KR102318801B1 - Driver traffic accident rate prediction system - Google Patents

Abstract

The present invention relates to a system for predicting a driver's traffic accident rate for predicting an accident rate that may occur in the future for each driver through personal information possessed by an individual, past driving-related violation information, and accident information for an arbitrary driver.
Accordingly, the present invention is expected to have a public benefit effect of improving normal driving habits through driving records generated in the daily life of the driver, thereby improving the economic loss caused by traffic accidents through the improved driving habits of the driver, and the insurance company side It has the effect of providing the possibility of providing various services to the driver by measuring the insurance loss ratio using the driver's accident rate.

Description

Driver traffic accident rate prediction system

The present invention relates to a system for predicting a driver's traffic accident rate, and more particularly, a driver who predicts an accident rate that may occur in the future for each driver through personal information of an individual for any driver, violation information related to past driving, and accident information of the traffic accident rate prediction system.

As the number of vehicles increases due to the development of the industry, the number of drivers increases and traffic accidents occur frequently, as well as violations of traffic laws. Although driving experience is being converted into digital data, there is a problem that such meaningful data of drivers cannot be utilized so highly.

In addition, the technology that utilizes the driver's past data is rare, and it is difficult to find it easily in the field of artificial intelligence.

On the other hand, Republic of Korea Patent Publication No. 10-2019-0059723 "Artificial intelligence-based traffic accident prediction system and its method" recognizes an object from an image captured by a camera and uses an artificial intelligence-based accident probability algorithm for the recognized object. A vehicle system that analyzes to predict the probability of an accident, and if an accident is predicted as a result of the prediction, the vehicle system that transmits the accident prediction data to the control center, and analyzes the accident prediction data transmitted from the vehicle system It consists of a control center that transmits and learns about the accident probability using the accident prediction data, and transmits the learning result to the vehicle system to update the AI-based accident probability algorithm.

However, in the prior art, an accident can be prevented in advance by predicting vehicles, people, and roads with a high risk of an accident using an image recognition deep learning technique, and sending the prediction result to the control center to quickly take follow-up measures, There is a problem in that the rate of traffic accidents cannot be predicted using personal information of each driver, violation information related to past driving, and accident information.

Patent Document 1. Republic of Korea Patent Publication No. 10-2019-0059723 "Artificial intelligence-based traffic accident prediction system and method"

Accordingly, the present invention has been devised to solve the above problem, and an object of the present invention is to induce the driver's safe driving by quantitatively calculating the driver's accident probability.

Another purpose is to provide insurance companies with the accident rate that will occur in the future for each driver, and to use it as a data to improve the insurance loss ratio from the insurance company side along with the insurance system guidance.

The system for predicting a driver's traffic accident rate according to an embodiment of the present invention for achieving the above technical problem is data that collects personal attributes of an individual related to driving for each driver, violation attributes related to driving, and accident information Setbuwa; Through the accident information, the accident rate is generated by synthesizing the number of accidents already determined for each driver within a certain period, and the individual attributes and violation attributes of each driver are bundled for each attribute to form plastic groups. In combination with the number of times, each level of importance is set for each attribute group or a combination of attribute groups, and a first attribute value for accident prediction is given between each attribute group and the number of accidents for each attribute group, and from the first attribute value A second attribute value is generated by extracting attribute values corresponding to individual attributes and violation attributes of any driver among each driver, and the second attribute values are the second attribute values excluding the second attribute value for each driver. When combining with each other, the third attribute values for measuring the accident rate of an arbitrary driver corresponding to the upper level calculated by applying the importance appropriate to each combination are generated, and then the process of discarding the third attribute values less than the set value is repeated to obtain the highest level. The fourth attribute value most related to the accident is calculated for the remaining random drivers, and the other random drivers are repeated until the accident rate by combining the fourth attribute value and the accident rate algorithm reaches the previous accident rate within the set range. a predictive model unit; and generating a fifth attribute value by extracting attribute values corresponding to personal attributes and violation attributes of any new driver from the first attribute value when the accident rate reaches an accident rate within a set range, and the fifth attribute value is generated. After generating the sixth attribute values for measuring the accident rate of any new driver corresponding to the higher level calculated by applying the importance appropriate for each combination when combining the attribute values with the other fifth attribute values except for themselves, the sixth attribute values are generated. By repeating the process of discarding the sixth attribute values less than the set value, the seventh attribute value most related to the accident is calculated for any new driver remaining at the highest level, and the result is generated in combination with the accident rate algorithm to generate the accident rate of the new driver The one made up of the model part is taken as a solution.

