Data correctness optimization

1. A computer implemented method for improving data correctness of measured values using a ground truth dataset, comprising:

using a transceiver and at least one processor coupled to the transceiver for:

receiving a plurality of datasets from different data sources, the datasets comprising a plurality of data elements, wherein each data element includes an identifier and at least one measured value associated with the identifier;

determining data correctness values for the measured values using at least one of a panel and calibration measurements for checking correctness of the measured values, wherein a data correctness value is associated with a probability that a measured value is correct;

adding a data element including a single data correctness value for each measured value for each identifier with which a respective measured value is associated to the ground truth dataset, wherein the single data correctness value being based on the determined data correctness values for the measured values, whereby the data correctness values in the ground truth dataset define probability distributions of data correctness for the measured values; and

outputting the ground truth dataset, wherein the ground truth dataset is used to determine data correctness of measured values included in at least one new dataset by

receiving the new dataset from a data source, the new dataset comprising a plurality of data elements, wherein each data element of the new dataset includes an identifier and at least one measured value associated with the identifier; and

using respective data correctness values in the ground truth dataset for the measured values of respective data elements of the ground truth dataset determined as being with known identifiers, in determining data correctness values for measured values of data elements of the new dataset, wherein a known identifier is an identifier which is included in the ground truth dataset and in the new dataset, wherein determining data elements of the ground truth dataset with known identifiers comprising comparing identifiers included in the new dataset with identifiers included in the ground truth dataset, wherein data correctness values in the ground truth dataset are assigned for measured values of data elements of the new dataset based on overlapping of the measured values of data elements of the new dataset with measured values of data elements of the ground truth dataset with known identifiers.

2. The computer implemented method of claim 1 , wherein the at least one processor is further configured to determine a data correctness value for a measured value received from a respective data source for a subset of data elements received from the respective data source and to assign the data correctness value for the measured value determined for the subset of the data elements to the measured values of a same sensor information of the other data elements received from the respective data source.

3. The computer implemented method of claim 1 , wherein the at least one processor is used to determine the single data correctness value for the respective measured value for a respective identifier based on data correctness values for the measured value received from different data sources, when two or more data elements with the respective identifier and an identical measured value are included in the plurality of datasets received from the different data sources.

4. The computer implemented method of claim 3 , wherein the at least one processor is used to determine the single data correctness value for the respective measured value for the respective identifier by calculating

p

=

∏

i

n

p

i

∏

i

n

p

i

+

∏

i

n

(

1

-

p

i

)

wherein p is a single data correctness value, n is a number of the different data sources, and p i is a probability for the measured value received from data source i to be correct.

5. The computer implemented method of claim 1 , wherein the at least one processor is used to determine the single data correctness value for the respective measured value for a respective identifier based on calculating

p

i

=

∑

j

k

p

ij

k

wherein p i is a mean probability for the measured value received from the data source i to be correct, k is a number of probabilities for the measured value of the data element received from the data source i to be correct, and p ij is a j-th probability for the measured value of the data element of the data source i to be correct, when two or more data elements with the respective identifier and an identical measured value are included in one or more datasets received from the same data source.

6. The computer implemented method of claim 1 , wherein the at least one processor is further used to determine a data correctness value for at least one measured value of data elements of the new dataset with unknown identifier based on a probability distribution of data correctness for the respective measured value, when the new dataset comprises a threshold level of data elements including a known identifier, wherein an unknown identifier is an identifier which is not included in the ground truth dataset.

7. The computer implemented method of claim 6 , wherein the at least one processor is used for:

obtaining the probability distribution of data correctness for the respective measured value based on the ground truth dataset,

binning of data correctness values of the probability distribution of data correctness for the respective measured value, wherein each bin includes data elements of the ground truth dataset with data correctness values for the respective measured value within a range of data correctness values for the respective measured value,

extracting, from each bin, a number of samples each including a number of data elements,

determining, for each sample, a number of data elements with identifiers in a respective sample which are identical to identifiers in the new dataset, wherein the respective sample is discarded if the number of data elements with identifiers in the respective sample which are identical to identifiers in the new dataset is below a threshold identifier number and wherein the respective sample is further processed if the number of data elements with identifiers in the respective sample which are identical to identifiers in the new dataset is equal to or above the threshold identifier number,

determining for each sample a sample probability score based on the data correctness values for the respective measured value of the data elements with identical identifiers in the ground truth dataset and the new dataset,

determining for each bin a bin probability score of a respective bin based on the sample probability scores of the respective bin,

determining for each bin a weighted bin score by multiplying a respective bin probability score with a respective number of data elements with identifiers in the respective bin which are identical to identifiers in the new dataset,

determining for each bin a confidence score based on a respective weighted bin score and the respective number of data elements with identifiers in the respective bin which are identical to identifiers in the new dataset,

determining a final score by dividing a sum of the confidence scores by a sum of the number of data elements with identifiers in the bins which are identical to identifiers in the new dataset, and

assigning the final score as data correctness value to the measured value of the data elements of the new dataset with unknown identifier.

