JPH02213981A - Method for displaying information - Google Patents

Abstract

PURPOSE:To recognize a record indicating abnormal contents and the trend of a record group by finding the similarity between record attributes and arranging record information with higher similarity closer. CONSTITUTION:Attribute information regarding an (i)th record is denoted as a1, information represented as area is denoted as S1, and information represented by the color, pattern, etc., of a division surface is denoted as X1=(x1i, x2i,..., xni). A display image plane is divided into N to display N pieces of record information, similarities r(ai, aj) among the N pieces of record information are found, and the distances among the respective divided areas of the display image plane are shorter and shorter as the similarities are higher and higher. Then, a record group, e.g. product group with high similarities is displayed together in a specific place on a screen. Consequently, a large amount of record information is understood according to mutual position relation and the inclination of the whole, the trend of article groups, and abnormal values are easily recognized.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for collectively displaying a large amount of record information in a computer system on a screen.

[Conventional technology]

Conventional measurement and control Vo 1.27. &10 Facing Control Expert System, the method described by Tenma et al. visually displays record information in a computer system, such as sales information for each product, based on position and size information given in advance to each record. Met. For example, on a schematic image of a product shelf viewed from the front, a color scale is displayed according to product sales (red for high sales,
Low sales have been done by blue etc.). Additionally, sales and inventory efficiency for each department have been visually displayed on the floor layout diagram of a store by filling in the colors.

[Problem to be solved by the invention]

The above conventional technology displays record information based on random display positions, without considering the display method when record information to be displayed in a computer system is not given display position information on the display screen. Since then, there has been a problem in that users cannot recognize the existence of abnormal records or the trends in record groups.

An object of the present invention is to provide a method for displaying record information that can recognize abnormal records and recognize trends in record groups, taking into consideration the mutual positional relationship of record information, even if the position information of a large amount of record information is not given in advance. Our goal is to provide the following.

[Means to solve the problem]

To achieve the above objective, we calculate the degree of similarity between the /g characteristics of each record, divide the screen so that records with high similarity are placed close together, and fill each divided screen in accordance with the color scale of the displayed information. Display.

In particular, in a desirable embodiment of the present invention, in the process of dividing the screen so that records with high similarity are placed close to each other,
The hierarchical relationship between records is determined from the similarity between the attributes of each record, and the screen is divided sequentially from the top of the first hierarchy.

[Effect]

Assume that there are N pieces of record information on the computer system. Consider product information as an example of record information. Each record information includes attribute information and information to be displayed on the screen. The attribute information regarding the i-th record is ai, the information that should be expressed as area is SL, and the information that should be expressed as the color or pattern of the dividing surface is xL" (Xll HX2te "・p
Xn1).

The display screen is divided into N pieces and N pieces of record information are displayed. Similarity r(al, aJ) between N pieces of record information
The distance between each divided area after the display screen is divided is determined so that the higher the similarity, the closer the distance between the divided areas. As a result, a group of records with a high degree of similarity, such as a group of products, will be clustered and displayed in a specific location on the screen. Products with the highest degree of similarity are displayed in areas adjacent to each other. Therefore, it is possible to understand the record information of large areas displayed in patterns based on their mutual positional relationships, and it becomes easy to recognize trends in all items, trends in product groups, and abnormal values.

Furthermore, for each record, the record with the highest degree of similarity is determined, and the hierarchical relationship between the records is thereby determined, and the screen is sequentially divided from the top of the hierarchy.

Figure 3 shows how to find hierarchical relationships between records.
If each record is represented by a node and a directed arc is drawn for the node of record j that gives the largest similarity 1' (am, aa) for each record i, several connected spring digraphs are obtained (the third %(a)), when the whole is connected as one directed graph, find the unique node pair that exists and can be said to be the center of the graph and have the maximum mutual similarity, and use one of them as the root to obtain a tree structure. If the whole is represented by two or more connected graphs (subclusters), the above procedure is repeated only between the roots of each connected graph to obtain a higher-level tree structure. This procedure is repeated until the whole is represented in the tree structure of - (FIGS. 3(a) to 3(d)).

Once a tree-like hierarchical structure can be created among all records, one display screen can be divided into smaller parts by subdividing one display screen from the top of the first hierarchy in accordance with the branching of the tree. do.

