JP5530961B2 - Work shift table creation support system - Google Patents

Description

  The present invention relates to a work shift table creation support system that can extract shift conditions that should be set in advance when creating a work shift plan table from a past or teacher work shift table.

  Conventionally, in hospitals, nursing homes, call centers, stores, etc., work styles are not fixed according to the day of the week, but so-called shift work is performed in which a plurality of forms are provided every day or every certain period.

  In the workplace that employs this shift work, the work shift plan table is created in advance at the beginning of the month, etc., and the worker will work according to it, so there is a device for easily creating the work shift plan table. It has been developed (see Patent Documents 1 and 2).

  When using such an apparatus for creating a work shift plan table, it is necessary to set a work day desired by the worker in advance, or to set a work pattern to be prohibited such as not to make night shifts continuous.

  In view of this, Patent Document 1 discloses a work shift creation support device that can extract shift conditions such as a day of work for each worker and a day when the worker cannot work from a past work shift table.

  Further, Patent Document 2 discloses an apparatus that can determine the operation results for each worker from the past sales results for each day and the information of the workers who worked on that day, and reflect them in the creation of the work shift plan table. Is disclosed.

JP 2008-269256 A JP 2007-26361 A

  However, the work shift creation support device disclosed in Patent Document 1 can reflect the past work record, but corresponds to the shift condition that can be adopted as a rule whether the past information to be reflected is accidental. Since it is not verified whether it is a thing, there is a possibility that the extracted shift condition is a cobblestone mixture. In Patent Document 1, the worker can set the priority of the shift condition. However, when many shift conditions of various levels are extracted, it is difficult to set the priority. ,It takes time and effort.

  In addition, the work shift plan table creation apparatus disclosed in Patent Document 2 can reflect past work results, but extracts and reflects data related to operation results such as worker evaluation and sales result information. Only.

  Therefore, the present invention provides a work shift table creation support system capable of accurately extracting shift conditions to be applied when actually creating a work shift plan table from a past or teacher work shift table. It is intended to provide.

  In order to achieve the above object, a work shift table creation support system according to the present invention is a work shift table creation support system that extracts shift conditions necessary for creating a work shift plan table, and includes a date direction associated with a date, Reading means for fetching a past work shift table having a worker direction associated with a worker or a work shift table serving as a teacher, and data of each cell of the work shift table fetched by the reading means for the worker direction or the Preprocessing means for organizing in association with data in the date direction, temporary shift condition setting means for selecting candidates for the shift condition from the processing data arranged by the preprocessing means as temporary shift conditions, Performs a statistical test based on the data and extracts the temporary shift condition that satisfies the specified condition as the main shift condition A shift condition extraction means, characterized in that an output means for outputting the extracted the shifted condition.

  Here, the preprocessing means generates array data in which the data of each cell and the data of the cells adjacent to the worker direction are associated with each other, and the temporary shift condition setting means generates predetermined data in the array data. Arrangement data that appears more than the number of times is selected as a temporary shift condition, and the shift condition extraction unit can be configured as a main shift condition when the temporary shift condition is established as a shift condition by comparison with a significance level. .

  Further, the pre-processing means totals the cell data in the date direction as a daily total number for each data type, and the temporary shift condition setting means temporarily compares the daily total number with a predetermined value. A shift condition can be selected, and the shift condition extraction unit can be configured to use the shift condition when the temporary shift condition is established as a shift condition by comparison with a significance level.

  Further, in the preprocessing means, the data of the worker direction cell is totaled as a total number of worker units for each data type, and in the temporary shift condition setting means, the total number of worker units is calculated with a predetermined value. The shift condition extraction unit can select a condition that becomes a temporary shift condition by comparison, and can set the shift condition as a main shift condition when the temporary shift condition is established as a shift condition by comparison with a significance level.

  Further, the preprocessing means calculates an appearance interval for each data type in the worker direction as a data interval value, and the temporary shift condition setting means calculates the temporary shift condition by comparing the data interval value with a predetermined value. The shift condition extraction means can be configured as the main shift condition when the temporary shift condition is established as the shift condition by comparison with the significance level.

  Further, the pre-processing means generates combination data associating the data of each cell with the data of another worker's cell in the date direction, and the temporary shift condition setting means is a predetermined data among the combination data. The combination data that appears more than the number of times is selected as a temporary shift condition, and the shift condition extraction unit is configured to use the shift condition as a main shift condition when the temporary shift condition is satisfied as a shift condition by comparison with a significance level. it can.

  Another work shift table creation support system according to the present invention is a work shift table creation support system that extracts shift conditions necessary for creating a work shift plan table, and associates a date direction associated with a date with a worker. Reading means for importing a past work shift table having a worker direction or a work shift table serving as a teacher, and organizing the data of each cell of the work shift table read by the reading means in association with the worker direction data Preprocessing means, conversion processing means for converting the processing data arranged by the preprocessing means into conversion data, clustering means for performing clustering using the conversion data, and shifting clusters classified by the clustering means Output means for outputting as a condition.

  The work shift table creation support system of the present invention configured as described above captures a past or teacher work shift table by a reading means, selects a temporary shift condition from the captured data, and sets the temporary shift condition as a main shift. Statistically test whether the condition should be met.