As described above, according to the present invention, a public benefit effect of improving normal driving habits is expected through driving records generated in the daily life of the driver, and thereby the economic loss due to traffic accidents is improved through the improved driving habits of the driver. In addition, the insurance company has the effect of providing the possibility of providing various services to the driver by measuring the insurance loss ratio using the driver's accident rate.

1 is a configuration diagram of a system for predicting a driver's traffic accident rate according to an embodiment of the present invention;
2 is an exemplary diagram showing attributes and article height rates according to an embodiment of the present invention;
3 is an exemplary view showing the importance of each attribute according to an embodiment of the present invention;
4 is a diagram illustrating an example of applying an attribute value according to an embodiment of the present invention;
5 is a diagram illustrating an accident rate setting according to an embodiment of the present invention;
6 is an exemplary view of individual driver accident rate measurement according to an embodiment of the present invention;

Hereinafter, the configuration and operation will be described with reference to the accompanying drawings for the optimal embodiment of the present invention, and the embodiment of the present invention may be modified in various forms, and the shape of the components in the drawings will be emphasized more clearly In order to do so, detailed descriptions of well-known functions and configurations that may be exaggerated and may unnecessarily obscure the gist of the present invention will be omitted.

1 is a configuration diagram of a driver's traffic accident rate prediction system according to an embodiment of the present invention, wherein the driver's traffic accident rate prediction system 100 includes a data set unit 110, a prediction model unit 120, and a result derivation model unit ( 130).

In more detail, the data set unit 110 collects and dataizes personal attributes of an individual related to driving for each driver, violation attributes related to driving, and accident information.

For example, the personal attributes are related to driving such as gender, age, occupational group, residential area, working area, marital status, child age, number of children, driving experience, cumulative driving distance, vehicle type, mileage, driving time, parking lot, etc. It corresponds to each attribute possessed by an individual.

In addition, the violation attributes correspond to attributes possessed by the driver related to driving, such as speed violations, driver's license penalty points, parking and stopping violations, speeding violations, and signal violations center line violations.

In addition, the accident information is information on how many times an accident occurred during a certain period for each driver.

Therefore, the data set unit 110 collects personal attributes, violation attributes, and accident information for each driver and builds big data of hundreds of thousands or more of drivers, not hundreds of drivers, in order to predict the accident rate in advance.

At this time, it is desirable to collect information about the violation attributes for each driver through e-Pain (National Police Agency), each local government enforcement complaint site, and the Insurance Development Institute.

The predictive model unit 120 generates an accident rate by synthesizing the number of accidents already determined within a certain period for each driver through the accident information, and sets individual attributes and violation attributes of each driver for each attribute. After forming plastic groups by grouping, they are combined with the number of accidents to set the importance of each attribute group or a combination of attribute groups, and a first attribute value for accident prediction is given between each attribute group and the number of accidents for each attribute group. and extracting attribute values corresponding to individual attributes and violation attributes of any driver among the drivers from the first attribute value to generate a second attribute value, wherein the second attribute values are their own for each driver. When combining with the remaining second attribute values except for , the third attribute value below the set value is generated after generating the third attribute values for measuring the accident rate of any driver corresponding to the higher level calculated by applying the importance appropriate for each combination. The fourth attribute value most related to the accident is calculated for any driver remaining at the top by repeating the discarding process, and when the accident rate combining the fourth attribute value and the accident rate algorithm reaches the previous accident rate within the set range Repeat with any other drivers until .

For example, the predictive model unit 120 generates an accident rate by synthesizing the number of accidents already determined within a certain period for each driver through the accident information. Xn is each driver, Y1 to Yn are individual attributes and violation attributes.

After all, the accident rate is determined by the number of accidents that have already been determined and converted into data.

Next, the individual attributes and violation attributes possessed by each driver are grouped by each attribute to form plastic groups, and then combined with the number of accidents to set the importance of each attribute group or attribute group combination, see FIG. 3 . In detail, X1 to Xn are each driver, and each driver has personal attributes and violation attributes such as marital status, driving history, speeding, and signal violation, and the importance is set for these attributes.

That is, for each driver, the marital status attribute shown in Fig. 3 is grouped from X1 to Xn, the driving experience attribute is bundled from X1 to Xn, the speeding attribute is bundled from X1 to Xn, and the signal violation attribute is bundled from X1 to Xn. to form attribute groups, and for each attribute group, for example, in the case of the speeding shown in FIG. 3, the importance is set through a causal relationship between the speeding and the number of accidents.