Note that since the end of the tree corresponds to N records and has information Si that should be expressed as an area, the total display area of that branch is also given to the branch node of each school. By giving the total area sum ΣS1 to the root and dividing it by the ratio of the total display area of the lower branch nodes when dividing one display screen, it is possible to perform division considering both similarity and display area. Zero division should be done with vertical or horizontal straight lines, and it is best to choose vertical or horizontal lines so that the length of the area after division will be short. Which lower branch nodes are associated with each side of the dividing straight line is determined based on the degree of similarity with records representing one surrounding already divided area.

Note that when there are three or more lower branch nodes, the total display area of each branch node is arranged in descending order, and the display area is divided starting from the largest branch node.

〔Example〕

An embodiment of the present invention will be described below with reference to FIG. To initialize the table, read the number of records N and the display area of each record (Si3) from the record information file 2, and write them in the area column of the hierarchical structure creation table 3 (step 1
), Figure 1 shows the table format of record information file 2 in Figure 7. Record information file 2 has N
The area Si that corresponds to records and should be given when displaying
1 Attribute information a+=(
atl+ at! , "'+a+M)-, display information x1 to be displayed by color, etc. is stored.

In step 4, details of the 6-layer structure creation table 3, where i is the record number, are shown in FIG.

Attribute information for each record, such as product brand, classification code, and color, from the record information file.

Read attribute information (al) such as price, and calculate mutual similarity r
(al+aJ) is calculated using the following formula, and j (=m (i)) giving the highest degree of similarity is determined as the end of the arc, and written in the end column of Table 3.

22, k is the attribute number 2M is the number of attributes, α1 is a constant,
ω8 is a load.

Step 4 further includes the terminal m (i
), set i as the starting point information in Table 3.
Write it in the starting point field. In the example shown in Figure 3, record 1
A directed arc is heading from record 2. There is no arc that goes into Record 1.

Also, record 1 is given as the starting point for record 2. Also, read Table 3 for sub-cluster center detection, which also increments the pointer column by 1 to manage the case where multiple starting points occur, and initializes the flag to indicate that the arc should be passed for total area calculation.

A pair of records such that i=m (m (i) > [i,
m(i)] and set 1 in each upper flag field (step 5) 6 Calculation of area sum Step 6 finds the record whose pointer is O.

Add that total display area to the total display area of the end record and 5
This pointer is set to -1, and the pointer of the end record is also -1.1 or more processing is performed until there are no more records whose pointers are 0. For sub-cluster center records with an upper flag of 1, the one with the larger total display area is selected from the pair, and the total display area is the sum of both.

The smaller higher-order flag is set to O.

To determine the number of sub-clusters, the number of sub-clusters is determined from the sum of the numbers of 1's in the upper flag column (step 7).
In the example of FIG. 3, the number of subclusters is two. If the number of subclusters is greater than 1, initialization for the upper layer is performed, and those with an upper flag of 1 are found in Table 3, and added as subclusters from the N+1mth onward (step 8). That is, the total display area is newly set as the area S i', and the record # is set as the lower order #. In the example of FIG. 3, the total display area for record 3 (S1+S2+ S s+
Sa) is determined to be larger than that regarding record 4 (S40Ss), and the first subcluster 3' has an area Sa'''(Sx+Sz+Ss+S<+83+5l
1) is set in the hierarchical structure creation table 3, and the second sub-cluster 7' is also set in the same way. After initializing one or more upper layers, table 3 is set again from step 4 with the setting of the valid a. The writing process is repeated.

When the number of subclusters is determined to be 1 in step 7, at that stage the table 3 has the tree structure shown in the example in FIG. 3(d).

Therefore, the display screen is divided by starting from the row of table 3 of 3#, which is the root, and tracing the tree based on the lower order # and the starting point record number. That is, the display area shape set sets the vertex coordinates of the area shape of the display screen in the territory apex locus iIAm of Table 3 (Figure 2) (step 9).
1 display area division.

Look at the lower # or starting record number for the root record. If only the lower # is meaningful, the same area apex coordinates are set in the area apex coordinate field of the lower # record. If there is a start point record number, consider a vertical or horizontal straight line to divide the total display area of the start point record from the total display area of the record, and calculate the apex coordinates of the two new display areas (step 10) For the 0-divided area shape set, write the vertex coordinates of the display area obtained in step 10 in the area vertex coordinate wsm of table 3 (step 11). In the search for records that require division, the contents of the region apex coordinates in Table 3 are checked from below, and records that should be further divided are searched (step 12). If there is an area that should be further divided as a result of the search, the process is performed according to (step 13). , the repetition from step 10 is performed. If there is no region to be divided into regions, N pieces of region vertex information in table 3 are sequentially filled in by color filling processing for the divided regions (step 14).The color to be filled is determined by the display information x1 read from the record information file 2. is converted using a color scale. Note that instead of the color scale, 1.
A pattern scale such as an I4 pattern may also be used. The shape data of each filled area is sent to the display bag [15], and when it is determined in step 6 that the display of all eight divided areas has been completed, the entire process ends. In FIG. 4, according to the example of FIG.

