JP2003044646A - Business situation warning system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a system, based on management data inputted every second, that sends a managerial warning if the data are in a statistically abnormal situation or exceed a predetermined variation level. SOLUTION: A necessary statistical procedure (c) is executed for inputted time-series data (b), and, if some trend is found, the magnitude of the trend is detected to send a warning (h).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is a system capable of automatically finding a tendency from various management materials to be recorded daily and displaying it as a warning to be noted in management.

[0002]

2. Description of the Related Art Conventionally, software is widely used in which various kinds of management data to be recorded daily can be input into a computer every moment and the latest totalized result can be immediately output.

However, these conventional techniques merely record and store the input management data as they are and totalize them, and can detect an abnormal state from such management data and provide it as a warning as management data. It wasn't something.

[0004]

SUMMARY OF THE INVENTION The present invention is based on management data input every moment, and when it is in a statistically abnormal state or exceeds a preset variation level. It seeks to provide a system for issuing management warnings.

[0005]

[Means for Solving the Problems] [1] (1) A trend analysis is performed by performing necessary statistical processing on input time series data. , A step of examining the presence or absence of a tendency, (2) a step of ending the work when a tendency is not found, and a step of detecting a trend change by creating a regression equation when a tendency is found, (3) A business condition warning system characterized by comprising a step of creating a warning when the detected tendency is statistically probable, [2] When the detected tendency exceeds a range determined in advance. [3] The business condition warning system according to [1] above, which comprises a step of creating a warning.
When the detected tendency exceeds the range determined in advance, a warning is given also to the upper management person in charge according to the number of times and the degree of the exceeded [2]. The business condition warning system described.

[0006]

BEST MODE FOR CARRYING OUT THE INVENTION A specific configuration of the present invention will be described with reference to FIG. FIG. 1 is an example of a flow sheet of the method of the present invention, the basis of which is input initialization 1
The system is composed of the steps of (1), data input (2) and alarm report file (3).

First, a customer of this software inputs a reference value, which is a limit of alarm transmission determined by the customer through a communication line, and is stored as an initial setting reference 5. This criterion serves as a basis for determining the alarm determined by the customer in the alarm presence / absence process 8.

The customer sets the items to be input through the communication line 1 and makes the input initial setting. This is, for example, whether or not the initial setting items are theoretically sufficient.
It is set through.

For each of the initially set items, the management data is sent from the communication line 2 every moment, and the correct data is input through the input data check step 6.

This data is used in the alarm presence / absence processing step 8,
For example, statistical processing is performed.

The necessary statistical processing is to remove cyclical fluctuations, seasonal fluctuations and irregular fluctuations by a moving average such as an annual moving average (annual + weekly), and then trend analysis by Kendall as a trend analysis. The existence of temporal dependence is tested by the formula of or the trend test formula by the total number of runs, and the correlation analysis formula and regression analysis formula.

If there is no dependency, no alarm is necessary and the process ends.

When dependence is found, a trend line (for example, regression line) is determined by a moving average, a least squares method, a spline function, etc., and fluctuations are calculated.

This is compared with a reference value, and when the value exceeds a preset value, a signal for alarm creation is generated from the alarm 9, and in the alarm report creating step 10, the alarm report created by the alarm report processing 11 is output. And the report is saved in alarm file 3.

In step 11 of alarm report processing, an alarm report is created in accordance with preset alarm report criteria.

This is the first warning. Even if some measures are taken in accordance with this warning, if the abnormality is not resolved for a certain period of time, a multi-level alarm presence / absence processing 13 is performed through a multi-level alarm presence / absence checking step 12. The presence or absence of an alarm is checked according to the alarm criteria set here. If there is no alarm 14, the process ends. If there is an alarm 14, it is sent and an alarm report is generated according to the alarm report criteria in the multiple alarm report processing stage 16 through the alarm report generation stage 15.

If this multi-level alarm stage warning is reported to a higher level organization than the stage alarm of the first alarm report 10, the management can contact the warning in time.

[0018]

[Embodiments] As an embodiment, a method of judging whether or not a warning can be issued based on the daily sales amount will be specifically described below with reference to the drawings and tables.

As shown in FIG. 2, the daily sales amount transmitted (a) from the customer is recorded and accumulated as the actual result (b).

As the actual result (b), the actual results column in Table 1 below shows the actual sales results for each of the products A, B and C.

In order to eliminate fluctuations such as seasonal fluctuations and weekly fluctuations, the annual sales average (day) and annual and weekly moving average (day) of each product are calculated [process (c) in FIG. 2].