  For this reason, the shift conditions to be applied when actually creating the work shift plan table can be accurately extracted. In addition, since it is not necessary for the user to determine whether or not to use this shift condition, and shift conditions such as implicit rules that were not recognized can be extracted from the past work shift table, Even a user who is not skilled in creating a work shift plan table can easily create a work shift plan table.

  In addition, by generating the arrangement data in which the data of the taken work shift table is associated with the worker direction, it is possible to extract a shift condition for designating or prohibiting the arrangement of continuous work forms on a plurality of days.

  Furthermore, the shift conditions relating to the required number of people per day can be extracted by counting the data of the captured work shift table for each data type in the date direction.

  Moreover, the shift condition regarding the frequency | count that a certain type of work form is performed within a period can be extracted by totaling the data of the taken work shift table | surface for every data type in the direction of a worker.

  Furthermore, by calculating the appearance interval for each data type as the data interval value from the data in the worker direction of the taken work shift table, it is possible to extract a shift condition regarding the interval for performing a certain type of work style.

  In addition, by generating combination data that correlates the data of the imported work shift table with the data of other workers in the date direction, extract shift conditions related to worker combinations such as working or not working on the same day can do.

  In addition, if clustering is performed by the clustering means using the imported work shift table data and the finally classified clusters (groups) are extracted as shift conditions, only workers will be involved in creating the work shift plan table. It is possible to make adjustments in groups that do not depend on

It is explanatory drawing which showed the outline | summary which produces the work shift plan table using the work shift table preparation assistance system of embodiment of this invention. It is a block diagram explaining the structure of the work shift table creation assistance system of embodiment of this invention. It is the figure which illustrated the past work shift table. It is a flowchart explaining the flow of the process which extracts shift conditions by a statistical test. It is a flowchart explaining the flow of the process which extracts shift conditions by clustering. It is a figure explaining the method of clustering using the data taken in from the past work shift table, (a) is a figure explaining the setting of the first division axis, (b) is the second division axis. It is a figure explaining a setting.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 illustrates a process of creating a work shift plan table using the rule extraction device 2 as the work shift table creation support system of the present embodiment. The rule extraction device 2 described here is a computer in which a computer program for executing various means described later is installed. However, the work shift table creation support system of the present embodiment may be a computer program itself. Good. In FIG. 1, the rule extraction device 2 and the work shift table creation device 3 are shown separately, but may be an integrated computer program or a computer on which it is installed.

  This rule extracting device 2 is a device for reading a past work shift table 1 serving as a teacher or a teacher by a reading means 21 and extracting shift conditions for creating a future work shift plan table therefrom. Then, the work shift table creation device 3 is set using the shift condition 270 output from the output means 27, and the work shift plan table can be created.

  As shown in FIG. 1, the past work shift table 1 includes dates arranged in the horizontal direction in the top row of the table, and worker names arranged in the vertical direction in the leftmost column of the table. The shift work schedule of actual shift work is recorded in each cell. Here, the direction (the vertical direction in the table) that continues to the lower side of each day is the date direction, and the direction (the horizontal direction in the table) that continues to the right side of each worker is the worker direction.

  The data entered in each cell of this work shift table 1 indicates the work style of each worker on that day. For example, “day shift” indicates daytime work, “night shift” indicates night shift, “night shift” indicates day shift after night shift, and “rest” indicates a holiday. In addition to this, “early attendance” with early work hours, “half work” with half day work, “late work” with late work hours, “pay leave” instead of holiday work, “training” such as vacation, It can be arbitrarily set to suit the workplace where shift work is performed, such as “special leave” such as special leave.

  Then, on July 1 (Thursday), the data in the direction of date is data that is connected to “night light”, “day shift”, “night shift”, “day shift”, and “day off”. In the case of “Suzuki” who is a worker, data connected to “Dawn”, “Closed”, and “Day shift” is data for the worker.

  The data of this work shift table 1 is taken into the rule extracting device 2 by the reading means 21 as shown in FIG. The reading means 21 may be configured by combining a scanner and character recognition software, but the data of the work shift table 1 is converted into a hard disk or USB (Universal Serial Bus) as a text data file in CSV (Comma-Separated Values) format or the like. If it is stored in a recording medium such as a memory, the configuration may be such that only the file is fetched.

  The rule extracting device 2 selects a provisional shift condition from the reading means 21, the preprocessing means 22 for organizing the data of each cell fetched by the reading means 21, and the processing data arranged by the preprocessing means 22. It mainly includes temporary shift condition setting means 23, shift condition extraction means 24 for extracting a shift condition as the main shift condition from temporary shift conditions, and output means 27 for outputting shift condition 270.

  In addition, as another processing system after the preprocessing unit 22, a conversion processing unit 25 that converts the processing data arranged by the preprocessing unit 22 into conversion data that can be clustered, and a clustering unit 26 that performs clustering using the conversion data. And.

  The preprocessing means 22 performs processing for organizing the data of each cell of the work shift table 1 taken in by the reading means 21 in association with the worker direction or date direction data. The processing in the preprocessing unit 22 varies depending on the type of shift condition to be extracted as described later. However, when a plurality of types of shift conditions are extracted from one work shift table 1, the preprocessing unit 22 also includes them. Various processes are performed together.

  Further, the temporary shift condition setting means 23 performs processing for selecting, as the temporary shift condition, data that can be used as a shift condition by using the data subjected to the statistical processing or the like in the preprocessing means 22.

