JP6740860B2 - Safety stock determination device, method and program - Google Patents

Description

The present invention relates to objects (raw materials, materials, products, products, etc.) delivered from one or more shipping sources (production sites, etc.) to multiple delivery destinations (factories, commercial facilities, inventory sites, etc.). The present invention relates to a safety stock determination device, method, and program suitable for use in determining a safety stock to be held in advance.

At steel manufacturers, chemical manufacturers, oil manufacturers, electric power companies, etc., raw materials, materials, products, products, etc. are imported from overseas by ships and stored as inventory. Too much inventory will slow down cash flow and increase management costs, so it is desirable to keep a small inventory. On the other hand, if the inventory is too small, it is not possible to fully absorb the fluctuations in supply and demand, which may lead to production stoppages due to shortages. Therefore, in order to keep the out-of-stock rate of objects below a certain level, it can be said that it is essential to predict fluctuations in supply and demand and set the average inventory (appropriate inventory) as a management target.
However, when the target is imported by ship from all over the world, the lead time from ordering to delivery is very long, and it is easily affected by variations in supply and demand. For this reason, inventory tends to fall into excess or understatement, and once it falls into that state, losses are large. In addition, the supply of objects varies greatly due to the occurrence of sudden interruptions due to unseasonable weather, equipment troubles, and frequent fluctuations in the operation of ships that carry enormous amounts of cargo. Therefore, it was difficult to mathematically derive an appropriate inventory.

Non-Patent Document 1 presents a method of giving an appropriate inventory by adding cycle inventory and safety inventory as an appropriate inventory setting method for predicting fluctuations in supply and demand. In FIG. 15, the horizontal axis represents time, and the vertical axis represents the amount of inventory, showing the transition of inventory and showing the relationship between cycle inventory, safety inventory, and proper inventory. The cycle inventory is a running inventory that has a constant fluctuation in inventory, and the safety inventory is an inventory that does not run out of stock even if supply and demand vary. In Non-Patent Document 1, the cycle inventory is given by half of the average value of the incoming lot, and the safety inventory is calculated by "safety factor x maximum lead time ^1/2 x standard deviation of demand amount". Here, the maximum lead time is given by “average lead time+safety factor×standard deviation of lead time”. The safety factor is a parameter that determines how much delay is allowed, and the practitioner determines it according to the situation.

In Patent Document 1, the supply lead time fluctuates such that the order interval is an exponential distribution, the order delivery date is a normal distribution, the order quantity is a Γ distribution as the demand information, and the supply lead time is a Γ distribution as the supply information. Discloses a method of formulating an inventory plan that takes into consideration the trade-off relationship between improved delivery date compliance and inventory cost control.

JP, 2014-119768, A

Takao Suguro, "Appropriate Inventory Concepts and How to Find Them", Published September 10, 2003, Nikkan Kogyo Shimbun Emery N. Brown et al., The Time-Rescaling Theorem and Its Application to Neural Spike Train Data Analysis, Neural Computation 14, 325-346, 2001 Yosihiko Ogata, On Lewis' Simulation Method for Point Processes, IEEE Transactions On Information Theory, Vol. 27, No.1, January 1981

By the way, since the objects supplied by sea transportation are delivered from various regions outside the country, the lead time differs depending on the order. However, Non-Patent Document 1 considers only when the lead time is uniform, and when the lead time is different for each order, when the arrival (entry) of the ship is delayed much longer than the set lead time, This leads to a situation where the stock runs out.
Further, since the target object is transported from a distant foreign country by ship, the lead time from ordering to delivery is very long, from one week to several months. In addition, departure times vary greatly at loading points due to unseasonable weather, equipment failure, and interchange with ships of other companies. In addition, at the landing site, inventory supply timing and departure time will fluctuate due to the effects of port unloading capacity necks, warehouse capacity necks, offshore vessels, and unloading equipment failures.
Therefore, if the originally scheduled departure/arrival time of a ship is significantly changed, in order to prevent excess/understock at multiple delivery destinations, the delivery/delivery destinations should be reviewed, including updating the delivery time. Delivery plans are constantly reviewed and updated. In this way, if the ordering source or the delivery destination frequently changes after the order is placed once, the relationship between the ordering point and the delivery time necessary for calculating the lead time becomes unknown. That is, the lead time determined by using the relationship between the order point and the delivery time cannot be calculated because the order point itself is unstable.
Note that it is possible to compare the schedule with the actual result before a certain period of time and calculate the lead time from the difference. However, the variation of the lead time calculated by this method is unstable because it changes depending on how to select the scheduled period, and the lead time from ordering to delivery disclosed in Non-Patent Document 1 and Patent Document 1 was used. The safety stock calculation method causes excess/understock.
Furthermore, both Non-Patent Document 1 and Patent Document 1 only consider replenishment of an object to one base or a plurality of bases and the inventory thereof from one production base. It is not described when the target is replenished at any one of one base or multiple bases.

The present invention has been made in view of the above points, and has a long lead time from ordering to delivery of an object delivered from one or more shipping sources to a plurality of delivery destinations, and is unstable. The objective is to be able to appropriately determine the safety stock that each delivery destination should have, even under these circumstances.

