Information processing device, information processing method, and program

Abstract

An information processing device 1 comprises: a prior quantification unit 11 that finds a scalar vij obtained by quantifying a cross-correlation function between time-series data of items; a similarity calculation unit 12 that finds a semantic similarity uij indicating semantic proximity between items; and an unexpectedness calculation unit 13 that finds the degree of unexpectedness of a combination of items on the basis of the position of a point indicating the combination of the items on a plane where the axes are the scalar vij and the semantic similarity uij.

Description

[Technical field]

[0001] The present invention relates to an information processing device, an information processing method, and a program. [Background technology]

[0002] One of the roles of data science is to extract business intelligence from data. In order for data scientists to be able to make better proposals to customers, it is necessary to support them in acquiring a wide range of knowledge. In other words, it is expected that data scientists will be able to extract objective evidence from data that they would not intuitively think of, enabling them to derive unexpected business intelligence. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6620950 Summary of the Invention [Problem to be solved by the invention]

[0004] For example, electricity prices tend to rise and fall in tandem with gasoline prices several months later. This relationship between electricity prices and gasoline is obvious, but there is a possibility that precedence relationships exist between items that are not obvious, that is, items that are far apart in meaning. By finding unexpected precedence relationships that people would not think of or would find difficult to find, it is expected that this information can be used to formulate unexpected joint sales plans and pricing strategies.

[0005] Cross-correlation functions (CCFs) are a method for expressing precedence relationships between time series variables. In Patent Document 1, cross-correlation is used to learn word vectors from past data to be analyzed, but it does not find unexpected precedence relationships that people would not think of or would find difficult to find. In other words, it does not simultaneously consider heterogeneous things such as time series and word meanings.

[0006] The present invention has been made in view of the above, and has an object to extract a combination of items that unexpectedly have a preceding relationship in a chronological order. [Means for solving the problem]

[0007] An information processing device according to one aspect of the present invention includes a precedence quantification unit that calculates a scalar that quantifies a cross-correlation function between time-series data of items, a similarity calculation unit that calculates a semantic similarity that indicates semantic closeness between items, and an unexpectedness calculation unit that calculates an unexpectedness of a combination of items based on a position of a point that indicates the combination of items on a plane whose axes are the scalar and the semantic similarity. and an item extraction unit that extracts items by calculating a score based on the degree of surprise for each item. Equipped with.

[0008] In one aspect of the information processing method of the present invention, a computer calculates a scalar that quantifies a cross-correlation function between time series data of items, calculates a semantic similarity that indicates semantic closeness between items, and calculates a degree of surprise of a combination of items based on a position of a point that indicates the combination of items on a plane whose axes are the scalar and the semantic similarity. For each item, a score based on the unexpectedness is calculated to extract the item. . Effect of the Invention

[0009] According to the present invention, it is possible to extract a combination of items that unexpectedly have a preceding relationship in a time series. [Brief description of the drawings]

[0010] [Figure 1] FIG. 1 is a functional block diagram showing an example of the configuration of an information processing device according to the present embodiment. [Diagram 2] FIG. 2 is a flowchart showing an example of a process flow of the information processing device. [Diagram 3] FIG. 3 is a diagram illustrating an example of time-series data. [Figure 4] Figure 4 shows the time series data when the lag is set to -2, plotted on a plane. [Diagram 5] FIG. 5 is a diagram showing an example of the calculated cross-correlation function. [Figure 6] Figure 6 shows a graph in which the strength of correlation and semantic similarity are plotted on a plane. [Figure 7] FIG. 7 is a diagram showing an example of calculating the degree of surprise using the inner product of vectors. [Figure 8] FIG. 8 is a diagram illustrating an example of a hardware configuration of an information processing device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0012] [Configuration of information processing device] An example of the configuration of the information processing device of this embodiment will be described with reference to Fig. 1. The information processing device 1 is a device that extracts, from among a large number of items, items that move ahead even if their meaning is distant. The information processing device 1 includes a precedence quantification unit 11, a similarity calculation unit 12, an unexpectedness calculation unit 13, an item extraction unit 14, and a user interface 15.

[0013] The precedence quantification unit 11 obtains a scalar (representative value) that quantifies the precedence between the time series data of the items. More specifically, the precedence quantification unit 11 obtains the cross-correlation function of the time series data x and y of the items i and j, respectively, and obtains the representative value v of the obtained cross-correlation function. ij Calculate the representative value v ij is an arbitrary statistic of the cross-correlation function, which represents the strength of correlation between items i and j. Below, the representative value v ij to a scalar v ij or the strength of the correlation v ij It is also sometimes called.

