JP6595884B2 - Data evaluation apparatus, data evaluation method, and program - Google Patents

Description

  The present invention relates to a technique for evaluating and improving the reliability of data acquired from a plurality of data sources.

  Quality information (eg, throughput) in the network can be obtained in various ways. For example, a measurement device can be connected to a communication device inside the network, and quality information can be acquired as a measurement result by the measurement device. In addition, for example, a measurement result of a communication speed by an external research organization can be acquired as quality information.

  As described above, it is possible to acquire quality information (referred to as data) from various information sources (referred to as data sources).

JP 2008-31720 A

  However, data obtained from all data sources is not always reliable. Therefore, it is necessary to check whether the data obtained from the data source is reliable. For example, for data reliability check, data obtained from multiple data sources are compared, and if the correlation value between the data is high, it is determined that any data is reliable, and If the correlation value is low, it may be determined that at least one of the data is not reliable.

  In addition, data such as quality information is generally obtained as time-series data. However, data at all times in the time-series data is not always reliable, for example, only data in a specific time zone can be trusted. May be data. As an example, quality information cannot be stably measured in a time zone where the network is congested such as at night, and reliable quality information is not always obtained. Conversely, it can be expected that reliable network quality information can be acquired in a specific time zone where the network is not so congested.

  So, for example, when comparing raw data obtained from multiple data sources to check the reliability of the data, the correlation is low, but the correlation is high when viewed in a specific time zone. May occur. However, there has been no conventional technique that can evaluate the reliability of data in consideration of the characteristics of such time-series data.

  Note that the time associated with time-series data is one of the attributes of the data. The problem as described above is a problem that may occur without being limited to data whose data attribute is time.

  The present invention has been made in view of the above points, and provides a technique that makes it possible to evaluate the reliability of data obtained from a certain data source in consideration of the characteristics relating to the attributes of the data. With the goal.

According to the embodiment of the present invention, a first network quality information and the second network quality information obtained from different data sources, and the first network quality information having a signal strength which is a duplicate attribute Input means for inputting second network quality information ;
Based on the signal strength , each of the first network quality information and the second network quality information is divided into a plurality of sections, and the same in the first network quality information and the second network quality information . A calculation means for calculating a correlation coefficient between the network quality information of the sections;
Among correlation coefficients obtained for each section in the plurality of sections, a correlation coefficient satisfying a predetermined condition is determined, a section corresponding to the correlation coefficient is selected, and information indicating the section is output. A data evaluation apparatus comprising: an evaluation means.

Moreover, according to the embodiment of the present invention, a data evaluation method executed by the data evaluation device,
A first network quality information obtained from different data sources and the second network quality information, input for inputting the first network quality information and a second network quality information having a signal strength which is a duplicate attribute Steps,
Based on the signal strength , each of the first network quality information and the second network quality information is divided into a plurality of sections, and the same in the first network quality information and the second network quality information . A step of calculating a correlation coefficient between network quality information of the sections, and determining a correlation coefficient satisfying a predetermined condition among the correlation coefficients obtained for each section in the plurality of sections; There is provided an evaluation step including selecting an interval corresponding to a number and outputting information indicating the interval.

  According to the embodiment of the present invention, there is provided a technique that makes it possible to evaluate the reliability of data obtained from a certain data source in consideration of characteristics related to the attribute of the data.

1 is an overall configuration diagram of a system in an embodiment of the present invention. It is a block diagram of a data evaluation apparatus. It is a flowchart which shows the procedure of the process which a data evaluation apparatus performs. It is a figure which shows the example of input data. It is a figure which shows the example of the time-sequential data production | generation process according to an interval, and a score calculation process. It is a figure which shows the example (every day + restrictions) of the time series data generation process with restrictions, and a score calculation process. It is a figure which shows the example (every week + restriction | limiting) of the time series data generation process with restrictions, and a score calculation process. It is a figure which shows the example (every month + restrictions) of the time series data generation process with restrictions, and a score calculation process.

  Embodiments of the present invention will be described below with reference to the drawings. The embodiment described below is only an example, and the embodiment to which the present invention is applied is not limited to the following embodiment.