The gist of the present invention for solving the above problems is as follows.
[1] A safety stock determination device that determines a safety stock to be held at each delivery destination for an object delivered from one or more shipping sources to a plurality of delivery destinations,
For an object for which a safety stock is desired to be determined, as past performance information, delivery time information, and an input means for capturing usage amount information that is information on the usage amount of the object for each delivery destination,
Based on the delivery time information captured by the input means, it is assumed that a plurality of delivery destinations included in the delivery time information are one delivery destination (hereinafter, referred to as all delivery destinations), and delivery intervals to all the delivery destinations. A first probability distribution calculating means for obtaining a probability distribution of
Based on the usage amount information taken in by the input means and the probability distribution of the delivery intervals to all the delivery destinations obtained by the first probability distribution calculation means, the delivery destination information contained in the delivery time information is delivered to each delivery destination. Second probability distribution calculating means for obtaining a probability distribution of delivery intervals;
Based on the usage amount information taken in by the input means and the probability distribution of the delivery intervals to the delivery destinations obtained by the second probability distribution calculation means, the safety stock to be held at each delivery destination is calculated. A safety stock determination device comprising: safety stock calculation means.
[2] The second probability distribution calculation means uses the object for each delivery destination for delivery to the delivery destination i (i is a symbol representing the delivery destination) among the number of deliveries to all the delivery destinations. Given once every N _i times according to the quantity, give the delivery interval to the delivery destination i by the probability distribution obtained by convoluting the probability distribution of the delivery intervals to all the delivery destinations N _i times. The safety stock determination device according to [1].
[3] the second probability distribution calculating means, delivery times q _i to delivery destination i, as delivery times Q to all delivery destination, and characterized in providing a number N _i of the convolution with Q / q _i The safety stock determination device according to [2].
[4] A delivery delay allowable value setting means for setting the delivery delay allowable value α% is provided,
The safety stock calculation means further covers the probability distribution of the delivery intervals to each of the delivery destinations by 1-α% based on the delivery delay tolerance α% set by the delivery delay tolerance setting means, and The difference from the average of the probability distribution of the delivery intervals to each delivery destination is taken as the safety stock days, and this safety stock days is multiplied by the average usage amount for each delivery destination to calculate the safety stock that each delivery destination should have. The safety stock determination device according to any one of [1] to [3].
[5] The delivery to all the delivery destinations includes delivery directly from a shipping source and delivery via another delivery destination, [1] to [4] Safety stock determination device described in 1.
[6] The first probability distribution calculation means assumes that the probability distribution of delivery intervals to all the delivery destinations follows an exponential distribution or a gamma distribution,
The second probability distribution calculation means is such that the probability distribution of the delivery intervals to each of the delivery destinations follows a gamma distribution, and the safety stock determination according to any one of [1] to [5]. apparatus.
[7] The first probability distribution calculating means uses the probability distribution of delivery intervals to all the delivery destinations as an average number of deliveries per unit time (hereinafter, referred to as an average delivery rate) to all the delivery destinations. The exponential distribution or gamma distribution
The safety stock determination device according to [6], wherein the average delivery rate is treated as constant.
[8] The first probability distribution calculating means uses a probability distribution of delivery intervals to all the delivery destinations as an average number of deliveries per unit time (hereinafter referred to as an average delivery rate) to all the delivery destinations. The exponential distribution or gamma distribution
The safety stock determination device according to [6], wherein the average delivery rate is modeled as a function that changes with time.
[9] The safety stock determination device according to [8], wherein the first probability distribution calculation means determines the parameter of the function based on a predetermined information amount criterion.
[10] The safety stock determination device according to [8] or [9], wherein the function is represented by an exponential Fourier series.
[11] The first probability distribution calculating means obtains a future average delivery rate, which is a target period for determining safety stock, as a constant multiple of the past average delivery rate [8] to [10]. ] The safety stock determination device described in any one of.
[12] The first probability distribution calculating means obtains the constant, which is the constant multiple, as a value obtained by dividing the average of future average delivery rates by the average of past average delivery rates,
The safety stock determination according to [11], wherein the average of the future average delivery rate is estimated by a single regression analysis using past performance information, as being proportional to the cumulative arrival amount of the object. apparatus.
[13] The object is delivered from the shipping source to the delivery destination by sea transportation by ship,
Using the arrival time information as the delivery time information,
Using the probability distribution of the arrival interval of the ship to all the delivery destinations as the probability distribution of the delivery interval to all the delivery destinations,
Safety stock determination according to any one of [1] to [12], characterized in that a probability distribution of ship arrival intervals to each destination is used as a probability distribution of delivery intervals to each destination. apparatus.
[14] A safety stock determination method for determining a safety stock to be held at each delivery destination for an object delivered from one or more shipping sources to a plurality of delivery destinations,
Input means, for the object for which safety stock is desired to be determined, as past performance information, delivery time information, and a step of capturing usage amount information which is information on the usage amount of the object for each delivery destination,
The first probability distribution calculation means assumes that a plurality of delivery destinations included in the delivery time information is one delivery destination based on the delivery time information fetched by the input means (hereinafter, referred to as all delivery destinations). , A step of obtaining a probability distribution of delivery intervals to all the delivery destinations,
The second probability distribution calculation means, based on the usage amount information taken in by the input means, and the probability distribution of the delivery intervals to all the delivery destinations obtained by the first probability distribution calculation means, the delivery time. Determining a probability distribution of delivery intervals to each delivery destination included in the information,
The safety stock calculation means has each of the delivery destinations based on the usage amount information taken in by the input means and the probability distribution of the delivery intervals to the delivery destinations obtained by the second probability distribution calculation means. And a step of calculating a safety stock to be stored.
[15] A program for determining a safety stock to be held at each delivery destination for objects delivered from one or more shipping sources to a plurality of delivery destinations,
For an object for which a safety stock is desired to be determined, as past performance information, delivery time information, and an input means for capturing usage amount information that is information on the usage amount of the object for each delivery destination,
Based on the delivery time information captured by the input means, it is assumed that a plurality of delivery destinations included in the delivery time information are one delivery destination (hereinafter, referred to as all delivery destinations), and delivery intervals to all the delivery destinations. A first probability distribution calculating means for obtaining a probability distribution of
Based on the usage amount information fetched by the input means and the probability distribution of the delivery intervals to all the delivery destinations obtained by the first probability distribution calculation means, the delivery destination information contained in the delivery time information is delivered to each delivery destination. Second probability distribution calculating means for obtaining a probability distribution of delivery intervals;
Based on the usage amount information taken in by the input means and the probability distribution of the delivery intervals to the delivery destinations obtained by the second probability distribution calculation means, the safety stock to be held at each delivery destination is calculated. A program that causes a computer to function as a safety stock calculation means.

According to the present invention, regarding an object delivered from one or more shipping sources to a plurality of delivery destinations, the lead time from ordering to delivery is long, and even in an unstable situation, each delivery destination has it. It is possible to properly determine the safety stock to be used.

It is a figure which shows the function structure of the safety stock determination apparatus which concerns on 1st Embodiment. It is a figure which shows typically the concept of the shipment and delivery of the target object in 1st Embodiment. It is a flowchart which shows the safety stock determination process by the safety stock determination apparatus which concerns on 1st Embodiment. It is a figure which shows the example of the arrival time information for every delivery destination about a certain target object. It is a figure which shows the example of the amount-of-use information for every delivery destination about the product name X by day. It is a figure which shows the example of the histogram of the arrival interval of the ship to all the delivery destinations, and the approximate curve by exponential distribution. It is a figure which shows the example of the histogram of the arrival interval of the ship to the delivery destination B, and the approximate curve by a gamma distribution. FIG. 7 is a characteristic diagram showing an example of a result of calculating an inventory transition by giving an initial inventory by the safety inventory determination method according to the first embodiment. It is a figure which shows the function structure of the safety stock determination apparatus which concerns on 2nd Embodiment. It is a flowchart which shows the safety stock determination process by the safety stock determination apparatus which concerns on 2nd Embodiment. It is a figure which shows the example of the arrival time information for every delivery destination about a certain target object. It is a characteristic view which shows the estimation result of average arrival rate (lambda) (s), and the example of the graph which displayed the line segment at the arrival time to all the delivery destinations. It is a figure which shows the example of the graph which compared the arrival interval of the ship to each delivery destination with the sample randomly generated by the dilution algorithm. It is a figure which shows the example of the result of having calculated the safety stock days for every delivery destination. It is a figure for explaining the way of thinking of stock.

Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings.
[First Embodiment]
First, an outline of a safety stock determination method to which the present invention is applied will be described.
The present inventor, in order to suppress out-of-stock of objects (raw materials, materials, products, commodities, etc.) supplied by sea transportation, the safety stock is absorbed by the variation in arrival intervals of ships supplying the objects. I thought I should have. That is, safety stock should be given not from variations in lead time but from the difference between the average arrival interval of ships and the maximum arrival interval of ships by modeling the arrival intervals of ships at the delivery destination with a probability distribution. Thought. If the safety stock is set in this way, even if the ships arrive at different intervals from the average arrival interval, it is possible to manage the stock at the delivery destination (factory, commercial facility, stock base, etc.) so as to prevent the target from being out of stock.

For the above calculation, it is necessary to model the probability distribution of the arrival intervals of the ship. Therefore, as detailed below, we attempted to construct a probability distribution of ship arrival intervals for each delivery destination by analyzing ship dispatching work and statistically modeling ship dispatching.
First, the present inventor analyzed the arrival interval of each ship to each delivery destination, and found that the arrival interval had a probability distribution close to a gamma distribution. The gamma distribution is a distribution that represents a time interval until an arrival event that occurs according to an exponential distribution occurs a plurality of times, and is often used to calculate a life time distribution of equipment and a failure rate in a practical problem.

Based on this knowledge, the present inventor considered that the arrival phenomenon of a ship at each delivery destination may be caused by a phenomenon that follows a specific probability distribution such as an exponential distribution. Therefore, assuming a plurality of delivery destinations as one delivery destination (hereinafter, referred to as all delivery destinations), paying attention to the arrival at all delivery destinations of the ship, and calculating the arrival interval. As a result, it was found that the arrival intervals of ships to all delivery destinations tended to follow a highly random probability distribution such as an exponential distribution or a gamma distribution. In this case, the arrival interval of the ship to all the delivery destinations is not only the arrival from the shipping source to the first delivery destination, but also when the ship arriving in Japan goes to multiple destinations and goes to multiple destinations. Obtained by treating arrival at the port as one arrival at all delivery destinations.
I thought that the reason why the ships randomly arrived at all the delivery destinations was that they were affected by various influences in the process of arrival at all the delivery destinations. For example, the departure time of a ship that is transported from a distant place in the world varies due to accommodation with other companies' ships, waiting for loading, equipment troubles, and the like. Furthermore, ships have multiple lead times and vary due to unseasonable weather and ocean currents. In addition, even after a ship arrives at a base in Japan, the weather will not only affect it, but congestion at the port and inventory conditions at the unloading destination will also cause the ship to stay at sea. The time varies. In this way, when ships affected by various variations are mixed in the process of arriving at all delivery destinations, it is thought that the randomness will eventually become stronger and the randomness will be closer to a strong probability distribution.

In consideration of the above, the inventor of the present invention does not carry out the ship-dispatching work of the target object not between the two bases independently, but as shown in FIG. We considered that ships that arrived at all delivery destinations according to a highly probable probability distribution could be regarded as the task of allocating to each delivery destination according to their usage. Then, based on this idea, the arrival of the ship is divided into two stages, "arrival of all destinations of the ship" and "arrival of each destination of the ship", and the arrival interval of each destination of the ship The probability distribution of was modeled by the convolution integral of the probability distribution of the arrival intervals of all ships, and the method to calculate the safety stock that should be held at each destination was established.

Hereinafter, the safety stock determination method according to the first embodiment to which the present invention is applied will be described.
FIG. 1 shows a functional configuration of a safety stock determination device 100 according to the first embodiment. The safety stock determination device 100 is used to determine the safety stock to be held at each delivery destination for an object delivered from one or a plurality of shipping sources to a plurality of delivery destinations. For example, iron ore and coal in a steel maker, crude oil in a chemical maker and an oil maker, and LNG and coal in an electric power company are delivered from domestic shipping destinations to domestic destinations using ships. In such a case, the present invention is widely applicable to calculate the safety stock to be held at each delivery destination.

Reference numeral 300 is a database in which performance information is accumulated and saved. The record information includes port arrival time information for each object/delivery destination, and usage information for each object/delivery destination/day.

Reference numeral 101 denotes an input unit which is an input means, and from the database 300, regarding the object (product name X) for which safety stock is to be determined, arrival time information by delivery destination and usage information by delivery destination/day Take in and. It should be noted that the actual result information of the entire period stored in the database 300 may be fetched, or the actual result information of the analysis target period designated by the user may be fetched.
FIG. 4 shows an example of port arrival time information for the product name X for each delivery destination. The arrival time information for each delivery destination is given as a matrix of the name of the delivered delivery destination, the arrival time at the delivery destination, and the arrival amount [t]. The arrival time information is an example of “delivery time information” in the present invention.
FIG. 5 shows an example of the usage amount information of the product name X by delivery destination and by day. The usage amount information for each delivery destination and each day is given as a matrix of the delivery destination name used, the date used, and the usage amount [t].

Reference numeral 102 denotes a delivery delay allowable value setting unit which is a delivery delay allowable value setting means, and sets a delivery delay allowable value. The delivery delay allowable value is set, for example, as a delivery delay ratio [%]. The delivery delay allowable value can be appropriately set by the user via the input device 107, and may be set for each object and each delivery destination. Alternatively, default values may be used.

Reference numeral 103 denotes a first probability distribution calculation unit which is a first probability distribution calculation unit, and based on the arrival time information for each delivery destination captured by the input unit 101, a plurality of ports included in the arrival time information for each delivery destination. Assuming that each of the delivery destinations is one delivery destination (hereinafter, referred to as all delivery destinations), the probability distribution of the arrival intervals of the ship to all delivery destinations is obtained. In the example of FIG. 4, the delivery destinations are C, C, B,... In chronological order, but the delivery destinations are not identified, and the ship sends all delivery destinations to “2015/3/10 0 Then, the probability distribution of the interval is calculated assuming that the arrival has been made in the order of ":00", "2015/3/19 14:27", "2015/3/20 5:20",.... The probability distribution of arrival intervals of ships to all delivery destinations is an example of the “probability distribution of delivery intervals to all delivery destinations” in the present invention.