  Further, the shift condition extracting unit 24 performs a statistical test on the temporary shift condition selected by the temporary shift condition setting unit 23, and performs a process of setting as the main shift condition when a predetermined condition is satisfied.

  Further, the output unit 27 performs processing for outputting the main shift condition set by the shift condition extracting unit 24 as the shift condition 270. The output means 27 may be any configuration such as a means for writing data to a recording medium, a printer, or a monitor.

  Next, using the sample data of the work shift table 1A shown in FIG. 3, the processing flow of the rule extraction device 2 of the present embodiment will be described with reference to the flowchart of FIG.

  In the shift shift table 1A of FIG. 3, “early departure” is “early”, “day shift” is “working”, “late” is “late”, “night shift” is “night”, “nightlight” is “light”. " Usually, a work shift table for a period of one month is often used, but here, in order to simplify the explanation, the work shift table 1A for a period of 16 days will be described as an example.

  First, in step S1, the past work shift table 1A shown in FIG. 3 is read. Although FIG. 3 shows a state printed in the table format, the data of the work shift table 1A is stored in the recording medium in the CSV format, and is read by giving a reading instruction to the reading means 21. be able to.

  Subsequently, in step S2, the pre-processing means 22 generates aligned data. For example, the day after “Night (night shift)” will be “Ming (night shift)”, “Night (night shift)” will not be two consecutive days, etc. And when it is desired to find a shift condition related to the order (hereinafter, sometimes referred to as “shift arrangement rule”), the preprocessing means 22 generates arrangement data in which data of cells adjacent in the work direction are associated with each other.

  The work shift table 1A shown in FIG. 3 will be described as an example. For the worker “Sato”, the arrangement data such as “late → early”, “early → off”, “off → off”, “off → night”, etc. Generated. For the worker “Suzuki”, side-by-side data such as “night → light”, “light → leave”, “leave → work”, “work → work” is generated. Here, when creating a shift arrangement rule for the entire workplace, it is not necessary to associate and arrange workers and arrangement data.

  In step S3, line-up data whose number of appearances is equal to or greater than the preset number n is found. For example, if the number of occurrences of “night → light” and “light → break” is n times or more, the arrangement order of these two types of arrangement data is selected as a temporary shift condition (step S4).

  Since the two temporary shift conditions selected in step S4 are selected based on only the number of appearances, the shift condition extraction unit 24 should perform a statistical test to extract these shift conditions. It is determined whether or not it is a rule (step S5).

  In the following, the test performed for the temporary shift condition “night → light” will be described. In this test, a hypothesis that “there is no rule that“ the day after “night” becomes “light” ”” is established, and if the hypothesis is rejected, it is determined that there is a rule “night → light”.

  First, the probability P (A) that “night” appears can be obtained by counting the number of times “night (night work)” appears in the period (16 days) that is the subject of the work shift table 1A.

P (A) = NA / S
Here, S indicates the number of cells in the work shift table 1A (= 16 days × 8 workers).

  Subsequently, if the number of times “night” appears on the day before “Ming (Dawn)” is NB, the probability of “night” appearing NB times without any rule is the NB of P (A). To the power.

  On the other hand, if the significance level (1%, 5%, etc.) is set as the predetermined condition and compared with the NB power of P (A), if the NB power of P (A) is smaller than the significance level, An event with a probability that is less than the significance level has occurred, the hypothesis that was initially set up, that is, the "no rule" from night to light "is rejected, and the rule" from night to light " There is.

  Therefore, in step S6, the temporary shift condition “night → light” is set as the main shift condition. It should be noted that the same test is performed for the temporary shift condition “bright → off” to determine whether the shift condition can be set.

  In step S7, the output unit 27 outputs the rule 271 extracted as the main shift condition as the shift condition 270 (see FIG. 1). The output by the output means 27 is set in conjunction with the work shift table creating apparatus 3 regardless of whether it is a means that can be visually recognized by the user, such as printing by a printer, or a means for writing to a recording medium. Any means may be used.

  So far, the method for extracting the shift arrangement rule has been described, but subsequently, a method for extracting another type of rule 271 as the shift condition 270 will be described. The rule extraction device 2 can be set so that only a specific type of rule is extracted, or can be set so that a plurality of types of rules are extracted at a time.

  First, the case where the minimum number or the maximum number of workers assigned to one day is extracted as a shift condition (hereinafter sometimes referred to as “necessary number rule”) will be described.

  In this case, the pre-processing means 22 totals the cell data in the date direction as the daily total number for each data type. For example, in the work shift table 1A shown in FIG. 3, the work is such that one person is "late" on the day (Thursday), two are "work", two are "off", and two are "night". The number of daily aggregations for each form (type) can be aggregated. In addition, on the 4th (Sunday), the number of daily counts can be tabulated such that “work” is one person and “rest” is four. In addition, on the 5th (Monday), the number of daily counts may be tabulated such that “late” is 1 person, “work” is 2 people, “off” is 3 people, and “night” is 1 person. it can.

  Thus, the number of workers to whom each work style is assigned per day is not necessarily constant, and there may be a minimum or maximum number of workers per day as a rule depending on the work style.