[0022]

[Table 1]

FIG. 3 shows the results in the form of a chart for the product A, and FIG. 4 shows a diagram in the form of the product B.
FIG. 5 is a diagram showing the above.

Table 2 shows a trend test file by run prepared based on the data in Table 1 above.

[0025]

[Table 2]

The result obtained by this process (c) in FIG. 2 is tested for the presence or absence of a trend change by establishing a hypothesis that there is no trend change and an alternative hypothesis that there is a trend change in the trend change hypothesis test (d) stage. If no trend change is found, the work ends with no trend change (e).

When the result of the trend change existence (f) is obtained in the hypothesis test (d) stage, the predictive value of the volatility rate, the variable sales, etc. is calculated in the trend change calculation (h) stage through the regression analysis stage (g). Then, after the alarm report preparation step (i), it is transmitted (j), and the work is completed. The result is stored in the alarm file (3) of the multi-level alarm processing flow of FIG.

Table 3 below is a table showing predicted values such as trend lines based on the result of regression analysis of sales data of the product A.

Table 4 shows similar predicted values for the product B and Table 5 shows similar predicted values for the product C.

FIG. 7 is a moving average graph of the sales data of the product A and a trend line based on Table 3.

FIG. 8 is a moving average graph of sales data of the product B and a trend line based on Table 4, and FIG. 9 is a moving average graph of sales data of the product C and a trend line based on Table 5.

Table 6 below is a table showing forecasts of changes in annual sales of products A, B, and C.

[0033]

[Table 3]

[0034]

[Table 4]

[0035]

[Table 5]

[0036]

[Table 6]

From the record of the content and number of alarms stored in the alarm file 3 shown in FIG. 6, it is checked 12 whether or not an alarm (multi-level alarm) is required for the upper manager. The standard of this check is based on the alarm standard 13.

If there is no need for a multi-level alarm, the work ends there.

An example of the above multi-level alarm criteria is illustrated in Table 7 as the multi-level alarm criteria.

[0040]

[Table 7]

The difference / change rate adjustment is an adjustment% value with respect to the alarm reference value, and is an example of the adjustment value% of the alarm reference value for each hierarchy in the level division.

The planned standard difference rate is an example of the allowable range of the alarm with respect to the standard value determined in advance.

The tendency significant change rate is an example of the alarm tendency change rate standard when it is determined to be significant by the trend test.

The responsibility level is an example of a reference value (when an alarm is generated in this table) of how many alarms are to be issued within a plurality of management periods.

Lower 4CS (control span:
The abbreviation of management time width) is a period of one management period that should be checked for an alarm in the lower stage of management responsibility. Needless to say, this standard can be freely set according to the actual conditions of each business entity.

According to this standard, the presence or absence of a multi-level alarm is checked, and when an alarm occurs, an alarm report 15 is created according to the alarm report standard 16, which is saved 17 and is sent to the customer 18 This completes the work.

[0047]

As described above, according to the present invention,
Since statistical data that could not be processed without complicated work in the past can be processed in an extremely short time, it is possible to quickly show a warning that should be noticed in management as a warning. In addition, depending on the importance of matters to be noted in management, it is possible to give a warning to senior management.

[Brief description of drawings]

FIG. 1 is a flow sheet showing an outline of an alarm system of the present invention.

FIG. 2 is a flow sheet illustrating an example of an alarm according to the present invention.

FIG. 3 is a daily sales moving average (annual + weekly) graph of product A.

FIG. 4 is a graph of daily sales moving average of product B (annual + weekly).

FIG. 5 is a graph of daily sales moving average of product C (annual + weekly).

FIG. 6 is a multi-level alarm processing flow.

FIG. 7 is a sales forecast graph of product A.

FIG. 8 is a sales forecast graph for product B.

FIG. 9 is a sales forecast graph of a product C.

Claims (3)

[Claims] 1. A step of (1) performing necessary statistical processing on input time-series data to perform a trend analysis, and examining the presence / absence of a trend. (2) Processing if no trend is found. When the trend is found, a regression formula is created to calculate the trend fluctuation (3) A warning is created if the detected trend is statistically probabilistic. And a business condition warning system. 2. The business condition warning system according to claim 1, further comprising the step of generating a warning when the detected tendency exceeds a range determined in advance. 3. When the detected tendency exceeds a range determined in advance, a warning is given also to the upper management person in charge according to the number of times and the extent to which the tendency has been exceeded. The business condition warning system according to claim 2.