Reference numeral 104 denotes a second probability distribution calculation unit, which is a second probability distribution calculation unit, and is calculated by the first probability distribution calculation unit 103 and the usage amount information for each delivery destination captured by the input unit 101. Based on the probability distribution of the arrival intervals of all the ship's delivery destinations, the probability distribution of the arrival intervals of each of the ship's delivery destinations (a plurality of delivery destinations included in the arrival time information for each delivery destination) is obtained. The probability distribution of the arrival interval of the ship to each delivery destination is an example of the "probability distribution of the delivery interval to each delivery destination" in the present invention. As shown in FIG. 2, the ship loaded with the product of the product name X arrives at all the delivery destinations Q (that is, the number of deliveries to all the delivery destinations is Q), and the amount of use according to the delivery destinations is changed. I think it will be dispatched as q _A , q _B ,.... Here, it is assumed that a delivery destination i is regularly dispatched once every N _i =Q/q _i (that is, if delivery to the delivery destination i is once every N _i) Assuming that the arrival interval of the ship to the delivery destination i is a probability distribution obtained by convoluting the probability distribution of the arrival intervals of the ship to all the delivery destinations N _i times. i is a symbol representing the delivery destination, and i=A, B,..., Z in the example of FIG.

Reference numeral 105 denotes a safety stock calculation unit, which is a safety stock calculation means, and includes usage information for each delivery destination/day captured by the input unit 101, a delivery delay allowable value set by the delivery delay allowable value setting unit 102, and The safety stock to be held at each delivery destination is calculated based on the probability distribution of the arrival intervals of the ship to each delivery destination obtained by the probability distribution calculation unit 104 in FIG.

An output unit 106 is an output unit that outputs the result of the safety stock calculated by the safety stock calculation unit 105. For example, the result is displayed on the display 108, or the result is sent to an external device of the device 100.

Reference numeral 107 is an input device such as a pointing device or a keyboard, and 108 is a display.

Next, a safety stock determination method by the safety stock determination device 100 will be described.
FIG. 3 is a flowchart showing a safety stock determination process by the safety stock determination device 100 according to the first embodiment.
In step S1, the input unit 101 determines, from the database 300, the arrival time information for each delivery destination (see FIG. 4) and the delivery destination/day for the object (product name X) whose safety stock is to be determined. Usage amount information (see FIG. 5) and

Next, in step S2, the delivery delay allowable value setting unit 102 sets the delivery delay allowable value α%. The delivery delay allowable value α% is set as a delivery delay ratio [%].

Next, in step S3, the first probability distribution calculation unit 103 obtains the probability distribution of the arrival intervals of the ship to all the delivery destinations. It does not matter whether the ship arrives at all the delivery destinations, whether it is a direct arrival from the shipper or an arrival via another port (multiport unloading).
FIG. 6 is a diagram showing a histogram of arrival intervals of ships to all delivery destinations calculated from the arrival time information of FIG. 4 and an approximate curve by exponential distribution. As can be seen from FIG. 6, the arrival intervals of the ship to all delivery destinations follow an exponential distribution. Here, the exponential distribution is a probability distribution whose probability density function is given by Expression (1), and represents the arrival interval of randomly occurring discrete events. The reason why ships arrive at all delivery destinations at random is because ships shipped from a long distance overseas are greatly affected by bad weather and equipment troubles, so ships from various shipping sources arrive in Japan. It is considered that when they were mixed in the process of arrival, the randomness eventually became stronger and became closer to the exponential distribution. λ is the average arrival rate of all ships to the delivery destination, and represents the average number of arrivals of the ship per unit time. As a result of estimating λ using the maximum likelihood estimation method with respect to the arrival time information in FIG. 4, the value was 1.0 [ship/day]. The average arrival rate λ to all delivery destinations of the ship is an example of the “average delivery rate to all delivery destinations (average number of deliveries per unit time)” in the present invention.

In addition, the arrival interval of the ship to all the delivery destinations may follow the gamma distribution. When ships that generate according to the exponential distribution are shared by multiple companies, the company-to-company arrival intervals from a certain company may follow the gamma distribution.

Next, in step S4, the second probability distribution calculation unit 104 obtains the probability distribution of the arrival interval of the ship at each delivery destination.
As shown in FIG. 2, the ship loaded with the product of the product name X arrives at all the delivery destinations, Q ships, and is distributed as q _A ships, q _B ships,... Think ship. Here, it is assumed that a delivery destination i is always dispatched once every N _i =Q/q _i vessels. Under this assumption, the delivery destination i is dispatched once every N _i vessels arrive at all the delivery destinations. That is, the probability distribution of the arrival intervals of the ships to the delivery destination i is given as a distribution of the time until the number of the ships arriving at all the delivery destinations becomes N _i according to the exponential distribution. Time behavior according to the exponential distribution until the predetermined number of times occur is given by N _i times the convolution integral of the exponential distribution, the general gamma distribution. That is, assuming that the probability distribution of the arrival interval of the ship to the delivery destination i is g _i (t), it is possible to model with a gamma distribution as shown in Expression (2) by using two parameters of N _i and λ.

The number N _{i of} convolution integration is not limited to the above calculation method. For example, when it is desired to determine N _i in accordance with the usage rate of each delivery destination, the value obtained by multiplying the number Q of incoming ships at all delivery destinations by the usage rate of each delivery destination can be given as N _i . Specifically, two delivery destinations A and B are assumed, and it is assumed that the ratios of the respective usage amounts are 0.4 and 0.6. In this case, the delivery destination A, B each N _i is, Q × 0.4 for delivery destination A, if Q × 0.6 for delivery destination B, corresponding to the ratio of the amount _Ni is determined.

FIG. 7 is a diagram showing a histogram of arrival intervals of the ship to the delivery destination B and an approximate curve based on the gamma distribution created using Equation (2) by calculating λ and N _i for the delivery destination B, respectively. is there. The number of arrivals q _{i at} each delivery destination and the parameters λ and N _i of the gamma distribution were calculated from the arrival time information in FIG. 4, and the values shown in Tables 1 to 3 were given. A curve close to the actual result can be obtained only by using two parameters of λ and N _i .

In order to ensure the validity of this approximated curve, it is desirable to carry out a statistical test using the Kolmogorov-Smirnov test (KS test) or the like. The KS test is a method of establishing a null hypothesis that two populations or probability distributions follow the same distribution, and testing whether or not this null hypothesis holds. In the KS test, the probability p value that this null hypothesis holds is calculated. If the p value is 0.05 or less, the null hypothesis is rejected and the two populations or distributions do not follow the same distribution. Can be tested. This time, the KS test was performed on the histogram and the approximate curve in FIG. 7 with a significance of 5%. As a result, the p value was 0.62, which was a value sufficiently larger than 0.05, and the null hypothesis that the histogram and the approximate curve follow the same distribution was not rejected.

Next, in step S5, the safety stock calculator 105 calculates the safety stock that each delivery destination should have.
Under the probability distribution of the arrival interval of the ship to the delivery destination i given by the equation (2), in order to keep the delivery time achievement rate at 1-α% or more, the probability distribution g _i (t) is set at 1-α. % The difference between the coverage points and the average value of the probability distribution g _i (t) may be given as the safety stock days. Let Z [day] be the safety stock days, G _i (.) be the cumulative distribution function of the probability distribution g _i (t) of the arrival interval of the ship to the destination i, and let the inverse function of the cumulative distribution function be G ⁻¹ (・) is put, the number of days of safety stock that keeps the delivery rate at 1-α% or more is given by equation (3).

