential model is generated by performing normalization and one-hot encoding, and the sequential model is trained to generate an electric fire prediction model.

In addition, the model learning unit includes 'building specific number', 'usage_amount (KWh)_Mean', 'location code', 'individually announced land price', 'site area', 'building use name', 'building structure name', ' Building area', 'Height_consistency', 'Building-to-land ratio_consistency', 'Floor floor area_consistency', 'Number of floors above ground', 'Number of basement floors', 'Regional division', 'Total floor area', 'Building age', 'Type of business' In order to learn by setting 17 explanatory variables and the presence or absence of fire as dependent variables of can decide

In addition, the model learning unit may derive the fire risk index for each building through K-means clustering into five stages of safety, interest, caution, alertness, and risk for grading the prediction result.

On the other hand, the non-face-to-face reverse auction service method for fire insurance using AI building fire risk model according to an embodiment of the present invention for achieving the above object is an insurance brokerage server that collects data according to the request of a user terminal and concludes building fire insurance A fire insurance non-face-to-face reverse auction service method using an artificial intelligence building fire risk model, comprising the steps of: (a) transmitting, by a user terminal, executing an insurance application and receiving insurance target building information to the insurance brokerage server; (b) collecting, by the insurance brokerage server, data related to an insurance target building according to the insurance target building information; (c) generating, by the insurance brokerage server, a fusion data set by purifying and fusion of the collected data; (d) deep learning by the insurance brokerage server by calculating the fire prediction and fire risk of the building to be insured based on the convergence data set; (e) calculating, by the insurance brokerage server, a fire insurance premium for each insurance company according to the result of the deep learning learning and providing it to the user terminal; (f) selecting, by the user terminal, an insurance company and a fire insurance premium; and (g) the insurance brokerage server concluding a non-face-to-face building fire insurance contract with the user terminal.

In addition, in step (b), the insurance brokerage server collects spatial information-based public data for a building that is an object of prediction of a fire, and the spatial information-based public data includes building attribute information, building spatial information, and road name address information. , energy usage information and fire accident information, the building attribute information includes GIS building integrated information, individual official land price and building age, and the building spatial information may include a building name, a dong name, and a unique building number. .

In addition, in step (c), the insurance brokerage server refines and fuses the collected data according to the building unit standard including the building specific number to create a fusion data set, and the building unit standard is a building identification number and It may further include a unique lot number.

In addition, the step (c) may include: (c-1) generating, by the insurance brokerage server, the first fusion data by merging the building spatial information and the road name address information according to the building unique number; (c-2) according to the building identification number, the process of fusion of the first fusion data and the GIS building integrated information to generate second fusion data; (c-3) generating third fusion data by fusing the second fusion data and information on the age of the building according to the building identification number; (c-4) generating fourth fusion data by fusing the third fusion data and information on the individual official land price according to the lot unique number; (c-5) generating a fifth fusion data by fusing the fourth fusion data and the energy usage information according to the building identification number; and (c-6) generating the final fusion data by fusing the fifth fusion data and the fire accident information according to the building identification number.

In addition, in step (d), the insurance brokerage server designates a part randomly selected from the convergence data set as a first data group, and designates the remaining unselected data as a second data group, and the first data The group is trained using a random forest algorithm, the mean absolute error (MAE) and the correlation coefficient are calculated, and the possibility of fire is predicted using the second data group, and , using the fire possibility to classify into safety, concern, caution, alert, and risk, or by learning whether fire has occurred for the properties of a building based on the convergence data set, create a deep learning learning model, and create It is possible to estimate the fire risk of a building using the learned model.

And, in step (d), the insurance brokerage server performs undersampling, normalization, and one-hot encoding on the fusion data set to generate learning data, and between explanatory variables focusing on the fire presence/absence property included in the learning data. The number of attributes can be reduced by deriving the Pearson correlation coefficient.

The fire insurance non-face-to-face reverse auction service system and method using AI building fire risk model according to the present invention provides building risk rating information and attribute information according to the AI algorithm when a service user presents insurance target building information, Participating insurers can provide a non-face-to-face service that brokers contracts by presenting premiums.

In addition, the fire insurance non-face-to-face reverse auction service system and method using the AI building fire risk model according to the present invention provides data provision and verification for the non-life insurance company for the calculation of the non-life insurance rate, and provides the user with the fire insurance target It is possible to provide a service that connects the optimal insurance company by understanding the level of risk to the building.

In addition, the fire insurance non-face-to-face reverse auction service system and method using the AI building fire risk model according to the present invention intensively strengthens the fire risk for buildings such as single-family houses, villas, and multi-households subscribed for financial loans to users, and insurance It offers a lower insurance premium than a planner and can provide convenient services through an intuitive map-based service on a portable mobile terminal or computer used.

In addition, the fire insurance non-face-to-face reverse auction service system and method using the AI building fire risk model according to the present invention can intensively strengthen the publicity of safety-grade goods without insurance to the insurance company, and the safety grade is inexpensive and the risk grade It is possible to increase the rate of return by weighting the weight, and it is possible to commercialize and provide matching analysis data separately, and it has the effect of strengthening public relations through the association and non-face-to-face marketing by inviting the person in charge online.

In addition, the fire insurance non-face-to-face reverse auction service system and method using the AI building fire risk model according to the present invention is, By providing a safety level 5 map along with fire safety monitoring results and prediction information, it can be used in an early response system (safety map-disaster safety research institute system linkage) before an emergency such as a fire occurs, and is the basis for determining the fire safety prevention policy for industrial complexes It can be used as data.

In addition, the fire insurance non-face-to-face reverse auction service system and method using the AI building fire risk model according to the present invention is for private companies such as the public, non-life insurance companies, non-life insurance associations, universities and research institutes, together with fire safety monitoring results and prediction information By providing a rate calculation map, fire risk index, and risk data fused with fire insurance, it contributes to the activation of safety research (joint research support) through the secondary utilization of monitoring results and prediction information, and promotes the use of big data in regional industrial complexes, It is possible to develop a fire non-life insurance product using data, and it has the effect of promoting economic benefits to the public through differential application of non-life insurance.

In addition, the fire insurance non-face-to-face reverse auction service system and method using the AI building fire risk model according to the present invention supports citizens to set cheaper premiums to minimize damage from fire accidents, and at the same time provides insurance companies with insurance policyholders All stakeholders, including profit expansion, can expect profits from the expansion.

In addition, the fire insurance non-face-to-face reverse auction service system and method using the AI building fire risk model according to the present invention is an insurance product tailored to the risk level of a building using AI, taking into account the trend of increasing property damage and loss of life compared to fire accidents By designing this, it is possible to reduce insurance premiums and improve the profit structure of insurance companies, and to spread fire insurance in case of unexpected accidents.