  For example, in the present embodiment, data having time as an attribute, such as quality information, is used as data to be evaluated, but the present invention is not limited to data having time as an attribute, and is not related to time. It can be applied to various data.

(System configuration)
FIG. 1 shows an overall configuration diagram of a system according to the present embodiment. As shown in FIG. 1, the system in the present embodiment includes a data evaluation apparatus 100 and a plurality of data sources. FIG. 1 shows a data source 1 and a data source 2 in a plurality of data sources as an example.

  The data evaluation apparatus 100 in the present embodiment evaluates these data by comparing the data obtained from the data source 1 with the data obtained from the data source 2. Specific processing contents will be described later.

  Although the data to be evaluated in this embodiment is assumed to be time-series data such as network quality information, this is only an example, and the data to be evaluated by the data evaluation device 100 is specified. It is not limited to this type of data. It should be noted that “time-series data” in the present embodiment includes not only data arranged in time order at equal intervals but also data arranged in time order that is not equally spaced.

  Each of the data sources 1 and 2 is, for example, a measuring device that measures quality information of a certain network. That is, in this case, the data source 1 is a measuring device that measures the quality information of the network, and the data source 2 is another measuring device that measures the quality information of the network.

  Further, for example, the data source 1 may be a measuring instrument that measures the quality information of the network inside the network, and the data source 2 may be an external research organization for the network.

  The technique of this embodiment is, for example, when the data of one data source is reliable (eg, internal measurement data), and the reliability of the data of the other data source is unknown (eg, data from an external research institution) The reliability of data with unknown reliability can be evaluated by evaluating to what extent the data with unknown reliability is similar to the reliable data (whether there is a correlation).

  However, this is an example, and the reliability of both data may be unknown. Even if the reliability of both data is unknown, for example, if the correlation in a certain time zone is high, it can be estimated that the reliability of the data in that time zone is high, and the data in that time zone should be extracted And can be used for subsequent analysis. Further, the technique of the present embodiment can also be applied when it is considered that the reliability of both data is high. Even when the reliability of both data is considered high, the reliability is not always high at all times. By applying the technology of the present embodiment, for example, a highly reliable time zone or total period Etc. can be specified.

(Configuration example of data evaluation apparatus 100)
FIG. 2 shows a configuration example of the data evaluation apparatus 100 in the present embodiment. As shown in FIG. 2, the data evaluation apparatus 100 includes an input unit 101, a time-series data generation unit 102 by interval, a constrained time-series data generation unit 103, a score calculation unit 104, a score evaluation unit 105, an output unit 106, and A data storage unit 107 is provided. Details of the function of each unit will be described later as the contents of processing executed by each unit. The outline of each part is as follows.

  The input unit 101 inputs data from each data source. The time-series data generation unit by interval 102 generates time-series data by interval (eg, daily, weekly, monthly) from the data input by the input unit 100. The “interval” may be referred to as a “period”.

  The constrained time-series data generation unit 103 generates time-series data for each interval with a restriction (for example, a restriction that uses data at a specific time in a day) from the data input by the input unit 100. To do. In the present embodiment, the restriction is implemented based on a “time zone” such as a specific time zone and day of the week.

  The score calculation unit 104 uses the data source 1 for each of the time series data by interval generated by the time series data generation unit 102 by interval and the restricted time series data generated by the restricted time series data generation unit 103. The correlation with the data source 2 is calculated, and the obtained correlation coefficient (relation degree) is used as a score.

  The score evaluation unit 105 evaluates the score obtained by the score calculation unit 104 to determine an interval / constraint having a high correlation. The output unit 106 outputs information on the interval / restriction determined by the score evaluation unit 105, data corresponding to the interval / restriction, and the like. The data storage unit 107 temporarily stores data being processed by each unit, data of processing results, scores, and the like, and stores threshold values and the like used for score evaluation.

  The data evaluation apparatus 100 according to the present embodiment can be realized, for example, by causing one or a plurality of computers to execute a program describing the processing content described in the present embodiment. That is, the function of the data evaluation apparatus 100 is realized by executing a program corresponding to the process executed by the data evaluation apparatus 100 using hardware resources such as a CPU, a memory, and a hard disk built in the computer. Is possible. The above-described program can be recorded on a computer-readable recording medium (portable memory or the like), stored, or distributed. It is also possible to provide the program through a network such as the Internet or electronic mail.