However, the safety stock days calculated by the formula (3) are formulas for keeping the delivery delay at α% or less, and the stockout will not occur at this probability. This is because the safety stock days calculated by the equation (3) are based on the assumption that the amount of stock used is constant. In practice, if it is known in advance that the stock of a certain object is low, the production will be adjusted in advance by anticipating the expected arrival time of the ship and transferring it to other objects in advance so that the stock will not be out of stock. it can.

Table 4 shows the result of calculating the safety stock days [days] for each delivery destination by setting the delivery delay allowable value α% as 5%.
The safety stock is calculated by calculating the average usage [t/day] for each delivery destination from the usage information in FIG. 5 and multiplying it by the safety stock days [days] obtained in Table 4. Table 5 shows the calculation results of the average usage [t/day]. Table 6 shows the result of calculating the safety stock [t] by multiplying the safety stock days [days] by the average usage [t/day].

In this way, the safety stock can be calculated by using the convolution integral calculation by giving the probability distribution of the arrival intervals of the ship to all the delivery destinations and the number of ships assigned to each delivery destination of the ship.
FIG. 8 is a result of calculating the inventory transition by giving the initial inventory at the delivery destination B by the safety inventory determination method according to the present embodiment. If you use the safety stock calculated this time, you can see that there is no out-of-stock. Further, Table 7 shows the result of calculating the safety stock for a plurality of objects. In this way, by presenting the safety stock of various objects to the user for each delivery destination, the stock can be appropriately maintained.

Next, in step S6, the output unit 106 outputs the result of the safety stock calculated in step S5.

As described above, the lead time from ordering to delivery is long and unstable with regard to the objects delivered from one or more shipping destinations overseas or from multiple ports to multiple destinations by sea transportation. Even under the circumstances, it is possible to appropriately determine the safety stock to be held at each delivery destination so as to maintain the delivery rate achievement rate above a certain level. Thus, for example, by comparing the determined safety stock with the current safety stock, it becomes possible to determine the amount of inventory for each delivery destination, and a large inventory reduction is expected.
In addition, when determining safety stock, the arrival of the ship is divided into two stages: "arrival at all destinations of the ship" and "arrival at each destination of the ship", and The probability distribution of arrival intervals is modeled by the convolution integral of the probability distribution of arrival intervals to all delivery destinations of the ship. For example, it is possible to directly calculate the probability distribution of the arrival intervals of each ship from the actual information, but the accuracy is inferior to that of calculating the probability distribution of the arrival intervals of all the ship's destinations. .. On the other hand, in the method to which the present invention is applied, the arrival of a ship is divided into two stages, “arrival of all destinations of the ship” and “arrival of each destination of ship”, and the number of N is large. Since the probability distribution of the arrival intervals of the ship to all the delivery destinations is accurately obtained and the probability distribution of the arrival intervals of the ship to each delivery destination is obtained, the accuracy of determining the safety stock can be improved.

[Second Embodiment]
With the safety stock determination method described in the first embodiment, it is possible to determine the safety stock to be held at each delivery destination in a desired analysis target period for each object.
Here, the probability of occurrence of supply interruption of the object may vary depending on the season. It is known that the amount of supply fluctuates due to seasonal influences such as floods, typhoons, and cold waves because the target is transported from distant foreign countries. For example, in the case of an object shipped from Australia, there is a risk that the shipping of the object will be suspended due to the fragmentation of the onshore supply chain due to a natural disaster such as a flood during the rainy season in Australia in March to March. Therefore, it is necessary to anticipate a supply interruption in the period before 1 to 3 months and secure more safety stock than usual.

The safety stock determination method described in the first embodiment does not determine the safety stock in consideration of such seasonal variation.
In this case, it is conceivable that the analysis target period is subdivided according to the mesh in which the seasonal variation is taken into consideration, and the safety stock is determined for each subdivided period. However, in order to determine the safety stock with high accuracy, a sufficient amount of record information is required even during the subdivided period. In other words, it cannot be applied to objects whose delivery frequency is low. In particular, the objects for which safety stock is calculated by applying the present invention are assumed to be transported from a distant foreign country, and thus are often delivered in large lots and infrequent deliveries. In many cases, the performance information is not collected enough to accurately determine.

In order to solve the above-mentioned problems, the present inventor adds an improvement to the probability distributions of “the arrival intervals of ships to all delivery destinations” and “the arrival intervals of ships to each delivery destination” described in the first embodiment. We tried to model the probability distribution of the arrival intervals of ships to each destination, taking into account seasonal variations.
In order to add a seasonal variation element to the probability distribution of the arrival interval of a ship to each delivery destination, we came up with the idea that a model in which this probability distribution changes with time should be constructed, but the seasonal variation occurrence probability is It is very difficult to quantitatively set seasonal variation for each target, because the factors vary depending on the target and year, and the factors that cause it vary. Furthermore, the number of ship arrival times is small when viewed by destination, so more actual information was required to estimate seasonal fluctuations with high accuracy.

Therefore, paying attention to the arrival of ships at all the delivery destinations, using the arrival time of the ships at all the delivery destinations that have a large amount of actual information, the probability distribution of the arrival intervals of the ships at all the delivery destinations including seasonal fluctuations is calculated. It was thought that the probability distribution of the arrival interval of the ship to each delivery destination should be created using a model and using the modeled probability distribution.
Specifically, by modeling the average arrival rate λ, which has been treated as constant in the first embodiment, as a function λ(s) that changes with time at time s, all the deliveries of the ship can be considered in consideration of seasonal variations. Create a probability distribution of arrival intervals at the destination. Then, similarly to the first embodiment, if the probability distribution of the arrival intervals of the ship to each delivery destination is given by the convolution integral of the probability distribution of the arrival intervals of all the delivery destinations of the ship, seasonal fluctuations are taken into consideration. It was concluded that the arrival intervals of ships to each destination can be modeled.

The arrival process that takes into account the nonuniformity of the average arrival rate λ(s) is called a nonuniform Poisson process. The non-uniform Poisson process is a process in which individual events arrive independently, but the average arrival rate changes with time.
Next, specifically, a flow of determining λ(s) and safety stock when the arrival intervals of all ships to the delivery destination follow the non-uniform Poisson process will be described. However, the method described below is not limited to the case where the ship arrival intervals at all the delivery destinations follow the non-uniform Poisson process. For example, a counting process (nonuniform renewal process) in which the probability distribution of arrival intervals of ships to all delivery destinations follows an independent distribution such as a gamma distribution and the average arrival rate thereof changes with time may be used. The nonuniform renewal process can be treated in the same manner as the nonuniform Poisson process by using the time warping theory shown in Non-Patent Document 2.