  Therefore, the pre-processing means 22 extracts, for each work style, those that become the minimum value and the maximum value throughout the period (16 days) from the daily total number of each work style. Here, when creating a rule for the required number of people throughout a period, it is not necessary to organize the day or weekday, whether it is a holiday, or what day of the week, and the minimum value or maximum value in association with each other. .

  Subsequently, a value whose number of appearances is obvious is removed from the minimum value and the maximum value extracted for each work style. For example, a predetermined value such that the minimum value is 0 or the maximum value is equal to the total number of workers (eight persons) is self-evident, and there is no meaning to extract it as a shift condition. Then, the minimum value or the maximum value of each work style that is not obvious (not a predetermined value) is selected as a temporary shift condition of the minimum number or the maximum number of work styles (see step S4 in FIG. 4).

  Since the minimum or maximum number of persons selected as the temporary shift condition in this way is selected based only on the number of appearances in the work shift table 1A, a statistical test is performed in the shift condition extraction means 24. By doing so, it is determined whether or not the rule should be extracted as the present shift condition (see step S5 in FIG. 4).

  In the following, the test performed for the provisional shift condition that the maximum number of “late (late)” is one will be described. In this test, a hypothesis is made that “there is no rule that limits the number of people who can be assigned“ late ”to one person per day”, and if the hypothesis is rejected, the maximum number of people who are “late” in a day is 1 It is determined that there is a rule “is a person”.

  First, the probability P (A) that “late” appears can be obtained by counting the number of times NA that “late” appears in the period (16 days) that is the subject of the work shift table 1A.

P (A) = NA / S
Here, S indicates the number of cells in the work shift table 1A (= 16 days × 8 workers).

  Subsequently, based on the binomial distribution, the probability of “slow” appearing on a certain day is obtained for each number of appearances. The number of appearances ranges from 0 to the number of workers (8 people), and the probability can be obtained according to a binomial distribution formula. For example, here, the probability that “slow” appears 0 times is 0.455, the probability that it appears once is 0.377, the probability that it appears twice is 0.136, and the like.

  For this reason, the probability P (A, 0, 1) that “slow” appears only once or less (maximum number of people) per day is the sum of the probabilities from 0 to 1.

P (A, 0, 1) = 0.455 + 0.377 = 0.832
Assuming that there are NC days during the period, the probability that “late” will appear below the maximum number of people without any rule is the power of P (A, 0, 1) to the power of NC.

  On the other hand, when a significant level (1%, 5%, etc.) is set as a predetermined condition and compared with the power of NC of P (A, 0,1), P (A, 0,1) If the NC power is smaller, an event with a probability lower than the significance level has occurred, and there is no rule to limit the number of people who can be assigned “late” to one per person. The hypothesis is rejected, and "there is a rule that the maximum number of people who are" late "in a day is one."

  Therefore, a provisional shift condition that “the maximum number of people who are“ late ”in a day is one person” is set as this shift condition. If the provisional shift condition relates to the minimum number of persons, the probability obtained from the binomial distribution formula is summed from the minimum value to the total number of workers, and the result obtained by multiplying it by the NC power is tested. Just do it.

  Subsequently, regarding the number of work types of the same type that are assigned to one worker within a period (16 days), the minimum number or maximum number of times may be referred to as a shift condition (hereinafter referred to as “number of times within period” rule). ) Will be described.

  In this case, the pre-processing means 22 totals the worker-direction cell data as the total number of worker units for each data type. For example, in the work shift table 1A shown in FIG. 3, the work style (type) is such that “work” is 3 times, “absence” is 4 times, and “night” is 3 times during the period of the worker “Sato”. The total number of employee units for each employee can be aggregated. For the worker “Takahashi”, the total number of workers can be calculated such that “work” is 3 times, “leave” is 5 times, and “night” is 3 times. Furthermore, if the worker is “Tanaka”, the total number of workers can be counted such that “work” is 7 times, “leave” is 4 times, and “night” is 2 times.

  In this way, the number of times each work style is assigned to each worker within the period is not necessarily constant, and there may be a minimum or maximum number of times that assigns a work style common to each worker or all workers as a rule. There is sex.

  Therefore, the pre-processing means 22 extracts the minimum value and the maximum value for each work style from among the total number of worker units for each work style of all workers. Here, when creating the in-period rule for the entire workplace, it is not necessary to associate and organize workers and the total number of workers.

  Subsequently, a value whose number of appearances is obvious is removed from the minimum value and the maximum value extracted for each work style. For example, a predetermined value such that the minimum value is 0 or the maximum value is equal to the number of days in the period (16 days) is self-evident, and there is no point in extracting it as a shift condition. Then, the minimum value or the maximum value of each work style that is not obvious (not a predetermined value) is selected as a temporary shift condition of the minimum number or maximum number of each work style (see step S4 in FIG. 4).

  Since the minimum or maximum number of times selected as the temporary shift condition in this way is selected based only on the number of appearances in the work shift table 1A, the shift condition extracting means 24 performs a statistical test. By doing so, it is determined whether or not the rule should be extracted as the present shift condition (see step S5 in FIG. 4).

  In the following, a test performed for the provisional shift condition that the minimum number of times of “rest (holiday)” is four will be described. In this test, a hypothesis is made that “there is no rule that limits the number of occurrences of“ rest ”to 4 or more within a period”, and if the hypothesis is rejected, “the minimum number of“ rest ”within a period is 4 It is determined that there is a rule that “the time is”.