[0014] The similarity calculation unit 12 calculates the semantic closeness (semantic similarity) between items. More specifically, the similarity calculation unit 12 calculates the semantic vectors of items i and j, calculates the cosine similarity of the calculated semantic vectors, and calculates the semantic similarity u between items i and j. ij Let us assume that.

[0015] The surprise degree calculation unit 13 calculates the surprise degree between items from the strength of correlation between items and the semantic similarity. More specifically, the surprise degree calculation unit 13 calculates the strength of correlation between items i and j on a plane whose axes are the strength of correlation and the semantic similarity. ij and semantic similarity u ij The point (u ij ,v ij ) and plot the point (u ij ,v ij ) based on the position on the plane, the degree of surprise between items i and j is r ij For example, the unexpectedness calculation unit 13 calculates the center point μ(μ u ,μ v ) to point (u ij ,v ij ) based on the distance between items i and j, ij A cluster is a set of points obtained by plotting the strength of correlation and semantic similarity between many items. In this embodiment, for each combination of N items, the strength of correlation between items i and j, v ij and semantic similarity u ij , and draw a point (u ij ,v ij ) is plotted, where 1≦i, j≦N. Since the further away from the center of the group the more unexpected it is, the unexpectedness calculation unit 13 increases the unexpectedness as the distance from the center point increases.

[0016] The unexpectedness calculation unit 13 calculates the unexpectedness of the origin (0,0) or the center point μ(μ u ,μ v For example, the surprise degree calculation unit 13 extracts only points that have a positive correlation strength and a negative semantic similarity from the reference point.

[0017] The item extraction unit 14 calculates a score for each item based on the degree of surprise relative to other items, and extracts items with high scores.

[0018] The user interface 15 is equipped with a display means and an input means and provides an interface to the user. For example, it presents the degree of surprise calculated by the surprise degree calculation unit 13 to the user, accepts from the user a selection of a method for calculating the degree of surprise, displays the score calculated by the item extraction unit 14, and displays information on items extracted by the item extraction unit 14.

[0019] [Operation of information processing device] Next, an example of the flow of processing of the information processing device 1 of this embodiment will be described with reference to the flowchart of FIG.

[0020] In step S11, the leading quantification unit 11 converts the time series data x of item i and the time series data y of item j into a rate of change series x', y'. Time series data is a predetermined type of data for an item that varies along a time axis. The time series data is, for example, economic indicators such as prices. Economic indicators are often unit root processes, and there is a problem that spurious regression occurs when unit root processes are regressed against each other. To avoid this, the leading quantification unit 11 converts the original series x, y into a rate of change series x' t =(x t -x t-1 ) / x t-1 ,y' t =(y t -y t-1 ) / y t-1 Alternatively, the precedence quantification unit 11 converts the original sequence x, y into a difference sequence Δx t =x t -x t-1 ,Δy t =y t -y t-1 In this way, by considering the time series data in terms of the rate of change (difference), it is possible to detect items that undergo similar changes. Note that the leading quantification unit 11 may proceed to step S12 using the original time series data x, y as is without performing the process of step S11. The time series data may be an index other than an economic index. Hereinafter, the time series data x, y is assumed to be either the original series x, y, the rate of change series x', y', or the difference series Δx, Δy.

[0021] In step S12, the precedence quantification unit 11 calculates a cross-correlation function between the time series data x and the time series data y. xy (k) can be calculated using the following equation (1).

[0022]

number

[0023] Cross-correlation function R xy (k) is the correlation coefficient between the time series data x and the time series data y when the time series data y is shifted by time k. -1≦R xy (k)≦1. Unlike dynamic time warping (DTW), the cross-correlation function expresses leading and lagging, and is therefore directly linked to the predictability of time series. Therefore, the cross-correlation function indicates that the time series data y leads the time series data x by a significant amount (when k is negative and small, R xy It is also possible to extract those with large (k) values.

[0024] Now, referring to Figs. 3 to 5, the cross-correlation function R xy The calculation of (k) will be explained. The solid line in Figure 3 is the time series data x, and the dashed line is the time series data y. The cross-correlation function R when lag k = -2 xy When calculating (-2), as shown in Figure 4, x at time t t and y at time t-2 t-2 The point (x t ,y t-2 ) on the plane, i.e., the point (x 3 ,y 1 ), point(x 4 ,y 2 ), point(x 5 ,y 2 ) are plotted. These x t andy t-2 The correlation coefficient a is calculated using the following equation (2).