1 is a configuration diagram schematically showing the overall configuration of a fire insurance non-face-to-face reverse auction service system utilizing an AI building fire risk model according to an embodiment of the present invention.
2 is a block diagram schematically showing an internal configuration of an insurance brokerage server according to an embodiment of the present invention.
3 and 4 are diagrams illustrating a flow chart for explaining a fire insurance non-face-to-face reverse auction service method using an artificial intelligence building fire risk model according to an embodiment of the present invention.
5 is a flowchart illustrating a process of collecting, refining, and merging data based on insurance target building information in an insurance brokerage server according to an embodiment of the present invention to generate a convergence data set.
6 is a diagram illustrating a process of generating a convergence data set by collecting data in an insurance brokerage server according to an embodiment of the present invention.
7 is a diagram illustrating an example of providing a business function division menu in a service platform of an insurance brokerage server according to an embodiment of the present invention.
8 is a view showing the importance of predictive variables according to the result of predicting electric fire through the fire insurance non-face-to-face reverse auction service system utilizing the AI building fire risk model according to an embodiment of the present invention.
9 is a view showing the accuracy of predicting a building unit accident through a prediction model according to an embodiment of the present invention.
10 is a view showing an example of expanding the service for exclusive use of industrial complexes through the fire insurance non-face-to-face reverse auction service system utilizing the AI building fire risk model according to an embodiment of the present invention.
11 is an exemplary diagram illustrating fusion data generated by geo-coding in an insurance brokerage server according to an embodiment of the present invention.

Advantages and features of the present invention, and a method for achieving them will become apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but will be implemented in a variety of different forms, and only these embodiments allow the disclosure of the present invention to be complete, and those of ordinary skill in the art to which the present invention pertains. It is provided to fully inform the person of the scope of the invention, and the present invention is only defined by the scope of the claims. Accordingly, in some embodiments, well-known process steps, well-known device structures, and well-known techniques have not been specifically described in order to avoid obscuring the present invention. Like reference numerals refer to like elements throughout.

In order to clearly express various layers and regions in the drawings, the thicknesses are enlarged. Throughout the specification, like reference numerals are assigned to similar parts. When a part, such as a layer, film, region, plate, etc., is “on” another part, it includes not only cases where it is “directly on” another part, but also cases where there is another part in between. Conversely, when we say that a part is "just above" another part, we mean that there is no other part in the middle. Also, when a part of a layer, film, region, plate, etc. is said to be "under" another part, it includes not only the case where it is "directly under" another part, but also the case where there is another part in the middle. Conversely, when a part is said to be "just below" another part, it means that there is no other part in the middle.

Spatially relative terms "below", "beneath", "lower", "above", "upper", etc. It can be used to easily describe the correlation between an element or components and other elements or components. The spatially relative terms should be understood as terms including different orientations of the device during use or operation in addition to the orientation shown in the drawings. For example, when an element shown in the figures is turned over, an element described as "beneath" or "beneath" another element may be placed "above" the other element. Accordingly, the exemplary term “below” may include both directions below and above. The device may also be oriented in other orientations, and thus spatially relative terms may be interpreted according to orientation.

In the present specification, when a part is said to be connected to another part, this includes not only a case in which it is directly connected, but also a case in which it is electrically connected with another element interposed therebetween. In addition, when it is said that a part includes a certain component, this means that other components may be further included, rather than excluding other components, unless otherwise stated.

In this specification, terms such as first, second, third, etc. may be used to describe various components, but these components are not limited by the terms. The above terms are used for the purpose of distinguishing one component from other components. For example, without departing from the scope of the present invention, a first component may be referred to as a second or third component, and similarly, the second or third component may also be alternately named.

Unless otherwise defined, all terms (including technical and scientific terms) used herein may be used with the meaning commonly understood by those of ordinary skill in the art to which the present invention belongs. In addition, terms defined in a commonly used dictionary are not to be interpreted ideally or excessively unless clearly defined in particular.

Hereinafter, a fire insurance non-face-to-face reverse auction service system and method using an artificial intelligence (AI) building fire risk model according to the present invention will be described with reference to the drawings.

1 is a configuration diagram schematically showing the overall configuration of a fire insurance non-face-to-face reverse auction service system utilizing an AI building fire risk model according to an embodiment of the present invention.

Referring to FIG. 1 , the fire insurance non-face-to-face reverse auction service system 100 using the AI building fire risk model according to an embodiment of the present invention includes an insurance brokerage server 110 , an insurance database (DB; 120 ), and a user terminal 130 . ), a national spatial information portal server 140 and a road name address guide server 150 and the like.

In addition, it may further include a fire department server, a court registration department server, a city, gun-gu server, an accredited certification authority server, a credit rating agency server, and the like.

The insurance brokerage server 110 collects insurance target building-related data according to the insurance target building information input from the user terminal 130, refines and converges the collected data to set a data set, and provides insurance based on the data set After deep learning by calculating the fire prediction and fire risk of the target building, the fire insurance premium for each insurance company is calculated and presented. We can provide brokerage services.

The insurance DB 120 may store member information of the user terminal 130 , information on building fire insurance, and insurance contract information according to a building fire insurance contract. Here, the insurance database 120 may be configured as a Post GIS that can store spatial information and data types.

The user terminal 130 enters insurance target building information, selects one of the building fire insurance according to a number of non-life insurance companies according to the fire damage insurance brokerage service customized for each building, and concludes a building fire insurance contract non-face-to-face. .

The national spatial information portal server 140 may provide GIS building integrated information, individual public land price, and building age information to the insurance brokerage server 110 .

The road name address guide server 150 may provide the road name address and building information of the insurance target building.

The fire department server, according to the request of the insurance brokerage server 110 , may provide information on the fire of a building subject to insurance, or provide whether or not the firefighting law is complied with.

The court registration department server may provide, at the request of the insurance brokerage server 110 , registration data and registration information related to the insurance target building according to the insurance target building information.

The city, county, and gu server may provide, at the request of the insurance brokerage server 110 , information on a contract for an insurance target related to a building subject to insurance according to the information on the building subject to insurance, the official price of a building standard site, and tax information.

The credit rating agency server may provide personal credit information from the insurance brokerage server 110 to the insurance brokerage server 110 in response to a personal credit inquiry request of the user terminal 130 .

The accredited certification authority server may verify the user identity of the user terminal 130 and execute an electronic signature having legal effect according to the request of the insurance brokerage server 110 .

The insurance brokerage server 110 according to the present invention may collect data from databases of the Korea Electric Safety Corporation, the Ministry of Land, Infrastructure and Transport, the Ministry of Public Administration and Security, the National Statistical Office, and the Korea Meteorological Administration, in addition to the aforementioned organizations.

In the database of Korea Electric Safety Corporation, you can collect the customer master table and the inspection result date table, and in the database of the Ministry of Land, Infrastructure and Transport, information on the building permit book, outline information for each building, and information on the progress of the building permit book can be collected. have.

In addition, it is possible to collect necessary data from each institution, such as electric fire status information from the database of the Ministry of Public Administration and Security, administrative dong boundary information and demographic information from the database of the National Statistical Office, and weather information from the database of the Korea Meteorological Administration.

In particular, information on electrical safety inspection results obtained through the database of Korea Electric Safety Corporation, master information of end-of-day customers, and information on the current status of electrical fires identified annually by the Ministry of Public Administration and Security may be used as basic data.

On the other hand, general attribute information in the collected data includes customer information, electrical safety check information, building information, total building GIS, and 119 fire accident data information (119). fire accident data), climate information, administrative district information, and demographics information.

The customer information may include data such as a supply date, use or not (use_YN), and the electrical safety check information is a customer number (customer no.), diagnosis result (check). result), safety grade, and the like.

Building information may include usage, structure, completion date, coverage ratio, floor area ratio, etc., and total building information GIS) may include information such as a building position, a zip code, and an area.

The 119 fire accident data may include information such as a fire type, a cause, a location, an area, and the like, and the climate information may include a precipitation amount. It may include (precipitation), humidity (humidity), temperature (temperature), wind speed (wind speed) and the like.

Administrative district information may include administrative borders information, and demographic information may include information on age and gender region.

2 is a block diagram schematically showing an internal configuration of an insurance brokerage server according to an embodiment of the present invention.

Referring to FIG. 2 , the insurance brokerage server 110 according to an embodiment of the present invention includes a data collection unit 210 , a data preprocessing unit 220 , a model learning unit 230 , a communication unit 240 , and a control unit 250 . ), a service platform providing unit 260 , a storage unit 270 , and a display unit 280 .

The data collection unit 210 obtains GIS building integrated information, individual public land price, and building age information from the national spatial information portal server 140, or obtains a road name address and building information from the road name address guide server 150, or It is possible to obtain building energy information and fire information from a building energy DB and a fire DB operated by institutions, etc., and to refine and converge the acquired data to generate a convergence data set (Data Set).

The data pre-processing unit 220 derives a risk grade for the data collected through the data collection unit 210, obtains fire safety public data including fire data from the fire DB, and processes it as fusion data of building units and open data. can

The model learning unit 230 generates a fire prediction model for each building by learning the DNN-based fire accident information, and generates a building fire risk model to which a building fire risk grade is applied through a clustering technique, and the generated fire prediction model and building A fire hazard model can be trained.

The communication unit 240 includes the user terminal 130 through the communication network, as well as the national spatial information portal server 140, the road name address guide server 150, the fire department server, the court registration department server, the city/gun district server, the accredited certification agency server and credit evaluation. Data can be transmitted/received by communicating with an institution server or the like.

The control unit 250 controls each functional unit, but controls each functional unit to provide a non-face-to-face reverse auction service for fire insurance using the artificial intelligence building fire risk model according to the present invention.

When receiving insurance target building information from the user terminal 130, the service platform providing unit 260 calculates a fire insurance discount rate and fire risk grade based on a fire prediction model and a building fire risk model, and based on this, a plurality of Fire insurance premiums according to non-life insurance companies are calculated and presented to the user terminal 130, and a non-face-to-face non-face-to-face reverse auction service for fire insurance premiums tailored to each building can be provided by signing a building fire insurance contract non-face-to-face according to the selection of the user terminal 130. have.

In addition, the service platform providing unit 260 provides a Web GIS-based web service based on an open source-based Web GIS service architecture, and implements the spatial information as a web serviceable Geo server, and open layers (Open). Layers) provides visualization and functions by using the client library, and divides the business process into functional areas, functions, processes, and unit processes according to the business model, It is possible to derive a unit process for the system for providing rating information and provide a business function decomposition function.

At this time, the task function division menu includes a member management menu for member registration, modification, and withdrawal, a group management menu for group creation, editing, and deletion, an administrator member management menu for member approval and rejection, and a member management menu for group members. Permission management menu including administrator group management to add and delete; Fire risk level search and search menu for map-based building level search, area search, address (building) search, and fire risk level search; Fire safety inspection priority inquiry menu for regional inspection priority map-based inquiry, inspection priority list inquiry by region; and a fire safety inspection management menu related to fire safety inspection data registration, correction, deletion, and inquiry.

In addition, the model learning unit 230 performs under-sampling with respect to the fire prediction model at a ratio of 1:4 based on the fire accident in the fusion data set, and then converts it into the training data form of Tensorflow. For this purpose, a sequential model is generated by performing normalization and one-hot encoding, and by training this sequential model, it can be generated as an electric fire prediction model.

In addition, the model learning unit 230 includes 'building specific number', 'usage_amount (KWh)_Mean', 'location code', 'individually announced land price', 'site area', 'building use name', 'building structure name' ', 'Building area', 'Height_consistency', 'Building-to-land ratio_consistency', 'Floor floor space_consistency', 'Number of floors above ground', 'Number of basement floors', 'Region division', 'Total floor area', 'Building age', It can be learned by setting 17 explanatory variables of 'industry' and the presence or absence of fire as dependent variables.

In addition, the model learning unit 230 may select the number of clusters by using an elbow method to compare a certain number of clusters between clusterings in order to classify the fire risk in order to expand utilization.

For example, the model learning unit 230 may provide a graph form as shown in FIG. 3 as a result of applying the Elbow Method to the prediction result, and determine the optimal number of clusters to be 5.

In addition, the model learning unit 230 may derive the fire risk index for each building through K-means clustering in five stages of safety, interest, caution, alertness, and risk for grading the prediction result.

3 and 4 are diagrams illustrating a flow chart for explaining a fire insurance non-face-to-face reverse auction service method using an artificial intelligence building fire risk model according to an embodiment of the present invention.

3 and 4, in the fire insurance non-face-to-face reverse auction service system 100 using artificial intelligence building fire risk model according to an embodiment of the present invention, first, the user terminal 130 executes the building fire insurance application, It receives insurance target building information from the user and transmits it to the insurance brokerage server 110 (S310).

Here, the user terminal 130 executes the building fire insurance application according to the user's operation in a state in which the building fire insurance application is installed inside the terminal, and transmits and receives insurance-related information and data with the insurance brokerage server 110. can

Next, the insurance brokerage server 110 collects the insurance target building related data from the related servers according to the insurance target building information received from the user terminal 130 (S320).

In this case, the insurance brokerage server 110 may collect public data based on spatial information for a building that is a fire prediction target.