(Processing procedure executed by the data evaluation apparatus 100)
FIG. 3 is a flowchart showing a procedure of processing executed by the data evaluation apparatus 100. The processing contents executed by the data evaluation device 100 will be described in detail below along the procedure of the flowchart shown in FIG.

<Step S101: Data Input>
First, data is input from each of the data source 1 and the data source 2 by the input unit 101 of the data evaluation apparatus 100. Hereinafter, the data of the data source 1 is DATA ₁ and the data of the data source 2 is DATA ₂ .

Each of DATA ₁ and DATA ₂ is time-series data including, for example, network quality information and time information indicating the time when the quality information is acquired. The time information may be the actual time when the quality is measured. For example, in the case where “average quality every hour” is provided from the data source, the time information is the time (0:00, 1 o'clock, etc.). Information other than these may be used.

FIG. 4 shows an example of input data. FIG. 4 shows a case where both DATA ₁ and DATA ₂ are data for each hour from January 1 (1/1) to June 30 (6/30). As an example, in the column below “2” of 1/1, there are values such as average quality at 1/1 AM2 to AM3. In order to make the distinction between DATA ₁ and DATA ₂ easier to understand, DATA ₁ is indicated by a solid line table, and DATA ₂ is indicated by a dotted line table. The same applies to the other drawings.

  Alternatively, the input unit 101 or other functional unit may delete the singular value in the input data and use the input data from which the singular value is deleted for subsequent processing.

<Step S102: Generate Time Series Data by Interval>
Next, Interval time-series data generating unit 102, for each of the DATA ₁ and DATA ₂ input at step S101, generates the time-series data by Interval. The interval used is, for example, day, week, month, etc. That is, for each of DATA ₁ and DATA ₂ , time series data such as daily, weekly, monthly, etc. is generated as time series data by interval. As an example, in the case of every week, an average value (may be a representative value) of data for one week from the start time in DATA ₁ is calculated, and thereafter, an average value for each week is calculated to obtain time series data. . The same applies to DATA ₂ .

Here, in the above example, the type of interval is 3, but the number of types of interval is not limited to 3. Assuming that the number of interval types is N, the index indicating the interval type is k, and the time-series data by interval is expressed as ts _k , in step S102, for each of DATA ₁ and DATA ₂ , ts ₁ , ts ₂ , ... ts _N is generated. Here, for DATA ₁ , DATA ₁ ts ₁ , DATA ₁ ts ₂ ,. DATA ₁ represents a ts _N, for DATA _{2 are, DATA 2 ts 1, DATA 2} ts 2, .... DATA ₂ ts _N.

FIG. 5 shows an example of generated data when generating time-series data for each day, week, and month from each of DATA ₁ and DATA ₂ . As shown in FIG. 5, DATA ₁ ts ₁ and DATA ₂ ts ₁ are generated as time series data for each day, and DATA ₁ ts ₂ and DATA ₂ ts ₂ are generated as time series data for each week. DATA ₁ ts ₃ and DATA ₂ ts ₃ are generated as time series data.

<Step S103: Score Calculation for Time Series Data by Interval>
Next, the score calculation unit 104 compares the time-series data for each interval of DATA _{1 and} the time-series data for each interval of DATA ₂ for each interval used in step S102 (that is, calculates a correlation), A correlation coefficient is calculated, and a score is calculated from this. The calculated score is stored in the data storage unit 107. For the calculation of the correlation, a general correlation function for calculating the correlation between the two time series data can be used. It shows that the similarity between the time series data compared is so high that the calculated score (for example, the value of -1 to 1) is large.

More specifically, as shown in FIG. 5, daily, weekly, if the time-series data of each month, the DATA ₁ ts ₁ and DATA ₂ ts ₁ is a time series data of daily DATA ₁ and DATA ₂ Are compared to calculate a score (p1). Also, the score (p2) is calculated by comparing DATA ₁ ts ₂ and DATA ₂ ts ₂ which are the time series data of DATA ₁ and DATA ₂ for each week. Furthermore, the score (p3) is calculated by comparing DATA ₁ ts ₃ and DATA ₂ ts ₃ which are time series data of DATA ₁ and DATA ₂ every month.

  In FIG. 3 to FIG. 5, since the number N of types of intervals is 3, three scores p1, p2, and p3 are obtained. In general, N scores are obtained.

<Step S104: Constrained time series data generation>
Next, the constrained time series data generation unit 103 generates constrained time series data from the input data. Here, a common attribute between DATA ₁ of data source ₁ and DATA _{2 of} data source 2 is determined in advance, and when the constraint is applied to each of DATA ₁ and DATA ₂ using the attribute. Generate series data. For example, if DATA ₁ has 35 attributes, DATA ₂ has 12 attributes, and there are three common attributes, constrained time series data is generated for each of the three attributes. can do.

  Common attributes in the present embodiment are, for example, attributes related to time in time-series data. For example, time in a day, time zone in a day (eg, time zone every 6 hours), There are days of the week, weekdays / weekends, etc. Moreover, it is good also as an attribute based on the content of data. For example, when data has signal strength as quality, signal strength can be used as an attribute. In this case, for example, the strength of the signal strength is divided into stages (sections), and a chunk described later is generated. The common attribute may be an attribute included in data obtained from the data source, or may be an attribute added by processing data obtained from the data source.

More specifically, for example, when the common attribute is “time in one day”, the constrained time-series data generation unit 103 sets the data for each of DATA ₁ and DATA ₂ into segments for each hour. Divide. That is, in this case, data from AM 0 to AM 1, data from AM 1 to AM 2,. The data is divided into 24 segments, such as data from 10 PM to 11 PM and data from 11 PM to AM 0. If the entire data is, for example, data for 6 months, for example, “data from AM0 to AM1” includes “data from AM0 to AM1” for each day for 6 months. It will be.

As another example, for example, if the common attribute is “day of the week”, for each of DATA ₁ and DATA ₂ , the data is as follows: Monday data, Tuesday data,. Divided into 7 segments. If the entire data is, for example, data for six months, for example, “Monday data” includes data for each Monday for six months.

In the present embodiment, a segment obtained by dividing data as described above is called a “chunk” and is represented by c. And, for example, data from AM0 to AM1, data from AM1 to AM2,. When the data is divided into 24 chunks, such as data from PM10: 00 to PM11: 00 and data from PM11: 00 to AM0, these chunks are c ₁ , c ₂ ,..., C _{24 in} chronological order. It expresses. In general, when divided into m number of _chunks, c 1, _{c 2,} ..., it can be expressed as _{c m.}

As described above, after dividing each of DATA ₁ and DATA ₂ into chunks, aggregation is performed for each chunk. Note that the correlation between chunk data may be obtained without aggregation.

  Aggregation in the present embodiment is executed by generating time-series data by interval described in step S102. For example, assuming that the entire data is data for 6 months and the attribute is “time in a day” and the data is divided into 24 chunks, each chunk (eg, “AM0 hour to AM1 hour”) As shown in step S102, time-series data such as daily, weekly, monthly, etc. is generated.

For example, when the attribute is a "time of day", time-series data of daily (ts ₁₎ in the chunk _{c 1} DATA ₁ ( "o'clock AM0 o'clock ~AM1") _is, DATA 1 _ts 1 _{c 1} It is expressed as More generally, assuming that the interval type is N and the number of chunks is m, in step S104, the constrained time-series data generation unit 103 converts DATA ₁ to “DATA ₁ ts ₁ c ₁ , DATA ₁ ts ₁ c _2. , ... DATA ₁ ts _{1 cm} , DATA ₁ ts ₂ c ₁ , DATA ₁ ts ₂ c ₂ , ... DATA ₁ ts _{2 cm} , ... DATA ₁ ts _N c ₁ , DATA _{1 ts N c 2, .... DATA 1} ts N c m "generates, from DATA _{2" DATA 2 ts 1 c 1, DATA} 2 ts 1 c 2, .... DATA 2 ts 1 c m, DATA ₂ ts ₂ c ₁ , DATA ₂ ts ₂ c ₂ , ... DATA ₂ ts _{2 cm} , ... DATA ₂ ts _N c ₁ , DATA ₂ ts _N c ₂ , ... DATA ₂ ts _N c _m "the It is formed. The above example is a case where there is one attribute. However, when generating constrained time-series data for a plurality of attributes (M), M sets of the above data are generated.

6 to 8 show examples of constrained time series data when the attribute is “time in one day”. FIG. 6 shows time-series data aggregated every day after the chunk division. For example, DATA ₁ ts ₁ c ₁ is obtained by obtaining an average value for each day of data in the chunk of DATA ₁ from AM0 to AM1 and making it time-series data.

FIG. 7 shows time-series data aggregated every week after chunk division. For example, DATA ₁ ts ₂ c ₁ is obtained by obtaining an average value for each week of the data in the chunk of DATA ₁ from AM0 to AM1 and making it time-series data. FIG. 8 shows time-series data aggregated every month after chunk division. For _example, DATA _{1 ts} 3 _c 1 is obtained by the time-series data to determine the average value for each month for the data in the chunk at AM0 o'clock ~AM1 of DATA _1.

<Step S105: Score Calculation for Restricted Time Series Data>
Next, the score calculation unit 104 calculates the score by calculating the correlation between the data sources for each constrained time-series data obtained in step S104. In other _words, the score between the DATA _{1 ts} 1 _c 1 and DATA ₂ ts ₁ c _1, the score between the DATA 1 _ts 1 _{c 2} and DATA ₂ ts ₂ c _2, .... Score between DATA ₁ ts _{1 cm} and DATA ₂ ts _{1 cm} , score between DATA ₁ ts ₂ c ₁ and DATA ₂ ts ₂ c ₁ , ..., DATA ₁ ts _N c The score between _m and DATA ₂ ts _{N cm} is calculated. The correlation calculation method is the same as the calculation method in step S103.

Then, the chunk having the maximum score is obtained for each interval, and the score and the information of the chunk from which the score is obtained (information indicating which section of the chunk) is stored in the data storage unit 107. For example, the attribute is For a time of day, for each day, a score between DATA ₁ ts ₁ c ₁ and DATA ₂ ts ₁ c ₁ , a score between DATA ₁ ts ₁ c ₂ and DATA ₂ ts ₂ c ₂ , .... 24 scores are obtained, such as a score between DATA ₁ ts _{1 cm} and DATA ₂ ts ₁ c _24, and the maximum value of these is obtained (for example, chunk at 3 PM to 4 PM) And the information and the score thereof are stored in the data storage unit 107. The same applies to weekly data and monthly data.

  For example, as shown in FIG. 6, for the time-series data for each day, a score is obtained for each chunk to obtain a maximum value p4. In the example shown in FIG. 7, for the time-series data for each week, a score is obtained for each chunk to obtain the maximum value p5. Further, in the example shown in FIG. 8, a score is obtained for each chunk of the time series data for each month, and the maximum value p6 is obtained.

  In FIGS. 6 to 8, since the number N of types of intervals is 3, three maximum values p4, p5, and p6 are obtained. In general, N maximum values are obtained. N may be 1.

  In addition, the above example is a case where there is one attribute. However, when generating constrained time-series data for a plurality of attributes (M), since M score sets are generated, the maximum N × M value scores are obtained.

  The score calculation for the time series data by intervals described in step S103 may be performed together with the score calculation for the time series data with constraints after generating the time series data by intervals and the time series data with constraints. Moreover, it is good also as not performing the score calculation about the time-sequential data classified by interval demonstrated by step S103. In this case, in the example described in p1 to p6, the following evaluation is performed using only p4, p5, and p6. In this case, if N is 1, the following evaluation is performed only for one score (eg, p4).

<Step S106: Score Evaluation>
Next, the score evaluation unit 105 evaluates the score stored in the data storage unit 107 by the processing so far.

  As described above, the number of intervals is N (eg, daily, weekly, monthly, N = 3), and the number of attributes is M (eg, “time in one day”). In this case, if M = 1), N scores are obtained in step S103, and N × M scores are obtained as evaluation target scores in step S106. Therefore, the total number of scores is N + N × M (= N (1 + M)).

  The score evaluation unit 105 selects the maximum score from N (1 + M) scores. The selected score, and the interval and chunk information (constraint information) from which the score is obtained are output targets.

  Also, a certain threshold value (or benchmark) is determined, and it is determined whether or not the maximum score exceeds the threshold value. If so, the maximum score, the interval related to the score, and chunk information (constraint information) May be determined as an output target.

  In addition, N (1 + M) scores are sorted in descending order (from largest to smallest), and all scores that exceed the threshold (benchmark), and interval and chunk information (constraint information) related to the score are output. It may be determined as a target of. Alternatively, a predetermined number P may be determined in advance, and the top P scores among all scores exceeding the threshold (benchmark) may be determined as output targets.

  5 to 8 (interval = (daily, weekly, monthly), attribute = (time in one day)), p1, p2,. Since six scores of p6 (N (1 + M) = 3 (1 + 1) = 6) are obtained, for example, the maximum score (or the score exceeding the threshold (benchmark)) among these is the output target score. It is determined.

  In the above example, it is assumed that a score with a large positive correlation is output, but this is an example, and a score with a large correlation in the negative direction (that is, the absolute value of the negative correlation coefficient) A score with a large value) may be output.

<Step S107: Data output>
In step S107, the output unit 106 outputs the score determined as the output target in step S106, the interval at which the score is obtained, and the chunk information. In addition to these, time-series data from which the score is obtained may be output. In this case, for example, in the example shown in FIG. 5 to FIG. 8, when the score in the chunk from weekly PM3 to PM4 is determined as an output target, DATA ₁ ts ₂ c ₁₆ and DATA ₂ ts ₂ c ₁₆ is output. In this case, chunk data before aggregation for each week may be output. Note that since input data, time-series data obtained in the course of processing, and the like are stored in the data storage unit 107, they can be output by reading the data from here. However, unused data may be deleted from the data storage unit 107.

  In the processing described so far, when data is correlated between data sources (raw data, time-series data by interval, or time-series data with constraints) may be shifted in the time direction. In addition, when a high correlation coefficient (eg, a correlation coefficient exceeding the benchmark) is obtained, the shifted time length may be output.

Here, shifting in the time direction means that, for example, in the case of raw data, the time (time stamp) of one acquired data between the data sources to be compared is increased (or decreased) by a predetermined time by the input unit 101. The processing described so far is performed by using, as input data, data with a time stamp that is increased (or decreased) by the predetermined time. Further, in the case of time-series data by interval, the time-series data generation unit by interval 102 sets the time of one data to a predetermined time, for example, with an interval (eg, day, DATA ₁ ts _{1 in} FIG. 3) as a unit. The correlation coefficient is calculated between the data that has been increased (or decreased) and the data that has been increased (or decreased) and the other data. As an example, when the time of DATA ₁ ts ₁ in FIG. 3 is shifted by two days, the data in the 1/1 column before the shift is moved to the 1/3 column, and the data in the 1/2 column before the shift is , Move to the quarter column, and so on. Alternatively, the time may be changed such that 1/3 is changed to 1/1 without changing the data position. The same applies to other intervals and constrained time-series data.

  When including processing to shift in the time direction, for example, time lengths with a plurality of directions (for example, for data source 1, increase A time, decrease A time, increase / decrease 0 time, increase B time, decrease B time) Prepare and perform processing for each time length (increase / decrease in 0 hours, that is, including processing in the case of not shifting as described above), and perform steps for the entire score obtained by all processing The score evaluation process described in S106 is performed. As for the output, the time length used in the shifting process is output together with the interval and chunk information (constraint information) at which the score determined as the output target is obtained.

(Example when using 3 or more data sources)
Although the case where the number of data sources is two has been described so far, this is an example, and the number of data sources may be three or more. A processing example in this case will be described below.

Here, when there are X data sources, the number of combinations for selecting two from X (pairs to be compared) is _X C ₂ = X! / (X-2)! 2! It is. For example, if X = 4, the number is ₄ C ₂ = 4! / 2! 2! = 6. In the following, it is assumed that data of four data sources A, B, C, and D is used with X = 4.

  In this case, when the data evaluation apparatus 100 acquires A, B, C, and D, AB, AC, AD, BC, BD, and CD are selected as combinations of two selected from A, B, C, and D. Know that there is.