The average arrival rate λ(s) has to be determined based on the record information because the shape and the frequency component are different for each object. A method of estimating the average arrival rate λ(s) from the record information will be described below.
First, the average arrival rate λ(s) is modeled by the exponential Fourier series shown in Expression (4). Next, the parameters of this exponential Fourier series are determined based on a statistical information criterion, thereby determining the parameters that are suitable for each object.

In the exponential Fourier series of equation (4), A _i and B _i are variables that give a model shape, and P is a period. In the exponential Fourier series, by setting the period P, variations in various periods can be modeled. For example, if P=1 year, the seasonal variation of one year cycle can be modeled. Also, the variables A _i and B _i that give the model shape can model various frequency components depending on the number of parameters.
In the present embodiment, the parameter of the average arrival rate λ(s) is determined using the Akaike's Information Criterion (AIC) as shown in Expression (5). AIC is one of the evaluation indexes of the model, the first term of the equation (5) represents the log-likelihood function of the model, and the second term represents the number of parameters. The AIC is a criterion in which the number of parameters representing the complexity of the model is added as an evaluation index in addition to the maximum likelihood estimation method that determines the parameters by maximizing the likelihood function. In AIC, a plurality of models are compared with each other, and a model having the best balance between goodness of fit and complexity can be determined.

Here, the likelihood function of the time series {s _i} ⁿ _{i = 1} of Arrival time to all delivery destination according nonuniform Poisson process, the ship arrived event does not occur from time s _i-1 to s _i, time Equation (6) represents the probability that a ship arrival event will occur at s _i .

Assuming that each ship arrival event is independent, the likelihood function can be expressed in the form of a product, as in equation (7).

As described above, the average arrival rate λ(s) shown in equation (4) is substituted into equation (7) to create a likelihood function, and the parameter that maximizes the likelihood function is calculated by the maximum likelihood estimation method. If the obtained logarithmic likelihood and the number of parameters of the model are substituted into the equation (5) to obtain the AIC and the models are compared with each other, the model with the best fit can be determined from a plurality of models.

Hereinafter, a safety stock determination method according to the second embodiment to which the present invention is applied will be described. It should be noted that the description will focus on the differences from the safety stock determination method according to the first embodiment, and a detailed description of common points with the first embodiment will be omitted.
FIG. 9 shows a functional configuration of the safety stock determination device 200 according to the second embodiment. The input unit 201, the delivery delay allowable value setting unit 202, the second probability distribution calculation unit 204, the safety stock calculation unit 205, the output unit 206, the input device 207, and the display 208 are the input unit 101 and the delivery date in the first embodiment. The delay allowable value setting unit 102, the second probability distribution calculation unit 104, the safety stock calculation unit 105, the output unit 106, the input device 107, and the display 108 are the same as those of the first embodiment, and the description thereof will be omitted.
As a difference from the first embodiment, the first probability distribution calculation unit 203 includes an estimation unit 203a that estimates the average arrival rate λ(s).
Further, a cycle setting unit 209 for setting the cycle P is provided. The period P can be appropriately set by the user via the input device 207. Alternatively, default values may be used. From the viewpoint of considering seasonal fluctuations, in most cases, it is considered that P=1 year is set in order to model seasonal fluctuations in one year cycle.

Next, a safety stock determination method by the safety stock determination device 200 will be described.
FIG. 10 is a flowchart showing a safety stock determination process by the safety stock determination device 200 according to the second embodiment.
Steps S11 and S12 are the same as steps S1 and S2 in the first embodiment, and a description thereof will be omitted here.

In step S13, the estimation unit 203a of the first probability distribution calculation unit 203 estimates the average arrival rate λ(s) based on the arrival time information for each delivery destination imported in step S11. In this embodiment, the period P is one year, and the seasonal variation of the maximum one-year period is modeled.
FIG. 11 shows an example of port arrival time information for a certain object for each delivery destination. As described in the first embodiment, the delivery destinations are A, C, A,... In chronological order. However, if the delivery destinations are not identified, the ship will deliver to all delivery destinations in “2015/1. /6 15:00”, “2015/1/12 21:00”, “2015/1/14 6:00”, and so on.

12 and Table 8 show the results of determining the model using the AIC from the arrival times at all the delivery destinations, which are obtained from the arrival time information in FIG.
FIG. 12A is a graph showing the estimation result of the average arrival rate λ(s), and illustrates the estimation result when the number of parameters having the lowest AIC is 5. Further, FIG. 12B shows line segments at the time of arrival at all delivery destinations. As can be seen from FIG. 12, it can be seen that the ship arrival frequency was lowest around May 2015 to July 2015.
Table 8 shows the relationship between the number of parameters in equation (4), the parameter values, and the AIC. As can be seen from Table 8, in the port entry time information of FIG. 11, it can be seen that the AIC is the lowest when the number of parameters is 5. By thus determining the model using the AIC, it was possible to determine the model in consideration of the goodness of fit of the model and the balance of the number of parameters.

Next, in step S14, the first probability distribution calculation unit 203 obtains the probability distribution of the arrival intervals of the ship to all the delivery destinations. Step S14 is the same as step S3 except that the average arrival rate, which has been treated as constant in the first embodiment, is a function λ(s) that changes with time s, and the description thereof is omitted here.

Next, in step S15, the second probability distribution calculation unit 204 obtains the probability distribution of the arrival interval of the ship at each delivery destination. Step S15 is the same as step S4 except that the average arrival rate, which has been treated as constant in the first embodiment, is set to a function λ(s) that changes with time s, and the arrival of the ship at the delivery destination i. If the probability distribution of the interval is g _i (s, t), it is possible to model with a gamma distribution as in Expression (8) by using two parameters of N _i and λ(s).

The number of arrivals q _i to each delivery destination and the parameter N _{i of the} gamma distribution were calculated from the arrival time information in FIG. 11, and the values shown in Tables 9 and 10 were given.

As described above, by determining the average arrival rate λ(s) estimated based on the AIC and the parameter N _{i of the} gamma distribution, the non-uniformity is taken into consideration and the arrival interval of each ship to the delivery destination The probability distribution could be modeled.