  First, the probability P (A) of occurrence of “absence” can be obtained by counting the number NA of occurrences of “absence” in the period (16 days) that is the subject of the work shift table 1A.

P (A) = NA / S
Here, S indicates the number of cells in the work shift table 1A (= 16 days × 8 workers).

  Subsequently, assuming that there is no rule, the probability of “absence” appearing for one worker within a period is obtained for each number of appearances based on the binomial distribution. The number of appearances takes a value from 0 to the number of days in the period (16 days), and the probability can be obtained according to a binomial distribution formula. For example, when P (A) = 0.281, the probability of “rest” appearing 0 times in 16 days is 0.00507, the probability of appearing once is 0.0318, the probability of appearing twice is 0.0932, and the like.

  For this reason, the probability P (A, 4, 16) that “pause” appears 4 times (minimum number of times) or more per person is the sum of the probabilities from 4 times to 16 times.

P (A, 4, 16) = 0.216 + 0.203 + 0.146 + ... = 0.700
If the number ND of employees who have four or more “days off” during such a period is ND (8 = 8), then “days off” will be all jobs unless there are any rules. The probability that a person will appear more than the minimum number of times is the eighth power of P (A, 4, 16).

  On the other hand, when a significant level (1%, 5%, etc.) is set as a predetermined condition and compared with the eighth power of P (A, 4,16), P (A, 4,16) If the 8th power is smaller, an event with a probability less than the significance level has occurred, and there is no rule that limits the number of occurrences of “rest” within the period to 4 or more. "And the hypothesis is rejected, and" there is a rule that the minimum number of times of "rest" within a period is 4 ".

  Therefore, a provisional shift condition that “the minimum number of“ rest ”within the period is four” is set as the main shift condition. When the provisional shift condition is related to the maximum number of times, the probability obtained from the binomial distribution formula is calculated by summing the probabilities from 0 to the maximum number of times and performing the test on the ND raised to the ND power. Good.

  Further, a description will be given of a case where a minimum interval or a maximum interval is extracted as a shift condition (hereinafter sometimes referred to as a “shift interval rule”) for an interval until a certain work pattern appears next.

  In this case, the preprocessing means 22 calculates the appearance interval for each data type in the worker direction as the data interval value. For example, in the work shift table 1A shown in FIG. 3, the interval of “vacation” of the worker “Sato” is 0 days, 2 days, and 5 days, and the data interval values such as the minimum interval 0 days, the maximum interval 5 days, etc. Can be calculated. In addition, the interval of “rest” for the worker “Takahashi” is 0 days, 2 days, and 4 days, and data interval values such as a minimum interval of 0 days and a maximum interval of 4 days can be calculated. Further, the interval of “rest” for the worker “Tanaka” is 5 days, 1 day, and 5 days, and data interval values such as a minimum interval of 1 day and a maximum interval of 5 days can be calculated.

  Thus, the interval at which each work style is assigned to each worker is not necessarily constant, and there may be a minimum or maximum interval that assigns a work style common to each worker or all workers as a rule. .

  Therefore, the pre-processing means 22 extracts the minimum value and the maximum value among the data interval values of all work patterns of all workers for each work pattern. Here, when creating a shift interval rule for the entire workplace, it is not necessary to associate and organize workers and data interval values.

  Subsequently, a value (predetermined value) that becomes an interval of 0 days is removed from the minimum value extracted for each work pattern. In addition, a data interval value that is equal to or less than a preset interval d days (predetermined value) is found from the maximum values extracted for each work pattern. Then, the extracted data is selected as a temporary shift condition of the minimum interval or the maximum interval of each work style (see step S4 in FIG. 4).

  Since the minimum interval or the maximum interval selected as the temporary shift condition in this way is selected based only on the number of appearances in the work shift table 1A, the shift condition extracting means 24 performs a statistical test. By doing so, it is determined whether or not the rule should be extracted as the present shift condition (see step S5 in FIG. 4).

  In the following, the test performed for the provisional shift condition that the maximum interval of “rest (holiday)” is 5 days will be described. In this test, a hypothesis that “there is no rule that limits the maximum interval of“ rest ”to 5 days” is established, and if the hypothesis is rejected, there is a rule that “the maximum interval of“ rest ”is 5 days” Is determined.

  First, the probability P (A) of occurrence of “absence” can be obtained by counting the number NA of occurrences of “absence” in the period (16 days) that is the subject of the work shift table 1A.

P (A) = NA / S
Here, S indicates the number of cells in the work shift table 1A (= 16 days × 8 workers). In addition, when thinking about after “absence” appears in the work shift table 1A, the probability P (g, A) that “day off” appears after g days in succession, Can be obtained as follows.

P (g, A) = P (A) × (1−P (A)) ^g
In addition, when “absence” appears once in the direction of the worker and then “absence” appears again, it is counted as one, and the total number of occurrences for all workers during the period is the number NE. To do. This is the number of intervals between “leaves”, and is equal to the sum of all the numbers obtained by subtracting 1 from the number of times within the “leave” period of each worker. For example, in the work shift table 1A, “Sato” is 3 times, “Suzuki” is 3 times, “Takahashi” is 4 times, “Tanaka” is 3 times, “Watanabe” is 6 times, “Ito” is 3 times, “Yamamoto” is 3 times, “Nakamura” is 3 times, and NE is 28 times.

