JP2734622B2 - Tabular reverse inference method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to backward inference in tabular inference.

[Conventional technology]

For example, the conventional "TK! Solver" {Unibac Pamphlet} has a tabular interface, and mathematical formulas are used to link the grids in the table. "Variable a + variable b + variable c
Assuming that an expression in which the variable d on the right side is determined by the variables a, b, and c on the left side as in “= variable d”, (4), (5), and (6) in FIG. , b, c, (7)
Is a variable d on the right side, and is calculated by substituting the values “1”, “2”, and “2” into the three variables on the left side, and “5” is displayed as the variable d. When the variable c on the left side is unknown and its value is obtained, there is no need to modify the equation. When a value "10" is input to the variable d on the right side, it is already defined as shown in FIG. The value "7" of the variable c can be obtained by inversely calculating the expression and displayed. However, when there are two or more unknown variables, the reverse calculation cannot be performed.

[Problems to be solved by the invention]

Since the conventional tabular inference method operates as described above,
When the number of unknown variables is one, the inverse calculation can be performed. However, when there are two or more unknown variables, the value cannot be obtained in the opposite direction to the defined expression. Therefore, for example, when the total value increases, no inconsistency arises if one variable is corrected as an unknown variable, but the entire change is absorbed by one variable and cannot be assigned to a plurality of variables.

(1) Therefore, there is a problem that a great deal of trial and error is required before setting the values of all variables to appropriate values. (2)
In addition, it is not possible to define procedural contents (hereinafter called rules) that deal with non-numerical values, or to execute in the opposite direction (pretend pretend), so that it cannot support decision making other than numerical calculations. There were two issues.

The present invention solves the above two problems. Even when there are two or more unknown variables in a defined expression or rule, inference is made in the opposite direction using the values input so far. The purpose is to be able to do it.

[Means for Solving the Problems] A table-type backward inference method according to the present invention is a table-type inference method which defines a procedure relating to a numerical data relationship between cells constituting a table in table-type inference. Means, and inference means for determining one or more cell values constituting the left side in response to a change in the right cell defined by the inter-cell numerical data definition means, wherein the inference means comprises at least a change in a variable cell on the right side. The proportional distribution method distributes the minutes evenly to the floating cells on the left side, the proportional distribution method distributes changes in the variable cells on the right side in proportion to the data values in the floating cells on the left side, and changes in the variable cells on the right side Is allocated to the floating cells on the left side based on the designated ratio.

[Action]

According to the present invention, the value input to the right-hand side is replaced with all unknown variables by using any of proportional, equal, and ratio methods based on the values input up to that point, and In the rule, an overall search is performed using the value candidates that have already been set, and all the values of the unknown variables on the left side that match the values of the variables on the right side are found, and the one found first is shown. However, this is a pretend for all the variables on the left side, and if there is a variable whose value is to be stored, it is stored by setting the value of the variable to “fixed”.

〔Example〕

 An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a configuration diagram of a tabular inference method according to the present invention, wherein (1) is a knowledge base, (2) is an inference engine,
(3) is a tabular interface. The inference engine (2) makes inferences using the contents of the knowledge base (1) in which formulas and rules described between cells are stored, and displays them on a tabular interface (3). The inference engine (2) has a forward inference unit and a backward inference unit, and controls each of forward inference and backward inference.

As shown in FIG. 2 with the same reference numerals as in FIG.
In the forward mode, the operation is the same as in FIG. 5 described above. Here, when the mode of the backward inference is switched and “10” is input to the variable d, the variable a is set to match the value of the variable d as shown in FIG. ,
The values of the variables b and c are changed. If only the value of the variable c is to be changed as shown in FIG.
It can be realized by fixing the values of a and the variable b.