Here, the spatial information-based public data may include building attribute information, building space information, road name address information, energy consumption information, and fire accident information.

In addition, as shown in Table 1, the building attribute information includes GIS building integration information, individual official land price, and building age information, and the building space information may include a building name, a dong name, and a building unique number.

The GIS building integrated information is information provided by the national spatial information (national key data) portal server 140, and may include a GIS building integrated identification number, a unique number, a legal dong code, and the like.

In addition, the individual official land price information is also information provided by the national spatial information (national key data) portal server 140, and may include the individual official land price, whether a standard land price, land area, and the like.

In addition, building age information is information provided by the national spatial information (national key data) portal server 140, and may include a building height, a building age, an age group name, and a total floor area of a building.

In addition, the GIS building integrated information master is information provided by the national spatial information (open market) portal server 140, and may include a building name, a dong name, a UFID, and the like.

In addition, the road name address building information is information provided by the road name address guide server 150, and may include a building management number, a city name, a city, county name, and the like.

In addition, the building energy information is information provided by Seumteo, etc., and may include usage amount, new address_serial number, street address, and the like.

In addition, general fire information may include a lot address, a road name address, a building management number, and the like.

Then, the insurance brokerage server 110 generates a fusion data set by purifying and fusion of the collected data (S330).

That is, the insurance brokerage server 110 performs pre-processing, such as removing unnecessary attributes from the collected data, and as shown in FIG. 6 , purifies and converges data in units of buildings based on the unique building number to finally To create a fusion data set. 6 is a diagram illustrating a process of generating a convergence data set by collecting data in an insurance brokerage server according to an embodiment of the present invention. The building unit standard may include a building identification number and a lot unique number. In FIG. 6 , the tables constituting the fusion data set may include a building information table, a national key data table, a seumteo table, a fire information table, an open market table, and a road name address table. The building information table stores GIS building information, GIS building information master, road name address building information, and building age information. The joint key of the GIS building information master and the road name address building information is the unique building number (BD_MGT_SN), and the GIS building information, the GIS integrated building information master, and the road name address building information joint key are GIS. It is an integrated building identification number (UFID), and the joint key of GIS building information, GIS building information master, road name, address, building information, and building age information is a GIS building identification number (UFID). In addition, the joint key of GIS building identification number (UFID) and individual public price is land code (PNU), and the joint key of land code and building energy is unique building number (BD_MGT_SN), The joint key of building unique number and fire accident information is building unique number (BD_MGT_SN).

Next, the insurance brokerage server 110 analyzes data based on the convergence data set, and uses exploratory analysis, correlation analysis, and statistical analysis to calculate the fire prediction and fire risk of the building to be insured, and deep learning it ( S340).

At this time, the insurance brokerage server 110 may calculate regional characteristics, usage amount, building use, floor area ratio, building-to-land ratio and building age, etc. by executing exploratory analysis on the convergence data set (S332).

In addition, the insurance brokerage server 110 may generate a predictive model by performing correlation analysis and statistical analysis on the fusion data set (S334).

In addition, the insurance brokerage server 110, by designating a part selected in a random manner from the fusion data set as the first data group, and designating the rest not selected as the second data group, the first data group in a random forest (random forest) forest) algorithm, calculates the mean absolute error (MAE) and correlation coefficient, predicts the possibility of fire using the second data group, and uses the It can be classified into safety, concern, caution, vigilance, and danger.

In addition, the insurance brokerage server 110 generates a deep learning learning model by learning whether or not a fire occurs for the properties of a building based on the convergence data set, and calculates the fire risk of the building using the generated learning model. have.

The fire risk (R) can be calculated using information on potential hazards (P), active risks (A), basic measures (N), special measures (S) and fire resistance measures (F) as shown in Equation 1 below. can

The fire risk index is a model in the field of fire safety based on professional evaluation and past experience. The risk index is a relatively simple calculation system, and it is possible to derive a hazard score assignment and a relative risk score through various analysis processes.

The fire risk index method is used as a quantitative fire risk assessment in the insurance field and is also called a standard table rate. It can be calculated using a combination of functions.

The Fire Risk Index (FRI) can compare relative risks rather than absolute risks, so it is possible to calculate the comparative risk of various populations, and has the advantage of being particularly cost-effective. The National Fire Code (NFC) of the United States specified the evaluation method related to the fire risk level in 'NFPA 551'. , quantitative methods, and cost/benefit risk methods are classified into five categories.

The FREM fire risk level assessment developed including the concept of the Gretener Method, a fire risk index methodology, is composed of empirically derived figures, fire initiation and spread factors, and fire prevention factors. It is a method of creating an index by first classifying the use of a building and checking the evaluation factors that follow. . The 38 evaluation factors consist of 13 potential risks, 5 active risks, 7 factors such as presence or absence of basic countermeasures, 7 special countermeasure factors such as alarm and autonomous fire brigade, and 6 fire-resistance countermeasures.

In addition, with respect to the fire risk, an evaluation grade can be derived by applying a weight to each evaluation item of the fire risk index as shown in Table 2 below.

That is, a weight is applied to all 17 evaluation items, and the weight value is multiplied by the evaluation result grade to derive the evaluation grade. If we look at the weights, the sum of the weights is 1.0, the interior materials are given a weight of 0.0576 in 5 grades, and fire suppression equipment such as fire extinguishers and sprinklers are also graded 5, but the weight is calculated as 0.0668. The fire department's ability, arrival time and accessibility, and maintenance weighted 0.1681 (0.31 * ability + 0.47 * arrival + 0.22 * accessibility and equipment), 5 grades for fire protection with a weight of 0.0666, and a weight of 0.1675 ( Weights such as 0.35*heat insulation +0.28*fireproofing +0.24*penetrating part +0.13*combustibility), 0.0687 for doors, and 0.0473 for burning risk through windows can be assigned.

In addition, the insurance brokerage server 110 generates learning data by performing under-sampling, normalization, and one-hot encoding on the fusion data set, and the fire included in the learning data. The number of attributes can be reduced by deriving the Pearson correlation coefficient between explanatory variables based on the presence or absence attribute.

That is, in the insurance brokerage server 110, the model learning unit 230 performs under-sampling (S341) with respect to the fire prediction model at a ratio of 1:4 based on the fire accident in the fusion data set (S341) and then TensorFlow Normalization (S342) and One-Hot encoding (S343) are performed to convert to the training data form of (Tensorflow) to generate a sequential model (S344), and this sequential model is Through training (S345), verification (10 Cross validation; S346), fire risk is derived (S347), and graded (k-means clustering) in 5 steps (S348) to generate an electric fire prediction model can do.