  If there is no rule, the probability that all the intervals between “leaves” are equal to or less than the maximum interval (5 days) is (P (0, A) + P (1, A) + P (2, A) + P (3 , A) + P (4, A) + P (5, A)) to the NE power.

  On the other hand, when a significance level (1%, 5%, etc.) is set as a predetermined condition and compared with this probability, if it is less than the significance level, an event with a probability less than the significance level has occurred. That is, the hypothesis that “there is no rule that limits the maximum interval of“ rest ”to 5 days” is rejected, and “there is a rule that the maximum interval of“ rest ”is 5 days”.

  Therefore, the temporary shift condition “the maximum interval of“ rest ”is 5 days” is set as the main shift condition. When the temporary shift condition is related to the minimum interval, (1− (P (0, A) + P (1, A) +... + P (minimum interval−1, A))) raised to the NE power. The same test may be performed for.

  Furthermore, for example, for two workers, a combination that assigns a work form such as the same work form or not the same work form is extracted as a shift condition (hereinafter also referred to as “combination rule”). The case will be described.

  In this case, the preprocessing means 22 generates combination data in which data of a certain worker's cell is associated with data of another worker's cell in the date direction of the cell. For example, in the work shift table 1A shown in FIG. 3, on the 8th, 9th, and 12th when the work style of the worker “Sato” is “work”, the work style of the worker “Nakamura” is also “work”. It has become. Also, on the 3rd and 11th when the working style of the worker “Tanaka” is “night”, the working style of the worker “Yamamoto” is also “night”.

  In this way, there is a possibility that there is a rule between the two workers that the same work mode is set on the same day or not.

  Therefore, the pre-processing means 22 calculates the number of appearances and the appearance probability for each worker's work style.

  Then, for a work pattern in which the number of appearances is equal to or more than the preset nd day, a combination of workers assigned with the same work pattern on the same day is found to generate combination data. For example, as described above, the worker “Tanaka” and the worker “Yamamoto” have been assigned “night (night shift)” on the same day twice during the period (predetermined number). Therefore, the combination data of the two persons is selected as a temporary shift condition for the “night” duty pattern (see step S4 in FIG. 4).

  Since the temporary shift condition of the combination data selected here is selected based only on the number of appearances in the work shift table 1A, the shift condition extraction unit 24 performs a statistical test to perform the main shift. It is determined whether or not the rule should be extracted as a condition (see step S5 in FIG. 4).

  Below, the test | inspection performed with respect to temporary shift conditions that the worker "Tanaka" and the worker "Yamamoto" become "night" on the same day is demonstrated. In this test, the hypothesis that “There is no rule to assign“ Night ”to“ Tanaka ”and“ Yamamoto ”on the same day” is established, and if that hypothesis is rejected, “Tanaka” and “Yamamoto” It is determined that there is a rule “assign night”.

  First, the probability P (A) that “night” appears is obtained by counting the number of times “night” appears for the worker “Tanaka” in the period L (16 days) that is the subject of the work shift table 1A. Can do.

P (A) = NA / L
Similarly, the number of times “night” appears for the worker “Yamamoto” is counted. Here, since NB (3 times) is larger than NA (2 times), NB is considered as a reference.

  Subsequently, assuming that there is no rule, the probability that an event of probability P (A) occurs NA times or more in NB trials (three times) is obtained based on the binomial distribution. This can be determined according to the binomial distribution formula. For example, when P (A) = 0.125, the probability of the worker “Tanaka” appearing 0 times out of 3 times is 0.670, the probability of appearing once is 0.287, the probability of appearing twice is 0.041, 3 times The probability of appearing is 0.00195.

  If there is no rule, the probability of matching NA times (2 times) or more is the sum of all probabilities from NA times (2 times) to NB times (3 times).

  On the other hand, when a significance level (1%, 5%, etc.) is set as a predetermined condition and compared with this probability, if it is less than the significance level, an event with a probability less than the significance level has occurred. And the hypothesis that “There is no rule that assigns“ night ”to“ Tanaka ”and“ Yamamoto ”on the same day” is rejected, and “Tanaka” and “Yamamoto” have “night” on the same day. There is a rule to assign ".

  Therefore, a temporary shift condition that ““ Tanaka ”and“ Yamamoto ”assign“ night ”on the same day” is set as the main shift condition.

  In the above, the case where the rule 271 serving as the shift condition is extracted by the statistical test has been described. However, the case where the grouping 272 as the shift condition is performed using the clustering method will be described below (see FIG. 1). .

  The shift shift table 1A shown in FIG. 3 describes the shift work (duty pattern) of eight workers, but from the data of the duty pattern described in each worker's cell during the period, the worker May be classified into some groups.

  For example, if one person in a group is always “working” or “night” on any day, shift conditions can be set for each group instead of each worker. This group may result in a grouping 272 according to job title, but the rule extraction device 2 does not specifically recognize such a predetermined job title, but each duty during the period. The grouping 272 is performed only from the work style data described in the person's cell.

  First, when grouping 272, the past work shift table 1A is read as shown in the flowchart of FIG. 5 (step S11). The reading of the work shift table 1A is the same as in the case of the extraction of the rule 271 described above, and can be omitted if the above step S1 has already been executed.