  The data evaluation apparatus 100 calculates a score of N (M + 1) for each combination by performing the same process as the process described so far. Here, the values of N (number of intervals) and M (number of attributes) may be the same or different for each data source.

  Then, for each combination, the data evaluation apparatus 100 checks whether there is a score that exceeds a threshold (or benchmark, the same applies hereinafter) among N (M + 1) scores, and a combination having a score that exceeds the threshold To extract.

  For example, assuming that a score exceeding a threshold value is obtained in three combinations of AB, CD, AD among AB, AC, AD, BC, BD, and CD, the data evaluation apparatus 100 has already obtained a score for each combination. Output the explained data. For example, for the combination AB, an interval at which a score exceeding the threshold is obtained, and constraint information (chunk information) are output. In addition to these, time-series data of intervals / constraints from which the score is obtained may be output.

(Summary of embodiment)
As described above, according to the present embodiment, the first data and the second data obtained from different data sources, the first data and the second data having overlapping attributes are obtained. Based on the input means for inputting and the attribute, each of the first data and the second data is divided into a plurality of sections, and data in the same section in the first data and the second data And calculating a correlation coefficient between the plurality of sections, determining a correlation coefficient satisfying a predetermined condition from the correlation coefficients obtained for each of the plurality of sections, and corresponding to the correlation coefficient There is provided a data evaluation device including an evaluation unit that selects a section and outputs information indicating the section.

  For example, for each of the first data and the second data, the calculation unit aggregates a plurality of data included in each section obtained by dividing the plurality of sections into a plurality of sections and aggregates the data. The correlation coefficient is calculated between data.

  The calculating means uses a plurality of types of periods as the predetermined period, calculates a correlation coefficient for each section in the plurality of sections for each type of period, and the evaluating means calculates the period type In addition, among the correlation coefficients obtained for each section, a correlation coefficient satisfying a predetermined condition is determined, a period type and a section corresponding to the correlation coefficient are selected, and information indicating the period and section May be output.

  The calculation unit aggregates the first data and the second data for each predetermined period without dividing the first data and the second data, and aggregates the first data and the aggregated data. The evaluation means calculates the correlation coefficient obtained for each section, the first data on which the aggregation is performed, and the aggregation Of the correlation coefficients calculated with the second data that has been performed, a correlation coefficient that satisfies a predetermined condition may be determined.

  The predetermined condition is, for example, that the correlation coefficient is larger than a predetermined threshold. Further, the first data and the second data are data having time as an attribute, for example. The evaluation unit may output the first data and the second data of a portion corresponding to a section in which a correlation coefficient satisfying a predetermined condition is obtained.

(Section 1)
Input means for inputting the first data and the second data, the first data and the second data obtained from different data sources, having overlapping attributes;
Based on the attribute, each of the first data and the second data is divided into a plurality of sections, and the correlation coefficient between the data in the same section in the first data and the second data Calculating means for calculating
Among correlation coefficients obtained for each section in the plurality of sections, a correlation coefficient satisfying a predetermined condition is determined, a section corresponding to the correlation coefficient is selected, and information indicating the section is output. Evaluation means
A data evaluation apparatus comprising:
(Section 2)
The calculation means aggregates a plurality of data included in each section obtained by dividing the plurality of sections into the plurality of sections for each of the first data and the second data, and aggregates the data. The correlation coefficient is calculated between
2. The data evaluation device according to item 1, wherein
(Section 3)
The calculation means uses a plurality of types of periods as the predetermined period, calculates a correlation coefficient for each section in the plurality of sections for each type of period,
The evaluation means determines a correlation coefficient satisfying a predetermined condition among the correlation coefficients obtained for each type of the period and for each section, and determines the type and section of the period corresponding to the correlation coefficient. Select and output information indicating the period and section
The data evaluation apparatus according to item 2, characterized in that:
(Section 4)
The calculation unit aggregates the first data and the second data for each predetermined period without dividing the first data and the second data, and aggregates the first data and the aggregated data. The correlation coefficient is calculated with the second data for which
The evaluation means includes a correlation coefficient obtained for each section and a correlation coefficient calculated between the first data on which the aggregation is performed and the second data on which the aggregation is performed. To determine the correlation coefficient that satisfies the predetermined condition.
4. The data evaluation device according to any one of items 1 to 3, wherein
(Section 5)
The evaluation unit outputs the first data and the second data of a portion corresponding to a section in which a correlation coefficient satisfying a predetermined condition is obtained.
The data evaluation device according to any one of Items 1 to 4, wherein the data evaluation device is characterized in that
(Section 6)
A data evaluation method executed by a data evaluation device,
An input step of inputting first data and second data, which are first data and second data obtained from different data sources, having overlapping attributes;
Based on the attribute, each of the first data and the second data is divided into a plurality of sections, and the correlation coefficient between the data in the same section in the first data and the second data A calculation step for calculating
Among correlation coefficients obtained for each section in the plurality of sections, a correlation coefficient satisfying a predetermined condition is determined, a section corresponding to the correlation coefficient is selected, and information indicating the section is output. Evaluation steps and
A data evaluation method comprising:
(Section 7)
A program for causing a computer to function as each unit in the data evaluation device according to any one of Items 1 to 5.
The present invention is not limited to the above-described embodiments, and various modifications and applications are possible within the scope of the claims.