In this case, since the causal relationship between the number of accidents and the importance corresponds to an algorithm that can be implemented by various design changes rather than through a specific algorithm, further description is omitted.

In addition, the importance is set by the combination of attribute groups, such as the causal relationship between marital status and driving experience shown in FIG. 3, marital status, driving experience, causal relationship between speeding, driving experience, speeding, and the causal relationship of signal violation. .

Next, a first attribute value for accident prediction is given between each attribute group and the number of accidents for each attribute group, and from the first attribute value, the individual attributes and violation attributes of any driver among each driver are corresponding The generation of the second attribute value by extracting the attribute values to be obtained will be examined in detail with reference to FIG. 4 . For each attribute group, the marital status attribute and residential area attribute of each driver as shown in FIG. 4 (a). The first attribute value is assigned to each attribute, such as , work area attribute, speeding attribute, center line violation attribute, and signal violation attribute.

For example, as shown in (a) in FIG. 4 , the attribute A is assigned to the marital status attribute, the attribute B is assigned to the residential area attribute, the attribute C is assigned to the work area attribute, and the attribute D is assigned to the speeding attribute. In addition, the E attribute value is assigned to the center line violation attribute and the F attribute value is assigned to the signal violation attribute. corresponds to a numerical value.

After the attribute value is assigned in this way, the attributes of any driver among the respective drivers, for example, the marital status attribute, the residential area attribute, the speeding attribute, and the signal violation attribute as shown in FIG. From the first attribute value, the marital status attribute is extracted from the attribute A, the residential area attribute is the attribute B, the speeding attribute is the attribute D, and the signal violation attribute is the attribute F to generate second attribute values.

Next, for each of the second attribute values, a third attribute value for measuring the accident rate of an arbitrary driver corresponding to a higher level calculated by applying the importance appropriate for each combination when combining with the other second attribute values except for themselves The process of discarding the third attribute values less than or equal to the set value is repeated to calculate the fourth attribute value most related to the accident for any driver remaining at the highest level after generating the values. , as shown in FIG. 5 (a), assuming that the second attribute values are A, B, D, F, the remaining second attribute values excluding themselves, that is, in the case of A, the remaining B, D, F and When combining, if the importance shown in FIG. 3 is applied and B, D, and F are applied as in A, high-level third attribute values such as G, H, I, J, K, and L are generated.

After that, if it is assumed that the upper-level third attribute values are G, H, I, the remaining J, K, and L properties except for G, H, I among G, H, I, J, K, and L are set again. The highest fourth attribute value is calculated by repeating the above-described method by applying the importance of the group combination.

At this time, the importance applied to J is set by the correlation between the residential area B attribute and the speeding D attribute, and the importance applied to K is set by the correlation between the residential area B attribute and the signal violation attribute F, and applied to L The importance set by the association between the speeding attribute D and the traffic violation attribute F is applied in the same way until the fourth attribute value is calculated.

Next, the other arbitrary drivers are repeated until the accident rate obtained by combining the fourth attribute value and the accident rate algorithm reaches the previous accident rate within the set range.

For example, if the accident rate is 15% and the set range is 14.9% to 14.3%, if the accident rate calculated by the fourth attribute value and the accident rate algorithm is out of 14.9% to 14.3%, it is continued to perform as described above for other drivers. Repeat until 14.9% to 14.3% is reached.

In addition, since the accident rate algorithm can be applied to various calculation methods, further description is omitted.

The result deriving model unit 130 extracts attribute values corresponding to personal attributes and violation attributes of any new driver from the first attribute value when the accident rate reaches the previous accident rate within the set range to obtain a fifth attribute. value is generated, and when the fifth attribute values are combined with the other fifth attribute values except for themselves, the degree of importance appropriate for each combination is applied to measure the accident rate of any new driver corresponding to the higher level calculated After generating the sixth attribute values, the process of discarding the sixth attribute values that are less than or equal to the set value is repeated to calculate the seventh attribute value most related to the accident for any new driver remaining at the highest level, and combine it with the accident rate algorithm to create a new Generates the driver's accident rate.

For example, the result deriving model unit 130 is a sample for predicting the accident rate as the accident rate reaches the previous accident rate within the set range, so the attribute of the new driver is as shown in FIG. If it is 1 time and there are 4 signal violations, a 5th attribute value is generated by extracting from the first attribute value for the male attribute, the Seoul residence attribute, the speeding 1 time attribute, and the 4 signal violation attribute.