  Subsequently, in the pre-processing means 22, for example, as described in the in-period number rule, the cell data in the worker direction is totaled as the total number of worker units (total data) for each data type (step S <b> 12). . That is, if the worker is “Sato”, the total data includes “work” three times, “night” three times, “early” one time, and the like. For the worker “Suzuki”, “duty” is 2 times, “night” is 3 times, and “early” is 2 times. Further, for the worker “Takahashi”, “work” is 3 times, “night” is 3 times, and “early” is 1 time. If the worker is “Nakamura”, the total data is “work” 12 times, “night” 0 times, and “early” 0 times. The other workers are also counted in the same manner, but the explanation will be continued with these four persons as an example.

  In step S13, the totalized data is converted into the following vector values (converted data) by the conversion processing means 25.

Worker "Sato" = (work, night, early) = (3, 3, 1)
Worker "Suzuki" = (work, night, early) = (2, 3, 2)
Worker "Takahashi" = (work, night, early) = (3, 3, 1)
Worker "Nakamura" = (work, night, early) = (12,0,0)
As shown in FIG. 6A, for example, the aggregated data converted into vector values has the number of “work (day shift)” as the vertical axis and the number of “night (night shift)” as the horizontal axis. Can be plotted on a two-dimensional graph.

  Then, if the number of night shifts is selected as the division axis 4A in step S14, it can be divided into a cluster (group) of “Nakamura” and a cluster (group) of “Sato”, “Takahashi” and “Suzuki” ( Step S15).

  This division is performed by the clustering means 26, but actually, the vector values are not plotted on the graph, but the clustering is performed by calculating the distance between the position of the division axis 4A and the vector value of each worker. Here, the position of the dividing axis 4A can be set to a position where the sum of the error in distance is minimized by the method of least squares.

  In step S16, it is examined whether the clusters of “Sato”, “Takahashi”, and “Suzuki” can be further divided. For example, as shown in FIG. 6B, the vector values of three persons are plotted on a two-dimensional graph with the number of “night (night shift)” as the vertical axis and the number of “early (early departure)” as the horizontal axis. be able to.

  If the number of early appearances is selected as the division axis 4B, it can be divided into a cluster of “Suzuki” and a cluster of “Sato” and “Takahashi”. Such clustering is terminated when the division axis 4B is so close that it cannot be divided. Here, since “Sato” and “Takahashi” have the same number of “early” and cannot be divided by the division axis 4B having the number of early appearances, clustering is terminated.

  Actually, since there are more types of work patterns than the above three types, assuming that the number of work patterns (types) is NK, the vector values of each worker are plotted in the NK-dimensional hypercube. . .. Are used as the division axes 4A, 4B,... In order from the number of work modes (types) having a large variance in the number of appearances, and clustering is performed sequentially.

  In such clustering, if the difference in the number of appearances at which the division is stopped is dmin, the set of workers who finally fit in the NK-dimensional hypercube with one side dmin is the same group.

  When the workers listed in the work shift table 1A shown in FIG. 3 are clustered by the clustering means 26, cluster 1 = “Nakamura”, cluster 2 = “Sato, Takahashi, Tanaka”, cluster 3 = “Suzuki, Watanabe” , Ito, Yamamoto ”, grouping 272.

  Next, the operation of the rule extraction device 2 of the present embodiment will be described.

  The rule extraction device 2 of the present embodiment configured as described above reads the past work shift table 1A by the reading unit 21, selects a temporary shift condition from the acquired data, and sets the temporary shift condition as the main shift condition ( The rule 271) is statistically tested.

  For this reason, the shift condition 270 to be applied when actually creating the work shift plan table can be accurately extracted based on the probability. In addition, since it is not necessary for the user to determine whether or not the rule 271 should be changed, a shift condition such as an implicit rule that has not been recognized can be extracted from the past work shift table 1A. Even a user who is not skilled in creating a work shift plan table can easily create a work shift plan table.

  Moreover, the shift condition (shift arrangement rule) which designates or prohibits the arrangement | sequence of the work form of the continuous several days is extracted by producing | generating the arrangement | sequence data which linked | related the data of the taken work shift table 1A with a worker direction. be able to.

  Furthermore, by shifting the data of the captured work shift table 1A for each data type in the date direction, it is possible to extract a shift condition (necessary number rule) regarding the necessary number of persons per day.

  Further, by collecting the data of the taken work shift table 1A for each data type in the direction of the worker, it is possible to extract a shift condition (number of times rule within the period) regarding the number of times that a certain type of work mode is performed within the period. it can.

  Furthermore, by calculating the appearance interval for each data type as data interval value from the data for the worker direction in the captured work shift table 1A, a shift condition (shift interval rule) relating to the interval for performing a certain type of work style is extracted. can do.

  In addition, by generating combination data in which the data of the captured work shift table 1A is associated with the data of other workers in the date direction, shift conditions relating to worker combinations such as working or not working on the same day ( Combination rule) can be extracted.

  Further, if clustering is performed by the clustering means 26 using the data of the taken work shift table 1A and the finally classified clusters (groups) are extracted as shift conditions (grouping 272), the work shift plan table is obtained. It becomes possible to make adjustments in units of groups that do not rely on workers alone when creating Moreover, although the manager and the worker are not recognized, it is possible to find out the grouping 272 necessary for creating the work shift plan table.