DESCRIPTION OF SYMBOLS 1, 2 Data source 100 Data evaluation apparatus 101 Input part 102 Time-sequential data generation part 103 by intervals 103 Time series data generation part with restrictions 104 Score calculation part 105 Score evaluation part 106 Output part 107 Data storage part

Claims (7)

A first network quality information obtained from different data sources and the second network quality information, input for inputting the first network quality information and a second network quality information having a signal strength which is a duplicate attribute Means,
Based on the signal strength , each of the first network quality information and the second network quality information is divided into a plurality of sections, and the same in the first network quality information and the second network quality information . A calculation means for calculating a correlation coefficient between the network quality information of the sections;
Among correlation coefficients obtained for each section in the plurality of sections, a correlation coefficient satisfying a predetermined condition is determined, a section corresponding to the correlation coefficient is selected, and information indicating the section is output. A data evaluation device comprising: an evaluation means. It said calculation means for each of said first network quality information and said second network quality information, a plurality network quality information included in each interval obtained by dividing the plurality of sections for every predetermined time period The data evaluation apparatus according to claim 1, wherein the correlation coefficient is calculated between the aggregated network quality information . The calculation means uses a plurality of types of periods as the predetermined period, calculates a correlation coefficient for each section in the plurality of sections for each type of period,
The evaluation means determines a correlation coefficient satisfying a predetermined condition among the correlation coefficients obtained for each type of the period and for each section, and determines the type and section of the period corresponding to the correlation coefficient. The data evaluation apparatus according to claim 2, wherein the information is selected and information indicating the period and the section is output. The calculating means aggregates the first network quality information and the second network quality information for each predetermined period without dividing the first network quality information and the second network quality information . A correlation coefficient between the network quality information of the second network quality information and the aggregated second network quality information ,
The evaluation unit includes a correlation coefficient obtained for each of the sections, the phase calculated between the first second network quality information network quality information and the aggregate has been performed in which the aggregate is performed The data evaluation apparatus according to any one of claims 1 to 3, wherein a correlation coefficient that satisfies a predetermined condition is determined among the number of relations. The said evaluation means outputs the said 1st network quality information and the said 2nd network quality information of the part corresponding to the area from which the correlation coefficient which satisfy | fills a predetermined condition was obtained. 4. The data evaluation device according to any one of 4 above. A data evaluation method executed by a data evaluation device,
A first network quality information obtained from different data sources and the second network quality information, input for inputting the first network quality information and a second network quality information having a signal strength which is a duplicate attribute Steps,
Based on the signal strength , each of the first network quality information and the second network quality information is divided into a plurality of sections, and the same in the first network quality information and the second network quality information . A step of calculating a correlation coefficient between network quality information of the sections, and determining a correlation coefficient satisfying a predetermined condition among the correlation coefficients obtained for each section in the plurality of sections; An evaluation step of selecting an interval corresponding to the number and outputting information indicating the interval.   The program for functioning a computer as each means in the data evaluation apparatus of any one of Claims 1 thru | or 5.

Images