When the fifth attribute values are combined with the other fifth attribute values excluding themselves, the importance is applied as shown in FIG. 6 to generate higher-level attribute values of a, b, c, d, e, and f. Among the values b, c, d, e, and f, the discarding process is repeated to calculate the highest seventh attribute value and combine with the accident rate algorithm to generate an accident rate.

In order to help the understanding of the present invention, only a few of the attributes described above are applied, and it is preferable to apply several tens of the number of attributes.

On the other hand, as the result deriving model unit 130 applies the attributes of the arbitrary new driver to the dataset unit 110, the predictive model unit 120 newly sets the importance for each attribute group or a combination of attribute groups. it is preferable

For example, if the accident rate is generated for a driver who is a man, lives in Seoul, and has one speeding and four signal violations, the result derivation model unit 130 is a male attribute, Seoul resident attribute, The 1 overspeed attribute and 4 signal violation attribute are applied to the data set unit 110 , whereby the importance is newly set by each attribute group or a combination of attribute groups, thereby increasing the accuracy of the accident rate.

On the other hand, the result deriving model unit 130 changes the personal attributes and violation attributes of the arbitrary new driver as the predetermined period elapses after a predetermined period of time has elapsed based on the time when the accident rate of the arbitrary new driver is calculated. It is desirable that the accident rate be newly created according to the

For example, if driver A created the accident rate as of October 21, 2019, the natural change of driver A's attributes over a certain period of time, for example, the same driver due to childbirth, age change, change in the number of accidents, change of residence, etc. Even at A, a new accident rate is created.

As described above, the best embodiments of the present invention are disclosed in the drawings and detailed description, and the specific terms used above are only used for the purpose of describing the present invention, limiting the meaning or of the present invention described in the claims It is not used to limit the scope.

Therefore, various modifications and equivalent other embodiments are possible therefrom by those of ordinary skill in the art, and the true technical protection scope of the present invention should be defined by the technical spirit of the claims.

100: driver's traffic accident rate prediction system
110: data set part
120: predictive model unit
130: result derivation model unit

Claims (3)

a data set unit that collects driving-related personal attributes, driving-related violation attributes, and accident information for each driver;
Through the accident information, the accident rate is generated by synthesizing the number of accidents already determined for each driver within a certain period, and the individual attributes and violation attributes of each driver are bundled for each attribute to form attribute groups. In combination with the number of times, each level of importance is set for each attribute group or a combination of attribute groups, and a first attribute value for accident prediction is given between each attribute group and the number of accidents for each attribute group, and from the first attribute value A second attribute value is generated by extracting attribute values corresponding to individual attributes and violation attributes of any driver among each driver, and the second attribute values are the second attribute values excluding the second attribute value for each driver. When combining with each other, the third attribute values for measuring the accident rate of an arbitrary driver corresponding to the upper level calculated by applying the importance appropriate to each combination are generated, and then the process of discarding the third attribute values less than the set value is repeated to obtain the highest level. The fourth attribute value most related to the accident is calculated for the remaining random drivers, and the other random drivers are repeated until the accident rate by combining the fourth attribute value and the accident rate algorithm reaches the previous accident rate within the set range. a predictive model unit; and
When the accident rate reaches the previous accident rate within the set range, a fifth attribute value is generated by extracting attribute values corresponding to personal attributes and violation attributes of any new driver from the first attribute value, and the fifth attribute The values are set after generating the sixth attribute values for measuring the accident rate of any new driver corresponding to the upper level calculated by applying the importance appropriate for each combination when combining the values with the other fifth attribute values except for themselves. A result derivation model that generates the accident rate of a new driver by repeating the process of discarding the sixth attribute values below the value, calculating the seventh attribute value most related to the accident for any new driver remaining at the highest level, and combining it with the accident rate algorithm Driver's traffic accident rate prediction system, characterized in that consisting of parts.
The method according to claim 1, The result derivation model unit
The predictive model unit for each attribute group or a combination of attribute groups as the attributes of the arbitrary new driver are applied to the data set unit, and the driver's traffic accident rate prediction system, characterized in that the importance is newly set.
The method according to claim 2, The result derivation model unit
After a certain period of time has elapsed based on the time when the accident rate of the arbitrary new driver is calculated, the accident rate according to the change of the personal attributes and the violation attributes of the arbitrary new driver is newly generated as the predetermined period elapses. Driver's traffic accident rate prediction system.

Images