In addition, the model learning unit 230 may select the number of clusters by using an elbow method to compare a certain number of clusters between clusterings in order to classify the fire risk in order to expand utilization. For example, the model learning unit 230 may provide the result of applying the Elbow Method to the prediction result, and determine the optimal number of clusters to be 5. That is, the model learning unit 230 may derive the fire risk index for each building through K-means clustering in five stages of safety, interest, caution, alertness, and risk for grading the prediction result.

The model learning unit 230 of the insurance brokerage server 110 is an explanatory variable 'building specific number', 'usage_amount (KWh)_Mean', 'place code', 'individual announced land price', 'land area', 'building Name of use', 'Building structure name', 'Building area', 'Height_consistency', 'Building-to-coverage_consistency', 'Floor area ratio_consistency', 'No. 17 explanatory variables of 'total floor area', 'building age' and 'industry' and the presence or absence of fire can be set as dependent variables for learning.

In the insurance brokerage server 110 , the service platform providing unit 260 may divide the business process into a function area, a function, a process, and a unit process according to a business model. Here, the unit process is a process that defines, for example, input and output processing, etc. as the detailed lowest unit constituting the business process.

In addition, the service platform providing unit 260 in the insurance brokerage server 110 derives a unit process for the system for providing fire risk rating information and provides a Business Function Decomposition menu as shown in FIG. 7 . can do. 7 is a diagram illustrating an example of providing a business function division menu in a service platform of an insurance brokerage server according to an embodiment of the present invention. In FIG. 7, the business function division menu of the service platform largely includes an authority right menu, a fire risk level inquiry/search menu, a fire safety inspection priority inquiry menu, and a fire safety inspection management menu. The permission management menu includes the member management menu for member registration, modification, and withdrawal, the group management menu for group creation, editing, and deletion, the administrator member management menu for member approval and rejection, and the member management menu for adding and deleting members to and from the group. Includes group management of administrators. In addition, the fire risk level search and search menu provides menus such as map-based building level search, area search, address (building) search, and fire risk level search. In addition, the fire safety inspection priority inquiry menu includes menus such as a regional inspection priority map-based inquiry and a regional inspection priority list inquiry. In addition, the fire safety inspection management menu includes menus such as fire safety inspection data registration, correction, deletion, inquiry, and the like.

Next, the insurance brokerage server 110 calculates the fire insurance premium for each insurance company according to the result of deep learning learning and provides it to the user terminal (S350).

Accordingly, the user terminal 130 selects an insurance company and a fire insurance premium according to the user's selection operation (S360).

Accordingly, the insurance brokerage server 110 concludes a non-face-to-face building fire insurance contract with the user terminal 130 online (S370).

5 is a flowchart illustrating a process of collecting, refining, and converging data based on insurance target building information in an insurance brokerage server according to an embodiment of the present invention to generate a convergence data set.

Referring to FIG. 5 , the insurance brokerage server 110 according to an embodiment of the present invention generates first fusion data by fusion of building space information and road name address information according to a unique building number (S510).

Next, the insurance brokerage server 110 generates second fusion data by fusion of the first fusion data and the GIS building integration information according to the building identification number (S520).

Next, the insurance brokerage server 110 generates third fusion data by fusion of the second fusion data and the information on the age of the building according to the building identification number (S530).

Next, the insurance brokerage server 110 generates the fourth fusion data by fusion of the third fusion data and the information on the individual official land price according to the lot unique number (S540).

Next, the insurance brokerage server 110 generates the fifth fusion data by fusion of the fourth fusion data and the energy usage information according to the unique building number (S550).

Next, the insurance brokerage server 110 generates the final fusion data by fusing the fifth fusion data and the fire accident information according to the unique building number (S560).

8 is a view showing the importance of predictors according to the result of predicting electric fire through the non-face-to-face reverse auction service system for fire insurance using AI building fire risk model according to an embodiment of the present invention.

That is, FIG. 8 is an example of performing electric fire prediction for Daegu Metropolitan City by applying the fire insurance non-face-to-face reverse auction service system 100 using AI building fire risk model according to an embodiment of the present invention. Big data collection and collection Data set creation, algorithm execution, and verification were carried out in a fusion of the collected data.

In the big data collection stage, data managed by countries or public institutions was collected.

In the creation stage of the fusion data set, the key data, fire accident data (119 dispatch data) and electrical safety inspection data collected and managed by the Korea Electric Safety Corporation (KESCO) were secured, fused with address data, and meteorological data were added.

Next, the convergence data set was completed by merging the DB centered on the building utilizing public data. In the process of convergence data, address data was spatialized to transform it into coordinate data, and after DB pre-processing was performed, it was fused around the building unit (spatial information including building attribute information).

The algorithm stage is the stage of generating an electric fire prediction model, and risk prediction modeling was performed with the fire accident occurrence pattern for each building by using the electric fire accident presence data.

In the verification stage, the developed risk prediction model was evaluated and compared with the accuracy and actual accident data.

A learning model was created through the random forest algorithm by using 60% of the fusion data set as training data.

Finally, in order to verify the prediction model using electrical factors and environmental factors, the predicted result values were derived as shown in FIG. 9 through the verification data. 9 is a view showing the accuracy of predicting a building unit accident through a prediction model according to an embodiment of the present invention. If the predicted value was 0.5 or more, a fire occurred (1), and if it was less than 0.5, there was no fire (0), and a matrix chart was created with the actual accident. As shown in FIG. 9 , the accuracy of predicting a building unit accident as an accident was confirmed to be 74.7% through the prediction model according to the present invention.

10 is a view showing an example in which the service for exclusive use of industrial complexes is expanded through the fire insurance non-face-to-face reverse auction service system utilizing the AI building fire risk model according to an embodiment of the present invention.

Referring to FIG. 10 , the fire insurance non-face-to-face reverse auction service system 100 using AI building fire risk model according to an embodiment of the present invention includes factory data (production process, company information, etc. factory-on service of the Ministry of Commerce, Industry and Energy, etc. data) can be added to expand the service to industrial complexes.

That is, the fire prediction model dedicated to the industrial complex with the addition of the factory DB according to the present invention can generate a fusion data set by the following process. First, null check and unique check are executed on UFID and building identification number to remove nulls and duplicate instances. Next, a unique building number is created by joining the road name building information and the GIS building integration master. Next, UFID is created by combining the first convergence set and GIS building integration information. Next, the UFID is created by combining the secondary fusion set and building age information. Then, the PNU is created by combining the tertiary convergence set and the individual official price information. Next, a building unique number is generated by combining the fourth fusion set and electric energy information. Next, the main affiliation of the fire accident information is generated as a building-specific number derivative variable. Next, the final fusion set is created by combining the 5th fusion set and fire accident information to generate a unique building number.

11 is an exemplary diagram illustrating fusion data generated by geo-coding in an insurance brokerage server according to an embodiment of the present invention.

Referring to FIG. 11 , the data fused in units of buildings in the insurance brokerage server 110 according to an embodiment of the present invention includes building geographic information (Building GIS), electrical safety check information (Electrical safety check), and building completion. (Completion date), Fire accident causes, Building location, Building Usage, Fire accident location, Fire accident kind and Building area area), etc., can be seen in a fused state.

Therefore, more information can be fused and utilized as data to be analyzed.

In addition, general property information includes electrical fire accidents by building, age of buildings, total floor area, building-to-land ratio, use, structure, etc., electrical safety inspection information such as insulation resistance, ground resistance, leakage current, weather information such as temperature, humidity, and precipitation It may consist of properties such as buildings.

In addition, the insurance brokerage server 110 designates some of the convergence data as a first data group, and divides the rest into a second data group. In order to divide the data group, each of the learning process and the prediction process to be described later is performed. By dividing, the first data group is used for data analysis and the second data group is used for prediction. In this case, the partitioning method may be selected in a random manner in order to exclude data clustering and designated as the first data group, and the remaining unselected data may be designated as the second data group.

The insurance brokerage server 110 may learn the first data group using a random forest algorithm at the time of analysis, and calculate a mean absolute error (MAE) and a correlation coefficient. .

This is to find major factors related to electrical and environmental factors for the presence or absence of electrical fire accidents, to predict building unit fire accidents and to quantify the risk through a prediction model. To this end, it is to classify various factors using previously managed electrical safety inspection data, building registers, and fire accident data, and to find out the characteristics between regions where accidents occur and regions where there are not.

The insurance brokerage server 110 may perform learning through a random forest algorithm by using the first data group as learning data.

A random forest is a combination of a plurality of decision binary trees in an ensemble form, and trees are grown in a random way in each binary tree. Since the random forest is based on decision trees, it has a fast learning speed and the ability to process a large amount of data. That is, a random forest can create several decision trees and derive a final result through voting.

After configuring N sampling data sets (randomly sampling observations and features) through the bootstrap aggregating process, that is, bagging the data, each decision tree model is constructed. , it is possible to maintain low bias and reduce high variance by determining the prediction results in a voting method by individual prediction models.

1 data group is learned using a random forest algorithm, and for analysis according to it, mean absolute error (MAE) and correlation analysis can be calculated.

Mean absolute error (MAE) is the absolute value of the error, that is, a value obtained by adding up all absolute deviations and dividing it by the number of records, and is defined as in Equation 2 below.

는 실제 사고유무 값이고,

은 예측된 사고유무 값이며, N은 전체 레코드 수를 나타낸다.here,

is the actual value of whether there is an accident,

is the predicted accidental value, and N is the total number of records.

Correlation analysis is a statistical technique that analyzes how closely a linear relationship exists between two variables, and the strength of the relationship between two variables is called correlation.

Correlation (r) measures how similar two variables are. If the correlation is 0 < r ≤ +1, it is said to have a positive correlation, and if -1 ≤ r < 0, it is said to have a negative correlation, r = 0, it is said to be uncorrelated. However, uncorrelation does not mean that there is no correlation, but that there is no linear correlation.

When one variable increases, it is said to be a positive correlation in which the other variable increases, and when it decreases, it is said to be a negative correlation.

Correlation analysis can be performed to examine the similarity between an actual fire accident and a predicted fire accident by using the Pearson correlation coefficient used based on such a theory.

The Pearson correlation coefficient is defined as in Equation 3 below.

Here, r is the weight of similarity between the actual accident and the predicted accident, xi represents the actual accident or not, and yi represents the predicted accident or not.

In addition, the insurance brokerage server 110 may predict the possibility of a fire by using the second data group.

By using the second data group that was not previously used for modeling, the possibility of fire occurrence can be verified and predicted, and the risk level for each building can be derived and displayed on the map.

In order to verify the prediction model using the finally developed electrical and environmental factors, the predicted result values were derived using the second data group.

As described above, if the value of the probability of fire occurrence is 0.5 or more, fire occurs (1), and if it is 0.5 or less, fire does not occur (0).

Comparing the prediction results with the actual accident, according to the predictability of fire occurrence, there are cases in which the location where the accident has already occurred is indicated as dangerous. have.

Meanwhile, the user terminal 130 according to the present invention may include a display unit that displays the result of executing the building fire insurance application on a screen. In this case, since the display unit may be implemented as an organic light emitting display (OLED) device, the user terminal 130 may be referred to as an organic light emitting display device.

In the organic light emitting display device according to the present invention, a gate metal layer including a gate electrode and a signal wiring is formed on a substrate. Here, the substrate may be made of a flexible plastic material having a flexible characteristic so that the display performance can be maintained even when the organic light emitting display device is bent like paper.

In addition, the gate metal layer may be formed of any one of a first metal material having a low resistance characteristic, for example, aluminum (Al), an aluminum alloy (AlNd), copper (Cu), a copper alloy, molybdenum (Mo), and molybdenum (MoTi). It may be formed as a single-layer structure or a double-layer or triple-layer structure by being made of two or more first metal materials. This gate metal layer forms a lower electrode of the capacitor C1 and extends to form a gate electrode of the driving thin film transistor DRT.

In addition, a gate insulating layer and an etch stop layer made of an insulating material, for example, silicon oxide (SiO2) or silicon nitride (SiNx), which are inorganic insulating materials, are formed on the entire surface of the display area of the substrate including the gate metal layer.

Between the upper portion of the gate insulating layer and the etch stop layer, a semiconductor layer 121 made of any one selected from amorphous silicon, polysilicon, and semiconductor oxide is formed corresponding to each of the thin film transistors SWT, SST, and DRT. A portion of the semiconductor layer is exposed through a contact hole formed on the etch stop layer, and the source and drain electrodes of the driving thin film transistor (DRT) and the data signal (Vdata) are formed on the insulating layer including the contact hole and the etch stop layer. A first source and drain metal layer including an application wiring and a power supply voltage (ELVDD) application wiring is formed.

Here, the first source and drain metal layers include, for example, aluminum (Al), aluminum alloy (AlNd), copper (Cu), copper alloy, molybdenum (Mo), molybdenum (MoTi), chromium (Cr), and titanium (Ti). ) of any one or a combination of two or more substances.

In particular, a source electrode and a drain electrode made of a metal material are formed in the driving thin film transistor DRT, which are spaced apart from each other and contact the semiconductor layer exposed through the contact hole, respectively. Accordingly, the gate electrode, the gate insulating layer, the semiconductor layer, and the source and drain electrodes form one driving thin film transistor (DRT). In addition, in addition to the driving thin film transistor (DRT), the switching thin film transistor (SWT) and the sensing thin film transistor (SST) are also formed in the same stacked structure.