As for the validity of the estimation result, it is desirable to carry out a statistical test using a KS test or the like, as in the first embodiment. However, since the probability distribution of Expression (8) is a function that depends on the time s, it is difficult to directly compare it with the arrival time information for each delivery destination as in the first embodiment.
Therefore, the KS test is performed by generating a sufficiently large number of random numbers according to the probability distribution of Expression (8) within the safety stock calculation period and comparing the obtained sample with the arrival time information for each delivery destination. The random number first transforms the probability distribution of equation (8) into a non-uniform Poisson process by the time warping theory described in Non-Patent Document 2. Then, according to the algorithm described in Non-Patent Document 3, random numbers according to the non-uniform Poisson process were generated (diluted algorithm).
FIG. 13 is a graph comparing the arrival intervals of ships to each delivery destination and samples randomly generated by the dilution algorithm. Table 11 shows the results of carrying out a two-sided KS test at a significance level of 5% under the null hypothesis that there is no difference between the actual and model sample representative values for both samples. The model was generated 1685, 746, and 665 times for each of the delivery destinations A, B, and C, and the KS test was performed. As a result, the P value of the KS test exceeded the significance level of 5% at all sites. No hypothesis was rejected. In other words, it cannot be said that there is a difference between the actual results and the representative value of the sample of the model.

Next, in step S16, the safety stock calculator 205 calculates the safety stock that each delivery destination should have.
Under the probability distribution of the arrival interval of the ship to the delivery destination i given by the equation (8), in order to keep the delivery time achievement rate at 1-α% or more, the probability distribution g _i (s, t) is set to 1 The difference between the point covering -α% and the average value of the probability distribution g _i (s, t) may be given as the safety stock days. Let Z(s) [days] be the safety stock days, and let G _i (.) be the cumulative distribution function of the probability distribution g _i (s, t) of the arrival interval of the ship to the delivery destination i, and take the inverse function of the cumulative distribution function. Is set as G ^-1 (.), the safety stock days to keep the delivery rate at 1-α% or more is given by equation (9).

However, as described in the first embodiment, the safety stock days calculated by the expression (9) is a calculation expression that keeps the delivery delay to α% or less, and the stock out will not occur with this probability. .. This is because the safety stock days calculated by the equation (9) are based on the assumption that the amount of stock used is constant. In practice, if it is known in advance that the stock of a certain object is low, the production will be adjusted in advance by anticipating the expected arrival time of the ship and transferring it to other objects in advance so that the stock will not be out of stock. it can.

FIG. 14 shows the result of calculating the safety stock days [days] for each delivery destination with the delivery delay allowable value α% set to 5%. As shown in FIG. 14, by modeling the average arrival rate as a function λ(s) that changes with time at time s, it can be seen that the safety stock days can be calculated according to the season. In particular, as shown in FIG. 12, in the port arrival time information of FIG. 11, it was confirmed that the supply amount decreased from May to July, but as shown in FIG. 14, the above period is larger than usual. Results were obtained with a safety stock days.

The safety stock is obtained by calculating the average usage [t/day] for each delivery destination and multiplying it by the obtained safety stock days [days]. Table 12 shows the calculation results of the average usage [t/day]. Further, Table 13 shows the result of calculating the safety stock [t] by multiplying the safety stock days [days] by the average usage [t/day]. In this way, by modeling the probability distribution of the arrival intervals of ships to all delivery destinations using the non-uniform Poisson process, it was possible to calculate the safety stock in consideration of seasonal fluctuations.

Next, in step S17, the output unit 206 outputs the result of the safety stock calculated in step S16.

By the way, the safety stock obtained in the present embodiment is a value based on the past performance, and it is considered that the safety stock is determined in anticipation of future supply schedule.
Here, how to appear in the seasonal variation of the average arrival rate of the future, not the same as appear the way of seasonal variation of the average arrival rate, which was estimated from past experience λ (s), also unchanged from the past further convolution number N _i of the integration Suppose Then, the future average arrival rate λ future (s) can be calculated by multiplying the average arrival rate λ past (s) determined using the actual information by a constant number so as to match the future supply amount. The predicted value of) is obtained. Specifically, the average arrival rate determined using the actual information is set as λ past average, the average expected arrival rate in the future is set as λ future average, and the average arrival rate estimated using the actual information is set. Assuming λ past (s), the average arrival rate λ future (s), which represents the seasonal variation expected in the future, can be determined by the equation (10).

The average arrival rate is proportional to the cumulative value of the arrival amount of the target object during the analysis target period (hereinafter, the cumulative arrival amount). Therefore, the average value λ future average of the average arrival rate predicted in the future can be modeled by the formula (11) if the future predicted value of the cumulative arrival amount is known. In Expression (11), a and b are constants, and may be estimated from the performance information using single regression analysis or the like.

As described above, even if the future supply conditions are different from those in the past, the safety stock can be determined by providing the prediction model of the probability distribution of the arrival intervals of all ships to the delivery destination in advance. If the number of convolution integrals N _i is also expected to change in the future, first, as shown in equation (12), the λ future average obtained from equation (11) is multiplied by the analysis target period, and all delivery destinations are calculated. Calculate the expected number of ships entering the ship Q'.

Next, as shown in the equation (13), if the expected number of arrivals Q'to all the delivery destinations is multiplied by the ratio of the planned usage amount for each object/delivery destination to obtain N _i , the predicted value of N _i is obtained. Can be asked.

By using the N _i and λ future average calculated as described above, the safety stock can be determined in anticipation of future supply schedules.

The analysis period of the average arrival rate λ past (s) estimated using the past record information may be a single year or a plurality of years. For example, if the practitioner wants to model the average seasonal variation over the past multiple years, the average arrival rate λ past (s) may be estimated for the actual information for multiple years. If you want to set the safety stock to prevent the delivery-time compliance rate above a certain level even in the year with the greatest seasonal impact in the actual results, calculate the average arrival rate λ past (s) for that single-year data. You can estimate it. In this way, by modeling the seasonal variation by changing the analysis target period, it is possible to determine the safety stock quantity that more closely matches the actual requirements.

In this embodiment, the exponential Fourier series is used for the average arrival rate λ(s), but the present invention is not limited to this, and for example, Fourier series or polynomial may be used.
Further, although the AIC is used as the information criterion, the present invention is not limited to this, and it is possible to use other information criterion for determining the model. For example, an information criterion such as Bayes information criterion (BIC) may be used.

Although the present invention has been described along with the embodiments, the above embodiments merely show examples of embodying the present invention, and the technical scope of the present invention is construed in a limited manner by these. It must not be. That is, the present invention can be implemented in various forms without departing from its technical idea or its main features.
The safety stock determination device to which the present invention is applied is realized by, for example, a computer device including a CPU, a ROM, a RAM, and the like. Although the safety stock determination devices 100 and 200 are shown as one device in FIGS. 1 and 9, they may be configured by a plurality of devices, for example.
The present invention also provides software (program) that realizes the functions of the present invention to a system or apparatus via a network or various storage media, and the computer of the system or apparatus reads and executes the program. It is feasible.