This shows how a rectangular display screen is divided. Figure 4(a)
In this case, node 3'' representing the total area of the display screen is node 3'
In FIG. 4(b), the area represented by node 3' is the sum of the areas of nodes 1 and 2, and the area represented by node 3. FIG. 4(c) shows a state where node 7' is divided into node 5 and the sum of areas of nodes 4 and 6. FIG. 4(c) shows a state where node 7' is divided into node 7 and node 8.
d) shows a state in which all nodes 1 to 8 are divided. As described above, according to this embodiment, when a large amount of record information on a computer system is replaced with colors and patterns and the screen is divided and displayed, similar information is placed spatially close to each other, so that abnormalities can occur. It is possible to recognize records that indicate specific content and to recognize trends in record groups. Furthermore, information can also be assigned to area, increasing the types of information that can be obtained from - eyes.

Next, a second embodiment according to the present invention is shown in FIG.

For similarity calculation, the attribute information Ml (a-) is read from the record information file 18 in Figure 7, and the similarity calculated using equation (1) is performed (step 17).However, in the following processing, the similarity will not be used as is. Therefore, the information is converted to the dth distance, which has the same information but the magnitude relationship is reversed and is 0 between the same records.

d l J = (Σtak) r (ai,
aJ) ”・(2) Here, ω is (1
) is the load of the equation.

Below, when N records are placed at N points on a plane, the distance relationship in equation (2) is 2! J! Projection is performed through repeated calculations using the steepest descent method so that the error is minimized.First, before starting the iterative calculations, calculate the parameters used in the calculation and find the initial values of N coordinates on the plane. . That is, in distance parameter calculation step 19, the following is calculated. In step 20, N plane coordinates are set as (yi
t+yta)=(FI IIod(i, M), Q
trunc(i+M)) −(4) Here, M=
trunc(-7M+1), Q = L/M, and are set at equally spaced lattice points and stored in the plane coordinate table 21.

In steepest descent process step 22, D I J = (d, t a - d l J)/
(d□1d1.umbrella) (6) By p=1.2, the old coordinates (yt+(t
Update coordinates (yth (1+1), y
The table format of the plane coordinate table 21, which calculates iz(t+1)) and writes it into the plane coordinate table 21, is as follows.
As shown in FIG. 8, area center coordinates correspond to each record number.

Step 23 determines the difference between the old coordinates and the new coordinates, and if the difference is less than a certain value, it is determined that convergence has been achieved, and the iterative calculation ends. Step 24 of Voronoi diagram division is
As shown in the example in Figure 6, the convergence point (ytz* ytz)
A Voronoi diagram is obtained for each point, and a vertex coordinate string of the polygon of the divided region for each point is stored in the divided region data 25. In Figure 6, 2 of the convergence points of N=17 XEPs
The dimensional position coordinates and the surrounding area (A-E) are shown. Each line segment of the Voronoi diagram is a perpendicular bisector of the convergence point (yit+ysz) at the center of the divided regions on both sides. Therefore, from the points within each divided area, the corresponding convergence point is the closest convergence point.

As shown in FIG. 6, the number of vertices of each area j is −(y1+). For each corresponding divided area, the display information (X,) in the record information file 18 is read out, converted into a color or pattern based on a single color scale or pattern scale, the divided area is filled in, and sent to the display device 127. Due to the above processing.

The display area can be divided so that similar information is arranged spatially close to each other while preserving the similarity relationship to the maximum extent possible.

In the above embodiment, each record information is information such as product turnover rate, profit, sales growth rate, etc. Record attributes include each product's brand, store classification class, manufacturer,
color, size, price, material.

It is an attribute such as a type. It is possible to implement a product information display system that displays product information on thousands of items by subdividing the sales floor area on a sales floor layout diagram including aisles.

There, each product attribute is expressed numerically as a code system. When different numbers, even if they are 1, represent completely different things, it is preferable to choose a large constant αk in equation (1) related to the attribute, and when representing the proximity of numerical values and attributes, αk should be chosen small. Furthermore, by selecting a company with zero weight, it becomes possible to display information that suits the purpose of the user, such as a method of displaying product information that emphasizes brand relationships and a method of displaying product information that emphasizes store classification classes. Note that the shape of the sales floor is generally an irregular polygon, so the outer frame is not necessarily a rectangle as shown in Figure 4 in the first embodiment, but as long as it is a polygon, the area calculation is 3. Since it is easier to do so by calculating the sum of areas of a rectangle, the area division process in step 10 of FIG. 1 can be performed in the same way as for a rectangle.

〔Effect of the invention〕

According to the present invention, in a method of displaying record information to be displayed in a computer system by dividing the display screen corresponding to the records and displaying the information using a color scale, for example, position information on the display screen is not given. Even in the case of record attributes, the degree of similarity between record attributes is determined, and the record information with a higher degree of similarity is placed closer. Therefore, information users can simultaneously consider the mutual positional relationship of record information, and identify records with abnormal contents. of! ! ! This has the effect of providing a method of displaying record information that allows the user to recognize trends in record groups.

[Brief explanation of the drawing]

FIG. 1 is a processing flow diagram of an embodiment of the present invention, FIG. 2 is a configuration diagram of the table 3 shown in FIG. 1, and FIG. 3 is an example of similarity between records and creation of a hierarchical structure by creating a directed graph. Fourth
The figures are an explanatory diagram of sequentially dividing a rectangular display area according to the example of Fig. 3, Fig. 5 is a processing flow diagram of the second embodiment, and Fig. 6 is a processing flow diagram of the second embodiment.
The figure is an explanatory diagram of display area division using a Voronoi diagram. FIG. 7 shows the format of the record information file. FIG. 8 shows the format of the plane coordinate table. 9th
The figure shows the format of divided area data. Abusive picture book. 3! 3 (L) Zuyo 3 (/,) Figure 3 (c) Concave 3rd Biri Figure ■ ((Ugura

Claims (1)

[Claims] 1. With respect to a plurality of pieces of record information to be displayed, (a) the degree of similarity between the attributes of each record information and the attributes of other record information is determined, and (b) the degree of similarity is determined from among the degrees of similarity. Select record information with the highest degree of similarity to the attributes of the record information for each piece of record information; (c) obtain a directed graph in which each of the records is a node and the relationship between the record information with the highest degree of similarity is an arc; (d) Find a pair of records showing the highest degree of similarity from among the relationships between record information with the highest degree of similarity, and create the above directed graph in a tree structure with the node corresponding to one of the records in this pair as the root. (e) The root of the above tree structure is set as the upper level, and each time the tree branches from the upper level, the target area is divided and displayed using the area information held in the record information corresponding to each node. , A method of displaying information characterized by: 2. An information display method according to claim 1, characterized in that a display corresponding to each record is displayed in a manner different from that of an adjacent record. 3. An information display method as claimed in claim 1, characterized in that the area information is made to correspond to a specific attribute value of each record. 4. In the information display method described in claim 1, the distance between corresponding nodes is determined so that records with higher similarity are closer, and the display location corresponding to each record is determined according to the distance between each node. An information display method characterized by: 5. In the method for displaying information according to claim 1, each node is projected onto a point on the display surface so that the distance between the nodes is preserved, and the display is performed using a holon diagram regarding the projected point. A method of displaying information characterized by dividing a surface. 6. Consisting of a file storing product information for each product, a computer, and a display device, the computer (a) reads out predetermined attribute information for each product from the product information in the file; (b) reads out (c) Find the maximum similarity for each product from the above similarities; (d) Set each product as a node and the maximum similarity for each product as an arc. Obtain a directed graph, (e) convert the directed graph into a tree whose root is one of the nodes at both ends of the arc with the maximum similarity among the maximum similarities for each product, and (f) convert the root of the tree to the upper Then, starting from the top,
Each time the tree branches, a predetermined display range is divided in proportion to the value of the attribute of the product corresponding to each node, and (g) the divided display range is output to the display device. Features a product information display system. 7. The product information display system according to claim 6, wherein the display range is a sales floor layout figure of the product. 8. (a) Find the mutual similarity of the plurality of values, (b) Give a shorter distance to the one with the greater similarity, (c) Determine the area in proportion to each of the plurality of values. , (d) the distance between the center of the figure having the area and the center of the figure corresponding to another value corresponds to the distance corresponding to the degree of similarity, and (e) the plurality of values are represented by a figure. Graphical method.