Here, the gate electrode and the drain electrode of the switching thin film transistor (SWT) are connected to the scan line and the data line, respectively, and the source electrode of the switching thin film transistor (SWT) is electrically connected to the gate electrode of the driving thin film transistor (DRT). and source electrodes of the sensing thin film transistor SST and the driving thin film transistor DRT are connected to each other.

In addition, the data signal applying wiring in the first source and drain metal layers forms the upper electrode of the capacitor C1.

Meanwhile, although it is exemplified that both the first source and drain metal layers have a single-layer structure, a double-layer or triple-layer structure may be formed by a combination of two metal materials.

In addition, a passivation layer is formed on the first source and drain metal layers to cover the driving thin film transistor (DRT) and to expose a portion of the first source and drain metal layers. In particular, a portion of the passivation layer is etched to expose the power supply voltage application wiring of the lower first source and drain metal layers, and is brought into contact with the upper second source and drain metal layers.

A second source and drain metal layer is formed on the passivation layer. The second source and drain metal layers may be formed of the same material as the first source and drain metal layers, and in particular, patterned on the upper portion of the driving thin film transistor (DRT) to apply the same voltage as the gate electrode, thereby forming a dual gate (dual gate). ) and an auxiliary gate electrode forming the structure.

As the second source and drain metal layers extend to the exposed region of the power supply voltage application wiring of the first source and drain metal layers and come into contact with each other, a signal supplied therefrom is applied to the anode metal layer.

In addition, an interlayer insulating film is formed on the second source and drain metal layers. A first contact hole exposing the lower second source and drain metal layers is formed in a partial region of the interlayer insulating layer, and an anode metal layer having a shape separated for each pixel is formed on the upper portion of the interlayer insulating layer including the first contact hole. is formed.

Here, the region exposed by the first contact hole overlaps the capacitor C1 formed by the data signal application wiring of the gate metal layer and the first source and drain metal layers downward, and the second source and drain metal layers and the anode metal layer are in contact with each other. defined as the first region.

In the first region, the data signal applying wiring of the first source and drain metal layers and the second source and drain metal layers are insulated from each other by a passivation layer.

Although not shown, the anode metal layer constitutes the anode electrode of the organic light emitting diode. On the anode metal layer, an organic light emitting layer and a cathode electrode are formed in an organic light emitting pattern for emitting red, green, and blue, respectively. Accordingly, the anode metal layer, the cathode electrode, and the organic light emitting layer interposed between the two electrodes form an organic light emitting diode.

Here, the organic light emitting layer may be composed of a single layer made of an organic light emitting material, or a hole injection layer, a hole transporting layer, a light emitting material layer, to increase luminous efficiency. It may consist of multiple layers of an electron transporting layer and an electron injection layer.

In particular, the anode metal layer of the present invention extends in one direction and extends not to the first contact hole, but to the second contact hole of the interlayer insulating film where the second source and drain metal layers that do not overlap with the passivation film are exposed, so that the second source and drain It is characterized in that it is in double contact with the metal layer.

That is, in the interlayer insulating film included in one pixel, not only the first contact hole formed in the region overlapping the passivation film, but also the passivation film is etched to apply the power supply voltage of the first source and drain metal layers and the second source and drain metal layers A second contact hole is further formed to correspond to the contact area, and the anode metal layer extends to the second contact hole so that the second source and drain metal layers and the anode metal layer come into contact with each other in redundancy.

Here, the region exposed by the second contact hole is a second region in which the gate metal layer, the power supply voltage application wiring of the first source and drain metal layers, the second source and drain metal layers, and the anode metal layer are sequentially formed and in direct contact with each other. is defined

According to this structure, even if an open failure of the second source and drain metal layers occurs due to the step difference of the passivation film in the etching process of the second source and drain metal layers, the electrical connection between the anode metal layer and the second source and drain metal layers is there will be no change The signal supplied to the anode metal layer is the power supply voltage ELVDD applied through the driving thin film transistor DRT.

In addition, at least one passivation film, an organic film, and a protective film are further provided on the anode metal layer to prevent moisture penetration into the organic light emitting layer, thereby forming an organic light emitting display device.

As described above, according to the present invention, based on the AI model that is learned by fusion of all building unit attribute information and fire information across the country, Geo AI technology is applied to predict the level of risk for each building unit, and the user can provide insurance target building information can realize a fire insurance non-face-to-face reverse auction service system and method using AI building fire risk model that calculates cheap and reasonable fire insurance premiums and provides contract brokerage services by presenting the calculated premiums by multiple non-life insurance companies. have.

The present invention described above is not limited to the above-described embodiments and the accompanying drawings, and it is common in the technical field to which the present invention pertains that various substitutions, modifications and changes are possible without departing from the technical spirit of the present invention. It will be clear to those who have the knowledge of

100: Fire insurance non-face-to-face reverse auction service system using AI building fire risk model
110: insurance brokerage server
120: Insurance Database
130: user terminal
140: national spatial information portal server
150: road name address guide

Claims (10)

a user terminal for entering insurance target building information, selecting one of building fire insurance according to a number of non-life insurance companies according to a fire damage insurance brokerage service customized for each building, and concluding a fire insurance contract in a non-face-to-face manner;
According to the insurance target building information input from the user terminal, data related to the insurance target building is collected, the collected data is refined and fused to set the data set, and based on the data set, the fire prediction and the fire risk of the insurance target building are calculated. An insurance brokerage server that calculates and presents fire insurance premiums for each insurance company after calculating and learning deep learning, and concludes a building fire insurance contract non-face-to-face according to the selection of the user terminal, thereby providing a customized fire non-life insurance brokerage service for each building; and
an insurance database for storing member information of the user terminal, information on the building fire insurance, and insurance contract information according to the building fire insurance contract;
Fire insurance non-face-to-face reverse auction service system using artificial intelligence building fire risk model including
The method of claim 1,
a credit rating agency server that provides personal credit information from the insurance brokerage server to the insurance brokerage server in response to a personal credit inquiry request of the user terminal;
At the request of the insurance brokerage server, the court registration server for providing registration data and registration information related to the insurance target building according to the insurance target building information;
In response to the request of the insurance brokerage server, according to the insurance target building information according to the insurance target building-related probate contract information, the official price of the building standard land, and a server for providing tax information;
In response to the request of the insurance brokerage server, a certified authentication authority server to verify the identity of the user of the user terminal and to perform a legal digital signature;
In response to the request of the insurance brokerage server, a fire service server that provides information on the fire of the building subject to insurance or whether or not the firefighting law is complied with;
a national spatial information portal server that provides GIS building integration information, individual public land price, and building age information to the insurance brokerage server; and
a road name address guide server providing road name addresses and building information of the insurance target building;
Fire insurance non-face-to-face reverse auction service system utilizing artificial intelligence building fire risk model further comprising a.
The method of claim 1,
The insurance brokerage server,
Obtain GIS building integrated information, individual public land price, and age information from the national spatial information portal server, obtain street name address and building information from the street name address guide server, or obtain building energy information and fire information from the building energy DB and fire DB a data collection unit that acquires information, refines and fuses the acquired data to generate a fusion data set;
a data pre-processing unit for deriving a risk grade for the data collected through the data collection unit, and processing public fire safety data including fire data from the fire DB and processing it as building unit and open data fusion data;
By learning deep neural network (DNN)-based fire accident information, a fire prediction model is generated for each building, and a building fire risk model to which a building fire risk class is applied through a clustering technique is created, and the generated fire prediction model and a model learning unit that learns a building fire risk model; and
Upon receiving the insurance target building information from the user terminal, the fire insurance discount rate and fire risk class are calculated based on the fire prediction model and the building fire risk model, and fire insurance premiums according to multiple non-life insurance companies are calculated based on this a service platform providing unit for providing a non-face-to-face reverse auction service for each building by presenting it to the user terminal and concluding a building fire insurance contract non-face-to-face according to the selection of the user terminal;
Fire insurance non-face-to-face reverse auction service system using artificial intelligence building fire risk model including
4. The method of claim 3,
The insurance database is composed of Post GIS that can store spatial information and data types,
The service platform provider provides a Web GIS-based web service based on an open source-based Web GIS service architecture, implements the spatial information as a web serviceable Geo server, and uses an Open Layers client library. It provides visualization and functions by utilizing Provides a business function decomposition menu by deriving a unit process for the system,
The task function division menu includes member management related to member registration, modification, and withdrawal, group management related to group creation, editing, and deletion, manager member management related to member approval and rejection, and adding and deleting members to and from the group Permission management menu including administrator group management; Fire risk level search and search menu for map-based building level search, area search, address (building) search, and fire risk level search; Fire safety inspection priority inquiry menu for regional inspection priority map-based inquiry, inspection priority list inquiry by region; And fire safety inspection data registration, modification, deletion, and fire safety inspection management menu related to inquiry, using artificial intelligence building fire risk model fire insurance premium non-face-to-face reverse auction service system.
The method of claim 1,
The model learning unit,
For the fire prediction model, after performing under-sampling at a ratio of 1:4 based on the fire accident in the fusion data set, normalization and one-hot Encoding (One-Hot encoding) to generate a sequential model (Sequential Model), by training the sequential model (Training) to generate an electric fire prediction model,
'Building ID', 'Usage_Amount (KWh)_Mean', 'Location Code', 'Individual Official Land Price', 'Lot Area', 'Name of Building Use', 'Name of Building Structure', 'Building Area', 'Height' 17 explanatory variables of _construction, 'building-to-coverage-ratio_consistency', 'floor-floor ratio_consistency', 'number of floors above ground', 'number of basement floors', 'regional division', 'total floor area', 'building age', and 'industry' In order to learn by setting and learning the dependent variable of the presence or absence of fire, and to classify the fire risk, the number of clusters is selected using the Elbow Method for a certain number of clusters compared between clusterings, and K- for grading the prediction results. A non-face-to-face reverse auction service system for fire insurance using artificial intelligence building fire risk model that derives the fire risk index for each building through clustering into five levels of safety, interest, caution, alertness, and risk.
A non-face-to-face reverse auction service method for fire insurance using artificial intelligence building fire risk model of an insurance brokerage server that collects data at the request of a user terminal and concludes building fire insurance,
(a) transmitting the insurance target building information received by the user terminal by executing the insurance application to the insurance brokerage server;
(b) collecting, by the insurance brokerage server, data related to an insurance target building according to the insurance target building information;
(c) generating, by the insurance brokerage server, a fusion data set by purifying and fusion of the collected data;
(d) deep learning by the insurance brokerage server by calculating the fire prediction and fire risk of the building to be insured based on the convergence data set;
(e) calculating, by the insurance brokerage server, a fire insurance premium for each insurance company according to the result of the deep learning learning and providing it to the user terminal;
(f) selecting, by the user terminal, an insurance company and a fire insurance premium; and
(g) the insurance brokerage server concluding a non-face-to-face building fire insurance contract with the user terminal;
Fire insurance non-face-to-face reverse auction service method using artificial intelligence building fire risk model including
7. The method of claim 6,
In the step (b), the insurance brokerage server collects spatial information-based public data for a building that is a fire prediction target,
The spatial information-based public data includes building attribute information, building spatial information, road name address information, energy consumption information, and fire accident information, and the building attribute information includes GIS building integrated information, individual public land price and building age, The building spatial information includes a building name, a dong name and a unique building number, an artificial intelligence building fire risk model utilizing fire insurance non-face-to-face reverse auction service system.
8. The method of claim 7,
In the step (c), the insurance brokerage server refines and fuses the collected data according to the building unit standard including the building unique number to generate a fusion data set, and the building unit standard is the building identification number and the lot Further including a unique number,
(c-1) generating first fusion data by merging the building spatial information and the road name address information according to the building identification number;
(c-2) according to the building identification number, the process of fusion of the first fusion data and the GIS building integrated information to generate second fusion data;
(c-3) generating third fusion data by fusing the second fusion data and information on the age of the building according to the building identification number;
(c-4) generating fourth fusion data by fusing the third fusion data and information on the individual official land price according to the lot unique number;
(c-5) generating a fifth fusion data by fusing the fourth fusion data and the energy usage information according to the building identification number; and
(c-6) generating final fusion data by fusing the fifth fusion data and the fire accident information according to the building identification number;
Fire insurance non-face-to-face reverse auction service system using artificial intelligence building fire risk model including
7. The method of claim 6,
In step (d), the insurance brokerage server,
Designating a part randomly selected from the fusion data set as a first data group, designating the rest that are not selected as a second data group, learning the first data group using a random forest algorithm, and , Mean Absolute Error (MAE) and correlation coefficient are calculated, the probability of a fire is predicted using the second data group, and safety, interest, attention, classify as borderline, hazard, or
Fire insurance using artificial intelligence building fire risk model, which generates a deep learning learning model by learning whether a fire has occurred for a property of a building based on the convergence data set, and calculates the fire risk of a building using the created learning model Non-face-to-face reverse auction service system.
7. The method of claim 6,
In step (d), the insurance brokerage server,
By performing undersampling, normalization and one-hot encoding on the fusion data set to generate training data, and reducing the number of attributes by deriving a Pearson correlation coefficient between explanatory variables focusing on the fire presence or absence attribute included in the learning data, artificial intelligence Fire insurance non-face-to-face reverse auction service system using building fire risk model.

Images