[0025]

number

[0026] where x (bar above) is the average of x, and y (bar above) is the average of y. t-2 The correlation coefficient a is the cross-correlation function R xy (-2)=a. By changing the value of k and calculating the correlation coefficient for each k, the cross-correlation function R xy Find (k).

[0027] In step S13, the precedence quantification unit 11 obtains a representative value of the cross-correlation function. Since the cross-correlation function is a function of the lag k, any statistic of the cross-correlation function value in a predetermined interval (-L≦k≦+L) shown in any one of the following formulas (3) to (6) is calculated to obtain the representative value v of the cross-correlation function. ij Let us assume that.

[0028]

number

[0029] Equation (3) is the average value of Rxy(k) for -L≦k≦+L. Equation (4) is the maximum value of Rxy(k) for -L≦k≦+L. These average and maximum values ​​can be considered as simple representative values ​​of the relationship between time series data x and time series data y.

[0030] Equation (5) is the standard deviation of Rxy(k) for -L≦k≦+L. A small standard deviation suggests that there is a high correlation at a specific lag. In other words, if you shift the time series data y by k, you can capture data that roughly matches the shape of the time series data x. On the other hand, a relatively large standard deviation suggests that the time series data x and y both have waveforms that move with similar periods.

[0031] Equation (6) is the kurtosis of Rxy(k) for -L≦k≦+L. A large kurtosis suggests a high correlation at a specific lag k. In other words, if you shift the time series data y by k, you can capture data that is almost identical in shape to the time series data x.

[0032] It should be noted that a statistical value other than the above may be used as the representative value.

[0033] In step S14, the similarity calculation unit 12 calculates semantic vectors (distributed representations) of the items. For example, the similarity calculation unit 12 calculates semantic vectors of items i and j using Word2vec or ontology.

[0034] In step S15, the similarity calculation unit 12 calculates the similarity of the semantic vectors between the items, and sets this as the semantic similarity between the items. That is, the similarity u between items i and j ij can be calculated using the cosine similarity of the following equation (7). ij In addition to the cosine similarity, an index representing distance or similarity can be used.

[0035]

number

[0036] Here, P(→up) is the semantic vector of item i and Q(→up) is the semantic vector of item j.

[0037] The precedence quantification unit 11 and the similarity calculation unit 12 perform the process up to step S15 for each combination of N items, and calculate the strength of correlation v ij and semantic similarity u ij Request.

[0038] In step S16, the unexpectedness calculation unit 13 calculates the center point of the group. u ,μ v ) can be calculated using the following equation (8).

[0039]

number

[0040] Figure 6 shows a graph in which the strength of correlation and semantic similarity between pairs of items are plotted on a plane, with the horizontal axis representing semantic similarity and the vertical axis representing the strength of correlation, and the center point is determined.

[0041] In step S17, the unexpectedness calculation unit 13 calculates the unexpectedness of the pair of items based on the distance from the center point. The unexpectedness calculation unit 13 calculates the unexpectedness of the pair of items based on the distance from the center point. ij ,v ij ) and the center point μ(μ u ,μ v ) and calculate the unexpectedness r of items i and j. ij Let us assume that.

[0042] The Euclidean distance is calculated using the following formula (9).

[0043]

number

[0044] The Mahalanobis distance is calculated using the following formula (10).

[0045]

number

[0046] From the above, it is possible to extract pairs of items that are out of the center of the group as having a high degree of surprise. When extracting only pairs of items that are leading indicators despite having different meanings, the surprise calculation unit 13 calculates the number of items in the upper left quadrant (u ij <0 & v ij >0) or the upper left quadrant from the center point ((u ij -μ u ) / σ u <0 & (v ij -μ v ) / σ v) can be applied as a filter to extract only those items that are similar in meaning and have a time series correlation. The top right quadrant is an area where the meanings are similar and there is a time series correlation, while the bottom left quadrant is an area where the meanings are dissimilar and there is no time series correlation. Pairs of items that belong to either of these quadrants are obvious combinations. On the other hand, the bottom right quadrant is an area where the meanings are similar but there is no time series correlation, while the top left quadrant is an area where the meanings are dissimilar but there is a time series correlation. Pairs of items that belong to either of these quadrants are highly unexpected combinations. By filtering the pairs of items that belong to the top left quadrant, it is possible to extract combinations that are dissimilar in meaning but have a time series correlation.