100, 200: Safety stock determination device 101, 201: Input unit 102, 202: Delivery delay allowable value setting unit 103, 203: First probability distribution calculation unit 203a: Estimating unit 104, 204: Second probability distribution calculation unit 105, 205: Safety stock calculation unit 106, 206: Output unit 209: Cycle setting unit 300: Database

Claims (15)

A safety stock determination device for determining a safety stock to be held at each delivery destination for an object delivered from one or more shipping sources to a plurality of delivery destinations,
For an object for which a safety stock is desired to be determined, as past performance information, delivery time information, and an input means for capturing usage amount information that is information on the usage amount of the object for each delivery destination,
Based on the delivery time information captured by the input means, it is assumed that a plurality of delivery destinations included in the delivery time information are one delivery destination (hereinafter, referred to as all delivery destinations), and delivery intervals to all the delivery destinations. A first probability distribution calculating means for obtaining a probability distribution of
Based on the usage amount information fetched by the input means and the probability distribution of the delivery intervals to all the delivery destinations obtained by the first probability distribution calculation means, the delivery destination information contained in the delivery time information is delivered to each delivery destination. Second probability distribution calculating means for obtaining a probability distribution of delivery intervals;
Based on the usage amount information taken in by the input means and the probability distribution of the delivery intervals to the delivery destinations obtained by the second probability distribution calculation means, the safety stock to be held at each delivery destination is calculated. A safety stock determination device comprising: safety stock calculation means. The second probability distribution calculation means determines that among the number of deliveries to all the delivery destinations, delivery to the delivery destination i (i is a symbol representing the delivery destination) depends on the usage amount of the target object for each delivery destination. By assuming that the delivery interval is once in N _i times, the delivery interval to the delivery destination i is given by a probability distribution obtained by convoluting the probability distribution of delivery intervals to all the delivery destinations N _i times. The safety stock determination device according to claim 1. Claim wherein the second probability distribution calculating means, delivery times q _i to delivery destination i, as delivery times Q to all delivery destination, characterized by providing a number N _i of the convolution with Q / q _i The safety stock determination device described in 2. Equipped with delivery delay allowable value setting means for setting the delivery delay allowable value α%,
The safety stock calculation means further covers the probability distribution of the delivery intervals to each of the delivery destinations by 1-α% based on the delivery delay tolerance α% set by the delivery delay tolerance setting means, and The difference from the average of the probability distribution of the delivery intervals to each delivery destination is the safety stock days, and this safety stock days is multiplied by the average usage amount for each delivery destination to calculate the safety stock that each delivery destination should have. The safety stock determination device according to any one of claims 1 to 3, characterized in that. 5. The safety according to claim 1, wherein the delivery to all the delivery destinations includes direct delivery from a shipping source and delivery via another delivery destination. Inventory determination device. The first probability distribution calculation means assumes that the probability distribution of delivery intervals to all the delivery destinations follows an exponential distribution or a gamma distribution,
The safety stock determination device according to any one of claims 1 to 5, wherein the second probability distribution calculation means makes the probability distribution of delivery intervals to each delivery destination follow a gamma distribution. The first probability distribution calculation means uses a probability distribution of delivery intervals to all the delivery destinations as an exponential distribution using an average number of deliveries per unit time (hereinafter, referred to as an average delivery rate) to all the delivery destinations. Or expressed by gamma distribution,
The safety stock determination device according to claim 6, wherein the average delivery rate is treated as constant. The first probability distribution calculation means uses a probability distribution of delivery intervals to all the delivery destinations as an exponential distribution using an average number of deliveries per unit time (hereinafter, referred to as an average delivery rate) to all the delivery destinations. Or expressed by gamma distribution,
7. The safety stock determination device according to claim 6, wherein the average delivery rate is modeled as a function that changes with time. The safety stock determination device according to claim 8, wherein the first probability distribution calculation means determines the parameter of the function based on a predetermined information amount criterion. The safety stock determination device according to claim 8 or 9, wherein the function is represented by an exponential Fourier series. 11. The first probability distribution calculating means obtains a future average delivery rate, which is a target period for determining a safety stock, as a constant multiple of a past average delivery rate, according to any one of claims 8 to 10. The safety stock determination device according to item. The first probability distribution calculating means obtains the constant to be the constant multiple as a value obtained by dividing the average of future average delivery rates by the average of past average delivery rates,
The safety stock determination according to claim 11, wherein the average of the future average delivery rate is estimated by a single regression analysis using past record information, as being proportional to the cumulative arrival amount of the object. apparatus. The object is delivered from the shipper to the destination by sea transportation by ship,
Using the arrival time information as the delivery time information,
Using the probability distribution of the arrival interval of the ship to all the delivery destinations as the probability distribution of the delivery interval to all the delivery destinations,
The safety stock determination device according to any one of claims 1 to 12, wherein a probability distribution of a ship arrival interval to each of the delivery destinations is used as a probability distribution of a delivery interval to each of the delivery destinations. A safety stock determination method for determining a safety stock to be held at each delivery destination for objects delivered from one or more shipping sources to a plurality of delivery destinations,
Input means, for the object for which safety stock is desired to be determined, as past performance information, delivery time information, and a step of capturing usage amount information which is information on the usage amount of the object for each delivery destination,
The first probability distribution calculation means assumes that a plurality of delivery destinations included in the delivery time information is one delivery destination based on the delivery time information fetched by the input means (hereinafter, referred to as all delivery destinations). , A step of obtaining a probability distribution of delivery intervals to all the delivery destinations,
The second probability distribution calculation means, based on the usage amount information taken in by the input means, and the probability distribution of the delivery intervals to all the delivery destinations obtained by the first probability distribution calculation means, the delivery time. Determining a probability distribution of delivery intervals to each delivery destination included in the information,
The safety stock calculation means has each of the delivery destinations based on the usage amount information taken in by the input means and the probability distribution of the delivery intervals to the delivery destinations obtained by the second probability distribution calculation means. And a step of calculating a safety stock to be stored. A program for determining a safety stock to be held at each delivery destination for an object delivered from one or more shipping sources to a plurality of delivery destinations,
For an object for which a safety stock is desired to be determined, as past performance information, delivery time information, and an input means for capturing usage amount information that is information on the usage amount of the object for each delivery destination,
Based on the delivery time information captured by the input means, it is assumed that a plurality of delivery destinations included in the delivery time information are one delivery destination (hereinafter, referred to as all delivery destinations), and delivery intervals to all the delivery destinations. A first probability distribution calculating means for obtaining a probability distribution of
Based on the usage amount information fetched by the input means and the probability distribution of the delivery intervals to all the delivery destinations obtained by the first probability distribution calculation means, the delivery destination information contained in the delivery time information is delivered to each delivery destination. Second probability distribution calculating means for obtaining a probability distribution of delivery intervals;
Based on the usage amount information taken in by the input means and the probability distribution of the delivery intervals to the delivery destinations obtained by the second probability distribution calculation means, the safety stock to be held at each delivery destination is calculated. A program that causes a computer to function as a safety stock calculation means.

Images