8. The computer implemented method of claim 7 , wherein the at least one processor is used to determine the confidence score for each of the bins based on Wilson confidence interval.

9. The computer implemented method of claim 1 ; wherein the at least one processor is used to add, to the ground truth dataset, the data elements of the new dataset with unknown identifier for which the data correctness values for the measured values were determined.

10. The computer implemented method of claim 1 , wherein the at least one processor is used to compare the data correctness values for a respective measured value of respective data elements of the ground truth dataset to data correctness values for the respective measured value of the respective data elements determined using at least one of a panel and calibration measurements for checking correctness of the measured values in dependence of a triggering event.

11. The computer implemented method of claim 10 , wherein the at least one processor is used to adjust the data correctness values for the respective measured value of the respective data elements of the ground truth dataset based on a difference between the data correctness values for the respective measured value of the respective data elements of the ground truth dataset and the data correctness values for the respective measured value of the respective data elements determined using at least one of a panel and calibration measurements for checking correctness of the measured values.

12. The computer implemented method of claim 10 , wherein the at least one processor is used for:

obtaining a probability distribution of data correctness for the respective measured value based on the ground truth dataset,

binning of data correctness values of the probability distribution of data correctness for the respective measured value, wherein each bin includes data elements of the ground truth dataset with data correctness values for the respective measured value within a range of data correctness values for the respective measured value,

determining, for each bin, a bin data correctness value for the respective measured value using at least one of a panel and calibration measurements for checking correctness of the measured values,

learning an adjustment weight for the data correctness values for the respective measured value for the data elements for each bin based on minimizing a difference between a median data correctness value for the respective bin determined based on the data correctness values for the respective measured value in a respective bin weighted based on a current adjustment weight and the bin data correctness value for the respective bin, and

adjusting, for each data element, the data correctness value for the respective measured value in the ground truth dataset based on the respective adjustment weight.

13. The computer implemented method of claim 12 , wherein the at least one processor is used to perform determining, for each bin, a bin data correctness value for the respective measured value using at least one of a panel and calibration measurements for checking correctness of the measured values based on a subset of data elements selected from the respective bin.

14. The computer implemented method of claim 1 , wherein the measured values are received from one or more sensor devices.

15. A data correctness management system for improving data correctness using a ground truth dataset, comprising:

a transceiver; and

at least one processor coupled to the transceiver, the at least one processor executing a code comprising:

program instructions to receive, via the transceiver, a plurality of datasets from different data sources, the datasets comprising a plurality of data elements, wherein each data element includes an identifier and at least one measured value associated with the identifier;

program instructions to determine data correctness values for the measured values using at least one of a panel and calibration measurements for checking correctness of the measured values, wherein a data correctness value is associated with a probability that a measured value is correct; and

program instructions to add a data element including a single data correctness value for each measured value for each identifier with which a respective measured value is associated to the ground truth dataset, wherein the single data correctness value being based on the determined data correctness values for the measured values, whereby the data correctness values in the ground truth dataset define probability distributions of data correctness for the measured values, and

program instructions to receive a new dataset from a data source, the new dataset comprising a plurality of data elements, wherein each data element of the new dataset includes an identifier and at least one measured value associated with the identifier; and to

use respective data correctness values in the ground truth dataset for the measured values of respective data elements of the ground truth dataset determined as being with known identifiers, in determining data correctness values for measured values of data elements of the new dataset, wherein a known identifier is an identifier which is included in the ground truth dataset and in the new dataset, wherein determining data elements of the ground truth dataset with known identifiers comprising comparing identifiers included in the new dataset with identifiers included in the ground truth dataset, wherein data correctness values in the ground truth dataset are assigned for measured values of data elements of the new dataset based on overlapping of the measured values of data elements of the new dataset with measured values of data elements of the ground truth dataset with known identifiers.

16. The data correctness management system of claim 15 , the code comprising program instructions to determine a data correctness value for measured values of data elements of the new dataset with unknown identifier based on a probability distribution of data correctness for a respective measured value, when the new dataset comprises a threshold level of data elements including a known identifier, wherein an unknown identifier is an identifier which is not included in the ground truth dataset.

17. The data correctness management system of claim 15 , the code further comprising program instructions to compare the data correctness values for a respective measured value of respective data elements of the ground truth dataset to data correctness values for the respective measured value of the respective data elements determined using at least one of a panel and calibration measurements for checking correctness of the measured values and to adjust the data correctness values for the respective measured value of the respective data elements of the ground truth dataset based on a difference between the data correctness values for the respective measured value of the respective data elements of the ground truth dataset and the data correctness values for the respective measured value of the respective data elements determined using at least one of a panel and calibration measurements for checking correctness of the measured values.

18. The data correctness management system of claim 15 , wherein the measured values are received from one or more sensor devices.