232 extracts an invalid data part and an occurrence time of the corresponding part to generate side information.

When the invalid value of the data continues to exceed the noise determination reference interval, the abnormality sign monitor 231 determines that the invalid value of the data is not due to simple noise but is due to an abnormality sign. The invalid data start part detection and side information generator 233 extracts a start time of invalid data and generates side information.

When it is determined that the invalid value of the data is due to noise, the corresponding data is transmitted to a noise corrector 240. The noise corrector 240 extracts pre/post-data of the invalid data part based on side information received from the invalid data part detection and side information generator 232, and replaces a noise generating part with a corrected value through noise correction based on statistical characteristics, noise correction based on machine learning such as a recurrent neural network (RNN), or the like. The noise corrector 240 is optional and can be applied as needed, and is not applied to a system in which noise correction or deletion is unnecessary. When the noise corrector 240 is applied to correct noise, a valid data re-validator 250 validates the data validity once again to determine whether the invalid data due to the noise has been properly corrected, and determines whether the invalid data is transmitted to the valid data output unit 270 according to whether the invalid data has been properly corrected.

When it is determined that the invalid value of the data is due to the abnormality sign, the abnormality sign monitoring and determining unit 230 transmits the data to the abnormality sign output unit 260, and the abnormality sign output unit 260 displays the abnormality sign. The abnormality sign output unit 260 displays a value where the abnormality sign occurs and a comparison result with a normal category together, and thus, supports an administrator to more accurately determine whether there is the abnormality sign.

FIG. 5 is a diagram illustrating illustrates a data linkage validator according to an embodiment of the present invention.

The data linkage validator according to an embodiment of the present invention validates validity by linking various kinds of data. A priority determining unit 310 determines priority of data, assigns a higher priority to environmental and condition data and the like, and assigns a lower priority to operation data and the like.

For example, weather data is high-priority data, and high priority is assigned to environmental and condition data, such as whether there is precipitation and whether temperature is high or low.

For example, tourist destination floating population data is subordinated motion data, and when weather data of 20° C. without rain is selected as a priority condition and tourist destination floating population data is applied as a lower priority, a high value of floating population is generally obtained.

A priority data-based linkage validator 320 validates data by linking a plurality of data, and a decision tree method or the like is used for the validation.

As a result of the validation of the priority data-based linkage validator 320, when it is determined that there is no error in data, the data is transmitted to a valid linkage data output unit 370, and when it is determined that there is an error in data, the priority data-based linkage validator 320 transmits the corresponding data to the abnormality sign monitoring and determining unit 330.

The abnormality sign monitoring and determining unit 330 determines whether the result of data validity validated by linking multiple data is due to simple noise or an error based on a reason such as a failure or malfunction of an object data collection sensor.

When it is determined that the result of the data validity is due to the simple noise, the noise corrector 340 extracts the pre/post-data of the invalid data part, and replaces the noise generating part with the corrected value through noise correction based on statistical characteristics or noise correction based on machine learning such as a recurrent neural network (RNN).

When the noise corrector 340 is applied to correct noise, a valid data re-validator 350 validates the data validity once again to determine whether the invalid data due to the noise has been properly corrected, and determines whether the invalid data is transmitted to the valid linkage data output unit 370 according to whether the invalid data has been properly corrected.

When it is determined that the invalid value of the data is the abnormality sign, the abnormality sign monitoring and determining unit 330 transmits the corresponding data to the abnormality sign output unit 360, and the abnormality sign output unit 360 displays the value where the abnormality sign occurs and the comparison result with the normal category, and thus, supports the administrator to accurately identify whether there is the abnormality sign.

The linkage block integration validator 400 checks data errors and abnormality signs once again, checks the validity side information of each data, and finally monitors the decision tree values once again, thereby finally reviewing whether there are the invalid data, the abnormality signs, and the like.

FIG. 6 is a diagram illustrating a method of validating various kinds of data and a digital twin operation according to an embodiment of the present invention.

In step S610, data to be validated is selected from among input data. In step S620, data validity is validated for each data type. In step S630, validation is performed by linking various kinds of data using data validity results validated for each type. In step S640, linkage block integration validation is finally performed.

FIG. 7 is a diagram illustrating a process of validating data according to an embodiment of the present invention.

In step S621, validity is validated by comparing a current value of data and a validation reference value according to the data type, and it is checked whether there is an error in the data.

When it is determined that there is no error in the data in step S621, in step S625, valid data is output.

When it is determined that there is an error in the data in step S621, in step S622, it is determined whether the corresponding error is due to simple noise or an abnormality sign.

In step S622, when the invalid value of the data continues to exceed the noise determination reference interval, it is determined that the invalid value of the data is not due to the simple noise but is due to the abnormality sign. In this case, the side information is generated by extracting the time when the invalid data starts, and in step S624, the abnormality sign is output.

In step S622, when the invalid value of the data occurs for a very short time equal to less than the noise discrimination reference interval and then returns to the normal value, it is determined that the error is due to noise. In this case, the side information is generated based on the invalid data part and the occurrence time of the corresponding part.