[0047] In addition, the representative value of the cross-correlation function v ij By definition, -1≦u ij ≦1, -1≦v ij Since it is ≦1, preprocessing such as normalization or standardization is not required, so the shape of the population is not distorted and it is highly versatile.

[0048] In addition to the Euclidean distance and Mahalanobis distance calculated above, the unexpectedness calculation unit 13 may obtain the unexpectedness of the component in the 45 degree direction from the origin to the top left as shown in Fig. 7. Specifically, the unexpectedness calculation unit 13 may obtain the unexpectedness of the component in the 45 degree direction from the origin to the top left as shown in Fig. 7. Specifically, the unexpectedness calculation unit 13 may obtain the unexpectedness of the component in the 45 degree direction from the origin to the top left as shown in Fig. 7. ij ,v ij ) is the degree of surprise for items i and j, r ij Basically, -1≦u ij ≦1,-1≦v ij Assume ≦1.

[0049] In the example of Fig. 7, the unit vector e (up →) is a vector that starts at the origin and points 45 degrees to the upper left, but the unit vector e (up →) may be a vector with an angle θ that starts at any point (X, Y), for example, the center point of the group. The angle θ may be set arbitrarily by the user.

[0050] When the process up to step S17 is completed, the user interface 15 may present the user with a screen in which the strength of correlation and the semantic similarity of each pair of items are plotted on a plane. Both the surprise degree calculated using the Euclidean distance and the surprise degree calculated using the Mahalanobis distance may be presented to the user, and the selection of the surprise degree to be used in the item extraction unit 14 may be accepted from the user.

[0051] In step S18, the item extraction unit 14 calculates a score for each item based on the degree of surprise, and extracts items with high scores. i is calculated using the following formula (11). In addition, the item A with the highest score is extracted using formula (12).

[0052]

number

[0053] The user scores S i By referring to this, you can find out which items are leading indicators of many other items, even if their meanings are distant.

[0054] As described above, the information processing device 1 of the present embodiment uses a scalar v ij The semantic similarity u, which indicates the semantic closeness between items, is calculated by the precedence quantification part 11. ij A similarity calculation unit 12 for calculating a scalar v ij and semantic similarity u ij The system further includes a surprise degree calculation unit 13 that calculates the surprise degree of an item combination based on the position of a point indicating the item combination on a plane whose axis is . In this embodiment, by representing a function called a cross-correlation function by a scalar, it becomes possible to simply and quickly combine heterogeneous items, such as item time-series data and item meanings, and to detect items that move ahead even if their meanings are distant.

[0055] The information processing device 1 described above may be, for example, a general-purpose computer system including a central processing unit (CPU) 901, a memory 902, a storage 903, a communication device 904, an input device 905, and an output device 906, as shown in Fig. 8. In this computer system, the information processing device 1 is realized by the CPU 901 executing a predetermined program loaded onto the memory 902. This program can be recorded on a computer-readable recording medium such as a magnetic disk, an optical disk, or a semiconductor memory, or can be distributed via a network. [Explanation of symbols]

[0056] 1. Information processing device 11 Precedence Quantification Department 12 Similarity calculation part 13 Unexpectedness calculation part 14 Item Extraction Section 15 User Interface

Claims

1. a leading edge quantification unit that calculates a scalar that quantifies the cross-correlation function between time series data of items; a similarity calculation unit that calculates a semantic similarity indicating a semantic closeness between items; an unexpectedness calculation unit that calculates an unexpectedness of a combination of items based on a position of a point indicating the combination of items on a plane whose axes are the scalar and the semantic similarity; An item extraction unit that extracts items by calculating a score based on the degree of surprise for each item. Information processing device.

2. 2. The information processing device according to claim 1, The unexpectedness calculation unit calculates the Euclidean distance or Mahalanobis distance from a predetermined reference position to a point indicating the item combination, or a component in an arbitrary direction from the predetermined reference position, and determines the unexpectedness of the item combination. Information processing device.

3. 3. The information processing device according to claim 1, The precedence quantification unit converts the time series data into a rate of change series or a difference series to obtain the scalar. Information processing device.

4. The computer Calculate the scalar that quantifies the cross-correlation function between the time series data of items, Calculate the semantic similarity between items, which indicates the semantic closeness between items. calculating a degree of surprise of the combination of items based on a position of a point indicating the combination of items on a plane having axes of the scalar and the semantic similarity; A score based on the unexpectedness is calculated for each item, and the items are extracted. Information processing methods.

5. A program for causing a computer to operate as each part of the information processing device according to any one of claims 1 to 3.

Images