  The embodiment of the present invention has been described in detail above with reference to the drawings. However, the specific configuration is not limited to this embodiment, and design changes that do not depart from the gist of the present invention are not limited to this embodiment. Included in the invention.

  For example, in the above embodiment, the work shift tables 1 and 1A in which the worker direction is the horizontal direction and the date direction is the vertical direction have been described, but the present invention is not limited to this, and the worker direction is the vertical direction and the date. The present invention can also be applied to a work shift table in which the direction is horizontal.

  In the above embodiment, the case where the past work shift table 1A is read and the shift condition 270 is extracted has been described. However, when there is no past work shift table 1A, a work shift table serving as a teacher is provided. It may be created and read.

  Furthermore, although the period of the work shift table 1A described in the above embodiment is 16 days, it is not limited to this, and any one of 1 week, 1 month, 3 months, 6 months, 1 year, etc. A period can be set. Further, the number of workers is not limited to the number of workers described in the work shift table 1A.

  Moreover, although the said embodiment demonstrated the test method using a significance level, it is not limited to this. For example, it is possible to conduct a test using a hypothesis adopting method or a reject method defining a loss. This test method shifts after evaluating the loss L1 of “when it is determined that the existing shift condition does not exist by mistake” and the loss L2 of “when it is determined that the non-existent shift condition exists by mistake”. This is a method for determining whether to extract a condition. When the loss L1 is greater than the loss L2, the test condition is relaxed, and when the loss L2 is greater than the loss L1, the test condition is tightened.

  Furthermore, in the above-described embodiment, the arrangement data for two days is generated. However, the present invention is not limited to this, and arrangement data for three days or more can also be generated.

1,1A (past) work shift table 2 rule extraction device (work shift table creation support system)
21 Reading means 22 Preprocessing means 23 Temporary shift condition setting means 24 Shift condition extracting means 25 Conversion processing means 26 Clustering means 27 Output means 270 Shift condition 271 Rule (main shift condition)
272 Grouping (shift condition)

Claims (7)

A work shift table creation support system that extracts shift conditions necessary for creating a work shift plan table,
Reading means for capturing a past work shift table having a date direction associated with a date and a worker direction associated with a worker or a work shift table serving as a teacher;
Preprocessing means for organizing the data of each cell of the work shift table captured by the reading means in association with the data of the worker direction or the date direction;
Temporary shift condition setting means for selecting candidates for the shift condition from the processing data arranged by the preprocessing means as temporary shift conditions;
Shift condition extracting means for performing a statistical test based on the data of the work shift table and extracting a temporary shift condition satisfying a predetermined condition as the main shift condition;
A work shift table creation support system comprising output means for outputting the extracted main shift condition. In the pre-processing means, to generate array data that associates the data of each cell and the data of the cell adjacent to the worker direction,
The provisional shift condition setting means selects arrangement data that appears more than a predetermined number of times in the arrangement data as a provisional shift condition,
2. The work shift table creation support system according to claim 1, wherein the shift condition extraction unit uses the shift condition as a main shift condition when the temporary shift condition is established as a shift condition by comparison with a significance level. In the pre-processing means, the data of the cell in the date direction is totaled as a daily total number for each data type,
The provisional shift condition setting means selects a provisional shift condition by comparing the daily total number and a predetermined value,
2. The work shift table creation support system according to claim 1, wherein the shift condition extraction unit uses the shift condition as a main shift condition when the temporary shift condition is established as a shift condition by comparison with a significance level. In the pre-processing means, the data of the worker-oriented cells is totaled as the total number of workers for each data type,
The provisional shift condition setting means selects a provisional shift condition by comparing the total number of worker units and a predetermined value,
2. The work shift table creation support system according to claim 1, wherein the shift condition extraction unit uses the shift condition as a main shift condition when the temporary shift condition is established as a shift condition by comparison with a significance level. In the preprocessing means, the appearance interval for each data type in the worker direction is calculated as a data interval value,
The provisional shift condition setting means selects a provisional shift condition by comparing the data interval value with a predetermined value,
2. The work shift table creation support system according to claim 1, wherein the shift condition extraction unit uses the shift condition as a main shift condition when the temporary shift condition is established as a shift condition by comparison with a significance level. The pre-processing means generates combination data in which the data of each cell is associated with the data of another worker's cell in the date direction,
In the temporary shift condition setting means, combination data that appears a predetermined number of times or more in the combination data is selected as a temporary shift condition,
2. The work shift table creation support system according to claim 1, wherein the shift condition extraction unit uses the shift condition as a main shift condition when the temporary shift condition is established as a shift condition by comparison with a significance level. A work shift table creation support system that extracts shift conditions necessary for creating a work shift plan table,
Reading means for capturing a past work shift table having a date direction associated with a date and a worker direction associated with a worker or a work shift table serving as a teacher;
Preprocessing means for organizing the data of each cell of the work shift table captured by the reading means in association with the data for the worker direction;
Conversion processing means for converting the processing data arranged by the preprocessing means into conversion data;
Clustering means for performing clustering using the converted data;
A work shift table creation support system, comprising: output means for outputting clusters classified by the clustering means as shift conditions.