FIG. 4 is a flow chart in the backward inference mode. First, in S (41), when a value is input to the cell of the variable on the right side, then in S (42), the expression defined in the cell is an expression or a rule. Is determined, and if an expression is defined, first at S (43), an allocation method is selected. "Proportional allocation", "Equal distribution"
There are three methods of “ratio allocation”. Here, the proportional distribution means that when the value of the cell on the right side is changed, the difference value of the cell on the right side is determined by the difference between the value of the cell on the right side and the value of the variable on the right side. Is a method of proportionally distributing according to the value of. In addition, the equal distribution is a method of equally allocating the difference value of the right side cell to each floating cell on the left side, and the ratio distribution further assigns the difference value of the right side cell to the left side cell based on a separately specified ratio. Means to assign to Then, check whether the value has already been entered in the variable cell on the left side, and if not entered, distribute the difference value on the right side evenly to the non-input cells on the left side by "equal distribution". In the case where the backward inference has been performed once before, the same method as the previous one is used, and in the case of the first time, "proportional distribution" is selected in order to use the value originally input to the left side cell. Using this method, if there is a value input so far in S (44), the value is changed using various conditions such as referring to the value. On the other hand, in the case of a rule, the inverse execution cannot be performed using heuristics as in the case of a mathematical expression. Therefore, using the value candidate of the variable on the left side already set in S (45), S ( In step 46), all combinations are executed in the forward direction, and the one that matches the value of the input variable on the right side is searched.
Display the first one found in step 7).

Although the embodiment in one step of the backward inference method in the form of a table has been described, similar effects can be obtained for inference within a certain table or inference that continues without interruption until the end.

Also, in the above embodiment, the reverse inference in the form of a table in which mathematical expressions and rules can be described has been described. However, in the present invention, a spread sheet in which only mathematical formulas can be described is also used in a spreadsheet in which only rules are defined. The same effect is obtained in the expert system.

In the above embodiment, the value candidates are used for backward inference of rules. However, the same effect can be obtained by using value candidates for backward inference of mathematical expressions.

[Effects of the Invention] As described above, according to the present invention, in the tabular inference relating to the numerical data relationship between the cells constituting the table,
When the right side cell is changed, at least by applying each distribution method consisting of “equal”, “proportional”, and “ratio” independently of the state of the left side cell data such as uninput or input, or independent of the cell data. Since one or more floating cell values on the left side are inferred, there is an effect that detailed and efficient decision support can be performed in consideration of the value of the left side cell.

[Brief description of the drawings]

FIG. 1 is a block diagram of the present invention, FIG. 2 is a diagram showing an example of forward inference and an operation example for implementing backward inference, and FIG. 3 is an embodiment of the present invention (reverse inference). FIG. 4 is a flowchart of the present invention, and FIG. 5 is a diagram showing an example of forward inference and an example of an operation for performing a conventional inverse calculation.
FIG. 1 is a diagram showing one embodiment of a conventional inverse calculation. In the figure, (1) is a knowledge base, (2) is an inference engine,
(3) is a tabular interface. In the drawings, the same reference numerals indicate the same or corresponding parts.

 ────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-61-51267 (JP, A) Information Processing Society of Japan 34th (Early 1987) National Conference 763-764 Information Processing Society of Japan 35th Annual Meeting 1757-1758, p. 1897-1898 Information Processing Society of Japan 36th (Early 1988) 1897-1898

Claims (1)

(57) [Claims] 1. In a table format inference, an inter-cell numerical data defining means for defining a procedure relating to a numerical data relationship between cells constituting a table, and a right-side cell defined by the inter-cell numerical data defining means. An inference means for determining one or more cell values constituting the left side in response to the change, wherein the inference means at least uniformly distributes a change in the variable cell on the right side to the floating cell on the left side; , A proportional distribution method that distributes the change in the variable cell on the right side in proportion to the data value in the floating cell on the left side, and distributes the change in the variable cell on the right side to the floating cell on the left side based on the specified ratio A rational distribution method, and a tabular backward inference method characterized by comprising: